Cdc48, known as p97 or valosin-containing protein (VCP) in mammals, is an abundant AAA-ATPase that is essential for many ubiquitin-dependent processes. One well-documented role for Cdc48 is in facilitating the delivery of ubiquitylated misfolded endoplasmic reticulum proteins to the proteasome for degradation. By contrast, the role for Cdc48 in misfolded protein degradation in the nucleus is unknown. In the budding yeast Saccharomyces cerevisiae, degradation of misfolded proteins in the nucleus is primarily mediated by the nuclear-localized ubiquitin-protein ligase San1, which ubiquitylates misfolded nuclear proteins for proteasomal degradation. Here, we find that, although Cdc48 is involved in the degradation of some San1 substrates, it is not universally required. The difference in the requirement for Cdc48 correlates with the insolubility of the San1 substrate. The more insoluble the substrate, the more its degradation requires Cdc48. Expression of Cdc48-dependent San1 substrates in mutant cdc48 cells results in increased substrate insolubility, larger inclusion formation and reduced cell viability. Substrate ubiquitylation is increased in mutant cdc48 cells, suggesting that Cdc48 functions downstream of San1. Taken together, we propose that Cdc48 acts, in part, to maintain the solubility or reverse the aggregation of insoluble misfolded proteins prior to their proteasomal degradation.
Proteins degraded by the proteasome are initially targeted by ubiquitin-protein ligases, which mediate the covalent attachment of ubiquitin to their substrates for recognition by the proteasome. A number of ubiquitin-protein ligases have been shown to interact with the proteasome (Lee et al., 2005; Leggett et al., 2002; Panasenko and Collart, 2011; Um et al., 2010; Xie and Varshavsky, 2000; You and Pickart, 2001), suggesting that there is a direct route for delivery for the substrate from the ubiquitin-protein ligase to the proteasome. However, a key AAA-ATPase, known as p97 or valosin-containing protein (VCP) in mammals and Cdc48 in budding yeast (Saccharomyces cerevisiae), often acts between the ubiquitin-protein ligase and the proteasome to facilitate delivery of ubiquitylated substrates to the proteasome (Meyer et al., 2012). In yeast, Cdc48 was first discovered in a genetic analysis to identify proteins required for cell cycle progression (Moir et al., 1982) and was later found to be homologous to human p97/VCP (Fröhlich et al., 1991). Since its discovery, the function of Cdc48 has been extended to many ubiquitin-dependent processes, including protein quality control, membrane fusion and chromatin regulation (Dantuma and Hoppe, 2012; Meyer et al., 2012; Ye, 2006).
The involvement of Cdc48/p97/VCP in many essential cellular functions underscores its importance in the cell. This is further highlighted by the pathologies that are caused by or correlated with mutations in human p97/VCP (Nalbandian et al., 2011). For example, autosomal-dominant mutations in p97/VCP cause the devastating, age-dependent, degenerative disorder known as inclusion body myopathy associated with Paget's disease of bone and frontotemporal dementia (IBMPFD) (Watts et al., 2004). A subset of mutations in p97/VCP also account for 1–2% of familial amyotrophic lateral sclerosis (ALS) cases (Johnson et al., 2010). Augmented levels of p97/VCP are correlated with different types of human cancers and associated with poor prognoses (Laguë et al., 2012; Yamamoto et al., 2004; Yamamoto et al., 2005).
The main role of Cdc48/p97/VCP in the cell is thought to be as a segregase (Braun et al., 2002), whereby Cdc48 segregrates ubiquitylated proteins from protein complexes, chromatin or membranes (Jentsch and Rumpf, 2007). A well-studied segregase function of Cdc48 is its role in promoting protein quality control (PQC) degradation in the endoplasmic reticulum (ER) (Wolf and Stolz, 2012). Following the ubiquitylation of misfolded ER proteins, Cdc48 functions to extract ubiquitylated proteins from the ER membrane for degradation by cytoplasmic proteasomes. The PQC degradation role of Cdc48 has also been extended to the mitochondria (Taylor and Rutter, 2011). After oxidative stress, Cdc48 acts at the outer mitochondrial membrane to assist in removing ubiquitylated mitochondrial proteins for presentation to cytoplasmic proteasomes.
Although the cellular pool of proteasomes is cytoplasmic and nuclear in large part (Wójcik and DeMartino, 2003), the involvement of Cdc48 in the PQC degradation of misfolded proteins in the cytoplasm and nucleus is poorly understood. We previously identified the protein San1 in budding yeast as a nuclear-localized ubiquitin-protein ligase that acts with the ubiquitin-conjugating enzymes Cdc34 and Ubc1 to ubiquitylate misfolded nuclear proteins for proteasomal degradation (Gardner et al., 2005). After identifying the basic components of the San1 pathway, we turned our attention to discovering additional factors that function with San1. Here, we report that yeast Cdc48 is involved in San1-mediated nuclear PQC degradation but, unexpectedly, it is not universally required.
San1 substrates used to examine Cdc48 involvement in San1-mediated degradation
Previously, we characterized >40 substrates of San1, representing a broad variety of misfolded proteins likely to be encountered by San1 in the nucleus (Fredrickson et al., 2011; Gardner et al., 2005; Rosenbaum et al., 2011). The first group of San1 substrates comprises five proteins (Cdc13, Cdc68, Sir3, Sir4 and Ura3) that contain single missense mutations thought to cause their temperature-dependent misfolding (Fredrickson et al., 2011; Gardner et al., 2005). The second group of San1 substrates consists of 20 proteins with truncations thought to cause misfolding due to the deletion of key amino acid sequences (Rosenbaum et al., 2011). The third group comprises 16 small hydrophobic peptides that act as degrons and are thought to display the type of exposed hydrophobicity recognized by San1 in misfolded proteins (Fredrickson et al., 2011).
We wanted to use examples from each class of San1 substrate to explore whether Cdc48 is generally involved in San1-mediated degradation. For missense mutant proteins, we used Myc–Sir4-9, which has the point mutation S1132D (Gardner et al., 2005), and Myc–Cdc13-1, which has the point mutation P371S (Nugent et al., 1996). We also used variations of missense mutant proteins fused to the nuclear Gal4 activation domain (GAD) (Rosenbaum et al., 2011): GAD–Cdc68-1, which is the isolated N-terminal domain of Cdc68 (residues 2–468) carrying the structurally destabilizing G132D mutation (Evans et al., 1998), and GAD–Cdc13-1, which is the isolated N-terminal domain of Cdc13 (residues 1–600) carrying the structurally destabilizing P371S mutation (Nugent et al., 1996). For truncated proteins, we used GFPNLS–Tef2*, which contains residues 190–458 of the translation elongation factor Tef2 fused to GFP carrying an N-terminal nuclear localization signal (NLS), and GFPNLS–Bgl2*, which contains residues 20–313 of the endo-β-1,3-glucanase Bgl2 fused to GFPNLS (Rosenbaum et al., 2011). For the hydrophobic peptide degrons, we used peptide 6 (RDILVYTYILVYVYI) and peptide I (RETGWRVLLVVVGVVGIP) fused to GFPNLS (Fredrickson et al., 2011).
Cdc48 involvement in San1-mediated degradation varies among substrates
If Cdc48 were a general San1 pathway factor, we expected it would be required for the degradation of all tested San1 substrates to the same extent as San1. We found this was not the case and there was variability for the involvement of Cdc48 in the degradation of the nine San1 substrates we examined (Fig. 1A; supplementary material Fig. S1B). For some substrates, Cdc48 was essential for San1-mediated degradation, as was seen by the stabilization of GFPNLS–Tef2* and GFPNLS–Bgl2* in cdc48-3 mutant cells, similar to the stabilization seen in san1Δ cells. In other cases, Cdc48 was partially involved in San1-mediated degradation, as was seen by the lesser stabilization of Myc–Sir4-9, GFPNLS–Peptide-6 and GFPNLS–Peptide-I in cdc48-3 cells compared to the stabilization seen in san1Δ cells. Finally, some San1 substrates did not utilize Cdc48 for their San1-mediated degradation, as was seen by the lack of stabilization of GAD–Cdc68-1, Myc–Cdc13-1 and GAD–Cdc13-1 in cdc48-3 cells.
It is important to consider that the two substrates that showed strong dependency on Cdc48 for degradation, GFPNLS–Tef2* and GFPNLS–Bgl2*, were highly expressed from the strong galactose-inducible GAL1 promoter, and this might have forced a dependency on Cdc48 that would not be seen under lower expression conditions. Using reduced concentrations of galactose to induce lower expression (0.03% versus 3%), we observed that the degradation of GFPNLS–Tef2* and GFPNLS–Bgl2* was still dependent on Cdc48 (supplementary material Fig. S2A). GFPNLS–Bgl2* was less dependent on Cdc48 under lower expression conditions (compare supplementary material Fig. S2A with Fig. 1A), but the dependency was still equivalent to that observed in san1Δ cells (supplementary material Fig. S2A). Thus, it is possible that expression levels can influence the degree of Cdc48 requirement for some substrates. We do note that, at lower expression conditions, GFPNLS–Bgl2* was also less dependent on San1 (compare supplementary material Fig. S2A with Fig. 1A). Accordingly, it is possible that, when expressed at lower levels, GFPNLS–Bgl2* can be recognized by another degradation pathway.
The four substrates we examined that showed strong or partial dependency on Cdc48 for their degradation (GFPNLS–Tef2*, GFPNLS–Bgl2*, GFPNLS–Peptide-6 and GFPNLS–Peptide-I) were fused to GFPNLS. It is therefore possible that GFPNLS itself is degraded in a Cdc48-dependent manner and it imposes this requirement on the misfolded protein to which it is fused. We verified that GFPNLS is a stable protein (supplementary material Fig. S2B), eliminating this possibility.
Cdc48 is an essential cell cycle protein and cdc48-3 cells arrest in metaphase at the restrictive temperature of 37°C owing to activation of the spindle assembly checkpoint (Cao et al., 2003; Cheng and Chen, 2010). We did not observe stabilization of the Cdc48-dependent substrate GFPNLS–Tef2* after we initiated a similar metaphase arrest using nocodazole to prevent microtubule polymerization and activate the spindle assembly checkpoint (supplementary material Fig. S2C). Therefore, cell cycle arrest is not a contributing factor for the stabilization of select substrates in cdc48-3 cells.
Cdc48 possesses two ATPase domains, D1 and D2, that function in ATP binding and hydrolysis (Ye et al., 2003). The Cdc48-3 mutant protein contains two missense substitutions, P257L and R387K, in the D1 ATPase domain (Verma et al., 2011). These changes potentially alter both the structural integrity and the ATP binding and hydrolysis of the Cdc48 protein. Previously, mutants that are defective only in ATP hydrolysis were obtained by replacing conserved glutamate residues in the D1 (E305) or D2 domains (E578) of human p97/VCP with glutamine (Ye et al., 2003). These substitutions did not alter the binding of p97/VCP cofactors (Ye et al., 2003), indicating that the structural integrity of the mutant p97/VCP proteins was maintained. We used an analogous ATP-hydrolysis-deficient substitution in yeast Cdc48, E315Q (Ye et al., 2003), which we placed under control of the native CDC48 promoter in the previously constructed pRS316-Cdc48-Myc CEN plasmid (Tran et al., 2011). The mutation was incapable of complementing the cdc48-3 allele at the 37°C (Fig. 1C), as previously reported (Ye et al., 2003). We found that the Cdc48 ATPase mutant also did not complement the cdc48-3 allele for GFPNLS–Tef2* degradation (Fig. 1D), indicating that ATP hydrolysis is required for the function of Cdc48 in the degradation of select San1 substrates.
Loss of Cdc48 function leads to increased substrate ubiquitylation
Owing to the involvement of Cdc48 in the degradation of a subset of San1 substrates, we wanted to ascertain whether Cdc48 functions prior to or after San1-mediated ubiquitylation. Therefore, we performed assays to assess the ubiquitylation levels of Cdc48-dependent San1 substrates in parent, san1Δ and cdc48-3 strains. We observed enhanced ubiquitylation levels of both a completely (GFPNLS–Tef2*) and partially (GFPNLS–Peptide-6) Cdc48-dependent San1 substrate in mutant cdc48-3 cells compared with parent cells (Fig. 2A,B). To determine whether the increased ubiquitylation observed in cdc48-3 cells was San1 dependent, we examined the ubiquitylation of GFPNLS–Tef2* in cdc48-3 san1Δ cells. GFPNLS–Tef2* ubiquitylation was reduced in cdc48-3 san1Δ cells and was similar to that observed in san1Δ cells (Fig. 2C), indicating that the increased ubiquitylation after inhibition of Cdc48 function is due to San1 activity.
These results indicate that Cdc48 likely acts at a step after San1 ubiquitylation but prior to the proteasome. To see whether this was the case, we focused on GFPNLS–Tef2*. We introduced a pdr5Δ allele into parent, san1Δ and cdc48-3 cells to facilitate proteasome inhibition by MG132 (Chernova et al., 2003; Fleming et al., 2002). Proteasome inhibition by addition of MG132 for 1 hour led to greater levels of ubiquitylated GFPNLS–Tef2* in parent cells and the increase in ubiquitylation was comparable to what was observed in cdc48-3 cells without proteasome inhibition (Fig. 2D). We did not observe any increase in GFPNLS–Tef2* ubiquitylation in cdc48-3 cells with proteasome inhibition (Fig. 2D).
A difference in the solubility of substrates correlates with Cdc48 dependency
Because the Cdc48 requirement for San1 substrate degradation was variable, we wanted to understand the reason for this variability. Unlike PQC degradation in the ER, which requires Cdc48 as a means to extract ubiquitylated proteins from the ER membrane (Wolf and Stolz, 2012), San1 functions primarily in the nucleoplasm and nuclear substrate delivery to the proteasome would not have to contend with membrane extraction. Furthermore, the San1 substrates used in this study that require Cdc48, GFPNLS–Tef2* and GFPNLS–Bgl2*, are not chromatin-associated proteins and would not require extraction from chromatin as seen with RNA polymerase II or the transcriptional repressor alpha2 (Verma et al., 2011; Wilcox and Laney, 2009). Cdc48/p97/VCP does have chaperone activity in vitro and can prevent the aggregation of denatured proteins (Song et al., 2007; Thoms, 2002). Thus, we surmised that the requirement for Cdc48 in San1-mediated degradation might correlate with the insolubility of the San1 substrate. That is, the more insoluble the substrate the greater the need for Cdc48.
One way to examine substrate solubility is to perform sedimentation assays of lysates generated from cells expressing San1 substrates to determine how the substrates partition between soluble and insoluble fractions. Substrates that were completely dependent on Cdc48 for their degradation, GFPNLS–Tef2* and GFPNLS–Bgl2*, were the most insoluble of all substrates examined (Fig. 3A). GFPNLS alone was predominantly soluble (Fig. 3A). Substrates that had a partial dependency on Cdc48 for their degradation, Myc–Sir4-9, GFPNLS–Peptide-6 and GFPNLS–Peptide-I, exhibited an intermediate level of insolubility (Fig. 3A). Finally, substrates that had little, if any requirement for Cdc48 in their degradation, GAD–Cdc68-1, Myc–Cdc13-1 and GAD–Cdc13-1, demonstrated little, if any observable insolubility (Fig. 3A). Thus, the requirement for Cdc48 in San1-mediated degradation correlated with the general insolubility of the San1 substrate as measured in a cell lysate generated from parent cells grown at 30°C.
If the function of Cdc48 were to maintain the solubility of San1 substrates, we expected that a reduction in Cdc48 function might result in increased substrate insolubility compared to parent cells. When we examined the insolubility of GFPNLS–Tef2* and GFPNLS–Bgl2* in cdc48-3 cells at 30°C, which is a semi-restrictive temperature for Cdc48-3 function (Hsieh and Chen, 2011), we did observe that the proportion of each substrate increased in the insoluble fraction in cdc48-3 cells compared to parent cells (Fig. 3B). This indicates that Cdc48 function contributes to the solubility of Cdc48-dependent San1 substrates under semi-restrictive conditions where the substrates are stable in cdc48-3 cells (Fig. 3C).
We also examined the solubility of San1 substrates at 37°C, which is the fully restrictive temperature for cdc48-3. A potentially confounding effect at 37°C is that there is an upregulation of heat-shock proteins and chaperones at this temperature (Gasch et al., 2000; Lindquist and Kim, 1996). Increased expression of heat-shock proteins and chaperones has been shown to increase the solubility of prions and other aggregation-prone proteins in vivo (Duennwald et al., 2012; Newnam et al., 1999). Therefore, we first examined whether growth at 37°C altered the solubility of GFPNLS–Tef2* and GFPNLS–Bgl2* in parent cells. Consistent with an increase in heat-shock proteins and chaperones, we observed an increase in the solubility of GFPNLS–Tef2* and GFPNLS–Bgl2* in parent cells at 37°C compared with 30°C (Fig. 3D). However, we found that inactivation of cdc48-3 still resulted in a decrease in the solubility of GFPNLS–Tef2* and GFPNLS–Bgl2* compared with parent cells (Fig. 3E), indicating that Cdc48 functions to maintain the solubility of San1 substrates even at a temperature where the abundance of other chaperones is elevated. Taken together, at both semi-restrictive and restrictive temperatures that impair Cdc48-3 function, there is decreased solubility of the substrates in cdc48-3 cells relative to parent cells.
San1-dependent substrate ubiquitylation was increased in cdc48-3 cells (Fig. 2), indicating that Cdc48 works downstream of San1 and acts on ubiquitylated San1 substrates. If this is the case, we predicted that there should be no change in solubility between cdc48-3 san1Δ cells and san1Δ cells. Therefore, we examined solubility in parent, san1Δ, cdc48-3 and cdc48-3 san1Δ cells. One confounding feature is that we did observe decreased solubility in san1Δ cells compared with parent cells (Fig. 3F), which we think is due to the fact that San1 interacts with exposed hydrophobicity in misfolded proteins that causes them to aggregate (Fredrickson et al., 2013; Fredrickson et al., 2011). Thus, San1 interaction with substrates would help maintain their solubility to some degree by interacting with the feature that causes aggregation. We did not observe any additional decrease in the solubility of the substrates in cdc48-3 san1Δ cells relative to san1Δ cells (Fig. 3F), suggesting that intact San1 function is important for the Cdc48-dependent changes in solubility.
Loss of Cdc48 function leads to increased inclusion formation and reduced cell viability upon expression of insoluble, misfolded nuclear proteins
Because GFPNLS–Tef2* and GFPNLS–Bgl2* were more insoluble in cdc48-3 cells compared to parent cells, we wanted to determine whether there was an increase in the formation of nuclear inclusions of these substrates in cdc48-3 cells. Using fluorescence microscopy, we found that both GFPNLS–Tef2* and GFPNLS–Bgl2* formed nuclear inclusions in parent cells (Fig. 4A,B) and there was an increase in the size of the inclusions in cdc48-3 cells (Fig. 4C). By comparison, GFPNLS, which is predominantly a soluble protein (Fig. 3A), did not form inclusions in either parent or cdc48-3 cells (supplementary material Fig. S3B).
It was recently discovered that san1Δ cells are sensitive to the Hsp90 inhibitor geldanamycin (Theodoraki et al., 2012), indicating that San1 is crucial for optimal viability when the folding of Hsp90 client proteins is disrupted in the cell. We previously found that expression of many misfolded San1 substrates resulted in cellular toxicity in the absence of San1 function (Rosenbaum et al., 2011). Therefore, we expected that loss of Cdc48 function would also result in toxicity upon expression of San1 substrates that require Cdc48 for their degradation. We examined the growth of parent and cdc48-3 cells expressing either GFPNLS–Tef2* or GFPNLS–Bgl2* from an inducible promoter. Under these conditions, we found that expression of both substrates impaired cell viability in cdc48-3 cells at both the permissive temperature of 25°C and the semi-restrictive temperature of 32°C compared to parent cells (Fig. 4D). By contrast, GFPNLS alone did not cause any cellular toxicity in cdc48-3 cells (Fig. 4D). We were surprised to observe toxicity after expression of GFPNLS–Tef2* or GFPNLS–Bgl2* in cdc48-3 cells at the permissive temperature of 25°C, where cell growth is equivalent to parent cells without substrate expression. This suggested that Cdc48 function is hypomorphic for the San1 pathway in cdc48-3 cells at 25°C. To verify this, we conducted degradation assays of GFPNLS–Tef2* and GFPNLS–Bgl2* at 25°C. Consistent with their toxicity at this temperature, we found that both substrates were stabilized in cdc48-3 cells at 25°C. Thus, even at a permissive temperature for proper cell cycle progression, the cdc48-3 allele is hypomorphic for San1-mediated degradation. Taken together, reduction or loss of Cdc48 function results in reduced cell viability when insoluble, inclusion-forming San1 substrates consequently accumulate in the nucleus.
The increased inclusion formation of GFPNLS–Tef2* and GFPNLS–Bgl2* in cdc48-3 cells prompted us to explore whether the expression of GFPNLS–Tef2* and GFPNLS–Bgl2* altered Cdc48 localization in vivo. Using fluorescence microscopy, we found that Cdc48 was, in general, uniformly localized throughout the cell when GFPNLS was expressed (Fig. 5A). We observed enrichment of Cdc48 in the nucleus in ∼17% of cells expressing GFPNLS (Fig. 5B). By contrast, in cells expressing GFPNLS–Tef2* or GFPNLS–Bgl2*, we observed enrichment of Cdc48 in the nucleus in ∼75% of cells (Fig. 5A,B). Thus, Cdc48 concentrates within the nucleus upon expression of insoluble Cdc48-dependent San1 substrates, consistent with a potential chaperone role in maintaining San1 substrate solubility.
Cdc48 dependency can be changed by altering substrate solubility
The model emerging from these observations is that the requirement for Cdc48 in San1-mediated degradation correlates with the degree of misfolded substrate insolubility – the greater the substrate insolubility, the greater the need for Cdc48. A key test of this model would be to create a means to increase or decrease the solubility of a San1 substrate without affecting its San1 dependency, and then query how the solubility changes alter substrate dependency on Cdc48.
We reasoned this test could be accomplished through the use of the GFPNLS and GAD reporter proteins that we already used in this study. It has been previously shown that the folding of GFP can be negatively affected by its fusion to proteins that fold poorly or misfold (Waldo et al., 1999). Accordingly, fusion of GFPNLS to a misfolded protein might be expected to increase the insolubility of the GFPNLS-misfolded protein fusion owing to additional misfolding of GFPNLS. By contrast, transcriptional activation domains like the GAD are often intrinsically disordered when not bound to their protein targets (Dyson and Wright, 2005). In this case, a misfolded fusion protein cannot alter the folding of the GAD because it already lacks structure due to its inherent disorder. Furthermore, intrinsically disordered proteins are exceptionally soluble because they lack sufficient hydrophobicity to form hydrophobic cores (Tompa, 2002), and their fusion to proteins improves solubility in bacterial expression systems (Santner et al., 2012). As a result, fusion of the GAD to a misfolded protein might be expected to increase the solubility of the GAD-misfolded protein fusion.
To test the insolubility hypothesis further, we focused on GFPNLS–Tef2*, GFPNLS–Bgl2*, GAD–Cdc68-1 and GAD–Cdc13-1. We switched the fusion proteins to create the new substrates GAD–Tef2*, GAD–Bgl2*, GFPNLS–Cdc68-1 and GFPNLS–Cdc13-1. An important parameter of the swap strategy is that the change should not alter San1-dependent degradation. Deletion of SAN1 stabilized GAD–Tef2*, GAD–Bgl2*, GFPNLS–Cdc68-1 and GFPNLS–Cdc13-1 to a similar extent as GFPNLS–Tef2*, GFPNLS–Bgl2*, GAD–Cdc68-1 and GAD–Cdc13-1 (Fig. 6A; Fig. 1A). Therefore, swapping the fusion reporter did not alter San1-mediated degradation.
By contrast, the fusion reporter swap did alter the Cdc48 dependency for Tef2*, Bgl2* and Cdc68-1 as anticipated based on the solubility of the GAD and GFPNLS reporter proteins. The degradation of GAD–Tef2* and GAD–Bgl2* was now much less dependent on Cdc48 than the degradation of GFPNLS–Tef2* and GFPNLS–Bgl2* (Fig. 6A; Fig. 1A). In fact, we needed to use shorter time points in the degradation assay in order to observe any Cdc48 dependency for GAD–Tef2* and GAD–Bgl2*. By contrast, the degradation of GFPNLS–Cdc68-1 was now more dependent on Cdc48 than the degradation of GAD–Cdc68-1 (Fig. 6A; Fig. 1A). The degradation of GFPNLS–Cdc13-1 remained independent of Cdc48 (Fig. 6A). This is a key result because it means that GFPNLS itself does not dictate Cdc48 dependency.
If the hypothesis that Cdc48 dependency correlates with substrate insolubility is correct, then we expected appropriate changes in the solubility of GAD–Tef2*, GAD–Bgl2* and GFPNLS–Cdc68-1 that match their altered Cdc48 dependency. Consistent with their reduced dependency on Cdc48, GAD–Tef2* and GAD–Bgl2* were more soluble than GFPNLS–Tef2* and GFPNLS–Bgl2* (Fig. 6B). Consistent with an increased dependency on Cdc48, GFPNLS–Cdc68-1 was more insoluble than GAD–Cdc68-1 (Fig. 6B). The solubility of GFPNLS–Cdc13-1 was unchanged from GAD–Cdc13-1 (Fig. 6B), mirroring its Cdc48 independence. Taken together, the fusion reporter swap data provide important evidence supporting the model that the requirement for Cdc48 in the degradation of misfolded nuclear substrates correlates with substrate insolubility.
The components we previously found to be required for nuclear PQC degradation in yeast are the nuclear-localized ubiquitin-protein ligase San1, its cognate ubiquitin-conjugating enzymes Cdc34 and Ubc1, and the proteasome (Gardner et al., 2005). We now add Cdc48 as an additional, but not universal, participant in the San1 pathway. Our current working model for the function of Cdc48 in the San1 pathway is that Cdc48 helps to maintain the solubility of highly insoluble misfolded substrates after ubiquitylation and prior to proteasomal degradation (Fig. 7).
Cdc48 as a chaperone in nuclear PQC degradation
We initially predicted that San1 substrates would follow the same linear pathway to the proteasome. We discovered this was not the case and that some San1 substrates take different Cdc48-dependent and -independent routes correlated with their degree of insolubility (summarized in Table 1). Based on this correlation, the simplest explanation for the Cdc48 requirement in the degradation of San1 substrates is that Cdc48 uses a chaperone activity to maintain San1 substrate solubility, preventing aggregation prior to delivery to the proteasome. However, we cannot exclude that Cdc48 might also act to extract misfolded San1 substrates from aggregates should this occur en route to the proteasome. Cdc48/p97/VCP has been shown to prevent the aggregation of denaturated luciferase in vitro and in vivo (Song et al., 2007; Thoms, 2002). Furthermore, Cdc48/p97/VCP has been demonstrated to re-solubilize heat-denatured luciferase from insoluble aggregates as well as facilitate the clearance of pre-formed polyglutamine inclusions (Kobayashi et al., 2007). Based on these known chaperone activities, both scenarios for Cdc48 in San1-mediated nuclear PQC degradation could be operative (Fig. 7). Because the proteasome has difficulty degrading aggregated proteins (Holmberg et al., 2004; Snyder et al., 2003; Verhoef et al., 2002), maintenance of protein solubility upstream of the proteasome would be a crucial function for Cdc48 in a PQC degradation pathway.
The model we propose implies that substrate insolubility is a key parameter for Cdc48 dependency. The best test of the insolubility parameter was to find a way to alter the intrinsic solubility of Cdc48-dependent and -independent substrates. We accomplished this by the use of different proteins fused to each substrate. By swapping the GFPNLS and GAD fusions, we were able to alter the solubility of the San1 substrates in both directions, making insoluble substrates more soluble by use of the GAD fusion and making soluble substrates more insoluble by using the GFPNLS fusion. These solubility alterations corresponded with an appropriate change in Cdc48 dependency, with San1 substrates becoming more Cdc48 dependent as insolubility increased and less Cdc48 dependent as insolubility decreased. We do note that, whereas this was useful for demonstrating the correlation between Cdc48 dependency and substrate insolubility in our nuclear PQC studies, caution is needed in future studies exploring the Cdc48 dependency of protein degradation; the fusion protein used can influence the solubility of the protein to which it is fused, and correspondingly influence Cdc48 dependency.
Clarification on Cdc13 degradation
Of the San1 substrates we examined in this study, Cdc13-1 was one of the proteins that demonstrated no dependency on Cdc48 for its degradation. This was initially surprising because it was recently reported that wild-type Cdc13 is degraded in a Cdc48-dependent manner (Baek et al., 2012). However, we previously found that wild-type Cdc13 is stable (Gardner et al., 2005). We think the discrepancy between wild-type Cdc13 degradation in the study by Baek and colleagues and wild-type Cdc13 stability in our previous report can most likely be explained by the expression levels used in each study. We integrated a Myc–CDC13 at the endogenous CDC13 locus to allow expression from the endogenous CDC13 promoter (Gardner et al., 2005). In our case, Cdc13 protein levels would match the levels of its known partner protein Stn1 (Grandin et al., 1997). Baek and colleagues expressed 6His–GST–CDC13 from the strong GAL1 promoter on a plasmid. In this scenario, Cdc13 protein levels would be dramatically higher than the levels of Stn1, and it is known that stoichiometric imbalances can lead to the PQC degradation of the protein in excess (Hill and Cooper, 2000; Keppler and Archer, 2010; Mancini et al., 2000).
However, it is not clear why the overexpressed wild-type Cdc13 protein is degraded in a Cdc48-dependent manner, whereas the overexpressed Cdc13-1 protein is degraded in a Cdc48-independent manner. We think the proteins to which Cdc13 is fused in each study likely explains this difference. In the Baek and colleagues study, the authors used a Cdc13 protein fused to 6His–GST. Because GST forms a homodimer (Ji et al., 1992; Parker et al., 1990), its fusion with Cdc13 would result in the Cdc13 protein existing in a GST-dimer complex. Cdc48 is important for extracting proteins from complexes (Jentsch and Rumpf, 2007), and thus we think GST dimerization could force a dependency on Cdc48 that normally would not exist. The GAD and Myc epitope used in our studies here for Cdc13-1 do not impose any multimerization constraints on their fusion proteins. We think this must also be taken into consideration when using fusion proteins and epitope tags to explore dependency on Cdc48 for degradation.
Nucleus versus cytoplasm as sites of Cdc48 action on San1 substrates
All nuclear proteins are made by translation in the cytoplasm followed by import into the nucleus. Because Cdc48 is present in both the nucleus and cytoplasm, it is possible that the Cdc48 dependency we observe for insoluble San1 substrates is a function of Cdc48 acting both on the imported pool of substrate in nucleus and the transiting pool of protein in cytoplasm. Analysis of the substrate–GFP intensities in the nucleus and cytoplasm of the cells in Fig. 4 indicates that less than 5% of total visible San1 substrate is in the cytoplasm, making the cytoplasmic pool a very minor portion of the total cellular pool. Furthermore, inhibition of translation by addition of cycloheximide does not impede canonical nuclear import (Stathopoulos-Gerontides et al., 1999; van Drogen et al., 2001). Thus, it is probable that San1 substrates still present in the cytoplasm at the initiation of the degradation experiment will be rapidly and effectively transported into the nucleus after inhibition of translation. For these reasons, we think the stabilizing effects of cdc48 mutations on San1 substrates are primarily a function of Cdc48 action in the nucleus. Supporting this idea, we observed a general accumulation of Cdc48 in the nucleus of cells upon expression of insoluble San1 substrates.
Implications for loss of Cdc48/p97/VCP function in aggregation disorders
Human p97/VCP has been reported to colocalize with ubiquitylated aggregates in neurons taken from patients affected by many distinct types of neurodegenerative disorders including Alzheimer's, Parkinson's and Huntington's diseases and ALS (Hirabayashi et al., 2001; Ishigaki et al., 2004; Mizuno et al., 2003). Additionally, missense mutations in human p97/VCP underlie the etiology of IBMPFD (Watts et al., 2004) and might also be causative in some cases of ALS (Johnson et al., 2010). Loss of p97/VCP function in IBMPFD leads to the accumulation of ubiquitylated protein aggregates in affected cells (Hübbers et al., 2007; Janiesch et al., 2007; Ju et al., 2008; Weihl et al., 2006; Weihl et al., 2007). Here, we identified yeast Cdc48 as a crucial factor for the proteolysis of insoluble misfolded substrates in the nucleus. Furthermore, we revealed that the loss of Cdc48 function led to increased misfolded protein insolubility, in vivo inclusion formation and cell inviability when misfolded insoluble nuclear proteins were expressed. Further work will be required to determine whether possible failures of Cdc48/p97/VCP to clear insoluble misfolded proteins substantially contribute to the age-dependent pathology of neurodegenerative disorders associated with nuclear protein aggregation.
MATERIALS AND METHODS
Yeast strain constructions are as follows. Yeast strain RGY4461 (cdc48-3) was constructed by integrating pRG3026, which contains the cdc48-3 allele and the URA3 gene, next to the endogeneous CDC48 gene in BY4741 (Baker Brachmann et al., 1998). After growth of cells on medium containing 5-fluoro-orotic acid, to select for recombination between the tandem cdc48-3 allele and the CDC48 gene, correct cdc48-3 isolates were identified by their temperature-sensitive growth at 37°C. The correct cdc48-3 allele was confirmed by sequencing and complementation with a plasmid containing the wild-type CDC48 gene (Fig. 1C,D). Yeast strain RGY4463 (cdc48-3, san1Δ::KanMX) was constructed by deleting the SAN1 gene in yeast strain RGY4461 with the KanMX cassette amplified from pRS400 (Baker Brachmann et al., 1998). Correct san1Δ::KanMX integrants were confirmed by colony PCR. Yeast strains RGY4835 (parent), RGY4836 (san1Δ::KanMX), RGY4837 (cdc48-3) and RGY4944 (cdc48-3 san1Δ::KanMX) were constructed by deleting the SIR4 gene in yeast strains BY4741, RGY506, RGY4461 and RGY4463 respectively. After confirmation that the SIR4 gene was deleted by colony PCR, the plasmid pRG539 [PTDH3-1myc-sir4-9::LEU2 (Gardner et al., 2005)] was integrated at the SIR4 promoter. Yeast strain RGY5176 (PTDH3-3xHA-Ub::URA3, pdr5Δ::KanMX) was made by integrating the plasmid pRG145, which expresses a 3×HA epitope-tagged version of ubiquitin from the TDH3 promoter, at the TDH3 locus in the yeast strain RGY565 (BY4741 pdr5Δ::KanMX) that was obtained from the yeast deletion collection (Baker Brachmann et al., 1998). Yeast strain RGY5177 (PTDH3-3xHA-Ub::URA3, pdr5Δ::KanMX, san1Δ::NatMX) is congenic with RGY5176 except it has the SAN1 gene deleted with the NatMX cassette amplified from pRS40NAT (Andersen et al., 2008). Correct san1Δ::NatMX integrants in RGY5177 were verified by colony PCR and stabilization of San1 substrates. Yeast strain RGY5325 (PTDH3-3xHA-Ub::URA3, cdc48-3, pdr5Δ::KanMX) was constructed from RGY4461 by integrating the plasmid pRG145 as described above for yeast strains RGY5176 and RGY5177, and by deleting the PDR5 gene with the KanMX cassette from pRS400. Correct pdr5Δ::KanMX integrants were verified using colony PCR. Yeast strain RGY5416 (PTDH3-3xHA-Ub::URA3, cdc48-3, san1Δ::NatMX, pdr5Δ::KanMX) was constructed from RGY5325 by deleting the SAN1 gene with the NatMX cassette amplified from pRS40NAT. Correct san1Δ::NatMX integrants were verified by colony PCR. Yeast strains RGY5603 and RGY5682 (cdc48-3) were made by tagging the endogenous NUP60 gene in yeast strains BY4741 and RGY4461 with a 3′ integrating mCherry::HIS3MX6 cassette. Yeast strain RGY5624 was made by tagging the endogenous CDC48 gene in yeast strain BY4741 with a 3′ integrating mCherry::HIS3MX6 cassette.
Cycloheximide-chase degradation assays
Cycloheximide-chase degradation assays were performed as previously described (Fredrickson et al., 2011). Briefly, cells were grown in 3% raffinose medium to ∼1×107 cells/ml. Galactose was added to 3% and the cells were incubated 2 hours. Cycloheximide was added to 50 µg/ml and the cells incubated for 0–3 hours. Cells were lysed at the appropriate time point in 200 µl SUMEB buffer (8M Urea, 1% SDS, 10 mM MOPS, pH 6.8, 10 mM EDTA, 1 mM PMSF, 0.01% Bromophenol Blue). Proteins were resolved on SDS-PAGE gels, transferred to nitrocellulose and immunoblotted with anti-GFP (Sigma), anti-Myc (Sigma) or anti-Gal4AD (GAD) antibodies (Millipore). Equivalent loading and sample stability was measured by Ponceau S staining (Romero-Calvo et al., 2010).
Cells expressing 3HA–ubiquitin were grown in 3% raffinose medium to ∼1×107 cells/ml. Galactose was added to 3% and the cells were incubated 2 hours. Harvested cells were lysed in 200 µl SUME buffer (8M Urea, 1% SDS, 10 mM MOPS, pH 6.8, 10 mM EDTA, 1 mM PMSF). Lysates were clarified by centrifugation, and immunoprecipitation (IP) buffer (15 mM Na2HPO4, 150 mM NaCl, 2% Triton X-100, 0.1% SDS, 0.5% deoxycholate, 10 mM EDTA, pH 7.5) was added to each sample to a final volume of 1 mL. Anti-GFP or anti-Myc antibodies conjugated to Protein A Dynabeads (Dynal) were added to the lysates and incubated overnight at 4°C. The beads were washed once in IP buffer and twice in IP wash buffer (50 mM NaCl, 10 mM Tris-HCl, pH 7.5). For Fig. 2A–C, immunoprecipitated proteins were eluted in SUMEB buffer. For Fig. 2D, samples were divided in half and immunoprecipitated proteins were eluted in either SUMEB buffer (for lysate blot) or 100 mM acetic acid (for ubiquitin blot), which was subsequently neutralized 1∶1 in SUTEB buffer (8M Urea, 1% SDS, 100 mM unbuffered Tris-HCl, 10 mM EDTA, 0.01% bromophenol blue). All samples were incubated at 65°C for 10 minutes and then clarified for 5 minutes by centrifugation at 12,800×g. Proteins were resolved on SDS-PAGE gels, transferred to nitrocellulose and immunoblotted with anti-GFP (Abcam), anti-Myc (Sigma) or anti-HA antibodies (Sigma).
Sedimentation assays were adapted from a previously described protocol (Theodoraki et al., 2012). Cells were grown in liquid medium with 3% raffinose to ∼1×107 cells/ml. Galactose was added to 3% and the cells were incubated 2 hours. 5 ml of cells were harvested and lysed in lysis buffer (100 mM Tris-HCl pH 7.5, 200 mM NaCl, 1 mM EDTA, 1 mM DTT, 5% glycerol and 0.1% Nonidet P40) plus PMSF by vortexing for 5 minutes at 4°C with 100 µl of 0.5-mm acid-washed glass beads. To remove unlysed cells, lysates were centrifuged at 700 g for 1 minute at 4°C. 50 µl lysate, representing the ‘total lysate’, was removed and added to 50 µl SUMEB. 100 µl remaining lysate was centrifuged at 12,800 g for 15 minutes at 4°C. 100 µl supernatant, representing the ‘soluble fraction’, was added to 100 µl SUMEB. The pellet, representing the ‘insoluble fraction’, was resuspended in 100 µl lysis buffer and 100 µl SUMEB. All samples were incubated at 65°C for 10 minutes and then clarified for 5 minutes by centrifugation at 12,800 g. Proteins were resolved on SDS-PAGE gels, transferred to nitrocellulose and immunoblotted with anti-GFP (Sigma) or anti-Gal4AD (GAD) antibodies (Millipore).
Cells were grown in 3% raffinose medium to ∼0.5×107 cells/ml. Galactose was added to 3% and the cells incubated for 6 hours. Cells were harvested, fixed in 4% paraformaldehyde in 0.1 M sucrose for 15 minutes, washed in wash buffer (1.2 M sorbitol, 0.4 M KPO4), stained with DAPI for 10 minutes in wash buffer plus 2% Triton X-100 and washed two times in wash buffer. Cells were imaged on a Nikon Eclipse 90i with a 100× objective, filters for GFP [HC HiSN Zero Shift filter set with excitation wavelength (450–490 nm), a dichroic mirror (495 nm) and emission filter (500–550 nm)], mCherry [HC HiSN Zero Shift filter set with excitation wavelength (530–560 nm), a dichroic mirror (570 nm) and emission filter (590–650 nm)] or DAPI [HC HiSN Zero Shift filter set with excitation wavelength (325–375 nm), a dichroic mirror (400 nm) and emission filter (435–485 nm)], and a Photometrics Cool Snap HQ2 cooled CCD camera with NIS-Elements acquisition software.
All western blots were scanned using an Epson Perfection V350 Photo scanner at 300 dpi. All images were processed with a Mac iMac or Pro computer (Apple) using Photoshop CS (Adobe) and quantified using ImageJ.
We thank Randy Hampton (Division of Biological Sciences, University of California - San Diego, USA), Jeff Brodsky (Department of Biological Sciences, University of Pittsburgh, USA), Jeff Laney (Biochemistry and Cellular & Molecular Biology, University of Arizona, USA), Peter Kaiser (Department of Biological Chemistry, University of California - Irvine, USA), Ray Deshaies (Division of Biology, California Institute of Technology, USA) and Yihong Ye (National Institute of Diabetes and Digestive and Kidney Diseases, USA) for plasmids and yeast strains. We thank Ning Zheng (Department of Pharmacology, University of Washington, USA) and Rachel Klevit (Department of Biochemistry, University of Washington, USA) for scientific input and insights on differential requirements.
The authors declare no competing interests.
P.S.G performed the majority of the experiments, with contributions from S.V.C.C. and R.G.G. P.S.G and R.G.G. designed the experiments, performed the data analyses, generated the figures, and wrote the manuscript.
This worked was supported by the National Institute on Aging (National Institutes of Health) [grant number R01AG031136 to R.G.G.]; an Ellison Medical Foundation New Scholar Award in Aging (to R.G.G.); and a Marian E. Smith Junior Faculty Award (to R.G.G.). Deposited in PMC for release after 12 months.
Supplementary material available online at http://jcs.biologists.org/lookup/suppl/doi:10.1242/jcs.141838/-/DC1
- Received August 29, 2013.
- Accepted February 1, 2014.
- © 2014. Published by The Company of Biologists Ltd