Distinct Steps in Dislocation of Luminal Endoplasmic Reticulum-associated Degradation Substrates ROLES OF ENDOPLASMIC RETICULUM-BOUND p97/Cdc48p AND PROTEASOME*

Dislocation of endoplasmic reticulum-associated degradation (ERAD) substrates from the endoplasmic reticulum (ER) lumen to cytosol is considered to occur in a single step that is tightly coupled to proteasomal degradation. Here we show that dislocation of luminal ERAD substrates occurs in two distinct consecutive steps. The first is passage across ER membrane to the ER cytosolic face, where substrates can accumulate as ubiquitin con-jugates. In vivo , this step occurs despite proteasome inhibition but requires p97/Cdc48p because substrates re-main entrapped in ER lumen and are prevented from ubiquitination in cdc48 yeast strain. The second dislocation step is the release of accumulated substrates to the cytosol. In vitro , this release requires active proteasome, consumes ATP, and relies on salt-removable ER-bound components, among them the ER-bound p97 and ER-bound proteasome, which specifically interact with the cytosol-facing substrates. An additional role for Cdc48p subsequent to ubiquitination is revealed in the cdc48 strain at permissive temperature, consistent with our finding that p97 recognizes luminal ERAD substrates through multiubiquitin. BiP interacts exclusively with ERAD substrates, suggesting a role for this chaperone in ERAD. We propose a model that assigns the cytosolic face

The endoplasmic reticulum-associated degradation (ERAD) 1 is a quality control process that selectively eliminates malfolded proteins or unassembled subunits of oligomeric proteins in the secretory pathway (1,2). ERAD substrates are dislocated from the endoplasmic reticulum (ER) back to the cytosol via the Sec61 complex (3)(4)(5)(6). In the cytosol, ubiquitin is conjugated to the ERAD substrates that are degraded by the proteasome (7). The proteolytically active proteasome has been implicated in the dislocation of ERAD substrates by virtue of their scarcity in cytosol of proteasome-inhibited cells (8 -11). Although stabilization of proteins in the secretory pathway by proteasome inhibitors is generally accepted as an indication for ERAD, ubiquitination of such proteins is a direct evidence for the access of the substrate to the cytosol (7). In most cases, however, accumulation of the multiubiquitinated ERAD substrates to detectable levels requires inhibition of the proteasome. If the proteasome is indeed required for dislocation (8 -11), then ubiquitination of ERAD substrates would not be observed. This applies especially to luminal ERAD substrates, as membrane ERAD substrates inherently display cytosolic domains, which may be ubiquitinated irrespective of dislocation (11)(12)(13)(14)(15). Therefore, it is our view that luminal ERAD substrates, whose dislocation is an absolute prerequisite for ubiquitination, are the proteins of choice to address directly coupling among dislocation, ubiquitination, and proteasomal activity.
Among luminal ERAD substrates, the best characterized are CPY* in yeast (5,16,17) and immunoglobulins (Igs) in mammalian cells (8,9). Igs are particularly attractive because they are oligomeric proteins in the secretory pathway composed of heavy and light chains, both of which are subjected to proteasomal degradation. This is indicated by the stabilization of s chains, the heavy chains of secretory IgM (9,18,19) or of Ig light chains (8), upon proteasome inhibition. Under these conditions, s or accumulate in the microsomal fractions rather than the cytosol, implicating the proteasome in dislocation (8,9). Nevertheless, we detected ubiquitination of the luminal s when the proteasome proteolytic activity was inhibited (20), indicating that the proteasome was not required for s to gain access to the cytosol-facing ubiquitination machinery.
Using cell fractionation, protease protection assay, and salt wash in B lymphocytes and yeast, as well as in vitro dislocation experiments, here we dissect dislocation of luminal substrates into two consecutive steps: (i) passage across the ER membrane and (ii) release to the cytosol. We investigate the roles of the proteasome and of the recently discovered cytosolic ERAD chaperone p97/Cdc48p (21) in dislocation and reveal differential requirements for these components in the two dislocation steps. We define the cytosolic face of the ER membrane as a site to which luminal ERAD substrates are fully dislocated and cytosolic ERAD components are recruited, extending the ERbound ERAD machinery (7) to include the ER-bound subpopulation of the proteasome (22,23) as well as the ER-bound subpopulation of p97/Cdc48p.
Biosynthetic Labeling and Cell Incubation-Cells were starved for methionine for 45 min and either labeled for 45 min or pulse labeled for 15 min with [ 35 S]methionine (100 Ci/ml; 1,000 Ci/mmol) and then chased in the presence of 25 M ALLN, 5 M MG-132, or their combinations. When not labeled, cells were incubated for 3-4 h with ALLN, MG-132, or their combinations. Cells and medium were separated, and cells were either lysed as described previously (24) or subjected to cell fractionation (see below). Yeast cells were concentrated (5 A 600 ) and incubated for 4 h at permissive (30°C) or nonpermissive (37°C) temperatures. Cells were collected and either lysed as described previously (21) or subjected to fractionation (see below), and fractions were solubilized as described (24). To prevent deubiquitination, 20 mM N-ethylmaleimide was included in the lysis buffer.
For in vitro dislocation assays, 38C cells metabolically labeled for 45 min or unlabeled cells incubated for 3 h with or without various drugs were disrupted in 0.25 M sucrose and 10 mM potassium phosphate, pH 7.5. P10 microsomes were resuspended either in S10 cytosol from equivalent amounts of 38C or COS-7 cells (60 and 30 mg/ml protein, respectively), or in the disruption buffer, all containing 2 mM phenylmethanesulfonyl fluoride and 2 kallikrein inhibitor units/ml aprotinin. Resuspended microsomes, supplemented with an ATP-regenerating system (3 mM ATP, 5 units/ml creatine kinase, 10 mM creatine phosphate, 3 mM MgCl 2 ) or treated with 25 units/ml apyrase were incubated at 37°C in the absence or presence of MG-132, Z-L 3 VS (a kind gift from M. Bogyo, University of California San Francisco and H. Ploegh, Harvard University) or ALLN. Reactions were terminated by incubation on ice followed by centrifugation (10,000 ϫ g, 20 min) to separate microsomes and soluble fractions.
Protease Protection Assay and Salt Wash-The optimized reaction was found to be a 45-min incubation at 30°C with 5 g/ml trypsin for PNS of B cells (adapted from Ref. 8) and a 30-min incubation on ice with 0.5 mg/ml trypsin for PNS of yeast (17). Reactions were terminated by adding phenylmethanesulfonyl fluoride to 2.5 mM followed by centrifugation to fractionate PNS to microsomes and cytosol, as described above. Where indicated, microsomes were washed by resuspension in ice-cold 0.5 M KCl. Washed microsomes were recovered by centrifugation; wash supernatants were collected, and their proteins were precipitated by 10% trichloroacetic acid.
Immunoprecipitation and Immunodetection-IgM heavy chains, HA-tagged CPY*, BiP, or myc-tagged ubiquitin and their associated proteins were immunoprecipitated in each experiment from comparable amounts of lysed cells or fractions. For cytosolic fractions, an equal volume of lysis buffer was added. Antibodies used for immunoprecipitation: goat anti-mouse (SouthernBiotech), mouse anti-HA (clone 12CA5), mouse anti-myc (clone 9E10), were all followed by protein A-Sepharose (RepliGen); rabbit anti-mouse BiP (Affinity BioReagent) was followed by anti-rabbit IgG-Sepharose (Sigma). Total cell extract proteins and immunoprecipitated , HA-CPY*, BiP, or myc-ubiquitin were resolved by reducing or nonreducing SDS-PAGE, electroblotted onto nitrocellulose, and detected either by direct autoradiography or by immunoblotting. Antibodies used for immunoblotting were: horseradish peroxidase-conjugated anti-mouse (SouthernBiotech); biotin-conjugated mouse anti-BiP (SouthernBiotech) followed by horseradish peroxidase-conjugated avidin (Jackson); rabbit anti-s (21) and rabbit anti-proteasome ␣ subunit of Methanosarcina thermophila (Calbiochem) followed by horseradish peroxidase-conjugated anti-rabbit IgG (Jackson); mouse anti-HA (clone 12CA5), mouse anti-p97 (clone 58.13.3; ProGen), mouse anti-ubiquitin (clone P4D1, BabCO) used to detect ubiquitin in B cells and mouse anti-ubiquitin (clone Ubi-1, Zymed Laboratories Inc.) used to detect ubiquitin in yeast, followed by horseradish peroxidase-conjugated anti-mouse IgG (Jackson). The horseradish peroxidase was visualized by the enhanced chemiluminescence (ECL) reaction. The amounts of radiolabeled or immunoblotted proteins were determined by densitometry, and relative levels were calculated.

RESULTS
The luminal s heavy chain exhibits developmentally regulated stability; while being a stable protein that is secreted efficiently from plasma cells, it is degraded rapidly in the earlier differentiation stages of pre-B and B cells (19,24,30). Based on its stabilization by a variety of proteasome inhibitors, we and others have shown that s is degraded by the proteasome (9,18,19). Moreover, in B lymphocytes treated with proteasome inhibitors, we were able to detect ubiquitinated species of s ( Fig. 1A; see also Ref. 20). Therefore, we reasoned that, despite the inhibition of the proteasome, a part of the luminal s must be dislocated to gain access to the cytosolfacing ubiquitination machinery. To test this hypothesis, we fractionated IgM-expressing 38C B cells and determined the topology of s upon proteasome inhibition.
Enhanced BiP Association with s upon Proteasome Inhibition Suggests a Role for BiP in s Dislocation-Upon fractionation of 38C B cells, the majority (ϳ80%) of s was detected in the microsomal fraction, whereas only ϳ20% was recovered in the cytosolic fraction (Fig. 1, A and D). The BiP that coprecipitated with s was negligible and restricted to the microsomal fraction ( Fig. 1A, lanes 1 and 2). Upon blocking the proteasome, there was a ϳ60% increase in the cellular content of s, accompanied by 5-7-fold increase in the amount of coprecipitated BiP (Fig. 1A, lanes 1, 5, and 9; Fig. 1G). This increase in s was not recovered in the cytosolic fraction but rather in the microsomal fraction (Fig. 1, A and D). The enhanced coprecipitation of BiP could not be explained by accumulation of unassembled s heavy chains, the favored substrate of BiP (31), because none was detected (Fig. 1B, upper panel, lanes 1 and 2). Importantly, this increase was not accompanied by any rise in cellular BiP content (Fig. 1C, compare lanes 5 and 9 with lane 1), indicating that the unfolded protein response was not elicited. Also, in reciprocal experiments, increased amounts of s coprecipitated by an anti-BiP antibody (Fig. 1B, compare lanes 4 and 5 in the upper panel and lanes 5 and 6 in the lower panel). Thus, in proteasome-inhibited cells the physical stability of the BiP-s complexes is enhanced. Our findings concur with the previously reported enhanced BiP coprecipitation and increased physical stability of the BiP-light chain complexes in proteasome-inhibited NS1 cells (8). Interestingly, BiP pulled down s assembled with into 2 2 monomers but not unassembled s (Fig. 1B,   In agreement with the genetic data in yeast (32), our results provide the first biochemical evidence for involvement of BiP in ERAD of luminal substrates in mammalian cells.
Luminal s Crosses the ER Membrane and Accumulates at the ER Cytosolic Face Despite Proteasome Inhibition-The finding that s undergoes ubiquitination (20) prompted us to examine the distribution of ubiquitinated s in fractionated, proteasome-inhibited B cells. Heavily ubiquitinated s was observed mostly in microsomes, in correlation with the distribution of the s protein (Fig. 1A). These results demonstrated that ubiquitinated s was associated with the microsomes while facing the cytosol.
Protease protection assays demonstrated that the cytosolfacing ubiquitinated s represented a significant proportion of the s protein. In untreated cells, only ϳ8% of microsomal s was sensitive to trypsin, whereas in proteasome-inhibited cells, where microsomal s rose to ϳ160%, more than half of it was digested by trypsin ( Fig. 1A and quantified results in Fig. 1E). We conclude that upon proteasome inhibition microsomal s was redistributed between the two sides of the membrane so that the dislocated s increased by 7-fold, whereas cytosolic s hardly increased (Fig. 1D). As expected, all s ubiquitin conjugates were completely digested by trypsin ( Fig 1A). Importantly, the integrity of microsomal membranes was demonstrated by the complete protection of BiP from trypsin digestion (Fig. 1, A and C). The complete digestion of either dislocated s (Fig. 1A) or luminal s in Nonidet P-40-solubilized microsomes (data not shown) indicated that s could be digested by trypsin. An additional indication for the membrane integrity was the fate of m, the stable membrane isoform of heavy chains (24). Apart from the KVK sequence in its C terminus, which faces the cytosol, and its single transmembrane span, the rest of this protein is luminally oriented. Indeed, m was detected exclusively in microsomes and was fully protected from trypsin digestion regardless of proteasome inhibition (Fig. 1A).
Another evidence for the cytosolic orientation of microsomal s was afforded by the ability to remove it from the microsomal fractions by salt wash. When the proteasome was inhibited, the amounts of s removed by salt increased considerably (Fig. 1F, lanes 1 and 2), compared with the moderate increase in the amounts of s which remained in the washed microsomes ( Fig.  1F, lanes 3 and 4). Importantly, neither m nor luminal BiP was removed by salt wash, and both remained with the washed microsomes ( Fig. 1F, lower panel, lanes 3 and 4). Taken together, our findings define the cytosolic face of the ER membrane as a site to which the luminal ERAD substrate s is fully dislocated. Not only the arrival of s to this site occurs despite proteasome inhibition, but also the latter is a prerequisite for revealing this site. Dislocated s Interacts Exclusively with the Microsomebound Proteasome and Microsome-bound p97-Although the active proteasome appeared to play no role in s dislocation, coprecipitation of the proteasome with indicated their physical interaction, which was actually enhanced ϳ2.5-fold when the proteasome was proteolytically inactive (Fig. 1A, lanes 1, 5,  and 9; Fig. 1G). Remarkably, this interaction with was restricted to the proteasome subpopulation that was membranebound (Fig. 1A, lanes 1, 5, and 9). This finding suggests a role for the previously identified small ER-bound subpopulation of this cytosolically abundant ERAD component (22) and demonstrates, for the first time, interactions that involve this subpopulation. Moreover, these results are in agreement with the suggestion that the proteasome might be recruited to the ER membrane by multiubiquitinated ERAD substrates (23).
Recent findings have implicated the p97/Cdc48p as a novel cytosolic component essential for ERAD (16, 21, 26 -28), and we have shown that p97 physically interacts with heavy chains (21). Therefore, we investigated further the interaction of p97 with s in fractionated cells. Evidently, the ϳ20% cytosolic s (Fig. 1D) failed to coprecipitate p97 (Fig. 1A), although the pool of p97 was mostly cytosolic (90%; Fig. 1C, lanes 1 and 2). Conversely, p97 coprecipitated with s only in the microsomal fraction (Fig. 1A). Assuming that the observed interaction of p97 with s was direct, it was remarkable that as little as 6.5% of already dislocated s could efficiently precipitate a large proportion of the minute subpopulation of microsome-bound p97 (Fig. 1A). Proteasome inhibition enhanced ϳ2.5-fold the amounts of microsome-bound (Fig. 1C) as well as coprecipitated p97 (Fig. 1, A and G), whereas the p97 cytosolic form increased only marginally (Fig. 1C). As expected, p97 and the proteasome, which interacted with the cytosol-facing ubiquitinated microsomal s, were also sensitive to trypsin, whereas BiP, which interacted with the luminal s, was not (Fig. 1, A and G). We conclude that from the abundance of the proteasome and p97 found in the cytosol, only the minor membrane-bound subpopulations of these components are engaged with the ERAD substrate s.
Dislocation of Luminal Substrates Involves p97/Cdc48p-To test directly the suggested role of the essential protein p97/ Cdc48p in dislocation (16,21,26), we took advantage of yeast genetics, relying on the remarkable conservation of the ERAD pathway from yeast to mammalian cells. We chose the yeast strain carrying the temperature-sensitive cdc48-10 allele, for which we have demonstrated previously the marked stabilization of ERAD substrates at 37°C (21). Here, we biochemically monitored the distribution of the luminal ERAD substrate HA-CPY* with respect to the membrane of the microsomes. The majority of microsomal HA-CPY*, a luminal ERAD substrate in yeast, was protected from trypsin digestion when cells were grown at 37°C but not at 30°C (Fig. 2A, lanes 1, 2, 4, and 5).
In the wild-type CDC48 strain, similar levels of HA-CPY* were protected at both temperatures ( Fig. 2A, upper panel). These results implicate Cdc48p in the dislocation process. Interestingly, in the mutant cdc48 at the permissive temperature, higher levels of HA-CPY* were trypsin-sensitive, compared with wild-type CDC48 ( Fig. 2A). This suggested that events following dislocation were relatively slowed down in the mutant, resulting in substrate accumulation at the ER cytosolic face. The lingering of microsomal HA-CPY* protein in the lumen was corroborated further by salt wash experiments in mutant cdc48 cells that were shifted from 30°C to 37°C. There, the amounts of HA-CPY* removed by salt wash decreased significantly (Fig. 2B), in agreement with the decreased susceptibility of this luminal ERAD substrate to digestion by trypsin ( Fig. 2A).
p97/Cdc48p Acts before Multiubiquitination but Also Recognizes Substrates through Their Multiubiquitin Moieties-The interaction of p97 with dislocated microsomal s, which was enhanced upon proteasome inhibition and accumulation of multiubiquitinated s (Fig. 1, A and G), raised the possibility that p97 recognized the luminal ERAD substrate s through its attached multiubiquitin moieties. This is an attractive possibility because p97 directly interacted with multiubiquitin (29,33), and multiubiquitination was implicated previously in dislocation of several ERAD substrates (14,16). To examine in vivo the effect of multiubiquitination on the physical interactions between s and p97, we took advantage of the COS-7based transient expression system. First, we show that the fate of s in COS-7 cells was consistent with our findings in B cells because in both cell lines s was rapidly degraded, and this degradation was blocked by proteasome inhibitors (Fig. 3A; see also Ref. 20). Next, wild-type or mutant (K48R) myc-tagged ubiquitin was overexpressed in naïve or s-expressing COS-7 cells. To avoid a high background of ubiquitin signal, the following experiments were carried out in the absence of proteasome inhibitors. As shown in Fig. 3B, overexpression of wildtype ubiquitin resulted in abundance of myc-tagged ubiquitinconjugated proteins, which were immunoprecipitated by antimyc and probed with anti-ubiquitin antibodies (lane 6). The enhanced ubiquitination was also reflected by the slight increase in ubiquitin-conjugated s, which was immunoprecipitated by anti-and probed with anti-ubiquitin antibodies (com- pare lanes 3 and 9). Overexpression of K48R abolished the enhanced levels of myc-tagged multiubiquitinated proteins (compare lanes 6 and 8), as reflected also by dumping the slight increase in s ubiquitination (compare lanes 9 and 11). Consequently, s was stabilized (compare s in lanes 9 and 11), even if only marginally, similar to cystic fibrosis transmembrane conductance regulator (12). Next, we correlated the degree of ubiquitination to the binding of p97. The specific interaction of p97 with ERAD substrates was recapitulated when p97 was coprecipitated by an anti-antibody in s-expressing but not in mock-transfected COS-7 cells (compare lanes 1 and 3). Moreover, we reciprocated in vivo the in vitro findings of Dai and Li (33) by showing that regardless of s expression, p97 was pulled down by the abundant myc-tagged multiubiquitinated proteins (lane 6), but not by proteins decorated with myctagged K48R (lane 8). This demonstrated that the latter population, which was immunoprecipitated by an anti-myc antibody and therefore contained only proteins with impaired multiubiquitination, indeed could not bind p97.
Surprisingly, in s-expressing COS-7 cells, the interaction between s and p97 (monitored by the levels of p97 that were pulled down by an anti-antibody) was abolished by overexpression of wild-type ubiquitin but was restored by overexpression of K48R ubiquitin (Fig. 3B, compare lanes 3, 9, and 11). We interpreted these data as an indication for competition between the abundance of myc-tagged multiubiquitinated proteins and the multiubiquitinated s for binding to p97. Competition of multiubiquitin chains with cellular ubiquitinated proteins for FIG. 2. Dislocation and ubiquitination of CPY* require Cdc48p. A, yeast strains expressing wild-type CDC48 or temperature-sensitive cdc48 -10 ts mutant were transformed with HA-CPY*-encoding plasmid. After a 4-h incubation at 30 or 37°C, cells were disrupted, PNS was incubated with no additions, with trypsin alone or with trypsin in the presence of 1% Triton X-100 (TX-100), microsomes (P20) were recovered by centrifugation, lysed, resolved by reducing SDS-PAGE and immunoblotted (IB) with an anti-HA antibody. The blot is representative of three independent experiments. B, microsomes (P20) were recovered from yeast strain cdc48 -10 ts after a 4-h incubation at 30 or 37°C. Microsomes were washed by 0.5 M KCl, wash supernatant and washed microsomes were separated by centrifugation, resolved by reducing SDS-PAGE, and immunoblotted with an anti-HA antibody. C, the indicated yeast strains were incubated for 4 h at 30 or 37°C, cells were lysed, and HA-CPY* was immunoprecipitated (IP) with an anti-HA antibody. Immunoprecipitated HA-CPY* was resolved by reducing SDS-PAGE and immunoblotted with an anti-ubiquitin (anti-Ub) antibody and reprobed with an anti-HA antibody. Note that no ubiquitinated proteins were precipitated by the anti-HA antibody from yeast strains not expressing HA-CPY* (lanes 9 -12).
binding to p97 was reported previously (33). On the other hand, proteins in K48R-expressing cells with impaired multiubiquitination failed to compete s because they bind p97 poorly. In these cells, the diminished coprecipitation of p97 by the antimyc antibody, but not by the anti-antibody, probably reflected the mixed population of s, carrying not only short K48R-containing ubiquitin chains but also normal high molecular weight multiubiquitin moieties that bind p97 efficiently. Taken together, p97 appears to recognize the luminal ERAD substrate s through the same mechanism by which it recognizes many proteasomal substrates, namely through their multiubiquitin moieties.
In light of our observation that p97 recognized s through its multiubiquitin moiety, together with the suggested role of multiubiquitination in dislocation of ERAD substrates (14,16) we sought to pinpoint the function of Cdc48p with respect to multiubiquitination. Again, by using the cdc48-10 allele, immunoprecipitation of HA-CPY* with an anti-HA antibody followed by probing with an anti-ubiquitin antibody revealed that the substantial ubiquitination of HA-CPY* at 30°C was completely abolished at 37°C (Fig. 2C, lanes 3 and 4). Together with the nearly complete protection from trypsin digestion observed at 37°C, this further indicated the involvement of Cdc48p in dislocation. Interestingly, in the CDC48 wild-type strain, a   FIG. 3. s is recognized by p97 through its multiubiquitin moiety but dissociates BiP prior to its binding to p97. A, COS-7 cells transfected with -encoding plasmid were pulse labeled for 30 min with [ 35 S]methionine 44 h post-transfection and chased for the indicated time in the absence (f) or presence (Ⅺ) of ALLN. Cells were lysed, s was immunoprecipitated, resolved by reducing SDS-PAGE, and exposed to autoradiography (S 35 ). The expression of s was confirmed by anti-s antibody and endoglycosidase H treatment (18,21). Remaining s was calculated from densitometry as the percent of s level at the end of the pulse (100%). The presented data are representative of three independent experiments. B, COS-7 cells were transfected with plasmids encoding , wild-type myc-ubiquitin (WT Ub), K48R myc-ubiquitin (K48R Ub) or their combinations. Cells labeled for 4 h with [ 35 S]methionine 44 h post-transfection were lysed, and anti-or anti-myc antibodies were used for parallel immunoprecipitations (IP). Immunoprecipitated proteins resolved by reducing SDS-PAGE and electroblotted were exposed to autoradiography (S 35 ) and then probed (IB) with anti-ubiquitin (anti-Ub) and reprobed with anti-p97 antibodies. C, 38C cells incubated for 4 h with (ϩ) or without (Ϫ) ALLN were fractionated, P10 (P) and S200 (S) fractions were lysed, and antior anti-BiP antibodies were used for parallel immunoprecipitations. Immunoprecipitated proteins resolved by reducing SDS-PAGE and electroblotted were probed with anti-(identical to Fig. 1B, lower panel) and reprobed with anti-p97 antibodies.
substantial ubiquitination of HA-CPY*, reported also by others (16), was detected at 37°C rather than at the normal conditions (Fig. 2C, lanes 1 and 2). Because this phenomenon was observed also in other strains harboring wild-type CDC48 (i.e. PRE1 (34) in Fig. 2C, lanes 5 and 6), it indicated that at nonpermissive temperature, HA-CPY* degradation was delayed relative to its ubiquitination (see Fig. 5, step 5), whereas under normal conditions it was not. s Dissociates from BiP before Its Binding to p97-Being a cytosolic component, p97 may interact with s only once this luminal substrate emerges from the ER membrane. Yet, because of topological considerations, p97 cannot initiate the dislocation of such a substrate from within the ER lumen. Because BiP appeared to participate in the ERAD of s (Fig.  1B), it was of interest to examine whether p97 and BiP cooperated while operating at their respective cytosolic and luminal sides of the ER membrane. This was tested by looking for a ternary complex, in which s interacted with BiP and p97 simultaneously. Based on probing precipitated complexes with anti-and reprobing with anti-p97 antibodies, it was evident that although BiP pulled down microsomal s and the latter pulled down p97, BiP failed to pull down p97 via s (Fig. 3C).
Nevertheless, our data do not exclude the possibility that BiP and p97 cooperate in a push-pull mechanism in which BiP dissociation from s precedes p97 binding.

In Vitro Dislocation of s Utilizes ATP and Microsome-bound Cytosolic Components and Reveals Proteasome-independent and -dependent Consecutive
Steps-The coprecipitation of the microsome-bound proteasome and p97 with s implied that cytosolic components of the ERAD pathway were recruited to the ER cytosolic face, to operate in dislocation of luminal s. Therefore, we hypothesized that isolated microsomes would be sufficient to carry out s dislocation in vitro. First, when microsomes isolated from [ 35 S]methionine-labeled cells were incubated with cytosol obtained from unlabeled cells, radiolabeled s was recovered in the soluble cytosolic fraction in a time-and ATP-dependent manner (Fig. 4A). When we next replaced cytosol with buffer, s was still released in the same manner (Fig. 4B, upper and lower panels). BiP, serving as a control for membrane integrity, remained in the microsomes (Fig. 4B, middle panel). Note that the release was not significantly enhanced by the addition of ATP unless the reaction mixture without the added ATP was also treated with apyrase (Fig. 4B, compare lower and upper panels and graphs). This FIG. 4. In vitro reconstitution of s dislocation relies on microsome-bound components. P10 microsomes from 38C cells were incubated for the indicated time with either S10 cytosol or buffer, and the released s was recovered from the supernatant. As indicated, S10 or buffer was untreated (none) or treated with apyrase or supplemented with ATP. A, microsomes obtained from cells labeled for 1 h with [ 35 S]methionine were incubated with S10 from unlabeled 38C cells, released s was immunoprecipitated (IP), resolved by reducing SDS-PAGE, electroblotted and exposed to autoradiography (S 35 ). B, unlabeled microsomes were incubated with buffer, and released s was resolved by reducing SDS-PAGE, electroblotted and probed (IB) with anti-(upper panel) and reprobed with anti-BiP (middle panel) antibodies. Note that BiP was not released to the buffer. For comparison, a sample of microsomes (input) was included (lane 13). Where indicated, proteasome inhibitors (5 M MG-132, upper panel and graph; 50 M Z-L 3 VS (VS), lower panel and graph) were included in the ATP-supplemented buffer. Released s in untreated (E, upper graph) or apyrase-treated (E, lower graph), ATP-supplemented (q) or proteasome-inhibited (f) samples was quantified and is presented in graphs in arbitrary units. C, unlabeled microsomes were either washed in 0.5 M KCl or incubated directly in ATP-supplemented buffer or S10 cytosol obtained from COS-7 cells. Released s was resolved by reducing SDS-PAGE and immunoblotted (IB) with an anti-antibody. To visualize better the reconstituted release from washed microsomes, the blot was also overexposed (lower panel).
reflected the endogenous ATP that remained associated with the isolated microsomes. These findings indicated that indeed, microsome-bound cytosolic components were sufficient to support s dislocation. Moreover, s was no longer recovered when microsomes were salt washed prior to their incubation in buffer (Fig. 3C, lanes 1-4), suggesting that essential microsomebound components were removed by salt wash. Nevertheless, when these salt-washed microsomes were incubated with cytosol, the time-and ATP-dependent dislocation of s was restored (Fig. 4C, lanes 4 -7), indicating that freshly added cytosolic components allowed the resumption of s dislocation. It should be noted that the cytosol added in this experiment was prepared from COS-7 cells, which do not express s, suggesting that the relevant cytosolic components were not B cell-specific. Taken together, our data demonstrate the involvement of ERbound cytosolic components in s dislocation. Such components are recruited to the ER via interactions that endure microsome preparation but are susceptible to high ionic strength. Finally, the similar proportion (ϳ1/3) of s which was sensitive to either trypsin (Fig. 1, A and E) or salt wash (Fig. 1F) indicated that the salt wash completely removed s from the cytosolic face of the microsomes, suggesting that the entire dislocation of s, from within the ER lumen to the cytosol, was reconstituted in vitro.
The in vitro release of s to the ATP-supplemented buffer could reflect a step subsequent to passage across the ER membrane which could not be observed in vivo because of the rapid degradation of s by the proteasome immediately upon release. However, proteasome inhibition also did not result in s accumulation in cytosol, because s, which crossed the ER membrane, actually remained associated with the cytosolic face of this organelle (Fig. 1). This raised the intriguing possibility that the step that required the proteolytically active proteasome was the actual release to cytosol rather than the passage of s across ER membrane. Indeed, in the presence of either MG-132 (Fig. 4B, upper panel and graph) or Z-L 3 VS (Fig. 4B,  lower panel and graph), the release of s to the buffer ceased within 15 min of in vitro incubation compared with the continuous release of s in the absence of proteasome inhibitors. Our combined in vivo and in vitro findings indicate that dislocation of ERAD substrates occurs in two consecutive steps: (i) p97/ Cdc48p-dependent but proteasome-independent passage across ER membrane and (ii) proteasome-dependent release to cytosol. This explains why in proteasome-inhibited cells s accumulated at the cytosolic face of the ER rather than being fully released to the cytosol. The data also implicate the mem-brane-bound active proteasome in the release of s from the cytosolic face of the ER. In dissecting dislocation into these two consecutive steps, it is still not clear whether p97/Cdc48p plays a role also in the release to the cytosol, but it is evident that it does play a role subsequent to ubiquitination. DISCUSSION In this work, the lumen-to-cytosol dislocation is biochemically dissected for luminal ERAD substrates, and the proposed coupling between initial steps of ERAD is reevaluated. Coupling between ubiquitination and dislocation has been indicated for luminal CPY* (35,36), soluble truncated ribophorin I (37), and membrane major histocompatibility complex class I heavy chain (38). Multiubiquitination in particular is required for dislocation of class I heavy chain (14) and CPY* (16). It implicates multiubiquitin in a ratcheting mechanism or in active pulling out of ERAD substrates across the ER membrane. Likewise, the active proteasome has been postulated to play a role in extraction of several membrane ERAD substrates (10, 39 -41). It is noteworthy that the only luminal ERAD substrates for which the proteolytically active proteasome is implicated in dislocation are the s Ig heavy chain and Ig light chain (8,9). Based on the failure to detect these luminal ERAD substrates in the cytosol upon blocking the proteasome (Refs. 8 and 9 and this work), here we provide evidence for an alternative dislocation mechanism. We demonstrate that despite the proteasome inhibition, the Ig s heavy chain crosses the ER membrane and accumulates at the cytosolic face of this membrane as multiubiquitin conjugates.
Studying ERAD using luminal substrates allows us to define their release to the cytosol as a discrete step in ERAD (Fig. 5). Accordingly, dislocation of luminal substrates is dissected into two consecutive steps, which hitherto were not distinguished: (i) passage across the ER membrane and accumulation of substrates at the cytosolic face of this organelle, hence defining the first station in dislocation; (ii) release of these substrates to cytosol, where they are degraded by the proteasome, hence defining the second station. Actually, the release to the cytosol en route to proteasomal degradation has been demonstrated mostly for integral membrane proteins (e.g. major histocompatibility complex class I heavy chain), for which dislocation entails release from the membrane and keeping the substrate in the second station in a soluble state (14,26,42). Nonetheless, defining the first station for membrane substrates is ambiguous because such substrates are inherent membrane-spanning proteins and by that resemble luminal soluble substrates al- FIG. 5. Schematic model. The order of events is marked by numbered arrows.
Step 5 may represent two consecutive steps, one blocked in CDC48 at 37°C and the other in cdc48 -10 at 30°C. For step 6, the two alternative options represent degradation in association with the ER membrane (6) or after release to the cytosol (6a, 6b). ready in the course of dislocation. Luminal soluble proteins, on the other hand, lack any cytosolic domains and therefore can be detected at the first station once they turn into entirely peripheral proteins (Refs. 16 and 17 and this work). In this regard, CPY* in yeast appears to behave differently than s in mammalian B cells, although both are bona fide luminal ERAD substrates. As shown previously (16), as soon as CPY* crosses the ER membrane it tends to accumulate at the ER cytosolic face as multiubiquitin conjugates, consistent with our observation that under normal conditions a significant proportion of microsomal HA-CPY* is accessible to trypsin. Conversely, s behaves similarly only when the proteasome is blocked because otherwise it is immediately eliminated (Ref. 9 and this work). Thus, for CPY* in yeast the second dislocation step (release to the cytosol) is delayed relative to the first one (passage across the membrane), whereas for s in B cells these two steps are temporally coupled. Moreover, when the proteasome fails to function, s accumulates at the ER cytosolic face because of uninterrupted passage followed by inhibited release (Fig. 5), whereas CPY* accumulates in the cytosol because of uninterrupted release followed by inhibited degradation (16). However, in contrast to the previous report (9), here we show that the step in s dislocation which requires the proteasomal activity is not the passage across the ER membrane but the release to the cytosol.
Our results assign the cytosolic face of the ER a midpoint to which luminal ERAD substrates emerge and p97 and the proteasome are recruited, to operate in ERAD. This notion is based on several lines of evidence. First, the reconstituted in vitro dislocation of s in the absence of added cytosol relies on ER-associated components. Second, as shown here for the first time, a luminal ERAD substrate that crosses the ER membrane interacts with the proteasome and, in particular, only with the minor ER-bound subpopulation identified previously in yeast and mammals (43)(44)(45). Third, the recently discovered novel cytosolic ERAD component, the AAA-ATPase p97/Cdc48p (Refs. 16, 21, 26 -28; for review, see Refs. 46 and 47), which participates in a variety of cellular processes and may chaperone protein unfolding or disassembly (48 -51), has been shown to coprecipitate with the membrane ERAD substrate 6myc-Hmg2p in yeast and with the luminal ERAD substrate s in B cells (21), as well as with the luminal yeast HA-CPY*. 2 Here we demonstrate that only the minor ER-bound subpopulation of p97/Cdc48p, and not its vast soluble pool, interacts with s, analogously to the proteasome. p97/Cdc48p probably plays a role in the actual passage of luminal ERAD substrates across ER membrane, as demonstrated for yeast HA-CPY*, which lingers in the lumen if Cdc48p fails to function. It remains to be established whether the exclusively membrane-and substrateassociated p97/Cdc48p is the subpopulation specialized in ERAD and whether recruitment to membranes provides mechanistic advantages. If so, this phenomenon can be exploited to improve the efficiency of ERAD in handling aberrant proteins in conformational diseases. The increased levels of ER-bound p97 in proteasome-inhibited cells raise the question of how p97 is recruited to membranes. Being a multiubiquitin binding protein, p97 may be recruited by association with multiubiquitinated ERAD substrates. In addition, recruitment may be regulated by p97 phosphorylation-dephosphorylation cycle (52). The possibility that p97 is associated with the ER because of its high affinity to syntaxin 5 is unlikely because only the p97/p47 complex interacts with syntaxin 5 (53,54), and p47 has been reported to obstruct rather than assist ERAD (26). Moreover, syntaxin 5 and the p97/p47 complex participate in homotypic membrane fusion (53). Although it has been shown that vesicular trafficking, which may involve homotypic fusion, is essential for ERAD of luminal substrates, it is dispensable for ERAD of membrane substrates (20,(55)(56)(57). The p97/ Cdc48 Ufd1/Npl4 complex, on the other hand, is clearly essential for ERAD of both luminal and membrane ERAD substrates (16, 21, 26 -28). Regardless of the mechanism, the ER-bound p97 resembles the ER-bound proteasome in the ability of both to interact with luminal substrates only subsequent to their emergence from the ER membrane. Then, the once luminal substrates have no topological barrier to interact with cytosolic p97/Cdc48p.
Two possible modes of substrate recognition by p97/Cdc48p can be envisaged. In the first, p97/Cdc48p interacts directly with a polypeptide segment as soon as it emerges from the ER membrane. This possibility may be supported by our finding that multiubiquitination of HA-CPY* is abolished, and this substrate remains entrapped in the ER lumen when Cdc48p fails to function. Moreover, it has been reported that p97 recognizes nonubiquitinated ERAD substrates (65). In the second, the recognition signal is the multiubiquitin moieties attached to the ERAD substrate, as demonstrated here for the luminal ERAD substrate s and in agreement with the direct interaction of p97 with multiubiquitin (29,33). Although this interaction is relatively weak (29,33), it is proposed to be assisted by the Ufd1⅐Npl4 complex, which by itself can bind multiubiquitin via the Npl4 zinc finger domain (58,59). Indeed, multiubiquitinated ERAD substrates bind the p97/Cdc48 Ufd1/Npl4 complex (26). Interestingly, p97 in the context of p97/p47 complex, which functions in homotypic membrane fusion, binds monorather than multiubiquitin (58). Based on the complementary results presented here and by Ye and co-workers (65), a dual function for p97/Cdc48p in dislocation is proposed (Fig. 5). p97/Cdc48p functions during the passage across the ER membrane by recognizing nonubiquitinated polypeptide segments, as well as subsequent to extensive multiubiquitination by recognizing multiubiquitin chains, and delivers them from the ER membrane to the proteasome. The physical interaction of p97/ Cdc48p with the proteasome on one hand (60,61) and with multiubiquitinated ERAD substrates on the other hand (Ref. 21 and this work), may assign p97 the responsibility for targeting multiubiquitinated ERAD substrates to the proteasome, as was suggested recently (47). Combined, the significance of these findings is increased further by the fact that they have been obtained for topologically distinct ERAD substrates, membrane and luminal, and in two distinct systems, yeast and mammalian cells.
In light of the two-step dislocation process, it is interesting to reevaluate the role played by p97/Cdc48 Ufd1/Npl4 . p97 is implicated in the single dislocation step of membrane major histocompatibility complex class I heavy chain (26). On the other hand, Ufd1 is implicated in the second dislocation step of luminal CPY*, namely the release to the cytosol (16). Regardless of the substrate, recent reports agree that p97/Cdc48 Ufd1/Npl4 participates in pulling ERAD substrates from the ER, as indicated by the activation of the unfolded protein response when p97/Cdc48 Ufd1/Npl4 fails to function (16,21,26). Here p97/ Cdc48p is shown to be directly involved in the passage of luminal substrates across the ER membrane and probably also in the release of these substrates subsequent to their multiubiquitination. Although p97/Cdc48p cannot initiate this passage, this chaperone can interact with luminal substrates as soon as they emerge from the ER membrane. To that effect, it is interesting that analogously to other luminal ERAD substrates (32), BiP appears to initiate s dislocation. Nonetheless, s dissociates from luminal BiP before it associates with cytosolic p97 (Fig. 5, step 3).