TDI-011536

Phosphorylation and nuclear transit modulate the balance between normal function and terminal aggregation of the yeast RNA-binding protein Ssd1

Abstract
Yeast Ssd1 is an RNA-binding protein that shuttles between the nucleus and cytoplasm. Ssd1 interacts with its target mRNAs initially during transcription by binding through its N-terminal prion-like domain (PLD) to the C-terminal domain of RNA polymerase II. Ssd1 subsequently targets mRNAs acquired in the nucleus either to daughter cells for translation or to stress granules (SG) and P-bodies (PB) for mRNA storage or decay. Here we show that PB components assist in the nuclear export of Ssd1and subsequent targeting of Ssd1 to PB sites in the cytoplasm. In the absence of import into the nucleus, Ssd1 fails to associate with P-bodies in the cytoplasm but rather is targeted to cytosolic insoluble protein deposits (IPOD). The association of Ssd1 either with IPOD sites or with PB/SG requires the PLD, whose activity is differentially regulated by the Ndr/LATS family kinase, Cbk1: phosphorylation suppresses PB/SG association but enhances IPOD formation. This regulation likely accrues from a phosphorylation sensitive nuclear localization sequence located in the PLD. The results presented here may inform our understanding of aggregate formation by RNA-binding proteins in certain neurological diseases.

Introduction
Many proteins in the eukaryotic cell are targeted to specific subcellular compartments and this targeting is often critical for the execution of the proteins’ functions and for survival of the cell. Organelles, such as mitochondria, Golgi, secretory vesicles, etc., are separated from the cytoplasm by a closed intracellular membrane. However, other distinct cellular structures are not bounded by membranes but comprise specific cytoplasmic complexes of discrete proteins and RNAs. The best studied of these are processing bodies, or P-bodies (PB), and stress granules (SG) (Anderson and Kedersha, 2009; Decker and Parker, 2012). PB are conserved from yeast to mammals and contain a core of proteins consisting of the mRNA decapping machinery, including the decapping enzyme Dcp1/Dcp2 and the activators of decapping Dhh1, Scd6, Edc3 as well as mRNAs (Decker and Parker, 2012; Eulalio et al., 2007; Jain and Parker, 2013). These components function in both translation repression and mRNA degradation and compete with the assembly of translational factors (Bhattacharyya et al., 2006; Buchan, 2014). PB are present in cells during normal growth but increase in number in cells subjected to stress, such as nutrient deprivation, heat shock or other events that lead to a reduction in translation initiation. SG are present only in cells subjected to stress or in which translational initiation has been abrogated. Their composition overlaps that of PB but in addition includes translation initiation factors, poly(A) RNA-binding proteins and the 40S ribosomal subunit (Decker and Parker, 2012).

Both PB and SG are dynamic structures whose components rapidly exchange with their cytoplasmic counterparts. Recent evidence suggest that they may be liquid phase droplets that form as a result of a phase transition process driven by weak multivalent interactions among the components of the aggregates (Weber and Brangwynne, 2012). Both low complexity sequences, such as prion-like domains (PLDs), and RNA components of these aggregates contribute to the multivalent interactions, and the nature and extent of these components can dictate the location and physical properties of the aggregates (Guo and Shorter, 2015; Zhang et al., 2015). Consistent with this observation, RNA binding proteins (RBPs) comprise a significant proportion of these aggregates and such proteins often encompass a PLD. Moreover, a disproportionate number of such RBPs are associated with pathogenic aggregates in a variety of neurodegenerative diseases, such as Amyotrophic Lateral Sclerosis (ALS) and Parkinson’s disease (Ramaswami et al., 2013). More recent observations have added to the number of distinct non-membrane bound cytoplasmic compartments. Narayanaswamy et al. (Narayanaswamy et al., 2009) showed that a large fraction of normally soluble metabolic enzymes in Saccharomyces cerevisiae forms discrete aggregates in the cytoplasm upon nitrogen starvation and Shah et al. (Shah et al., 2014) showed that a significant number of kinases in Saccharomyces forms cytoplasmic aggregates upon transition into stationary phase. Some of these aggregates overlap with PB and SG, but others comprise different distinct foci. Like SG, these aggregates disperse upon the cells’ return to normal growth.

While the non-membrane bounded cytoplasmic compartments described above serve at least in part as transitory storage sites for mRNAs and proteins to be reused after cessation of stress, several discrete sites serve as cellular deposition sites for misfolded proteins that arise from proteotoxic stress. In Saccharomyces cerevisiae, several such sequestration sites exist: Insoluble Protein Deposits (IPODs), intranuclear or juxtanuclear quality control compartments (INQs/JUNQs), aggresomes and Cytoplasmic Q-bodies (CytoQs) (Kaganovich et al., 2008; Miller et al., 2015; Sin and Nollen, 2015; Wang et al., 2009). They are distinguished by their location in the cell, with IPODs located adjacent to the vacuole, INQs in the nucleus adjacent to the nucleolus and CytoQs as dispersed aggregates in the cytoplasm. The sites are also distinguished by the type of misfolded proteins and the specific chaperones associated with them. For instance, IPODs preferentially form from amyloidogenic proteins, such as prions, and recruit the chaperone system consisting of the Hsp70 member, Hsp104, and specifically the Hsp40 member, Sis1 (Miller et al., 2015). Aggresomes are specific sites at the spindle pole body, forming only upon expression of human huntingtin exon 1 with an expanded polyglutamine domain. What targets RBPs to their deposition sites versus the ribonucleoprotein (RNP) granules described above is not clear. Here we show that nuclear import and PLD phosphorylation combine to specify to which cytoplasmic particle the yeast RBP Ssd1 localizes.

Ssd1 is a nucleo-cytoplasmic shuttling protein with multiple functions in the life cycle of its bound mRNAs, particularly those that are destined for polarized localization and translation (Kurischko et al., 2011b). Ssd1 contains a single canonical nuclear localization signal (NLS) and once Ssd1 enters the nucleus it binds through its RNA binding domain (RBD) to a set of approximately 50 mRNAs, a majority of which encode cell wall proteins. Ssd1 exports these mRNAs out of the nucleus and delivers them to sites of polarized growth (Figure 1A) (Hogan et al., 2008; Jansen et al., 2009; Kurischko et al., 2011a; Kurischko et al., 2011b; Mitchell et al., 2013; Uesono et al., 1997). Upon stress Ssd1 and its bound mRNAs localize to SG and PB likely through direct interaction between Ssd1 and components of PB and SG (Kurischko et al., 2011a; Richardson et al., 2012; Tarassov et al., 2008; Zhang et al., 2014). Ssd1 is the essential substrate of Cbk1, a highly conserved tumor suppressor Ndr/LATS kinase. Deletion of CBK1 or elimination of the phosphorylation sites on Ssd1 targets it and its bound mRNAs irreversibly to PB and is lethal (Jansen et al., 2009; Kurischko et al., 2011a). Here we show that PB/SG components play a role in the nuclear export of Ssd1, indicating that interaction between Ssd1 and PB components occurs initially in the nucleus. Moreover, in the absence of association with these components in the nucleus, Ssd1 is targeted to IPODs. Both the association with PB/SG and with IPODs requires the N-terminal PLDs and these associations are inversely affected by Ssd1’s phosphorylation status (Figure 1B). Similar processes may contribute to aggregate formation of RBPs in certain neurological diseases such as ALS (Ramaswami et al., 2013).

Results
P-body and stress granule localization of yeast Ssd1 depends on its prion-like domains. PLDs potentiate protein-protein interactions often as a means of directing proteins to their appropriate functional sites within a cell (Hennig et al., 2015; March et al., 2016). The N- terminus of Ssd1 carries several potential PLDs, as determined by a hidden Markov Model for predicting prions (Fig 1B) (Alberti et al., 2009). Two of these, aa 1-23 and aa 48-154, are nearly contiguous and we consider this a single domain and refer to aa 1-162 as PLD1. PLD1 corresponds to the domain that binds the phosphorylated C-terminal domain of RNA polymerase II (Phatnani et al., 2004). A second potential PLD is predicted for aa 263-336. All but one (S228) Cbk1 phosphorylation sites in Ssd1 lie in these two PLDs (Jansen et al., 2009; Mazanka et al., 2008) and dephosphorylation of these sites results in constitutive PB association of Ssd1 (Kurischko et al., 2011a). To determine whether either or both PLDs are required for association of Ssd1 with PB and/or SG, we constructed Ssd1 variants deleted for aa 1-162 and aa 1-331, accordingly. Since Ssd1163-1250-GFP and Ssd1332-1250-GFP proteins expose an unfavorable N- terminal amino acid and are highly unstable in the cell (Bachmair et al., 1986), we inserted three glycine residues at the N-termini (Ssd1GGG163-1250, Ssd1GGG332-1250), which stabilized the proteins. As previously reported, a portion of wild type Ssd1 protein, expressed under control of the GPD1 promoter, associated with the PB/SG components even under favorable growth conditions (Kurischko et al., 2011a). We observed that following glucose starvation, essentially all Ssd1 protein formed cytoplasmic foci, many of which colocalized with PB component Edc3 (Fig 2). We then examined the localization of Ssd1 lacking one or both of the PLDs. No cytoplasmic foci were observed for Ssd1GGG163-1250-GFP in glucose medium. After shifting cells to glucose free medium, most of the Ssd1 remained dispersed in the cytoplasm although a few faint foci appeared, but only after longer exposure to glucose deprivation. Some of these foci co-localized or overlapped with Edc3 (Fig 2). PGPD1-SSD1GGG332-1250-GFP, which lacks both PLDs, showed similar behavior as observed for Ssd1GGG163-1250-GFP (Suppl. Fig S1). These experiments document that PLD1 plays a critical role in promoting association of Ssd1 with PB.

Aggregate formation and nuclear export of Ssd1 require PB components. Kurischko et al. (Kurischko et al. 2011a) previously showed that Ssd1 physically associates with PB components. Since certain PB components are crucial for both PB and SG formation, we asked if Ssd1 also depends on them for its association with these structures. Accordingly, we examined the localization of Ssd1 in strains lacking individual PB components. Deletion of EDC3, PAT1 or reduced expression of NOT1 (DAmP-not1) eliminates Ssd1 cytoplasmic foci formation although PB form, albeit at a reduced rate, in edc3Δ and pat1Δ strains (Decker et al., 2007; Teixeira and Parker, 2007) (Fig 3A). In addition, Ssd1 fails to form foci in cells lacking the SG component Pbp1. However, Ssd1 foci formation occurs normally in cells lacking the SG component Pub1 (Fig 3B). This indicates that association of Ssd1 with PB requires the presence of several PB and some SG components. Furthermore, as evident from Figure 3A, Ssd1 in a wild type background exhibits no detectable accumulation in the nucleus, but in those strains in which Ssd1 fails to form foci, a significant fraction of the protein resides in the nucleus. Thus, cytoplasmic foci formation correlates with export from the nucleus. While this correlation might point to a requirement for functional PB or SG to provide a sink to trap Ssd1 in the cytoplasm and thus extract it from the nuclei, a more compelling interpretation is that PB components associate with Ssd1 in the nucleus to facilitate its nuclear export and target it for cytoplasmic PB granules. Consistent with this hypothesis, Pat1, Not1 and Pbp1 are nucleo-cytoplasmic shuttling proteins (Collart and Struhl, 1994; Haimovich et al., 2013b; Kumar et al., 2002). Thus, we surmise that at least three PB components and one SG protein likely facilitate Ssd1’s transport out of the nucleus.

Cytoplasmically restricted Ssd1 forms IPODs. Kurischko et al. (Kurischko et al., 2011b) previously identified a single NLS within Ssd1 located at amino acids 417 to 427. They reported that the mutant protein in which all eleven of these amino acids were converted to alanines, Ssd1(417-427)11A, failed to enter the nucleus and accumulated in cytoplasmic aggregates. Although these aggregates colocalized with Edc3, they are morphologically distinct from PB or SG (Kurischko et al., 2011b) (Fig 4). Besides being larger than PBs and generally presenting as a single aggregate per cell, they often appear as rings or donut shapes. In some cases, they appear honeycombed, perhaps reflecting the compound aggregation of multiple rings. These aggregates form in the absence of PB components (Fig 4), further distinguishing them from PBs as well as from wild type Ssd1. Rather, these aggregates resemble previously described insoluble protein deposits, or IPODs. To investigate if Ssd1(417-427)11A aggregates comprise IPODs, we examined the colocalization of Ssd1(417-427)11A with Hsp104 and Sis1, a disaggregase and a cooperating protein chaperone that function in protein disaggregation in IPODs. As shown in Fig 5, Hsp104 presents a variety of localization patterns toward Ssd1(417-427)11A. Hsp104 either fills the cavity of a Ssd1(417-427)11A ring, surrounds large aggregates, or resides adjacent to smaller foci and rings. In some cases multiple small foci of Ssd1(417-427)11A surround a large “aggregate” of Hsp104 (see also Suppl. Fig S2 and Supplemental movies S1 to S4). Sis1 also localizes to Ssd1(417-427)11A aggregates, either surrounding or permeating them (Fig 5; Suppl. Fig S3; Suppl Movies S5 to S8). Finally, by staining cells expressing Ssd1(417-427)11A-GFP with FM4-64, we determined that these Ssd1 aggregates reside in the cytoplasm adjacent to but outside of vacuoles (Suppl. Fig S4), consistent with the previously reported localization of IPODs (Buchan et al., 2013; Miller et al., 2015; Petroi et al., 2012; Tardiff et al., 2013). In sum, these data are consistent with Ssd1(417-427)11A aggregates as IPODs.

PLD1 of Ssd1 is essential for IPOD formation. Given that IPODs are preferential sites for misfolded and/or amyloid proteins, we asked whether PLDs of Ssd1 are instrumental in forming IPODs. We constructed an Ssd1 variant that combined NLS(417-427)11A with the deletion of PLD1 (Ssd1GGG163-1250, (417-427)11A-GFP). When expressed from the GPD1 promoter in an ssd1 background, this protein localized uniformly in the cytoplasm of cells growing in glucose medium, with no detectable IPOD formation (Fig 6). From these results we conclude that PLD1 of Ssd1 promotes not only its association with PB/SG during stress but also its incorporation into IPODs when restricted to the cytoplasm.
Phosphorylation of PLDs regulates aggregate formation of Ssd1. Phosphorylation has been recently shown to affect the function of some prion-like domains (Gardiner et al., 2008; Hennig et al., 2015). The N-terminal 336 amino acids of Ssd1 that overlap the PLDs contain nine Cbk1 phosphorylation sites. We assessed the consequence of phosphorylation of Ssd1 on its PB/SG and IPOD formation. We examined the behavior of Ssd1 in which all the phosphorylation sites had been converted to phosphomimetic aspartate residues (Ssd1S/T9D). As evident in Figure 7, this mutant protein shows no PB association or other aggregate formation in unstressed cells, in stark contrast to wild type Ssd1. However, in conjunction with the NLS mutation, the mutant protein, Ssd1S/T9D, (417-427)11A, forms IPOD aggregates (Fig 8). Thus, we conclude that phosphorylation of the PLD in Ssd1 diminishes PB/SG association of the otherwise wild type protein but enhances IPOD formation of the cytoplasmically restricted protein.

Dephosphorylated Ssd1, arising either from inactivation of Cbk1 kinase or from elimination of all N-terminal phosphosites (Ssd1S/T9A), constitutively localizes to PB, where it traps mRNAs,and is highly toxic to the cell (Kurischko et al., 2011a). However, while Ssd1 lacking its NLS forms IPODs, we found that such a protein that also lacks the PLD phosphorylation sites fails to form IPODs and, as shown below, is no longer toxic. Rather, Ssd1S/T9A, (417-427)11A forms small uniform aggregates only incidentally coinciding with PB markers (Lsm1, Edc3 and Dcp2) both in unstressed and stressed cells, while significantly overlapping SG markers (Pab1 and Pub1) in nutrient stressed cells (Fig 9A and 9B), similar to the behavior of wild type Ssd1. These observations additionally indicate that Ssd1S/T9A, (417-427)11A retains an ability to respond to stress signaling. In sum, we conclude that dephosphorylation of the PLD sites enhances PB/SG localization but prevents IPOD formation in contrast to phosphorylation of the PLD sites, which suppresses PB/SG association but not IPOD formation.
Growth properties of cells with PLD and NLS mutations of SSD1. We analyzed the growth properties of cells expressing the various Ssd1 mutants described above (Fig 10A and B, Table 1). Cells lacking SSD1 (ssd1Δ) are mildly temperature sensitive and sensitive to the cell wall damaging agent calcofluor white. Cells carrying a mutant protein lacking a functional NLS (ssd1(417-427)11A) as the sole copy of the gene show the same phenotype, indicating that the NLS mutant fails to complement the deletion (Fig 10A). However, this mutant protein is not simply non-functional since its overexpression inhibits cell growth (Fig 10B), perhaps as a consequence of aggregate formation. The fact that deletion of PLD1 from the NLS mutant protein (Ssd1GGG163-1250,(417-427)11A) prevents aggregate formation, but only partially suppresses the overexpression growth inhibition (Fig 10B), supports the conclusion that both the lack of nuclear functions and aggregate formation contribute to the phenotypes of Ssd1(417-427)11A. Interestingly, deletion of one or both PLDs from the otherwise wild type Ssd1 (Ssd1GGG163-1250 and Ssd1GGG336-1250) exacerbates the calcofluor white sensitivity and the overexpression toxicity.

This observation indicates that the capacity of Ssd1 to associate with PBs fulfills essential functions. Conversion of the phosphosites to phosphomimetic residues combined with mutated NLS (Ssd1S/T9D,(417-427)11A) does not alter the phenotypes observed for cells expressing the NLS mutated protein. However, conversion of the phosphosites to non-phosphorylatable residues (Ssd1S/T9A,(417-427)11A) completely suppresses the temperature sensitivity and the calcofluor white sensitivity (Fig 10A). Thus, the double-mutant protein functions essentially as well as the wild type protein. This is consistent with the restoration of the subcellular localization properties of the double-mutant protein to those exhibited by the wild type protein. As previously reported and confirmed here, expression of Ssd1S/T9A completely inhibits cell growth, due apparently to irreversible sequestration of essential mRNAs in PB or SG (Kurischko et al., 2011a). This toxicity is partially rescued by the mutated NLS. We conclude that the 9A mutations not only suppress the NLS phenotypes but also that the NLS mutations suppress the 9A phenotypes.PLD1 contains a phosphorylation sensitive cryptic NLS. The results above indicate that the absence of N-terminal phosphorylation of Ssd1 suppresses the aggregation phenotype arising from mutation of the protein’s NLS. One possible explanation for suppression is that the N- terminal region possesses a phosphorylation sensitive NLS, such that the protein in its dephosphorylated state could gain entry to the nucleus. To test whether the N-terminal region could facilitate entry into the nucleus, we tagged endogenously expressed Ssd1 after aa 1-200 or aa 1-336 with GFP and analyzed the localization of the protein fragments. As evident from Figure 11A, these N-terminal fragments are sufficient to localize GFP to the nucleus, demonstrating that a functional NLS resides in the N-terminal domain.

These results demonstrate that the N-terminal domain of Ssd1 (aa 1-200) encompasses a functional NLS, while the results obtained with the phosphosite mutants suggest that the activity of this NLS is regulated by Cbk1-mediated phosphorylation. To test this assumption directly, we constructed strains carrying NIC96-mCherry as a nuclear marker and in which the only chromosomal copies of SSD1 are ssd11-200-GFP, ssd11-200, S5A-GFP or ssd11-200, S5D-GFP. The nuclear/cytoplasmic ratios of GFP in these strains confirmed the nuclear accumulation of the ssd1 N-terminal fragment. Moreover, we observed a significantly reduced level of nuclear accumulation of the N-terminal fragment carrying phosphomimetic variants (S5D), relative to either the wild type fragment or the non-phosphorylatable fragment (Fig 11B). These results confirm that phosphorylation of the N-terminal domain of Ssd1 reduces its ability to promote nuclear import.This conclusion gains support from a number of additional observations. First, Kurischko et al. (Kurischko et al., 2011b) reported that inhibition of Cbk1 kinase by treating a cbk1-as mutant with 1NA-PP1 enhanced the nuclear appearance of the N-terminal Ssd11-450 fragment. Second, nuclear localization of Ssd1 is crucial for suppressing the lethality of ssd1Δ sit4Δ (Fig 11C): expression of Ssd1 but not Ssd1(417-427)11A is sufficient to support the viability of an ssd1Δ sit4Δ strain. However, expression of the NLS mutated protein additionally lacking all nine Cbk1 sites (Ssd1S/T9A, (417-427)11A) does suppress the lethality of the strain. Consistently, conversion of phosphorylation sites to phosphomimetic residues or deletion of PLD1 (Ssd1S/T9D, (417-427)11A and Ssd1GGG163-1250, (417-427)11A) do not suppress the lethality, indicating a requirement for dephosphorylated PLD1 in suppressing the nuclear import defect. Third, Ssd1(417-427)11A fails tofully complement ssd1, due to the failure to enter the nucleus to acquire and deliver mRNAs and perhaps as a consequence of IPOD formation. However, Ssd1S/T9A, (417-427)11A promotes growth under normal and stressed conditions as well as wild type Ssd1 (Fig 10). These observations support our hypothesis above that activation of an N-terminal NLS through inhibition of Cbk1-mediated phosphorylation of the protein, by either inactivation of Cbk1 or elimination of its phosphorylation sites, provides an alternate means of importing Ssd1 into the nucleus.

Discussion
Protein trafficking and aggregate formation. Results presented here demonstrate that Ssd1 can reside in three distinct cytoplasmic compartments: either uniformly dispersed throughout the cytoplasm, incorporated into discrete P-body or stress granule aggregates or targeted for degradation/recycling in IPOD-like aggregates. Our results also clarify the molecular basis for targeting to the different locales. First, we find that the prion-like domains of Ssd1 are required for association of Ssd1 with PB and SG, consistent with previous observations (Reijns et al., 2008), but in addition, we find that the PLD is required for Ssd1 incorporation into IPOD-like aggregates. Previously described IPODs form around misfolded amyloidogenic proteins Ubc9ts, Rnq1 and Ure2 (Kaganovich et al. 2008). A previous study identified the N-terminal region of Ssd1 as a potential prion-like domain but failed to document that this region could form amyloid aggregates (Alberti et al. 2009). This is not inconsistent with our results, since we have shown here that amyloid formation requires phosphorylation of the Ssd1 PLD1 domain, which would not have occurred in the previous study. A second aspect of Ssd1 subcellular localization identified here is that nuclear import is required to prevent IPOD formation and promote association of Ssd1 with PB or SG. This suggests that Ssd1 can associate with PB components or mRNAs only within the nucleus and that such initial association in the nucleus is required for subsequent association with PB or SG structures.

Moreover, in the absence of nuclear import and subsequent association with mRNAs and/or PB components, Ssd1 forms aggregates, that coalesce into IPODs. This may serve as a fail-safe mechanism to insure that Ssd1 not properly engaged in its normal function is removed from the cytoplasm for recycling or degradation and thereby prevent cellular toxicity. Third, phosphorylation of the PLDs of Ssd1 significantly affects its subcellular localization. Phosphorylation of the domains diminishes association of Ssd1 with PBs/SGs but, in proteins restricted to the cytoplasm, promotes association with IPODs. In contrast, dephosphorylation of these sites enhances association with PB/SG but prevents association of the cytoplasmic restricted protein with IPODs. This could be due to a reduction in the aggregation tendency of the dephosphorylated protein or to activation of the cryptic NLS and restoration of nuclear entry for subsequent acquisition of mRNAs and PB components in the nucleus. The availability of a second NLS, activated upon dephosphorylation during stress, may provide an advantage to the cells. Ssd1 preferentially binds mRNAs encoding cell wall proteins and hence plays a significant role in cell wall integrity. Upon stress, a higher load of mRNAs and their storage in SG may lead to a faster recovery after cessation of the stress conditions.The nuclear-cytoplasmic cycle of Ssd1. Our studies of Ssd1 functional domains provide insight into the role of nuclear interactions in directing RBPs and their associated mRNAs to their appropriate cytoplasmic location in the cell. Under normal conditions Ssd1 enters the nucleus through the action of its single NLS417-427. We suggest that once Ssd1 enters the nucleus, PLDsin the N-terminus of Ssd1 promote its association with pCTD of pol II transcribing certain specific mRNAs. The interaction of Ssd1 with the pCTD would bring Ssd1 into proximity of mRNAs as they are being synthesized, thereby adding an Ssd1-targeted subcellular address to certain mRNAs at the time of initiation of synthesis. Among the mRNAs that Ssd1 binds are those that encode proteins required for cell wall stability and response to cell wall stress, some of which exhibit polarized localization to the daughter cell.

This model is consistent with previous data demonstrating an affinity of PLD1 of Ssd1 for the pCTD (Phatnani and Greenleaf, 2004).Moreover, the interaction between RBPs and pCTD of pol II may be a common mechanism for cotranscriptional RNA binding, as it was shown for Yra1, a nuclear poly(A) RNA-binding protein which is required for nuclear export of mRNA (MacKellar and Greenleaf, 2011).Ssd1 interaction with the pCTD may also promote its association with the PB complex. Recent data have documented that the PB complex interacts with the promoter region of certain genes (Dahan and Choder, 2013; Haimovich et al., 2013a; Haimovich et al., 2013b; Sun et al., 2012). Thus, Ssd1 through its association with pCTD of pol II could engage the PB complex as early as initiation of transcription. Deletion of PLDs eliminates association of Ssd1 with PB even though Ssd1 lacking PLD1 can still bind to PB components (data not shown). This suggests that interaction between Ssd1 and PB components in the nucleus as a consequence of the PLD mediated association with the pCTD may be a prerequisite for subsequent cytoplasmic localization of Ssd1.Our results further demonstrate that the interaction of Ssd1 with the components of the mRNA decay machinery in the nucleus are required for efficient export of Ssd1 and subsequent deliveryof Ssd1 and its cargo mRNA to the cytoplasm and to PBs and SGs. To the best of our knowledge, this is a novel nuclear role for PB components towards the nuclear export of RBPs. We found that either the absence of any of several PB components, the inhibition of Ssd1 nuclear import or the deletion of regions of the protein required for interaction with pCTD of pol II prevent subsequent association of Ssd1 with PBs. Moreover, the failure of Ssd1 to engage in this process by blocking nuclear import leads to PLD mediated aggregation of the protein into IPODs, which attract the chaperone complex comprising Hsp104 and the Hsp40 member, Sis1.Thus, the interaction of Ssd1 with the PB complex in the nucleus is permissive for subsequent association with PB/SG in the cytoplasm, but presentation of Ssd1 to the complex in the cytoplasm without prior interaction in the nucleus does not lead to productive association.

This may simply result from the contribution of mRNA acquired by Ssd1 in the nucleus to the multivalent interactions necessary for liquid phase aggregate formation resulting in PB or SG association. Alternatively, assembly of Ssd1 into PB complex may be an ordered process that is predicated on initial interaction in the nucleus. We summarize the model based on our results in Figure 12.Ssd1 as a model for neuropathic proteins. The dependency of aggregate formation on PLDs is well documented for human RBPs, like TDP-43, FUS and hnRNPA1 (Blokhuis et al., 2013; Kim et al., 2013). We hypothesize that IPODs have similarities with aggregates of FUS and TDP-43 in motor neurons of patients suffering from ALS. The majority of ALS related mutations in FUS, which promote formation of large cytoplasmic inclusions, occur in its C-terminal NLS. Moreover, deletion of the NLS from FUS in a mouse model leads to aggregate formation (Shelkovnikova et al., 2013a). Although these aggregates contain SG proteins, it is not clear ifsuch structures are related to SG (Bentmann et al., 2013; Blokhuis et al., 2013). Rather, nuclear import of FUS and its attendant binding to SG proteins may prevent it from association with large aggregates (Farrawell et al., 2015; Shelkovnikova et al., 2013b). By similarity with Ssd1, SG proteins may adhere to large aggregates of mutated FUS or TDP-43, when cells are subjected to stress imposed by the mutated FUS (Farg et al., 2012; Vance et al., 2013). Accordingly, further analysis of the structure-function relationship of Ssd1 could shed light on the pathology of neurotoxic proteins. Moreover, as we see with Ssd1, stimulating alternate trafficking of FUS or TDP-43 might alleviate aggregate formation with attendant therapeutic effect.Yeast strains and growth conditions. Standard yeast genetics and culture methods were used (Amberg et al., 2005).

For glucose depletion, 1.5 ml of cells were harvested by centrifugation for 30 sec and the cell pellet was resuspended in 1 ml glucose free medium. Cells were incubated for 1 min with repeated inversion of the Eppendorf tube. Cells were again spun down for 30 sec and resuspended in a small volume of the supernatant. Images were taken immediately and over a time course for up to 15 min.Strain FLY2184 (MATα his31 leu20 met150 ura30 ssd1Δ::NATMX) (Kurischko et al., 2011a) was derived from strain BY4741 (Brachmann et al., 1998) by replacement of the SSD1 coding region with the NATMX cassette. BY4741 based deletions of EDC3, PAT1, PBP1 and PUB1 were obtained from the yeast deletion collection. Strain DAmP-NOT1 was obtained from GE Healthcare Dharmacon Inc. BY4741 based strains with GFP-tags inserted after amino acids 200 (Y4183) and 336 (Y4184) of Ssd1 were constructed by transformation with PCR-based cassettes (Longtine et al., 1998). For strains carrying ssd11-200, S5A-GFP::KANMX or ssd11-200, S5D-GFP::KANMX as the only chromosomal copy of SSD1, the NAT cassette of FLY2184 was replaced with the corresponding fragments. These strains were crossed to Nic96- mCherry::URA3 (Dr. Yves Barral, Zurich, Switzerland) to incorporate a nuclear marker. Strain sit4Δ::KANMX ssd1Δ::NATMX [B2937] was constructed by crossing single deletion strains and transforming the diploid with B2937 (pRS426-sit4-102) before sporulation and tetrad dissection.Plasmids encoding RFP-tagged PB and SG proteins were obtained from Dr. Roy Parker (Univ. of Colorado, Boulder) and Dr. Charles Cole (Dartmouth Medical School, Hanover, NH).Plasmids carrying HSP104-mCherry and SIS1-mCherry were provided by Dr. S. Alberti, MPI Dresden, Germany. All plasmids are listed in Supplementary Table S1.Dilution series to define growth capacity. Cell densities of overnight cultures of strains BY4741 [pRS415], BY4741 [pRS416], and ssd1Δ containing the indicated pAG415-PGal1-SSD1- GFP or pAG416-PGal1-SSD1-GFP variants were counted and adjusted to equal numbers of cells/ml. 4 l of undiluted and serial 10x dilutions were spotted on agar plates containing either 2% glucose or 2% galactose. The plates were incubated for 3 days at 22oC. The same procedure was applied to overnight cultures of strains BY4741 [pRS415], ssd1Δ [pRS415] and ssd1Δ containing the indicated pAG415-PGPD1-SSD1-GFP variants.

4 μl of undiluted cultures and 10x dilutions were spotted for growth on SC-Leu at 22oC, SC-Leu at 37oC and SC-Leu at 22oC with 15 μg/ml calcofluor white (CFW) agar plates. Images were taken after 3 days. The same protocol was applied for the plasmid shuffle experiment to define the capacity of Ssd1 versions to support the viability of sit4Δ ssd1Δ cells. Deconvolution microscopy. Cells were imaged on a wide-field inverted microscope at Penn State College of Medicine on a Delta Vision; Applied Precision, Issaquah, WA with a charge- coupled device camera (CoolSNAP HQ; Roper Scientific, Tucson, AZ), using a 100× oil- immersion objective, or at Rockefeller University on a DeltaVision Image Restoration Microscope (Olympus IX70 inverted microscope with a pco. edge sCOS camera) using a 100x oil-immersion objective. The percentage transmittance for GFP and RFP channels and the exposure times were individually adjusted to the brightness of the signals, as all signals were plasmid based. Eighteen to twenty 0.2 μm z-stacks were taken of each focal plane using the DeltaVision softWoRx software. Where indicated, deconvolution of the z-stacks was performed. For each strain, at least 50 cells were imaged.Definition of nuclear-cytoplasmic ratio (N/C). The fluorescent intensities of individual nuclei and the adjacent cytoplasm were measured by Metamorph software as described previously (Kurischko et al. 2011b). In detail, a defined area in nuclei (N), identified by the nuclear pore marker Nic96-mCherry, was measured in a single plane of deconvolved z-stacks in the GFP channel. The same sized area was measured in the adjacent cytoplasm (C) and outside of cells (background B). The ratios of N-B/C-B were calculated for about 100 nuclei per strain and processed statistically by Wilcoxon rank sum test.FM4-64 staining of vacuole membranes. The procedure is a variation of previously published protocol (Vida and Emr, 1995). 5 ml cultures were spun down and resuspended in 500 μl YPD. 1 μl FM4-64 solution in DMSO (Molecular Probes Life Technology) was added to a final concentration of 32 μM and incubated in the darkness for 30’ at RT. Cells were spun down, washed to remove the excess of FM4-64 and resuspended in SC medium. Images were taken after 30’ to analyze TDI-011536 colocalization of membranes and Ssd1-GFP proteins.