Nucleolar localization of the yeast RNA exosome subunit Rrp44 hints at early pre-rRNA processing as its main function

The RNA exosome is a multisubunit protein complex involved in RNA surveillance of all classes of RNA, and is essential for pre-rRNA processing. The exosome is conserved throughout evolution, present in archaea and eukaryotes from yeast to humans, where it localizes to the nucleus and cytoplasm. The catalytically active subunit Rrp44/Dis3 of the exosome in budding yeast ( Saccharomyces cerevisiae ) is considered a protein present in these two subcellular compart-ments, and here we report that it not only localizes mainly to the nucleus, but is concentrated in the nucleolus, where the early pre-rRNA processing reactions take place. Moreover, we show by confocal microscopy analysis that the core exosome subunits Rrp41 and Rrp43 also localize largely to the nucleus and strongly accumulate in the nucleolus. These results shown here shed additional light on the localization of the yeast exosome and have implications regarding the main function of this RNase complex, which seems to be primarily in early pre-rRNA processing and surveillance.

Ribosome biogenesis involves the coordinated transcription, surveillance, modification, and processing of precursor rRNAs, which undergo modifications and several exo-and endonucleolytic cleavage reactions during its maturation process (13,14). In the canonical pre-rRNA maturation pathway, the RNA exosome complex is responsible for the degradation of the spacer sequence 5'-ETS after cleavage at A 0 , and for the 3'-5' end processing of 7S pre-rRNA to the mature 5.8S rRNA (Fig. S1A) (8,15). The 5'-ETS is released co-transcriptionally and completely degraded by the exosome, while SSU processome is formed by association of U3 snoRNP and other factors to the pre-rRNA being transcribed (16). Later during pre-rRNA processing, endonucleolytic cleavage at the C2 site in ITS2 of pre-rRNA 27S separates pre-rRNAs 7S (5.8S 1 5' region of ITS2) and 26S (3' region of ITS2 plus 25S) (17), which undergo exonucleolytic processing by the exosome and Rat1/Rai1, respectively, to generate mature rRNAs 5.8S and 25S (17,18). Mtr4 and the RNA exosome are essential for the ITS2 processing of 7S after cleavage at C2 (18), when the exosome subunit Rrp44 shortens 7S to the intermediate 5.8S 1 30, which is then handed over to Rrp6 that trims it to 6S pre-rRNA, which gives rise to the mature 5.8S rRNA after further processing in the cytoplasm (19). Exosome is also involved in quality control steps of rRNA processing, targeting 23S rRNA generated by direct cleavage at the A3 site for degradation, and unprocessed 35 S rRNA ( Fig. S1B) (20)(21)(22). Interestingly, despite the direct involvement of the exosome in 7S pre-rRNA processing and quality control, 60S subunits bearing 7S have been identified in polysomes of exosome cofactors mutants (23).
Although the structure and function of the exosome has been extensively studied in recent years, detailed information on the mechanisms responsible for the subcellular localization of its different forms is still lacking. Despite the identification of some of the yeast exosome subunits in the nucleolus (24), the core subunits of this complex are regarded as mainly present in nucleus and cytoplasm (3,9,24,25). Rrp4 and Rrp6 have been reported to be present in nucleus and nucleolus, in addition to the cytoplasm in the case of Rrp4 (9). We have recently identified the nuclear import mechanisms for Rrp6 and shown that this exclusively nuclear subunit has multiple transport pathways, being bound by the a-importin Srp1 and b-importins Kap95 and Sxm1 (12). To gather more information on the transport of the exosome to the nucleus, we investigated the nuclear import pathway of the other exosome catalytically active subunit, Rrp44, and show here that importins Srp1 (a) and Kap95 (b), respectively, are responsible for this transport, which occurs independently of Rrp6. Furthermore, we show that the exosome complex is highly concentrated in the nucleus, and more specifically in the nucleolus, where pre-rRNA is transcribed and where the early processing reactions take place.

Results
One of the essential functions of the exosome in yeast is the processing of pre-rRNA in the nucleus, which underlines the relevance of uncovering the mechanism of nuclear import that allows this protein complex to enter the nucleus. We have previously shown that one of the exosome active subunits, Rrp6, is transported to the nucleus by the karyopherins Srp1/Kap95 or Sxm1 (12), but information on the other catalytically active exosome subunit, Rrp44/Dis3, nuclear transport is lacking. Rrp44 has five distinct structural domains: an endonuclease PinC N-terminal domain, two CSD cold shock domains, an essential exoribonuclease domain, RNB, and an RNA-binding domain S1 (6). Based on software analyses (see "Experimental procedures"), we identified the presence of three putative nuclear localization signals (NLSs) in the amino acidic sequence of Rrp44 (named here NLS1, NLS2, and NLS3), each overlapping one of the functional domains, PinC, CSD2, and S1, respectively (Fig. 1).
NLS1 is a nuclear localization signal known as the PY-NLS, located in the PinC domain of Rrp44 between residues 172 and 188 ( 172 RAIRKTCQWYSEHLKPY 188 ). A weak consensus motif composed of an N-terminal hydrophobic or basic motif and a Cterminal R/H/KX 2-5 PY recognized by importins Kap104, Sxm1/ Kap108, Kap121, Kap114, Nmd5/Kap119, and Kap95 (26-28).  . Analysis of GFP-Rrp44 relative to DAPI using ImageJ is shown on the right. Cells expressing Rrp44 1-218 show two phenotypes, both depicted here, with this mutant localized to the nucleus of 67% of cells, and present in aggregates in 30% of cells. GFP-Rrp44 relative to DAPI by using ImageJ is shown on the right. Green lines represent GFP and blue lines represent DAPI. Images were acquired and edited separately. Quantification of different phenotypes is shown in Fig. S3.
NLS2 is located in the CSD2 domain between residues 370 and 401 ( 370 RRLLAKDAMIAQRSKKIQPTAKVVYIQRRSWR 401 ) and contains the consensus NLS, R/KX 2 LX n V/YX 2 V/IXK/RX 3 K/R, recognized by the importins Kap114, Kap95, Kap123, Pse1, and Kap104 (29). NLS3 is located at the C terminus of Rrp44 between residues 988 and 1001 ( 988 DPITSKRKAELLLK 1001 ). The prediction of this NLS is based on its similarity to the Drosophila Rrp44 (dDis3), which is recognized by importin a3 (30,31). These findings suggest that Rrp44 could be transported to the nucleus independently of other exosome subunits.

NLS1 is essential for nuclear import
To explore the importance of each of Rrp44 NLSs for its nuclear import, we constructed plasmids for episomal expression  (Fig. 1). Rrp44 variants were fused to GFP at the N terminus and expressed under control of the MET25 promoter, strongly induced in the absence of methionine. Analysis of the subcellular localization by fluorescence microscopy shows that despite being considered a nuclear and cytoplasmic protein (5,32), full-length GFP-Rrp44 was found mainly concentrated in the nucleus ( Fig. 2A), in agreement with high-throughput analyses (24).
Analysis of the subcellular localization of Rrp44 deletion and point mutants showed a variety of phenotypes depending on the presence of any of the putative NLS described above. Rrp44 mutants containing NLS1 in their sequence localized exclusively to the nucleus (Rrp44 1-985 , Rrp44 1-475 , Rrp44 R397A,R398T , Rrp44 P187A,Y188A ) or concentrated in the nucleus, but were also present in the cytoplasm (Rrp44 D364-407 , Rrp44 (1-363)(408-985) , and Rrp44 1-218 ) (Fig. 2). Mutant Rrp44 1-218 displayed two distinct phenotypes, some of the cells showed the protein concentrated in the nucleus, whereas others showed small cytoplasmic foci, which could correspond to degradation sites (Fig.  2B, Fig. S2). Despite the point mutations in PY-NLS, mutant Rrp44 P187A,Y188A still concentrates in the nucleus, indicating that the remaining sequence suffices for importin recognition. Mutants lacking NLS1, although present in the nucleus, showed strong signals in the cytoplasm (Rrp44 236-475 , Rrp44 475-1001 ). Mutant Rrp44 236-475 , which contains only NLS2, showed the same phenotype as GFP alone, whereas Rrp44 475-1001 was present in cytoplasm, but concentrated in the nucleus in some cells (Fig. 2B).
Mutants lacking NLS2, Rrp44 D364-407 , and Rrp44 (1-363)(408-985) localized to the nucleus, which could suggest that NLS2 is not important for Rrp44 nuclear import. However, despite being concentrated in the nucleus, these mutants showed detectable signals in the cytoplasm ( Fig. 2A), leading to the suggestion that in the absence of NLS2, Rrp44 is still transported to the nucleus but not very efficiently. Absence of the C-terminal NLS3, on the other hand, does not affect Rrp44 nuclear localization, because both mutants, Rrp44 1-985 and Rrp44 1-475 , which lack NLS3, localized exclusively to the nucleus, and Rrp44 (1-363)(408-985) , which lacks NLS2 and NLS3, was concentrated in the nucleus ( Fig. 2A).
Based on these results showing that deletion mutants containing NLS1 localize to the nucleus, whereas mutants lacking NLS1 are present in nucleus and cytoplasm (Fig. 2, Fig. S3), we can conclude that the presence of NLS1 plus either NLS2 or NLS3 is sufficient for Rrp44 nuclear localization. Hence, NLS1 would be the most important signal for Rrp44 nuclear import.
The expression of these deletion and point mutants of Rrp44 was analyzed by Western blotting of total cell extracts, which shows that most of the mutants are expressed at similar levels in yeast, and the protein bands detected have the expected molecular masses (Fig. 3). The ability of these mutants to complement growth of Drrp44/GAL::RRP44 cells was tested by expressing the GFP-fused Rrp44 mutants in this conditional strain, growing on galactose or glucose media (expression or depletion of Rrp44, respectively). Toxicity of GFP-Rrp44 variants was evaluated with the expression of WT Rrp44 (galactose) in the presence or absence of methionine for lowering or increasing the expression of the variants, respectively. The results show that point mutants Rrp44 R397A,R398T (NLS2 mutant) and Rrp44 P187A,Y188A (NLS1 mutant), which localize to the nucleus, fully complement growth of Drrp44/GAL::RRP44 on glucose, similar to GFP-Rrp44 (Fig. 4). The deletion mutant Rrp44 1-985 , which lacks NLS3, but localizes to the nucleus, also complements growth. These results reinforce those presented above, that the Rrp44 mutants transported to the nucleus are functional. Confirming the results of protein localization, mutants Rrp44 D364-407 and Rrp44 (1-363)(408-985) partially complement growth of Drrp44/GAL::RRP44 on glucose. Despite being transported to the nucleus, mutants with further deletions in Rrp44 sequence, resulted in nonfunctional proteins (Fig. 4). Interestingly, co-expression of GFP-Rrp44 at high levels (Gal) and functional Rrp44 mutants (-Met) led to slow growth, contrary to what was seen on glucose (Fig. 4). Addition of methionine to the medium alleviates this effect, increasing growth. These results suggest that high levels of Rrp44, combined with high levels of GFP-Rrp44, are deleterious to the cells. Mutant Rrp44 364-1001 , which showed dominant-negative effects, strongly inhibiting growth in all conditions, and displayed various localization phenotypes ( Fig. 4 and data not shown), and mutant Rrp44 D526-987 , which was not detected by Western blotting, were therefore not further analyzed.
These growth complementation results are consistent with the analysis of pre-rRNA processing on glucose, when GFPtagged Rrp44 variants are expressed in the absence of WT Rrp44. As a control, rRNA processing was analyzed in the presence of WT Rrp44 (galactose). In the absence of Rrp44, 7S rRNA accumulates, without accumulation of detectable 5.8S 1 30 species. In the absence of Rrp6, 7S processing by Rrp44 leads to strong accumulation of 5.8S 1 30 rRNA (detected with both probes P1 and P2). Point mutants Rrp44 P187A,Y188A and Rrp44 R397A,R398T fully complement the depletion of WT Rrp44, visualized by the efficient processing of 7S pre-rRNA and formation of mature 5.8S rRNA ( Fig. 5 and Fig. S9). Expression of the mutants Rrp44 1-218 (PIN domain), Rrp44 236-475 (CSD1 1 CSD2), or Rrp44 1-475 (PIN 1 CSD1 1 CSD2) leads to processing defects, with accumulation of pre-rRNA 7S and consequent decreased levels of mature 5.8S rRNA. Interestingly, mutant Rrp44 1-475 accumulates at low level and intermediate very similar to 5.8S 1 30, which is a characteristic of strains lacking Rrp6 (Fig. 5). Note that such rRNA species are better detected with . Analysis of growth of conditional strain Drrp44/GAL::RRP44 expressing Rrp44 mutants. Conditional strain was transformed with plasmids containing the Rrp44 mutants under control of MET25 promoter, and incubated in galactose (expression of WT Rrp44) or glucose (expression of Rrp44 mutants) medium, in the presence or absence of methionine (lower and higher levels of the mutants, respectively). Mutants Rrp44 D364-407 and Rrp44 (1-363)(408-985) partially complement growth in glucose, whereas Rrp44 1-985 , Rrp44 R397A,R398A , and Rrp44 P187A,Y188A fully complement growth in glucose, although overexpression of the latter inhibits growth in galactose.
P2, which hybridizes to the ITS2 region. In addition to 5.8S 1 30, mutants Rrp44 D364-407 and Rrp44 (1-363)(408-985) also accumulate intermediates longer than 5.8S 1 30 (Fig. 5). Mutant Rrp44 475-1001 (RNB 1 S1 domains), on the other hand, shows lower levels of pre-rRNA 7S and mature 5.8S (Fig. 5). These results suggest that Rrp44 475-1001 may not associate stably with Figure 5. Analysis pre-rRNA processing in Rrp44 mutants. Northern hybridization of total RNA extracted from Drrp44/GAL::RRP44/MET25::GFP 2 strains growing either in galactose (1 WT Rrp44) or glucose (2WT Rrp44) medium, and separated by electrophoresis on denaturing polyacrylamide gels. Precursor and mature rRNAs detected with the different probes are indicated on the right. 5S rRNA and scR1 were used as controls of RNAs not processed by the exosome. A representative of three independent experiments is shown. B, quantification of the bands detected by Northern hybridization. Signals of pre-rRNA 7S were corrected for scR1 signals and the plot shows the levels or 7S in the mutants relative to Rrp44 growing in glucose. C, schematic representation of the yeast 35 S pre-rRNA indicating the hybridizing positions of the probes. the exosome core, decreasing the efficiency of 7S processing, but may be free to degrade RNAs, leading to the decreased levels of both precursor and mature rRNAs. An overview of the phenotypes of the Rrp44 mutants relative to WT is shown in Fig.  S5.

Rrp44 is transported to the nucleus by karyopherins Kap95 and Srp1
To better understand the nuclear import pathway of Rrp44 and identify the karyopherins involved in this process, GFP-Rrp44 was expressed in strains with mutated karyopherin genes, chosen based on the putative NLSs present in the Rrp44 sequence. Deletion mutants of the nonessential karyopherins Sxm1, Kap123, and Kap114 were transformed with a plasmid coding for GFP-Rrp44 and subcellular localization of this protein was analyzed by fluorescence and confocal microscopy. The results show that none of these karyopherins affects Rrp44 localization ( Fig. S6 and data not shown). Lowering levels of Srp1 or Kap95 by incubating strains Dsrp1/GAL::SRP1 and Dkap95/GAL::KAP95 on glucose medium, on the other hand, strongly affects GFP-Rrp44 localization, leading to partial mislocalization of this protein to the cytoplasm, despite still being concentrated in the nucleus (Fig. 6). These results indicate that Kap95 and Srp1 are the main karyopherins involved in nuclear import of Rrp44.
Because Srp1 and Kap95 affected Rrp44 localization, physical interaction between these proteins was analyzed by co-immunoprecipitation after expression of GFP-Rrp44 and protein Atagged karyopherins in yeast. The results show that both A-Srp1 and A-Kap95 co-immunoprecipitate GFP-Rrp44 (Fig.  7A), confirming the interaction between these proteins and the involvement of Srp1 and Kap95 in the nuclear transport of Rrp44. Because the GFP tag alone was also detectable in the elution fractions, to exclude that interaction was due to GFP, an additional experiment was performed, in which extracts from yeast cells expressing GFP or GFP-Rrp44 were incubated with IgG-Sepharose without previous incubation with the karyopherins. After extensive washing, eluted proteins were analyzed by Western blotting, which shows that GFP binds unspecifically the resin with much higher affinity than GFP-Rrp44 (Fig. S4B), accounting for the background binding detected in the co-immunoprecipitation experiments (Fig. 7A). These results are in agreement with global protein interaction data that indicated protein complexes containing both Rrp44 and Srp1 (33,34).
To determine whether Srp1 and Kap95 were capable of interacting directly with Rrp44 in vitro, protein pulldown assays were performed. Recombinant GST-Rrp44 was immobilized on GSH-Sepharose beads and then incubated with His-Kap95 or His-Srp1. Fractions were separated by SDS-PAGE and subjected to Western blotting with antibodies against the His and GST tags. The results show the direct interaction between GST-Rrp44 and His-Kap95 (Fig. 7B). Although His-Kap95 is also pulled down with GST, it binds GST-Rrp44 more efficiently, as indicated by the intensity of the bands detected in elution fractions. His-Srp1, however, was not detected in the elution (Fig. 7B, lower panel). These results suggest that Kap95 binds Rrp44 NLS, being responsible for its nuclear transport. Srp1 may indirectly interact with Rrp44, depending on bimportins such as Kap95 to recognize this cargo protein. Combined, the results shown here strongly indicate that Rrp44 is transported to the nucleus by a Kap95-dependent import pathway, either by itself, complexed with Srp1, or with another b-karyopherin.

Nuclear transport of the exosome complex
Although Rrp6 has been shown to have multiple NLS and transport pathways (12), it is generally not considered to be responsible for the nuclear import of other exosome subunits because it is not an essential yeast protein. The lack of effect of Rrp6 on nuclear transport of core exosome subunits was confirmed here upon analysis of subcellular localization of Rrp44, Rrp41, and Rrp43 in Drrp6 cells (Fig. S7). Upon identifying NLS in Rrp44, we analyzed whether this essential exosome subunit influences the nuclear import of other exosome subunits. The episomal GFP-fused exosome subunits Rrp6, Rrp41, and Rrp43 were expressed in Drrp44/GAL::RRP44 either in the presence (Gal) or upon depletion of Rrp44 (Glu). Contrary to Rrp6 and Rrp44, Rrp41 and Rrp43, expressed under the strong MET25 promoter, are present in both nucleus and cytoplasm. Depletion of Rrp44 does not significantly affect the localization of these exosome subunits (Fig. 8), suggesting independent nuclear transport pathways. The nonaffected GFP-Rrp6 localization in the absence of Rrp44 was expected given the already described nuclear import pathways of Rrp6 (12). Control experiments in WT cells confirm that when overexpressed using episomal genes under control of MET25 promoter, Rrp41 and Rrp43 localize to the nucleus and cytoplasm (Fig.  S8). Overexpressed Rrp44 and Rrp6, on the other hand, show nuclear localization (Fig. S8).

Exosome subcellular localization
Curiously, as shown here, Rrp44 is mainly nuclear, whereas the exosome RNase PH ring subunits Rrp41 and Rrp43 show a nuclear and cytoplasmic localization. To determine whether this phenotype is due to the overexpression of the exosome subunits coded in multicopy plasmid under control of the strong MET25 promoter, chromosomal C terminally tagged GFP-fused exosome subunits genes, under control of their endogenous promoters, were constructed for Rrp44, Rrp6, Rrp41, and Rrp43. No growth defect was observed, showing that C terminally tagged proteins expressed at endogenous levels remain functional. Although the subunits of Exo10 are considered to be present both in nucleus and cytoplasm (9), the results shown here demonstrate that core subunits Rrp41 and Rrp43, and the catalytically active subunits Rrp44 and Rrp6 are all concentrated in the nucleus (Fig. 9). These proteins expressed at endogenous levels are visible almost exclusively in the nucleus, and more strikingly, are concentrated in the nucleolus (Fig. 9). Importantly, analyses using confocal microscopy and quantification of the sum of fluorescent signal, followed by projection along the z axis of all acquired confocal images and calculation of enrichment of the exosome signal in each cellular compartment (see "Experimental procedures"), corroborates the nucleolar localization of the exosome. Exosome concentration in the nucleoplasm is 6 to 9 times higher than in the cytoplasm, whereas nucleolar exosome concentration is 10 to 20 times higher than in the cytoplasm (Fig. 9). Quantification of the signal of these exosome subunits in each of the subcellular compartments (nucleolus, nucleus and cytoplasm) relative to the total signal in the cells shows that 70-80% of the exosome signal is in the nucleus (Fig. S10).
Careful inspection of individual z sections of the acquired fluorescent images, however, shows that despite being concentrated in the nucleolus, exosome subunits localization is slightly different from that of RNA polymerase I (Rpa190), as if in a different subnucleolar compartment (Fig. 10). These results strongly suggest that in the nucleolus, nascent rRNAs are exposed to a high concentration of exosome in a "processing compartment" adjacent, but distinct from RNA polymerase I, for processing and quality control of pre-rRNAs.

Discussion
The RNA exosome was first identified in S. cerevisiae as an RNase involved in maturation and quality control of stable RNAs (8,15). In the subsequent studies, it became clear that the exosome is a protein complex conserved throughout evolution, which is present in the nucleus and cytoplasm of eukaryotic cells, where it interacts with many cofactors and participates in different RNA processing and degradation pathways (3,35,36).
In yeast, Exo10, composed of the exosome core and Rrp44, is considered to be present both in nucleus and cytoplasm, participating in different reactions in each of these subcellular compartments (32). We have previously identified the nuclear import pathways of Rrp6 and shown that this nuclear exosome subunit has redundant mechanisms of transport to the nucleus, directly interacting with a-importin Srp1 and b-importins Kap95 and Sxm1 (12). To better understand the mechanism of exosome assembly and transport to the nucleus, where this complex participates in the essential process of ribosome maturation, we investigated here the nuclear import pathway of the other catalytically active exosome subunit, Rrp44.
We identified three putative NLSs in the primary sequence of Rrp44 and constructed deletion and point mutants to determine the Rrp44 sequences mediating its transport to the nucleus. We show evidence that NLS1 is the most important sequence, but not the only NLS, for the nuclear import of Rrp44, which suggests that different karyopherins may bind Rrp44 for directing it to the nucleus. Depletion of Kap95 strongly affects the nuclear localization of Rrp44, and accordingly, NLS1 contains the consensus sequence for Kap95 binding, a PY-NLS (37). However, point mutant Rrp44 P187A,Y188A still localizes to the nucleus, suggesting that additional signals are recognized by Kap95. Although Srp1 also affects Rrp44 transport to the nucleus, a classical NLS was not found in the Rrp44 sequence, suggesting that Srp1 could recognize a nonconsensus NLS, or that the transport is mediated by the Kap95/Srp1 heterodimer upon recognition of the NLS by Kap95, or by an additional b-karyopherin.
Rrp44 has endo-and exonucleolytic activities (5,38), and is involved in pre-rRNA processing in nucleus and nucleolus (8). The mutants that contain the exonuclease domain of Rrp44 and at least two of the putative NLSs, localize to the nucleus and complement growth of the Drrp44/GAL::A-RRP44 strain. Analysis of pre-rRNA processing in the mutants show that the mutants that do not localize to the nucleus, or that do not contain the RNB domain, display a defective pre-rRNA processing phenotype, with accumulation of intermediates with sizes in the range between pre-rRNAs 7S and 5.8S 1 30. Importantly, these intermediates are not substrates for Rrp6, which process the product of the Rrp44 reaction, 5.8S 1 30 (7). The intermediates accumulating in the presence of nonfunctional Rrp44 mutants have longer extensions at the 3' end than 5.8S 1 30, and have also been observed in mutants of core exosome subunits (8). Interestingly, mutant Rrp44 475-1001 (containing RNB 1 S1 domains, but lacking PIN and CSD domains) shows decreased levels of both precursor and mature rRNAs. These results are in agreement with Rrp44 structural data showing that PIN and CSD domains are responsible for the stable interaction between Rrp44 and the RNase PH ring of the exosome (39). Rrp44 lacking these domains may not associate stably with the exosome, decreasing the efficient processing of 7S, but instead, may be free and more active to degrade RNAs (40,41).
As shown here, depletion of Rrp44 does not affect nuclear localization of Rrp6, as expected, given that Rrp6 has its own NLSs (12), and does not directly interact with Rrp44 (39, 41). Interestingly, depletion of Rrp44 does not affect localization of episomal core exosome subunits either, suggesting a different mechanism of transport for the remaining subunits of this complex.
The most striking results coming from the analyses described here are the very low concentration in the cytoplasm of all the exosome subunits analyzed. Exosome is not only concentrated in the nucleus, but specifically in the nucleolus. This information is relevant in light of the function of the yeast exosome. In the cytoplasm, the exosome participates in the minor pathway of mRNA decay, degrading 3'-5' deadenylated mRNAs (42)(43)(44). The major pathway of cytoplasmic mRNA degradation in yeast does not involve the exosome, and starts with deadenylation of mRNAs by the Ccr4-Not complex (42,45,46), followed by decapping by Dcp1 (47, 48) and 5'-3' degradation by Xrn1 (49,50). Pre-rRNA processing and surveillance, on the other hand, starts in the nucleolus and continues to the nucleoplasm, as the pre-ribosomal particles are concomitantly transported toward the cytoplasm (51).
As soon as 5'-ETS is released after co-transcriptional cleavage of pre-rRNA at sites A 0 -A 1 , the exosome can degrade it in the nucleolus, after being recruited there by SSU processome factors (51)(52)(53)(54). The results shown here reinforces the importance of the exosome for pre-rRNA processing and quality control, by showing that the exosome is mainly concentrated in the nucleolus. The concentration of the exosome in the nucleolus also corroborate recent data showing the interaction of exosome subunits with pre-ribosomal particles 90S, pre-40S, and pre-60S, interactions that were stabilized by the inhibition of pre-rRNA processing by the depletion of pre-60S factor Nop53 (55).
As we also show here, the nucleolar concentration of the exosome subunits can only be appreciated when the GFP-fused proteins are expressed at endogenous levels. Overexpression of these proteins in plasmid-based systems lead to the visualization of weak signals in the cytoplasm and very strong nuclear signals, overshadowing the nucleolus.
Not only were the exosome subunits concentrated in the nucleolus, but careful high-resolution analysis of protein localization showed that when compared to RNA polymerase I subunit Rpa190 on individual z sections of spinning disk confocal microscopy, the exosome subunits do not exactly overlap Rpa190, suggesting a slightly different localization. Interestingly, recent data on mammalian cells show that the box C/D snoRNP subunit fibrillarin, which is also part of the SSU processome, does not exactly colocalize with RPA194, but rather shows a phase separation between rRNA transcription and processing (56). Our results could suggest that transcribed rRNA is co-transcriptionally processed in an associated compartment containing the exosome.
In summary, here we show that the exosome catalytic subunit Rrp44 is transported to the nucleus independently of other subunits. Importantly, we show that the yeast exosome is concentrated in the nucleolus, in a subcompartment flanking but different from RNA polymerase I, placing the exosome in the early phases of pre-rRNA processing.

Experimental procedures Construction of strains and yeast growth condition
Yeast maintenance and growth was performed in YPD medium (1% yeast extract, 2% peptone, and 2% glucose), or YNB medium (0.67% yeast nitrogen base, 0.5% (NH 4 ) 2 SO 4 , and 2% glucose or galactose) supplemented with the required amino acids. Plasmids constructed in this study, described in Table  1, were built according to standard cloning techniques and sequenced by the Big Dye method (PerkinElmer Life Sciences). Plasmids expressing the GFP fusions in yeast were constructed by inserting DNA fragments into pUG34 plasmid (57) using oligonucleotides with specific restriction sites Figure 8. Depletion of Rrp44 does not affect subcellular localization of core exosome subunits Rrp41 and Rrp43. Drrp44/GAL::RRP44 was transformed with plasmids expressing GFP-fused exosome subunits Rrp44, Rrp6, Rrp41, and Rrp43 and incubated in media containing either galactose (Rrp44 expression) or glucose (Rrp44 repression). All four exosome subunits are concentrated in the nucleus, although Rrp41 and Rrp43 are also visible in the cytoplasm. Images were acquired and edited separately. GFP-fused proteins relative to DAPI by using ImageJ is shown on the right. Green lines represent GFP and blue lines represent DAPI.
(sequences available upon request). Expression of these GFP fusions was regulated by MET25 promoter. Rrp44 point mutations were obtained by site-directed mutagenesis (Stratagene or In-Fusion-Takara) using plasmid pUG34-RRP44 as template and oligonucleotides containing the respective mutations. Plasmid pRS305-NUP57-tDimerRFP (58) was constructed as follows. NUP57 was amplified as two overlapping PCR fragments (primer pairs 1568/1570 and 1569/1571matrix strain BY4741). The two PCR fragments were mixed and used as matrix with primers 1568/1571 to amplify fulllength NUP57 with an internal BglII site. The NUP57-containing fragment was cloned in pRS305-NUP2-tDimerRFP (59) as a NotI-BamHI fragment. To construct a Nup57 t-dimer genomic-tagged strain, plasmid pRS305-Nup57-tDimer was linearized with BglII and inserted by homologous recombination in a BY4742 strain. It was selected for leucine prototrophy, and checked by fluorescence microscopy. Plasmid pFA6-mCherry-KlURA3 was constructed by cloning into pFA6-GST-KIURA3 PacI-Asc1 mCherry a fragment from pFA6-mCherry-HIS3.
Haploid strain Drrp44/YCplac33-GAL::A-RRP44 was obtained after sporulation of the diploid strain RRP44/Drrp44 previously transformed with plasmid YCplac33-GAL::A-RRP44 (Table 2). Strains bearing genomic insertion of GFP or mCherry were constructed by amplifying a PCR cassette containing the URA3 gene from Kluyveromyces lactis as a selectable marker and GFP tag or mCherry tag sequence from plasmids pFA6-GFP(S65T)-KIURA3 or pFA6a-mCherry-KIURA3, respectively. The PCR fragment was inserted by homologous recombination downstream from each protein gene bearing a TAP-tag (60) using the Swap-tag method (61). Transformants were selected for uracil prototrophy and checked by PCR of the targeted Figure 9. Endogenously expressed GFP-fused exosome subunits localize to the nucleolus. A, Z-projection of high resolution spinning-disk laser scanning confocal microscope images show the subcellular localization of GFP-tagged exosome subunits. Rpa190-mCherry was used as nucleolar marker. B, exosome subunits enrichment in different cell compartments. We calculated the enrichment of the GFP-tagged exosome subunits signals in each compartment over an homogeneous distribution of the cellular signal. For that, we calculated the mean intensity of each compartment and divided this value by the computed mean intensity (total cell signal/total cell surface). Blue dots represent cytoplasm enrichment, red dots represent nucleoplasm, and green dots represent nucleolus. Figure 10. Exosome subunits do not co-localize exactly with nucleolar RNA polymerase I subunit Rpa190. GFP-tagged images obtained by Z section of spinning disk confocal microscope shows that exosome subunits have slightly different localization from Rpa190-mCherry, used as nucleolar marker. Note that single Z step images are different from Z-projected images, which shows the sum of all detected signal of cells.
genomic loci, and analyzed by fluorescence microscopy. BY4742 strains bearing Rrp44 fused to GFP or mCherry were constructed as described (62). Similarly, Rpa190 fused to mCherry was constructed as described (62).
RPA190 was also genomic tagged with mCherry in exosome GFP-tagged strains and BY4742 strain. The same strategy was used by first amplifying the plasmid pFA6-mCherry-HIS3, then inserting it in the genome by homologous recombination and selecting for histidine prototrophy.

Fluorescence microscopy on fixed cells
For protein depletion based on GAL1 promoter, cells were inoculated either in selective galactose or glucose medium and incubated for 14 to 16 h to exponential phase. For microscopy analysis, cells were fixed in 70% methanol for 15 min, rinsed with cold PBS, and then treated with 1 mg/ml of RNase for 30 min. Cells were observed using a Nikon Eclipse Ti microscope equipped with filters for green fluorescence (GFP-3035B-000-ZERO, Semrock) and red fluorescence (Texas Red BrightLine set, TXRED4040-B, Semrock). The exposure times varied from 1 to 3 s. Images were processed and analyzed using the programs Nis Elements (version 3.07; Nikon) and ImageJ (National Institutes of Health, Bethesda, MD). Confocal images were captured in a 1024 3 1024-pixel format using a Zeiss LSM 780 confocal laser-scanning inverted microscope (Carl Zeiss, Germany) at the Research Facility Center (CEFAP-USP). Image stacks comprised eight images captured with a aPlan-Apochromat 3100/1.46 oil differential interference contrast M27 objective (Carl Zeiss), applying a zoom factor of 1.5.
Step intervals along the z axis ranged from 200 to 250 nm. Image processing was performed using Zen 2011 software (version 11.00.190; Carl Zeiss).

Fluorescence microscopy on living cells
Exponentially growing yeast were collected, resuspended in synthetic complete medium (DIFCO), and observed in the fluorescence microscope. Confocal microscopy was performed using a Nipkow-disk confocal system (Revolution; Andor) installed on an Olympus microscope (IX-81), featuring a confocal spinning disk unit (CSU22; Yokogawa) and a cooled electron multiplying charge-coupled device camera (DU 888; Andor). The system was controlled using IQ 2 software (Andor). Images were acquired using a 3100 Plan Apo 1.4 NA oil immersion objective and a 2-fold lens in the optical path. Single laser lines used for excitation were diode-pumped solid-state lasers exciting GFP fluorescence at 488 nm (50 milliwatt; Coherent) and mCherry fluorescence at 561 nm (50 milliwatt; Cobolt jive), and a Dichroic mirror Di01-T405/488/568/647-13x15x0,5 was used; Semrock was used and a bi-bandpass emission filter (FF01-512/ 630-25; Semrock) allowed collection of the green and red fluorescence. In our conditions, pixel size was 65 nm. For quantification of nucleolar volume, z stacks of 40 images with a 200-nm z step were used. Exposure times varied from 0.1 to 1 s. Digital pictures were processed using ImageJ (National Institutes of Health, Bethesda, MD) (RRID:SCR_003070).

Image analysis and quantification
Confocal images were imported into ImageJ, signal intensities of GFP-tagged proteins were measured and their subnuclear localizations were analyzed using a dedicated image analysis pipeline (ImageJ Macro). Cell area was determined based on transmission light thanks to the following steps: background removal (subtracting Gaussian Blur image -s = 40); segmentation of the cells with Moments threshold algorithm; determina-tion of the objects corresponding to the cells with Analyze Particle (ImageJ function); and separation of the cells in contact with Adjustable Watershed Plugin. Nucleolar and nuclear segmentation was achieved using the largest RNA polymerase I subunit Rpa190 tagged with mCherry, strongly enriched in the nucleolus, and detectable in the nucleoplasm. First, we applied a Sum Z-Projection to the initial mCherry Image. Then, to determine the nucleolus area in each cell, we used a Wavelet filter and finally segmented with a Yen threshold algorithm. To determine the nucleus area, we directly segmented on the Z-projected image a Triangle threshold algorithm. Cells were then divided in the nucleolus, nucleoplasm, and cytoplasm. ImageJ macro was used to quantify the GFP signal in each compartment. All quantifications were performed relative to background fluorescence, measured using control cells with Rpa190 tagged with mCherry, but with no GFP expression.
To quantify exosome signal enrichment in each compartment, the mean signal of the entire cell was calculated: ((intensity of gray level in the cellbackground)/cell surface area). The mean cellular signal was used as a proxy to a homogeneous concentration in the entire cell. Each compartment mean signal measure was then divided by the mean cellular signal in each cell. The concentration ratio shows the enrichment of exosome signal per compartment, all relative to the mean signal of the entire cell.

Protein pulldown
Cellular extracts (generated in 20 mM Tris, 150 mM NaCl, 1 mM EDTA, 0.8% Nonidet, 1 mM DTT) of Escherichia coli cells expressing either GST or GST-Rrp44 were incubated for 2 h at 4°C with 60 ml of GSH-Sepharose beads (GE Healthcare), the unbound material was collected and the beads washed with the same buffer. Beads were then incubated with cellular extracts containing His-Srp1 or His-Kap95, flow-through was collected, and beads were washed with the same buffer. Bound proteins were eluted with 50 mM Tris, pH 8.0, 10 mM reduced GSH.

Co-immunoprecipitation
Interaction of karyopherins Srp1 and kap95 with Rrp44 was tested by co-immunoprecipitation using strains Dkap95/GAL:: A-KAP95 and Dsrp1/GAL::A-SRP1 transformed with pUG34 or pUG34-Rrp44. For each strain, 2 liters of cells were grown to an OD600 ;1.0, and collected in co-IP buffer (0.1% Nonidet P-40, 150 mM NaCl, 5 mM EDTA, 50 mM Tris-HCl, pH 7.5, 0.53 Halt Protease Inhibitor mixture (Thermo Scientific) (modified from Ref. 64). Total yeast extracts were obtained with a Ball Mill device (Retsch, Mixer Mill MM 200 or Mixer Mill PM 100), and cleared by centrifugation at 40,000 rpm for 1 h at 4°C. Supernatant was then incubated for 2 h at 4°C with IgG-Sepharose 6 Fast Flow (GE Healthcare) previously equilibrated with co-IP buffer. The resin was washed four times with 500 ml of co-IP buffer (twice with 150 mM NaCl and twice with 250 mM NaCl), followed by a further wash with 100 mM ammonia acetate, 0.1 M magnesium chloride. Elution of the specific karyopherin was performed by incubation with 500 mM ammonium hydroxide for 20 min (65).

Northern hybridization
Total RNA was isolated from yeast cells by a modified hot phenol method (66). The RNA extract was quantified and 6 mg of denatured total RNA was loaded on gel. RNAs were separated by electrophoresis on 7% denaturing polyacrylamide gel, and transferred to Hybond nylon membrane (GE Healthcare). Hybridization was performed using biotin-labeled or fluorescent probes specific to rRNAs. 5S rRNA and scR1 were used as controls. Quantification of bands from Northern hybridizations was performed with ImageJ software.

Data availability
All data are contained within the manuscript.
Acknowledgments-We are grateful to all members of the Oliveira laboratory for help, reagents, and discussion, especially L. P. P. Cepeda. We thank Frederico Gueiros Filho and members of his laboratory for the use of the fluorescence microscope. We also thank Mario Costa Cruz from CEFAP-USP for help with confocal microscopy and discussion. We are very grateful to Sylvain Cantaloube from the LITC imaging platform of Toulouse TRI and Image processing platform for his assistance and development of the image analysis pipeline. We acknowledge members of the Gadal's lab for help, advice and discussion, especially Isabelle Leger-Silvestre, Christophe Dez, and Lise Dauban. Conflict of interest-The authors declare that they have no conflicts of interest with the contents of this article.