Known activities of the ubiquitin‐selective AAA ATPase Cdc48 (p97) require one of the mutually exclusive cofactors Ufd1/Npl4 and Shp1 (p47). Whereas Ufd1/Npl4 recruits Cdc48 to ubiquitylated proteins destined for degradation by the 26S proteasome, the UBX domain protein p47 has so far been linked exclusively to nondegradative Cdc48 functions in membrane fusion processes. Here, we show that all seven UBX domain proteins of Saccharomyces cerevisiae bind to Cdc48, thus constituting an entire new family of Cdc48 cofactors. The two major yeast UBX domain proteins, Shp1 and Ubx2, possess a ubiquitin‐binding UBA domain and interact with ubiquitylated proteins in vivo. Δshp1 and Δubx2 strains display defects in the degradation of a ubiquitylated model substrate, are sensitive to various stress conditions and are genetically linked to the 26S proteasome. Our data suggest that Shp1 and Ubx2 are adaptors for Cdc48‐dependent protein degradation through the ubiquitin/proteasome pathway.
The abundant, highly conserved AAA ATPase Cdc48 (p97 in mammals) is involved in a variety of important cellular processes including membrane fusion, ubiquitin‐mediated protein degradation and mitotic spindle disassembly (Woodman, 2003; Cheeseman & Desai, 2004). The common function of Cdc48 underlying these processes is believed to be the ATP‐dependent disassembly of oligomeric substrate proteins (Zhang et al, 2002), whereas the specificity of this ‘segregase’ activity is determined by distinct, mutually exclusive cofactors (Meyer et al, 2000). The heterodimeric Ufd1/Npl4 cofactor is required for Cdc48 functions related to the ubiquitin/proteasome pathway, including endoplasmic reticulum (ER)‐associated degradation (ERAD; Hampton, 2002), mobilization of the processed transcription factor SPT23 (Rape et al, 2001) and protein degradation through the ubiquitin fusion degradation (UFD) pathway (Johnson et al, 1995; Rape et al, 2001). The alternative cofactor p47 recruits Cdc48 to the postmitotic fusion of Golgi and ER membranes (Kondo et al, 1997; Roy et al, 2000). Interestingly, nuclear envelope assembly in Xenopus egg extracts requires p97p47 and p97Ufd1/NpL4 activities (Hetzer et al, 2001). Recently, p47 was shown to bind ubiquitylated proteins through an N‐terminal ubiquitin‐associated (UBA) domain (Meyer et al, 2002), a ubiquitin‐binding module also present in proteins involved in the ubiquitin/proteasome pathway, including Rad23 and Dsk2 (Hofmann & Bucher, 1996; Buchberger, 2002). The UBA domain of p47 is required for efficient p97p47‐mediated Golgi membrane fusion in vitro (Meyer et al, 2002), suggesting the existence of a critical, ubiquitylated substrate(s) in that assay. It was therefore postulated that Ufd1/Npl4 targets to Cdc48 multiubiquitylated substrates destined for proteasomal degradation, whereas p47 recruits monoubiquitylated substrates that are not to be degraded, perhaps even protecting them from multiubiquitylation (Meyer et al, 2002; Wang et al, 2004).
p47 and its yeast homologue Shp1 possess a ‘ubiquitin regulatory X’ (UBX) domain. The UBX domain displays high structural similarity with ubiquitin and defines a large, evolutionarily conserved protein family (Buchberger et al, 2001). Despite their wide distribution, functional data on UBX proteins are very limited. In particular, a potential involvement of UBX proteins other than Shp1/p47 in Cdc48‐mediated processes has not been addressed systematically.
In an effort to elucidate cellular functions of UBX proteins, we started to characterize the seven family members of Saccharomyces cerevisiae, which we propose to name as Shp1 (Ubx1) and Ubx2 through Ubx7 (Fig 1). Here, we establish that all seven UBX proteins bind Cdc48, suggesting that the UBX domain is a general Cdc48‐binding module. Moreover, a subset of UBX proteins possessing an amino‐terminal UBA domain binds ubiquitylated proteins in vivo. Importantly, we show that Shp1 and Ubx2 are involved in proteasomal protein degradation, thus demonstrating for the first time that they are adaptors for degradative Cdc48 activities.
All yeast members of the UBX family interact with Cdc48
Shp1 has been shown to bind Cdc48 (Braun et al, 2002). To test whether Cdc48 binding is a general property of UBX domain proteins, we performed two‐hybrid assays with all seven yeast UBX proteins. Indeed, all seven proteins interacted with Cdc48, as indicated by the ability of the tester yeast strain to grow on selective medium lacking histidine (Fig 2A). To confirm the interaction with Cdc48 at endogenous expression levels, we performed immunoprecipitation experiments with strains expressing chromosomally epitope‐tagged versions of UBX proteins under control of their own promoters. Consistent with the two‐hybrid data, Cdc48 was co‐precipitated with all seven UBX proteins (Fig 2B). Taken together, our data from two different experimental approaches indicate that Cdc48 binding is a common feature of UBX proteins.
The UBX domain is a key determinant of Cdc48 binding
The UBX domain is the only sequence motif shared by all seven UBX proteins, and thus a likely general Cdc48‐binding module. To test this hypothesis, carboxy‐terminally truncated mutants of Shp1, Ubx2, Ubx3 and Ubx5 lacking the UBX domain were analysed for Cdc48 interaction in two‐hybrid assays (Fig 3A). In comparison with the full‐length protein, Shp1ΔUBX showed significantly weaker but detectable binding to Cdc48, as indicated by growth on medium lacking histidine, but not under more stringent selection on medium lacking adenine. In contrast, the truncated versions of the other three UBX proteins completely lost their ability to bind Cdc48. Next, we tested the ability of isolated UBX domains to interact directly with Cdc48 in GST pulldown assays using recombinant proteins (Fig 3B). The UBX domains of Shp1, Ubx5 and Ubx7 bound Cdc48 very strongly, whereas that of Ubx4 showed only a weak interaction (Fig 3B, right panel). Unfortunately, we could not test the UBX domains of the other three UBX proteins in this assay because we were unable to express them in a soluble form. In summary, we show that the UBX domain is sufficient for binding of Shp1, Ubx4, Ubx5 and Ubx7 to Cdc48, and necessary in the case of Ubx2, Ubx3 and Ubx5, suggesting that the UBX domain is a general binding module for Cdc48.
UBA/UBX proteins bind ubiquitylated proteins in vivo
Shp1, Ubx2 and Ubx5 have an N‐terminal, variant UBA domain that is functional in ubiquitin binding in vivo and in vitro, with an apparent preference for oligo‐ and multiubiquitin chains (supplementary Fig S1 online). We tested the ability of full‐length Shp1, Ubx2 and Ubx5 at endogenous expression levels to interact with ubiquitylated proteins in vivo. Upon immunoprecipitation of the three UBA/UBX proteins, a high‐molecular‐weight ubiquitin‐reactive smear indicative of ubiquitylated proteins was co‐precipitated with Shp1, Ubx2 and Ubx5 (Fig 4A). In a reciprocal experiment, we used strains expressing a myc‐tagged version of ubiquitin. Again, Ubx2, Shp1 and Ubx5 co‐precipitated with ubiquitylated proteins (Fig 4B).
To extend our analysis to a defined ubiquitylated substrate, we performed immunoprecipitation experiments in strains expressing ubiquitin–proline–β‐galactosidase (Ub–P–βGal), a well‐established model substrate of the UFD degradation pathway (Johnson et al, 1995). Shp1 and Ubx2 were co‐precipitated with Ub‐P‐βGal and vice versa (Fig 4C,D), whereas no interaction could be detected between Ubx5 and Ub‐P‐βGal (data not shown). Together, these results establish that the UBA/UBX proteins Shp1 and Ubx2 bind ubiquitylated proteins in vivo.
Shp1 and Ubx2 are involved in protein degradation
The interaction of Shp1 and Ubx2 with bona fide multiubiquitylated proteins and with the UFD substrate Ub–P–βGal suggests a role for these UBA/UBX proteins in protein degradation through the ubiquitin/proteasome pathway. To directly test this hypothesis in vivo, we performed pulse‐chase experiments to compare the half‐life of Ub‐P‐βGal in wild‐type, Δshp1, Δubx2 and ufd1‐2 strains (Fig 5A). In the wild type, Ub‐P‐βGal was short‐lived with an apparent half‐life of 9 min. In contrast, in cells harbouring a classical mutation in the UFD pathway, that is, ufd1‐2, Ub‐P‐βGal was strongly stabilized with an apparent half‐life of more than 100 min even at permissive temperature. Importantly, deletion of SHP1 or UBX2 led to a significant stabilization of Ub‐P‐βGal (apparent half‐life 20 and 25 min, respectively). To exclude the possibility that this stabilization is caused by a general proteolysis defect of Δshp1 and Δubx2 cells, we analysed the degradation of R‐βGal, a short‐lived substrate of the N‐end rule pathway (Bachmair et al, 1986) that does not require Cdc48 (Ghislain et al, 1996). R‐βGal showed identical degradation kinetics in wild‐type, Δshp1 and Δubx2 cells (Fig 5B). In summary, Shp1 and Ubx2 are involved in the degradation of a Cdc48‐dependent, but not of a Cdc48‐independent model substrate, consistent with a role as cofactors of Cdc48 in ubiquitin‐dependent protein degradation.
Links to stress tolerance and proteasomal degradation
The role of Shp1 and Ubx2 in Ub‐P‐βGal degradation prompted us to investigate whether Δshp1 and Δubx2 mutants possess defects under stress conditions generating elevated levels of aberrant proteins. Indeed, Δshp1 cells were hypersensitive towards elevated temperature, cadmium and the amino‐acid analogue fluoro‐phenylalanine, whereas Δubx2 cells exhibited a less pronounced but significant sensitivity towards cadmium and fluoro‐phenlyalanine (Fig 6A). In order to genetically link SHP1 and UBX2 to the 26S proteasome, we constructed double mutants lacking the nonessential proteasomal subunit Rpn10 and Shp1 or Ubx2. Whereas Δrpn10 cells grew like wild type, Δshp1Δrpn10 and Δubx2Δrpn10 double mutants exhibited strong synthetic phenotypes under normal and stress conditions (Fig 6A). Furthermore, Δshp1Δubx2 double mutants displayed synthetic lethality (Fig 6B). Together, our phenotypic characterization shows that Shp1 and Ubx2 possess important, overlapping functions in cellular stress tolerance that are linked to proteasomal protein degradation.
In this study, we identify UBX domain‐containing proteins as a new family of Cdc48 interactors. On the basis of our data, we postulate that all UBX proteins are specificity factors recruiting Cdc48 to (as yet unknown) cellular targets. UBX proteins could confer specificity by recognition sites outside the UBX domain, and/or by distinct temporal and spatial distribution patterns inside the cell. Besides p47, VCIP135 and the yeast UBX proteins investigated in this study, further examples of UBX proteins with at least partly characterized function are the Fas‐associated factor FAF‐1, possessing as yet unclear roles in apoptosis and NF‐κB activation (Chu et al, 1995; Park et al, 2004), and TUG, a factor regulating transport of the GLUT4 glucose transporter to the plasma membrane by an unknown mechanism (Bogan et al, 2003). Our data suggest that p97 might also be involved in these processes, thus further broadening the spectrum of p97 functions.
A particularly interesting subset of UBX proteins, including Shp1 (p47), Ubx2 and Ubx5, possesses an N‐terminal UBA domain. We show for the first time that these UBA/UBX proteins bind ubiquitylated proteins in vivo. Importantly, we demonstrate a role for Shp1 and Ubx2 in protein degradation, as Δshp1 and Δubx2 mutants are partly defective in Ub‐P‐βGal degradation, sensitive to stress conditions generating aberrant proteins and synthetic defective with Δrpn10. Our data for Shp1 appear to contradict conclusions from previous work claiming that rat p47 binds and acts on monoubiquitylated proteins (Meyer et al, 2002; Wang et al, 2004). However, Meyer et al did not directly compare binding of p47 to proteins carrying ubiquitin chains of varying length. In fact, they show that immobilized p47 binds multiubiquitylated proteins from HeLa cell extracts, albeit less efficiently than Ufd1/Npl4 (Meyer et al, 2002). Furthermore, a recent study showed that downregulation of p47 in mammalian cells results in slightly elevated levels of ubiquitylated proteins (Wojcik et al, 2004). Finally, while this manuscript was under revision, an involvement of the fission yeast Shp1 homologue in ubiquitin‐dependent protein degradation was reported (Hartmann‐Petersen et al, 2004). Taken together, it appears that Shp1/p47 can recruit ubiquitylated proteins to degradative as well as nondegradative activities of Cdc48.
Whereas our work concentrated on the two major UBA/UBX proteins in yeast, Shp1 and Ubx2, a recent study by Decottignies et al (2004) on Ubx4, Ubx6 and Ubx7 showed, in agreement with our results, that all three UBX proteins bind Cdc48, and that the UBX domain of Ubx7 is sufficient for Cdc48 binding. Whereas single and double mutants in these UBX proteins did not exhibit significant phenotypes, a Δubx4Δubx6Δubx7 triple knockout strain showed a pronounced sporulation defect. Interestingly, the Δubx4Δubx6Δubx7 mutant was also found to stabilize Ub‐P‐βGal, suggesting some redundant role of Ubx4, Ubx6 and Ubx7 in proteasomal degradation. As none of the three UBX proteins possesses a known ubiquitin‐binding motif, their role in Ub‐P‐βGal degradation remains unclear. It is possible that these UBX proteins recruit further ubiquitin‐binding proteins to Cdc48. However, indirect effects of the triple mutation on proteasomal degradation cannot currently be excluded.
Given the accumulated evidence for the involvement of UBX proteins in the ubiquitin/proteasome pathway, Ufd1/Npl4 can no longer be considered the only adaptor for Cdc48 activities involving protein degradation. Nevertheless, a principal role of Ufd1/Npl4 is suggested by the fact that UFD1 and NPL4 are essential genes in S. cerevisiae, and that much higher amounts of multiubiquitylated proteins are bound to Ufd1/Npl4 as compared to p47 (Meyer et al, 2002). Also, the defect in Ub‐P‐βGal degradation is much stronger in ufd1‐2 cells than in Δshp1 or Δubx2 cells (Fig 5A). We interpret this fact to indicate that Ufd1/Npl4 is an essential component of the UFD pathway, whereas the two UBX proteins play auxiliary roles by recruiting excess Ub‐P‐βGal to Cdc48 in an experimental setting that likely saturates the Cdc48Ufd1/Npl4 complex. Specific cellular targets of UBX proteins remain to be identified.
Given the large number of UBX proteins and the existence of additional, unrelated cofactors of Cdc48 such as UFD2, UFD3 and SVIP, a highly variable regulatory network for degradative as well as nondegradative Cdc48 functions must be expected to be operative in vivo. The identification of regulatory mechanisms and further substrates will be the next step towards a better understanding of Cdc48 functions in the cell.
Cloning and yeast techniques. Standard protocols were followed for general yeast techniques, two‐hybrid assays and strain constructions. Yeast and Escherichia coli expression plasmids were constructed using standard cloning techniques; details are available upon request from the authors. Strains and plasmids used in this study are listed in supplementary Tables 1 and 2 online, respectively.
Immunoprecipitations. Yeast cells (100 ml, OD600=0.8) expressing 3HA‐tagged UBX proteins and/or myc‐tagged ubiquitin were harvested and lysed using glass beads in buffer A (50 mM Tris (pH 7.5), 100 mM KCl, 5 mM MgCl2, 0.1% NP‐40, 10% glycerol, 2 mM phenylmethylsulphonyl fluoride (PMSF), 2 mM benzamidine, 20 mM N‐ethyl‐maleimide (NEM) and 20 μM MG132). Lysates were incubated for 3 h in buffer A containing 0.3% NP‐40 with 2 μg monoclonal anti‐HA antibody or polyclonal anti‐myc antibody (Santa Cruz) coupled to protein A Sepharose beads (Amersham). For immunoprecipitations with strains expressing Ub‐P‐βGal, lysates were prepared in buffer B (50 mM Tris (pH 7.5), 100 mM NaCl, 1.5 mM MgCl2, 0.1% NP‐40, 2 mM PMSF, 2 mM benzamidine, 20 mM NEM and 20 μM MG132), precleared for 1 h with protein A sepharose beads, and incubated overnight with 4 μg anti‐βGal antibody (Promega) coupled to protein A Sepharose beads. Bound proteins were analysed by immunoblotting against Cdc48 (rabbit anti‐Cdc48, gift from S. Jentsch), ubiquitin (monoclonal anti‐Ub, Santa Cruz), βGal, or the respective epitope tag used for precipitation.
In vitro Cdc48‐binding assay. Cdc48 and GST fusion proteins of UBX domains were purified from E. coli according to standard protocols. Glutathione beads (10 μl) (Amersham) in buffer C (1 × TBS, 0.1% Triton X‐100) were saturated with GST–UBX domains at 4 °C and washed with buffer C. Cdc48 (10 μg) was incubated with the loaded beads for 1 h at 4 °C. The beads were extensively washed with buffer C, and bound Cdc48 was detected after SDS–PAGE by Coomassie staining or anti‐Cdc48 immunoblot.
Pulse‐chase experiments. To determine the in vivo half‐life of βGal at 30 °C, cells (15 ml, OD600=1.0) were harvested, resuspended in 1 ml SCGal‐Met‐Ura medium and labelled with 150 μCi [35S]methionine for 4 min. Cells were washed twice with 1 ml SCGal medium and resuspended in 5 ml SCGal containing 0.5 mg/ml cycloheximide. Aliquots of 1 ml were harvested after chase times of 0, 10, 20 and 30 min, and subjected to immunoprecipitation with monoclonal βGal antibodies. After SDS–PAGE, [35S]βGal was detected by autoradiography and quantified on a PhosphoImager.
Supplementary information is available at EMBO reports online (http://www.nature.com/embor/journal/vaop/ncurrent/extref.7400203‐s1.pdf).
We thank S. Jentsch for continued support and for sharing strains, plasmids and antibodies; S. Köglsberger for excellent technical assistance; and S. Jentsch, S. Müller and O. Stemmann for critical reading of the manuscript. H.R. and S.R. are PhD students with S. Jentsch. A.B. is funded by DFG (Emmy Noether grant Bu 951/1) and GIF (Young Scientist's grant 2049).
- Copyright © 2004 European Molecular Biology Organization