Endoplasmic reticulum (ER)-associated protein degradation requires the dislocation of selected substrates from the ER to the cytosol for proteolysis via the ubiquitin–proteasome system. The AAA ATPase Cdc48 (known as p97 or VCP in mammals) has a crucial, but poorly understood role in this transport step. Here, we show that Ubx2 (Sel1) mediates interaction of the Cdc48 complex with the ER membrane-bound ubiquitin ligases Hrd1 (Der3) and Doa10. The membrane protein Ubx2 contains a UBX domain that interacts with Cdc48 and an additional UBA domain. Absence of Ubx2 abrogates breakdown of ER proteins but also that of a cytosolic protein, which is ubiquitinated by Doa10. Intriguingly, our results suggest that recruitment of Cdc48 by Ubx2 is essential for turnover of both ER and non-ER substrates, whereas the UBA domain of Ubx2 is specifically required for ER proteins only. Thus, a complex comprising the AAA ATPase, a ubiquitin ligase and the recruitment factor Ubx2 has a central role in ER-associated proteolysis.
ER-associated protein degradation (ERAD) is part of a quality control system that ensures the delivery of only correctly folded secretory proteins to their final destinations. Recognition of malfolded polypeptides occurs in the ER lumen; however, their proteolysis takes place in the cytosol. As a result, a transport step from the ER to the cytosol must precede their degradation. This process, termed protein dislocation, most probably occurs through an aqueous channel1,2.
Protein dislocation requires conjugation of the transported polypeptides with ubiquitin3,4. In yeast, two ubiquitin ligases are central to this process. Hrd1 is an integral membrane protein that exposes a RING-finger domain on the ER surface5,6. It associates with another ER membrane protein, Hrd3, which contains a large ER lumenal domain. Both domains are essential for the function of the ligase7. The main ubiquitin-conjugating enzymes that function in the Hrd1 pathway are Ubc1 and Cue1-assembled Ubc7 (refs 5, 8). Doa10 is another RING-finger ubiquitin ligase of the ER membrane9. Genetic evidence indicates that Doa10 functions along with Ubc6 and Cue1-assembled Ubc7 (ref. 10). Both ligases seem to recognize distinct sets of substrates and contribute to the ubiquitin–proteasome system (UPS) activity at the ER surface. Notably, substrates of Doa10 are not exclusively lumenal or membrane-bound but also include cytosolic or nuclear proteins, such as the transcriptional regulator Mat-α2. Mat-α2 contains a degradation signal termed Deg1 that is recognized by Doa10 (ref. 10).
In addition to ubiquitin conjugation, the action of the AAA ATPase complex Cdc48, which consists of Cdc48, Npl4 and Ufd1, is necessary for dislocation4,11,12,13,14. Although an interaction between p97 and the mammalian ubiquitin ligase gp78 has recently been described15, the function of Cdc48 in ERAD is not yet clear. This ATPase may act as a motor that actively pulls dislocated molecules out of a putative export channel16. Alternatively, it may mobilize already dislocated and ubiquitin-conjugated molecules that are bound to the ER surface, possibly by ubiquitin-binding domain proteins4,13. A hint for such a mobilizing function was provided by the analysis of npl4-1 and ufd1-1 mutants. In these cells, ubiquitin-conjugated and completely dislocated forms of the ERAD substrate CPY* are found attached to the cytosolic face of the ER membrane4. Irrespective of the precise function, binding of the Cdc48 complex to the ER membrane is required for release of substrates16. Recently, VIMP has been suggested to serve as a membrane receptor for the p97 complex in mammals17. However, the function of VIMP in ERAD is not fully understood and most importantly, yeast does not contain a homologue of VIMP. This is surprising because most known ERAD components are highly conserved from yeast to humans2.
The data presented here demonstrate that yeast Ubx2 is associated with both the Cdc48 complex and with ERAD ubiquitin ligases. Cells that lack Ubx2 exhibit severe defects in ERAD of ER lumenal and membrane-bound substrates. Moreover, the interaction of the Cdc48 complex with membrane-bound ubiquitin ligases is disturbed in Δubx2 cells. Intriguingly, our data indicate an important role for the UBA domain of Ubx2 in ERAD.
We were interested in how the Cdc48 complex is integrated into the yeast ERAD pathway. Therefore, we searched for an ER membrane protein that could bind to the Cdc48 complex. Potential candidates should contain an UBX domain, which binds Cdc48 (refs 18,
To test whether Ubx2 is involved in the turnover of ERAD substrates, we determined the breakdown of CPY* by pulse-chase analysis. CPY* is a mutant, and thus malfolded, version of carboxypeptidase Y, which is retained in the ER lumen. Its degradation depends on a functional Cdc48 complex4,11,14. Loss of Ubx2 impaired CPY* proteolysis significantly (Fig. 1a). In similar experiments, we analysed the breakdown of membrane-bound Sec61-2, a short-lived form of the central subunit of the translocation pore22. Also the degradation of this substrate of the Hrd1 ligase is delayed in Δubx2 cells (Fig. 1b). We also investigated the turnover of Ubc6, a carboxy-terminal anchored ER membrane protein that is degraded via the Doa10 ubiquitin ligase9,23. Ubc6 breakdown is delayed in cdc48-1 and ufd1-1 mutants (Fig. 1c). Accordingly, the absence of Ubx2 results in a stabilization of Ubc6 comparable to that in Δdoa10 cells (Fig. 1d). Taken together, these results indicate an involvement of Ubx2 in Hrd1- as well as Doa10-dependent ERAD pathways. By contrast, deletions of ubx6 and ubx7 have no influence on CPY* turnover (data not shown).
Next, we investigated whether Ubx2 is membrane associated. To this end, we replaced the endogenous Ubx2 with fusion constructs of this protein that express a HA tag at the C- and N termini, respectively. When crude extracts of such cells were divided in soluble and membrane fractions by centrifugation, either Ubx2 fusion protein was found in the membrane pellet. Moreover, both versions were resistant to extraction with agents that strip peripheral proteins from membranes. Cdc48 and Ubc7 were washed off efficiently by urea, whereas Ubx2 behaves like the integral membrane protein Sec61 (Fig. 2a). Next we investigated the orientation of Ubx2 in the membrane. Upon addition of protease to crude cellular extracts, the HA tag was efficiently degraded, irrespective of its localization at the N- or C terminus of Ubx2 (Fig. 2b). Taken together, these data suggest a loop insertion of Ubx2 in the membrane with both termini exposed to the cytoplasm (Fig. 2c). This is in agreement with predicted transmembrane segments in the Ubx2 sequence.
To determine the physical interactions of Ubx2 we tagged the protein at the C terminus with either Protein A (Ubx2–ProtA) or glutathione S-transferase (Ubx2–GST). Both constructs were fully active with respect to Ubc6 turnover (data not shown). When Ubx2–ProtA was immunoprecipitated, Cdc48 was copurified as expected. Interestingly, both Doa10 and Hrd1 were also found in the precipitate (Fig. 3a). Such an association is in agreement with a direct role for Ubx2 in the breakdown of substrates of both ubiquitin ligases. Conversely, we were able to precipitate Ubx2 with a c-Myc-tagged version of Doa10 (Fig. 3b), which is also active with respect to Ubc6 degradation (data not shown). A RING-finger mutant of Doa10 still interacted with Ubx2–ProtA (Fig. 3c), albeit to a much lower extent. This indicates that the Doa10–Ubx2 complex is stabilized by interactions with ubiquitin-conjugated substrates. Of course, we cannot exclude that the weakening of Ubx2 binding is caused by structural changes in the cytoplasmic domain of mutant Doa10.
To analyse the function of Ubx2, we precipitated the ubiquitin ligases or the Cdc48 complex from wild-type and Δubx2 cells. Immunoprecipitation of a functional C-terminally HA-tagged version of Npl4 (Npl4–HA; data not shown) revealed that it is not only associated with Cdc48, as expected, but also with Hrd1 and Hrd3 (Fig. 3d). Notably, neither Hrd1 nor Hrd3 copurified with Npl4 in Δubx2 cells, indicating that Ubx2 mediates the interaction of this ligase complex with the assembled Cdc48 complex (Fig. 3d). Consistently, a pull-down of Doa10 from wild-type cells contained Cdc48, whereas that of Δubx2 cells did not (Fig. 3e). The physical interaction of Doa10 with Ubc6 and Cue1-assembled Ubc7 did not depend on Ubx2. This strengthened the idea that yeast Ubx2 mediates interaction of the Cdc48 complex with both ERAD ubiquitin ligases. The cytosolic amount of both Cdc48 and Npl4 remained unchanged in these experiments (Fig. 3f).
If Ubx2 has a general effect on proteolysis of ER proteins, the ubx2 deletion should increase malfolded proteins in the ER, which in turn should activate the unfolded protein response (UPR). The UPR is a signalling cascade that measures the content of aberrant proteins in the ER and induces the expression of the transcription factor Hac1 by unconventional RNA splicing24. Hac1 in turn increases the transcription of genes that contain a UPR response element in their promoters. Hence, we determined splicing of HAC1 RNA by northern analysis (Fig. 4a). In unstressed wild-type cells, only the unspliced inactive variant HAC1u is detectable (Fig. 4a; lane 1). After treatment of yeast cells with DTT, which increases the load of unfolded proteins in the ER, the faster migrating spliced variant HAC1i appears (Fig. 4a; lane 2). As expected, the splicing of HAC1 RNA depends on the presence of the UPR signalling kinase IRE1 (Fig. 4a; lanes 5 and 6). Cells lacking Hrd1 behave similarly to wild-type cells (Fig. 4a; lanes 3 and 4). However, in Δubx2 cells, HAC1i mRNA is already detectable under normal growth conditions (Fig. 4a; lanes 7 and 8), indicating a UPR induction. Interestingly, cells expressing HA–Ubx2 showed an increased UPR, comparable to Δubx2 cells (Fig. 4a; lanes 9 and 10). During our studies we noticed that HA–Ubx2 partially corrected the growth deficiencies of Δubx2 cells (data not shown) but still interfered with the turnover of ER proteins (Fig. 5b).
In addition, we measured the UPR using a β-galactosidase reporter gene under the control of a hybrid promotor containing a UPR response element25. Also in this assay, cells expressing HA–Ubx2 showed a strong induction of the UPR (Fig. 4b). For unknown reasons, we were unable to measure the UPR in the β-galactosidase assay in Δubx2 cells.
We further explored the function of the N terminus of Ubx2. The modification with the HA tag was close to the N-terminally located UBA domain and obviously affects the function of Ubx2 to some extent. Because Cdc48 is also involved in proteolysis of soluble non-ER proteins, we investigated a possible function of Ubx2 in this process. To this end, we determined the turnover of the Deg1 degradation signal fused to GFP26. Degradation of this hybrid protein relies on Doa10 (data not shown) and also depends on the Cdc48 complex, because breakdown of this reporter is delayed in cdc48-1, cdc48-3 and ufd1-1 mutants (Fig. 5a and data not shown). Deg1–GFP degradation is also abolished in Δubx2 cells, as is that of CPY* and Ubc6 (Fig. 5b). Surprisingly, cells that express HA–Ubx2 are able to degrade Deg1–GFP with wild-type kinetics, whereas the turnover of the ERAD substrates CPY* and Ubc6 is strongly impaired (Fig. 5b). This suggests that the N terminus, containing the UBA domain, is crucial for ERAD but dispensable for the degradation of soluble cytoplasmic or nuclear proteins.
We have identified a new component that is necessary for proteolysis of yeast ER proteins. This factor, Ubx2, recruits the Cdc48 complex to the ER-membrane-bound ubiquitin ligases Doa10 and Hrd1. In Δubx2 cells, both the association of Cdc48 with Npl4 and the interaction of Doa10 with the involved ubiquitin-conjugating enzymes remain intact. Thus, it seems likely that Ubx2 mediates the interaction of the assembled Cdc48 complex with the ERAD ubiquitin ligases Doa10 and Hrd1. However, cells lacking Ubx2 exhibit no major reduction in the amount of membrane-bound Cdc48, as is indicated in the membrane preparations shown in Fig. 3d, e. This indicates that other membrane-recruitment factors for Cdc48 may exist in addition. Because we observed a reduced turnover of membrane-bound and lumenal ERAD substrates in Δubx2 cells, we conclude that the Ubx2-mediated interaction of the Cdc48 complex with the ubiquitin ligases is crucial for dislocation of ER proteins. This view is further strengthened by the fact that Δubx2 cells show an increased UPR (Fig. 4) and a synthetic lethality with deletions of ire1, the sensoring kinase of the UPR (data not shown). Furthermore, UBX2 expression is controlled by the UPR27. Ubx2 deletion mutants were originally described to exhibit increased export of a secretory protein and the corresponding gene was therefore named SEL1 (secretion lowering)28. In the light of our results, we assume that in Δubx2 mutants, the ER quality control system is overloaded and that malfolded polypeptides manage to escape from this compartment. However, it seems that Ubx2 function is not restricted to ERAD but also concerns other UPS substrates as well18. The slow growth phenotype of cells lacking Ubx2 that does not correlate with the phenotypes of Δhrd1 and Δdoa10 mutants is in line with this observation.
UBX domain proteins have been characterized in mammalian cells, in which they seem to function in the UPS21. Yeast Ubx2 is unique, as it comprises both a UBX and a UBA domain that are separated by predicted transmembrane segments. It has been shown that UBX domains directly bind to Cdc48 (refs 18,
Taken together, we propose that Ubx2 acts upstream of Cdc48 in the degradation of UPS substrates. In addition, the UBA domain of Ubx2 may be specifically required for the release of ERAD substrates from the membrane. Ubiquitin-binding domains are common motifs of ERAD components. Cue1 was the founding member of a family of ubiquitin-binding proteins, one of which is the mammalian ERAD ubiquitin ligase gp78 (ref. 29). Rad23 and Dsk2, which seem to function downstream of the Cdc48 complex, both contain a UBA domain30,31. Recently, it has been shown that ubiquitin-conjugated substrates may be 'escorted' to the proteasome by these factors31. Thus, it is tempting to speculate that ERAD substrates are guided from a dislocation channel to the proteasome by a cascade of ubiquitin-binding factors, including Ubx2. These successive interactions may protect the substrate from premature de-ubiquitination and in addition contribute to the directionality of the transport process.
Yeast and plasmids.
Yeast-rich and minimal media were prepared as described, and standard genetic methods were used32. Yeast strains used in this study are listed in the Supplementary Information, Table S1. Gene deletions as well as epitope tagging in YUL26, YTX439, YON12, YON19, YON20 and YON29 were generated by homologous recombination as described33. Transformation of YWO1 with a doa10::TRP1 deletion module resulted in YUL25. YON11 was obtained by transformation of diploid strain DF5 (ref. 10) with a sel1::TRP1 deletion module and subsequent tetrad dissection. YON13, YON14, YON18 and YON29 were prepared by transformation of YTX140, YTX439, YUL26 and YJU26, respectively, with a sel1::kanMX6 deletion module. YWO1 was transformed with the respective tagging modules to obtain YUL26, YON12 and YON19. YUL26 served as a parent strain for epitope tagging in YON12 and YON20, whereas YTX439 is based on YTX140. N-terminal epitope tagging of SEL1 in YWO1 by the described method generated YON16 (ref. 34). YON22 was obtained by SEL1 epitope tagging in YUL26 as described35. YTX139 was generated by transformation of YWO1 with a ire1::URA3 deletion module. YJU96 was obtained by crossing a hrd1::TRP1 strain with a sec61-2 strain and subsequent tetrad analysis was performed. The prc1-1 allele was introduced into YON16 as described3 to generate YON28. The 112-amino-acid RING-finger motif of DOA10 was amplified by PCR from yeast genomic DNA and cloned into pQE31 (Qiagen, Hilden, Germany) using KpnI–SacI restriction sites. The resulting plasmid was used for mutagenesis utilizing the 'QuikChange' site-directed mutagenesis kit (Strategene, La Jolla, CA) to obtain a cysteine-to-serine exchange at amino-acid position 93. The modified insert was cloned by KpnI–SacI into pRS406. The resulting plasmid pON4 was linearized at a NarI restriction site and transformed into YUL26. After 5′-fluoro-orotic acid (FOA) selection, the resulting strain, YON26, was transformed with a SEL1–ProtA tagging module to generate YON27.
Antibodies specific for GST and GFP epitope were from Rockland (Gilbertsville, PA) and Santa Cruz (Santa Cruz, CA), respectively. Antibodies specific for CPY and Pgk1 were from Molecular Probes (Eugene, OR). For detection of the c-Myc and HA epitopes, 9E10 and 12CA5 antibodies were used, respectively. Specific antibodies against Ubc6, Cue1 and Sec61 were as described3,22,23. Specific antibodies against Cdc48, Hrd3 and Ubc7 were generated by E. Jarosch. Specific antibodies against Hrd1 and Kar2 were generous gifts from D. H. Wolf and T. A. Rapoport, respectively.
Membrane fractions were prepared from crude cell extracts essentially as described22. Microsomes were sedimented by ultracentrifugation (200,000g for 20 min in a TLA100.3 Beckman rotor) and the supernatants were considered as cytosolic fractions.
Protein degradation assays.
Pulse-chase experiments were performed as described22. Analysis of radiolabelled probes was performed using a Fujifilm FLA-3000 PhosphorImager. Cycloheximide decay assays were essentially performed as described23. pUL28 (ref. 26) expressed Deg1–GFP. For measuring Deg1–GFP turnover, cells were grown in synthetic minimal medium to early log phase and expression of Deg1–GFP was induced by addition of 0.1 mM CuSO4 for 2 h. Membrane fractions were prepared and immunoblotted for Ubc6, whereas cytosolic fractions were precipitated with trichloroacetic acid (TCA) and immunoblotted for Deg1–GFP.
Crude yeast extracts were prepared and, when indicated, incubated with 0.4% Triton-X100 on ice for 10 min. For protease digestion, proteinase K was added to a final concentration of 0.3 mg ml−1 and incubated for 15 min on ice. The reaction was stopped by adding 4 mM PMSF and subsequent TCA precipitation. Samples were analysed by immunoblotting.
Immunoprecipitation, immunoblotting and northern blotting.
Immunoprecipitation experiments were essentially performed as described22 except that only membrane fractions were used. Membranes were prepared in buffer containing 50 mM Tris pH 7.5, 200 mM NaAc and 10% glycerol and solubilized with 1% digitonin (Calbiochem, San Diego, CA). The digitonin lysate was cleared by centrifugation at 16,000g for 10 min prior to immunoprecipitation. During immunoprecipitation, the digitonin concentration was lowered to 0.5%. Standard methods for SDS–PAGE, western blotting and immunodecoration were used32.
Northern blotting was essentially done as described8 except that cells were grown in yeast-rich media.
Note: Supplementary Information is available on the Nature Cell Biology website.
The authors thank D. H. Wolf and T. A. Rapoport for the generous gift of strains and plasmids. We appreciate the helpful comments of the Sommer group on the manuscript. This work was partially supported by grants from the Deutsche Forschungs Gemeinschaft to T.S.
Supplementary Table S1
About this article
Scientific Reports (2015)