Cdc48 regulates intranuclear quality control sequestration of the Hsh155 splicing factor in budding yeast

Protein quality control (PQC) refers to the process of triage for non-native, unassembled or mislocalized proteins through a combination of sequestration, degradation and refolding (Hartl et al., 2011). PQC is a part of normal cellular homeostasis and is particularly important during stress, when non-native proteins assemble in visible aggregate structures at different subcellular locations. Various proteotoxic stresses elicit such a coordinated PQC response, with the best characterized of these being heat stress (Wallace et al., 2015) and expression of amyloid-forming proteins (Kaganovich et al., 2008). PQC responses depend on networks of molecular chaperones, on the action of stress-activated protein aggregase and disaggregase activities (Hartl et al., 2011) and on protein degradation by the ubiquitin–proteasome system and autophagy (Kaganovich et al., 2008; Kumar et al., 2016).

During stress in yeast, specific types of aggregate structures with different components and behaviors form (Saarikangas and Barral, 2016). In the cytoplasm, aggregates at the vacuolar membrane, called the insoluble protein deposit (IPOD), are repositories for irreversibly aggregated proteins destined for degradation, such as amyloids (Kaganovich et al., 2008). Smaller dynamic cytoplasmic aggregates called Q-bodies (quality control bodies) have also been described for protein aggregate reporters such as the temperature-sensitive mutant Ubc9-ts and VHL (Escusa-Toret et al., 2013; Ruan et al., 2017; Zhou et al., 2014). Finally, at the nuclear envelope, the juxtanuclear quality control site (JUNQ) forms peripheral to but outside the nucleus, whereas the intranuclear quality control site (INQ) forms peripheral to and inside the nucleus (Kaganovich et al., 2008; Miller et al., 2015). Most yeast studies of PQC focus on aggregation of the replication fork checkpoint proteins (i.e. Mrc1; Gallina et al., 2015), the activation of DNA repair proteins (i.e. Mus81; Saugar et al., 2017) or rewiring of the spliced transcriptome (i.e. Hsh155; Mathew et al., 2017).

INQ formation is driven by the small heat shock protein Hsp42 and the aggregase protein Btn2 (Ho et al., 2019), whereas INQ dissolution is regulated by Hsp104 and the Hsp70 system (Miller et al., 2015; Saugar et al., 2017). Here, we show that Cdc48 is also a stress-inducible component on INQ structures and that it plays an enzymatic role in INQ turnover. Cdc48 plays numerous normal and stress-responsive roles in cells, including in nuclear PQC (Gallagher et al., 2014; Maric et al., 2017; Pantazopoulou et al., 2016; Verma et al., 2011). Here, we show that Cdc48 is concentrated at INQ following DNA damage, and that its activity is important to suppress the spontaneous aggregation of splicing factor Hsh155 in INQ. Cdc48 physically interacts with Hsh155, even under non-stress conditions, and promotes its assembly with partners in the SF3B complex of the spliceosome. Taken together, these results demonstrate a previously unrecognized role for Cdc48 in protein triage to INQ and suggest unappreciated roles for Cdc48 in splicing.

We previously screened hundreds of genome stability regulatory proteins for relocalization in response to DNA damage, and identified uncharacterized stress-induced relocalization of Cdc48–GFP (Mathew et al., 2017). Cdc48 is widely distributed in unstressed cells, but is enriched at a nuclear periphery/ER localization due to its role in ER-associated degradation (ERAD) (Fig. 1A). Following methyl methanesulfonate (MMS) treatment, Cdc48 relocalized to nuclear and non-nuclear foci (Fig. 1A; Fig. S1A). To test whether these foci were INQ, we expressed Cdc48–GFP together with Hos2–mCherry, a known marker of INQ (Gallina et al., 2015; Tkach et al., 2012). Cdc48–GFP foci colocalized with Hos2–mCherry foci, suggesting that Cdc48 was accumulating at INQ following MMS treatment (Fig. 1B,C). Cdc48 also colocalized with Hos2 at peripheral aggregates, which may be cytoplasmic PQC (CytoQ) sites (Fig. 1B,C). We next ablated INQ formation by deleting the chaperones HSP42 and BTN2, which are essential for INQ protein localization (Gallina et al., 2015; Mathew et al., 2017; Miller et al., 2015). Cdc48–GFP foci were strongly diminished in MMS-treated hsp42Δ or btn2Δ cells, showing that INQ formation was essential for Cdc48 foci (Fig. 1D; Fig. S1B). In contrast, deletion of the Hsp40 homolog Apj1, which has previously been implicated in disaggregation and turnover of INQ-resident proteins, dramatically increased the frequency of Cdc48–GFP foci (Fig. 1D; Fig. S1B) (den Brave et al., 2020). Hos2–mCherry was used as a control, and it showed the expected decrease of INQ formation in cells lacking HSP42 or BTN2 and increased INQ frequency in cells lacking APJ1 (Fig. S1B). Taken together, the results of these experiments suggest that Cdc48 foci mark INQ structures induced by MMS treatment and are influenced by INQ regulatory chaperones.

Because Cdc48 functions in protein biogenesis and turnover, we suspected that it might have a functional role at INQ. We assembled a set of CDC48 temperature-sensitive alleles, which grew robustly at the permissive temperature of 25°C but exhibited a range of fitness defects (Fig. S1C) (Li et al., 2011). Examining the localization of an INQ substrate in the most severe allele, cdc48-4601 (hereafter called cdc48-4) (Fig. S1E), revealed spontaneous Hsh155–GFP and Hos2–GFP foci at nuclear and cytoplasmic aggregate sites (Fig. 2A; Fig. S1D). MMS treatment further increased the proportion of cells with visible aggregates beyond that seen for MMS-treated wild-type (WT) cells (Fig. 2B). We confirmed that other CDC48 temperature-sensitive alleles also increased spontaneous INQ formation to varying degrees (Fig. 2C). Consistent with a functionally important role, cdc48-4 alleles were significantly impaired in recovery from a 2 h treatment of MMS (Fig. 2D). Although there are many possible mechanisms for the MMS sensitivity of a CDC48 temperature-sensitive allele, our data show that Cdc48 function is important for regulating the INQ localization of Hsh155 in normal and DNA-damaging conditions. Finally, because Cdc48 uses its ATPase activity to disassemble or unfold proteins, we tested the ability of enzymatically active or dead alleles to rescue the cdc48-4 phenotype. Whereas WT CDC48 plasmids suppressed Hsh155 foci, an ATPase-deficient CDC48^E315Q allele could not (Fig. 2E,F) (Gallagher et al., 2014). Thus, Cdc48 ATPase activity is essential to suppress aggregation of Hsh155–GFP into INQ structures.

Previous studies have linked Cdc48 and its co-factors to proteostasis of nuclear proteins via the San1 E3 ubiquitin ligase and the inner nuclear membrane proteins Asi1–Asi3; however, we saw no significant effect of SAN1 deletion on INQ (Fig. S1F) (Gallagher et al., 2014; Pantazopoulou et al., 2016). In addition, Cdc48 is involved in regulation of protein–DNA complexes, including the removal of RNA polymerase from damaged DNA (Verma et al., 2011), the removal of replisomes during DNA replication termination (Maric et al., 2017), at protein–DNA crosslinks (Amunugama et al., 2018) and during the release of condensin complexes (Thattikota et al., 2018). Thus, Cdc48 is well-positioned to regulate protein sequestration pathways such as INQ formation following DNA damage detection. Other groups have noted the importance of SUMO to INQ formation (Gallina et al., 2015), and we also found that reducing SUMOylation with ubc9-ts (SUMO E2) or SIZ1 (SUMO E3) deletion reduced INQ localization of Hsh155–GFP in MMS-treated WT or cdc48-4 cells (Fig. 3A,B). Indeed, mutation of the SUMO-interacting motif of Cdc48 (CDC48^SIM) impairs its ability to rescue cdc48-4 cells, relative to rescue by WT CDC48 (Fig. 3C) (Bergink et al., 2013). Because we used Hsh155–GFP as a marker for INQ, we elected to directly test whether Hsh155 is SUMOylated using a Smt3–His×7 purification scheme. Whereas SUMOylation of the Rfa1–GFP control was easily detectable upon MMS treatment, no detectable pulldown of Hsh155–GFP with an Smt3–His×7 tag was observed with or without MMS (Fig. 3D). Thus, the identity of any SUMO targets important for regulating INQ will require additional study beyond the scope of this work. However, our findings again highlight the importance SUMO in regulating INQ.

If Cdc48 is important for the prevention or dissolution of INQ structures, then it could regulate the stability of INQ substrate proteins under stress. Consistent with this idea, Hsh155–GFP protein was more abundant in the cdc48-4 mutant, and more stable in cycloheximide chase experiments, than it was in WT (Fig. 4A). Because cycloheximide treatment can itself affect protein sequestration (Zhou et al., 2014), we also monitored the impact of Cdc48 on Hsh155 lifetime with an orthogonal approach using in vivo tandem fluorescent timer (tFT) fusions (Khmelinskii et al., 2012). The ratio of GFP to mCherry fluorescence in Hsh155–tFT fusions was significantly red-shifted in cdc48-4 cells, indicating a longer relative protein lifetime (Fig. 4B; Fig. S1G,H). These data are consistent with INQ serving a protective role for substrate proteins, because Hsh155 both becomes more abundant and aggregates in cdc48-4 cells. Cdc48 could affect the stability of Hsh155 through direct protein–protein interactions. Indeed, we observed co-precipitation of Hsh155–GFP with Cdc48–TAP and via reciprocal pulldown (i.e. Cdc48–GFP with Hsh155–TAP) under both untreated and MMS-treated conditions (Fig. 4C,D).

Hsh155 is a splicing factor that exists in a complex called SF3B with Hsh49 and Cus1. We previously found that Hsh155 disassembles from the SF3B complex upon MMS treatment (Mathew et al., 2017). To assess the potential impact of Cdc48 on SF3B complex stability, we monitored pulldown of Hsh155–GFP in WT and cdc48-4 cells with or without MMS treatment using co-immunoprecipation. Loss of Cdc48 function greatly diminished the presence of the Hsh155–Cus1 complex, even in untreated cells (Fig. 4E). This suggests that Cdc48 may play a role in the integrity of the SF3B complex even in the absence of DNA-damage signals. Whereas cdc48-4 grew too poorly to easily monitor splicing activity, the cdc48-3 allele that showed an intermediate induction of Hsh155 foci (Fig. 2C) did show reduced splicing flux in a LacZ-based splicing efficiency assay (Fig. 4F) (Galy et al., 2004; Tam et al., 2019).

Spatial control of PQC is a fundamental feature of stress responses in eukaryotic cells (Sontag et al., 2014). Here, we show that Cdc48, a key regulator of numerous cellular PQC activities, is also associated with INQ, where it interacts with, reduces the accumulation of, or speeds the turnover of the INQ substrate Hsh155 (Fig. 4G). This defines a novel and complementary function for Cdc48 in nuclear protein quality control. Loss of SUMO-conjugating enzyme function or E3 SUMO ligase function reduced INQ formation, consistent with a role for this modification in organized INQ structures. Considerably more work is needed to understand the regulation of INQ substrate protein localization and turnover. Our data specifically implicate Cdc48 in regulating the sequestration, complex formation and stability of the spliceosomal protein Hsh155 within the SF3B complex. We previously found that Hsh155 sequestration may regulate ribosomal protein gene expression through reduced splicing efficiency of ribosomal protein gene transcripts (Mathew et al., 2017). In this regard, it is notable that Cdc48 is also a critical regulator of ribosomal quality control mechanisms and can sense ribosomal stress (Brandman et al., 2012; Defenouillère et al., 2013). It is possible that these responses are coordinated during the DNA-damage response. Regardless, our data define a new function for Cdc48 in regulating the deposition of proteins at the INQ and potentially in the regulation of splicing complexes.

There is currently little evidence linking Cdc48 to splicing. Interestingly, a study of motor neuron transcriptome dynamics in induced pluripotent stem cell differentiation models of amyotrophic lateral sclerosis (ALS), where VCP (the human ortholog of CDC48) was mutated, showed that abnormal intron retention events are increased in VCP-mutated cells (Luisier et al., 2018). Thus, a role for Cdc48/VCP in splicing across species is possible, and we hope that additional research will elucidate how Cdc48 may affect splicing. In this regard, it is notable that the existence of an orthologous human intranuclear quality control response could involve disease genes, including the ALS gene VCP and the Hsh155 ortholog SF3B1, which is frequently mutated in various types of cancer (Yoshida and Ogawa, 2014). Given the proximity of the INQ to the nucleolus, we speculate that human studies highlighting the nucleolus as a phase-separated quality control compartment may reflect a conserved phenomenon homologous to the INQ in yeast (Frottin et al., 2019; Tkach et al., 2012). If so, then studies in model organisms should provide insight to nuclear protein quality control pathways that could modulate human disease.

Yeast (Saccharomyces cerevisiae) strains used in this study, including database IDs and genotypes, primers and plasmids, are listed in Table S1. All strains were in the s288c background and were grown under nutrient rich YPD medium (Sherman, 2002) or synthetic medium lacking amino acids (SC; Sherman, 2002) unless otherwise indicated. Schematics of the CDC48 mutant alleles are shown in Fig. S1E. Sequence information came from in-house sequencing or from publications (Simões et al., 2018; Verma et al., 2011). Serial dilution assays and growth curve analyses were performed as described previously (Mathew et al., 2017; Stirling et al., 2011). Briefly, cells with identical optical density (OD) were serially diluted tenfold and spotted on YPD plates using a 48-pin replica pinning manifold then incubated at the indicated temperatures for 72 h. Growth curves were analyzed in a Tecan M200 plate reader monitoring OD600 nm every 30 min for 48 h at 30°C. For all MMS treatment conditions cells were exposed to 0.05% MMS (∼99%; Sigma) in synthetic complete medium for 2 h unless otherwise noted. For growth curves, MMS was then washed out and replaced with fresh medium.

Imaging and subsequent analysis were performed as described previously (Mathew et al., 2017). Log-phase yeast were mounted on slides pre-treated with concanavalin A, in SC growth medium. Immobilized cells were imaged using an Objective HCX PL APO 1.40 NA oil immersion 100× objective on an inverted Leica DMi8 microscope with a motorized DIC (differential interference contrast) turret (for DIC imaging) and a filter cube set for FITC/TRITC (for GFP and mCherry fluorescence imaging). The images were captured at room temperature using a scientific complementary metal oxide semiconductor (sCMOS) camera (ORCA Flash 4.0 V2; Hamamatsu), collected using MetaMorph Premier acquisition software and post processed [including gamma adjustments, counting of cells with/without foci and foci intensity measurements for tandem fluorescent timer (tFT) experiments] using ImageJ (National Institutes of Health). For all microscopy experiments, the significance of the differences was determined using Prism version 5 or higher (Graphpad). For foci intensity measurements in tFT experiments, samples were compared with t-tests or ANOVA; Prism performs F-tests for variance as part of this analysis. For comparisons of proportions, Fisher tests were used, and P-values were Holm–Bonferroni corrected in the event of multiple comparisons. Sample sizes and specific statistical details for each image analysis are listed in the figure legends.

Co-immunoprecipitation was performed using yeast strains containing TAP tagged Cus1, Cdc48 and/or GFP tagged Hsh155, treated with or without MMS. TAP-tagged protein was captured using IgG sepharose fast flow beads (Sigma) and proceeded as described previously (Leung et al., 2016). Immunoblotting was carried out with mouse anti-GFP (1:500; ThermoFisher; Cat#. MA5-15256) and rabbit anti-TAP (1:1000; ThermoFisher; Cat#. CAB1001) antibodies. For western blotting, whole-cell extracts were prepared using tricholoroacetic acid (TCA) extraction and blotted with mouse anti-GFP, rabbit anti-RFA (1:1000; Agrisera; Cat#. AS07 214) or mouse anti-PGK1 (1:500; Santa Cruz; Cat#. Sc130335) antibodies, as described previously (Gallina et al., 2015).

Overnight cultures of the indicated strains were diluted to below an OD600 nm of 0.2 and allowed to progress into logarithmic phase before collection. The cells were treated with or without cycloheximide (CHX) at a concentration of 200 μg/ml for the indicated times (0–120 min). The final samples of 2×10⁷ cells were collected by centrifugation, and whole-cell extracts were prepared by TCA extraction and used for immunoblotting.

The splicing assay protocol was adapted and performed as described previously (Galy et al., 2004). All measurements were taken with three individual transformants per replicate for a total of five replicates. Cells were struck as a patch on SC –leucine plates, then replica plated to glycerol–lactate-containing SC medium (1.6% v/v glycerol and 1.5% w/v lactate) without leucine (GGL −leu). Cells from each patch were inoculated in liquid GGL −leu medium for 2 h at 34°C and induced with galactose (2% final concentration) for 1.5 h before treatment with MMS (0.05% final concentration) for 30 min. Cells were lysed and assayed for β-galactosidase using a Gal-Screen β-galactosidase reporter gene assay system for yeast or mammalian cells (Applied Biosystems), as per the manufacturer's instructions, and read with SpectraMax i3 (Molecular Devices). Relative light units were normalized to cell concentration as estimated by measuring OD600.

SUMO-conjugated proteins were detected by performing protein pulldowns under denaturing conditions, as described previously (Ulrich and Davies, 2009). In brief, strains of interest were transformed with a plasmid containing His-tagged SUMO (Smt3–His×7; Addgene) under the control of a copper-inducible promoter. Logarithmically growing cells were harvested at an OD600 nm of 0.6–0.8, and a total of 10⁹ cells were collected by centrifugation (4000 rpm, 3724 g; 5 min at 4°C). Pelleted cells were washed and resuspended in 5 ml of pre-chilled water. Cells were then lysed with 800 µl of 1.85 M NaOH containing 7.5% (v/v) 2-mercaptoethanol, followed by 20 min incubation on ice. Protein precipitation was carried out by adding 800 µl 55% TCA on ice for 20 min. Precipitates were collected by centrifugation (8000 g, 20 min, 4°C) and were resuspended in 1 ml Buffer A (6 M guanidine hydrochloride, 100 mM sodium phosphate and 10 mM Tris-HCl, pH 8.0) and incubated on a rotating block at room temperature for 1 h to solubilize the precipitate completely. The resulting solution was then centrifuged (16,000 g, 10 min, 4°C), and supernatant was transferred to tubes containing 30 µl precleared Ni-NTA agarose beads (Qiagen) in the presence of 0.05% Tween-20 and 15 mM imidazole. These were incubated overnight on a rotating block at room temperature. Following binding, the beads were washed twice with Buffer A containing 0.05% Tween-20 and four times with Buffer C (8 M urea, 100 mM sodium phosphate and 10 mM Tris-HCl, pH 6.3) containing 0.05% Tween-20. Beads were centrifuged (200 g, 15 s) and supernatant was completely removed. His–SUMO conjugates on the beads were eluted by adding 30 µl HU sample buffer (8 M urea, 5% SDS, 200 mM Tris-HCl, pH 6.8, 0.1 mM EDTA, 0.5% Bromophenol Blue and 15 mg/ml dithiothreitol) and heating at 70°C for 10 min. Resultant protein extracts were subjected to standard western blotting and probed for SUMO and other proteins of interest using antibodies as described above.

The peer review history is available online at https://jcs.biologists.org/lookup/doi/10.1242/jcs.252551.reviewer-comments.pdf