Polyubiquitin Serves as a Recognition Signal, Rather than a Ratcheting Molecule, during Retrotranslocation of Proteins across the Endoplasmic Reticulum Membrane

Polyubiquitination is required for retrotranslocation of proteins from the endoplasmic reticulum back into the cytosol, where they are degraded by the proteasome. We have tested whether the release of a polypeptide chain into the cytosol is caused by a ratcheting mechanism in which the attachment of polyubiquitin prevents the chain from moving back into the endoplasmic reticulum. Using a permeabilized cell system in which major histocompatibility complex class I heavy chains are retrotranslocated under the influence of the human cytomegalovirus protein US11, we demonstrate that polyubiquitination alone is insufficient to provide the driving force for retrotranslocation. Substrate release into the cytosol requires an additional ATP-dependent step. Release requires a lysine 48 linkage of ubiquitin chains. It does not occur when polyubiquitination of the substrate is carried out with glutathione S -transferase (GST)-ubiquitin, and this correlates with poly-GST-ubiquitin not being recognized by a ubiquitin-binding domain in the Ufd1-Npl4 cofactor of the ATPase p97. These data suggest that polyubiquitin does not serve as The binding experiments were carried out at 4 °C in 0.3 ml of 50 m M Hepes, pH 7.3, 150 m M potassium chloride, 2.5 m M magnesium chloride, 5% glycerol, 2 m M (cid:1) -mercaptoethanol, 0.1% Triton X-100, 1 mg/ml bovine serum albumin. 20- (cid:4) l samples of polyubiquitination reactions were incubated with 1 (cid:4) g of the various purified recombinant proteins. The proteins were then precipitated with specific antibodies, and bound ubiquitin chains were detected by immunoblotting with ubiquitin antibodies. Miscellaneous Methods— For ATP depletion, the cytosol was incu- bated for 20 min at 37 °C with 0.1 unit/ (cid:4) l hexokinase and 20 m M glucose. Where indicated, 3 m M AMP-PNP and 3 m M MgOAc were added during the chase period. Wild-type HA/A2 and K-R HA/A2 constructs and cell lines were described previously (7).

In eukaryotic cells, a quality control system in the endoplasmic reticulum (ER) 1 ensures that only proteins with a native, folded conformation leave the organelle for other destinations, such as the plasma membrane. Misfolded proteins that cannot reach their native state are retrotranslocated from the ER into the cytosol, where they are subsequently degraded by the proteasome (1)(2)(3). This cellular pathway is hijacked by certain viruses, such as the human cytomegalovirus. Two human cytomegalovirus proteins, US2 and US11, are able to direct newly synthesized major histocompatibility complex (MHC) class I heavy chains into the ER degradation pathway (4 -7). Human MHC class I heavy chain is a 43-kDa type I transmembrane protein with a large luminal/extracellular domain and a short cytosolic tail. The protein is initially inserted into the ER membrane and glycosylated, but under the influence of either US2 or US11, it is rapidly moved into the cytosol, where its N-glycan is cleaved off and the polypeptide chain is degraded by the proteasome (5). The absence of MHC class I molecules at the cell surface allows the virus to propagate without the infected cell being detected by cytotoxic T cells.
Most substrates destined to be degraded are polyubiquitinated while undergoing retrotranslocation. Our previous results showed that polyubiquitination is not only required for degradation by the proteasome but also for retrotranslocation per se (8). Lysine 48-linked polyubiquitin chains are required for protein degradation by the proteasome (9,10), but given that other linkages are required for other processes (11)(12)(13)(14)(15), it would be important to know which kind of polyubiquitin chains are involved in retrotranslocation. The role of polyubiquitination in retrotranslocation also remained unclear. In one model, polyubiquitin provides the driving force for moving the substrate into the cytosol by acting as a ratcheting molecule (for discussion, see Ref. 3). The attachment of the bulky polyubiquitin moiety to a polypeptide segment on the cytosolic side of the membrane would bias random movements of the substrate across the ER membrane by preventing it from sliding back into the ER lumen. This model would be analogous to the one describing posttranslational translocation in the forward direction. In this case, the binding of the luminal chaperone BiP prevents the translocating polypeptide chain from moving back into the cytosol, eventually resulting in its complete transport into the ER lumen (16). If a polyubiquitin-mediated ratcheting mechanism functioned in a similar way in retrotranslocation, polyubiquitination alone should be sufficient to move a substrate into the cytosol. Alternatively, it is possible that the polyubiquitin chain serves as a recognition signal for a downstream component. One candidate for this downstream component is the ATPase p97 (called Cdc48 in yeast), which together with its cofactor Ufd1-Npl4 has been implicated in moving polypeptides from the ER membrane into the cytosol (17)(18)(19)(20)(21).
In this study, we have addressed the role of polyubiquitination in retrotranslocation. We show that polyubiquitination is not sufficient to move a substrate into the cytosol. Rather, a subsequent ATP-dependent step is required, during which polyubiquitin in a lysine 48 linkage serves as a recognition signal. Our in vitro experiments suggest that the downstream component may be the p97-Ufd1-Npl4 complex.

EXPERIMENTAL PROCEDURES
Pulse-Chase Analysis with Permeabilized Cells-Control and US11expressing U373-MG astrocytoma cells (4) were cultured as described previously (5). The cells were detached from tissue culture flasks with trypsin and incubated in suspension in methionine-and cysteine-free Dulbecco's minimal essential medium for 1 h at 37°C. Where indicated, 50 M of proteasome inhibitor MG-115 (Calbiochem) was present during incubation. The cells resuspended at 1 ϫ 10 7 /ml were pulse-labeled for 3-5 min at 37°C in 290 Ci/ml [ 35 S]methionine and [ 35 S]cysteine ( 35 S-Protein Express Labeling Mix; New England Nuclear). They were then placed on ice and washed once with phosphate-buffered saline supplemented with 0.9 mM CaCl 2 and resuspended at 1.6 ϫ 10 7 cells/ml in PB (25 mM Hepes, pH 7.3, 115 mM potassium acetate, 5 mM sodium acetate, 2.5 mM MgCl 2 , 0.5 mM EGTA) containing 0.04% digitonin (Merck; purified as described in Ref. 22) and protease inhibitor mix (10 g/ml leupeptin, 5 g/ml chymostatin, 3 g/ml elastatinal, and 1 g/ml pepstatin). The samples were centrifuged, and the pellets were washed in PB without digitonin and resuspended in cow liver cytosol. Where indicated 1 M ubiquitin aldehyde, an inhibitor of deubiquitinating enzymes, was added. The samples were then chase-incubated at 37°C for different time periods. After centrifugation in a microcentrifuge at 14,000 rpm at 4°C for 10 min, the supernatant of the samples was removed and saved, whereas the pellet fraction was resuspended in PB. The lysates were made from the samples, and the immunoprecipitations were carried out as described below. Cow liver cytosol was prepared as described previously (8).
Lysate Preparation and Immunoprecipitations-The lysates were made from samples corresponding to 1-2 ϫ 10 6 cells in 0.5% Nonidet P-40 (Igepal; CA-630; Sigma), 50 mM Tris, pH 8, 150 mM NaCl, and 10 mM MgCl 2 . The buffer also contained 1 mM phenylmethylsulfonyl fluoride and the previously mentioned protease inhibitor mix. The samples were agitated for 20 min at 4°C and then clarified by centrifuging in a microcentrifuge at full speed for 10 min. The resulting supernatant was used for immunoprecipitation after SDS and dithiothreitol were added to final concentrations of 0.1% and 0.2 mM, respectively. Denaturing SDS lysates were made by resuspending cell pellets or cell fractions in 100 l of 2% SDS and 5 mM dithiothreitol. The samples were heated to 95°C for 5 min, cooled to room temperature, agitated vigorously, and diluted into Nonidet P-40 buffer so that the final concentrations in the lysate used for immunoprecipitation were ϳ0.2% SDS, 0.5 mM dithiothreitol, 0.5% Igepal, 50 mM Tris, pH 8, 150 mM NaCl, and 10 mM MgCl 2 . All of the immune complexes were recovered by precipitation with fixed Staphylococcus aureus bacteria (Staph A).
Anti-heavy chain serum and antibodies against bovine ubiquitin were described previously (7). The monoclonal antibody 12CA5, which recognizes the influenza hemagglutinin (HA) epitope, was purified from tissue culture cell supernatants by standard methods (23) or purchased from Sigma. Immunoprecipitation with heavy chain antibodies and reprecipitation with ubiquitin antibodies were done as described (7).
Ubiquitin Depletion-Ubiquitin was depleted from cow liver cytosol as described, using a recombinant GST-tagged Ubc2 that carries a serine instead of a cysteine in the active site (8). The depletion reaction mixture contained 16 M GST-tagged Ubc2 Ser and an ATP-regenerating system (24). The reaction was allowed to proceed for 10 min at 37°C. The reaction mixture was subsequently transferred to glutathione-Sepharose 4B beads and incubated at 4°C for 15-30 min. The nonbound material was frozen at Ϫ80°C.
Ubiquitin Reagents-Bovine ubiquitin was purchased from Sigma. For the experiment in Fig. 2, ubiquitin was labeled using the following procedure. Bovine ubiquitin was iodinated using Iodogen Iodination Reagent (Pierce). Each 90-l iodination reaction mixture contained 8.9 mg/ml ubiquitin and 22.2 mCi/ml Na 125 I (PerkinElmer Life Sciences) in 44.4 mM potassium phosphate, pH 7.6. The reaction tube was coated with 100 l of 1 mg/ml Iodogen in chloroform. The reaction was carried out on ice for 30 min, after which the reaction was stopped by the addition of 100 l of 0.4 mg/ml tyrosine and 16 l of 0.1 M nonradioactive NaI. Iodinated ubiquitin was purified by passing it through a Sephadex G-25 spin column equilibrated with PB. The final concentration was estimated at 300 M, labeled to 16,500 cpm/pmol. 125 I-Ubiquitin was added to the cow liver cytosol at a concentration of 20 M.
The pGEX-Ubiquitin plasmids coding for GST-ubiquitin and GST-K48R were provided by T. Sommer. All GST-ubiquitin proteins were expressed in BL21 (DE3) Escherichia coli. The cells were grown to an A value of ϳ1.0 and induced with 0.4 mM isopropyl-1-thio-␤-D-galactopyranoside for 6 h. They were lysed and centrifuged at 42,000 rpm in an ultracentrifuge to remove nonsoluble material. The fusion protein was purified from the supernatant by glutathione affinity chromatography.
Binding of p97 to Heavy Chains in Permeabilized Cells-The experiments were performed essentially as described (18). The cell fractions were solubilized in 1% Deoxy Big CHAP ([N,NЈ-bis(3-D-gluconamidopropyl)deoxycholamide]), 30 mM Tris/HCl pH 7.4, 150 mM potassium acetate, 4 mM magnesium acetate, 1 mM ATP, and protease inhibitors. Immunoprecipitations were carried out with His antibodies followed by a second precipitation with HA antibodies.
Miscellaneous Methods-For ATP depletion, the cytosol was incubated for 20 min at 37°C with 0.1 unit/l hexokinase and 20 mM glucose. Where indicated, 3 mM AMP-PNP and 3 mM MgOAc were added during the chase period. Wild-type HA/A2 and K-R HA/A2 constructs and cell lines were described previously (7).

Retrotranslocation Requires Polyubiquitination and a Subsequent ATP-dependent
Step-To study the mechanism of retrotranslocation we used a permeabilized cell system in which native cytosol can be replaced with cow liver cytosol (8), allowing for convenient manipulation of cytosol. Human astrocytoma cells, stably expressing the human cytomegalovirus protein US11, were pulse-labeled with [ 35 S]methionine and [ 35 S]cysteine. The cells were permeabilized by addition of a low concentration of digitonin, pelleted, and resuspended in cytosol from cow liver in the presence of ATP and an ATP-regenerating system. After chase incubation at 37°C, one portion of the samples was analyzed directly, and the other was separated into membrane pellet and supernatant fractions. The samples were solubilized in detergent, subjected to immunoprecipitation with antibodies to MHC class I heavy chains (␣HC), and analyzed by SDS-PAGE and autoradiography (Fig. 1A, lanes 9 -12). As reported previously (5, 7), the heavy chains were degraded during the chase period (lane 10 versus lane 9). In addition, the characteristic deglycosylated, faster migrating heavy chain species, which fractionated with the cytosol, appeared (lane 11 versus lane 12). To detect polyubiquitinated heavy chains, a portion of the samples was subjected to a second round of immunoprecipitation with ubiquitin antibodies (lanes 29 -32). Polyubiquitinated heavy chains, appearing as a smear of high molecular mass species, accumulated over time (lane 30 versus lane 29). Most of the polyubiquitinated heavy chains were found in the cytosolic fraction (lane 31 versus lane 32), indicating that they had been released from the ER membrane. In control cells not expressing US11, the heavy chains were stable (lanes 1-8), and no polyubiquitination was observed (lanes [21][22][23][24][25][26][27][28]. It should be noted that previous experiments showed that polyubiquitination is required for retrotranslocation (8), implying that the majority of the retrotranslocated heavy chains, visible in the cytosolic fraction as the deglycosylated species, must have undergone deubiquitination, even in the presence of 1 M ubiquitin-aldehyde.
When permeabilized cells expressing US11 were incubated with ATP-depleted cow liver cytosol (⌬ATP), the degradation of the heavy chains was greatly reduced, and the majority of the material stayed on the membrane (lane 16 versus lane 15). As expected from the ATP requirement of the ubiquitin-activating enzyme, no polyubiquitination was observed (lanes 33-36). The addition of AMP-PNP to the ATP-depleted sample did not increase the degradation of heavy chains (lanes [17][18][19][20], but it partially restored polyubiquitination (lane 38), in agreement with the expectation that the ubiquitin-activating enzyme can utilize the ATP analog because it hydrolyzes ATP into AMP and PP i (27,28). The polyubiquitinated chains that were ]cysteine, permeabilized, and incubated in cow liver cytosol either in the presence of ATP and an ATP-regenerating system (ARS), in the absence of ATP (⌬ATP), or in the absence of ATP but the presence of AMP-PNP (⌬ATP ϩ AMP-PNP). After incubation for 0 or 45 min at 37°C, one portion of the sample was directly analyzed, the other was fractionated into supernatant (S) and membrane pellet (P). The samples were subjected to immunoprecipitation (IP) with ␣HC and to a second round of immunoprecipitation with ␣Ub. Analysis was done by SDS-PAGE and autoradiography. The band migrating around 43 kDa is possibly heavy chain, deubiquitinated during sample preparation (*). HC Ϫ CHO and HC ϩ CHO indicate heavy chains without or with carbohydrate chain, respectively. 1 M ubiquitin-aldehyde was present during incubation. B, US11 cells stably expressing either HA-tagged wild-type or K-R heavy chains were treated as in A, except that cells were preincubated with proteasome inhibitor and that recovery of the heavy chains from the lysates was carried out with ␣HA. 1 M ubiquitin-aldehyde was present during incubation.
formed fractionated with the membrane rather than with the cytosol (lane 40 versus lane 39). The apparent incomplete restoration of polyubiquitination by AMP-PNP may in part be due to deubiquitination occurring during sample preparation (a band at the approximate size of unmodified heavy chains (indicated by an asterisk) was consistently generated even in the presence of protease inhibitors).
To show that this polyubiquitination occurs in a domain of the heavy chain that was originally in the ER lumen, we employed a mutant (K-R) heavy chain, in which the lysines in the C-terminal tail were mutated to arginines. Because ubiquitination can only take place on free amino groups (internal lysine residues or the extreme N terminus), the cytoplasmic domain of this mutant heavy chain cannot be ubiquitinated. For these experiments, we used stable cell lines that express HA-tagged MHC class I heavy chains in addition to US11. In the presence of ATP, a fraction of both the wild-type and the mutant (K-R) heavy chains were deglycosylated during the chase period (  8 versus lanes 1-4  and lanes 13-16 versus lanes 9 -12), which may reflect the existence of an ATP-dependent deubiquitinating step. Taken together, these results indicate that polyubiquitination occurs on a segment that was originally in the ER lumen and that modification of these residues alone is insufficient to move the substrate into the cytosol. An additional ATP-dependent step that requires hydrolysis of the ␥-phosphate of ATP appears to be involved.
To provide further evidence that an additional ATP-dependent step is required for the release of heavy chains into the cytosol, we performed experiments in the absence of AMP-PNP, thus excluding the possibility that the ATP analog blocked a step following polyubiquitination. Permeabilized cells labeled with [ 35 S]methionine and [ 35 S]cysteine were resuspended in cow liver cytosol in the absence of ATP and chase-incubated for 20 min. Then either buffer or 0.1 mM ATP was added, and the incubation was continued for another 10 min (30 min total). The polyubiquitinated heavy chains formed during the labeling period stayed on the membrane in the absence of ATP (Fig. 3, lanes 1-7, P). However, when ATP was added, about half of the polyubiquitinated chains were released into the cytosol (lane 13 versus lane 14). In the presence of ATP, some shorter polyubiquitinated species appeared in the membrane fraction, either because existing chains were deubiquitinated or because new polyubiquitin chains were formed. These results confirm that polyubiquitination alone is insufficient and that ATP is required for the release of modified substrate from the ER membrane into the cytosol.
Lysine 48 Linkage of Polyubiquitin Chains Is Required for Retrotranslocation-Different lysines in the ubiquitin molecule can serve for chain elongation, resulting in polyubiquitin chains with distinct roles in the cell. For example, polyubiquitin chains with lysine 48 linkages are required for proteasomal degradation (9), whereas chains with lysine 63 linkages are implicated in nonproteolytic events (11,(13)(14)(15). We therefore wished to determine whether a specific linkage is required for retrotranslocation. US11 cells were pulse-labeled with [ 35 S]methionine/[ 35 S]cysteine, permeabilized, and resuspended in ubiquitin-depleted cytosol replenished with buffer, wild-type ubiquitin, or ubiquitin mutants lacking specific attachment sites (K48R or K63R ubiquitin). The cells were incubated for 30 min at 37°C, and then a portion of the samples was analyzed directly, whereas the other was separated into a membrane pellet and a cytosolic supernatant fraction. After solubilization in detergent, the samples were subjected to immunoprecipitation with ␣HC and analyzed by SDS-PAGE and autoradiography. In the absence of ubiquitin (Fig. 4, lanes 1-4) These results indicate that polyubiquitin chains with lysine 48 linkages are required for retrotranslocation.
Poly-GST-Ubiquitination of MHC Class I Heavy Chain Prevents Its Retrotranslocation-Because polyubiquitination with a specific linkage is required for retrotranslocation but on its own is insufficient, we tested whether polyubiquitin chains may function as a recognition signal. Specifically, we investigated whether modification of polyubiquitin chains by the addition of a GST moiety to the N-terminal part of ubiquitin would influence retrotranslocation. We reasoned that GST-ubiquitin molecules may still be linked with one another to form poly-GST-ubiquitinated heavy chains but that these chains may no longer be recognized by a downstream component.
We incubated 35 S-labeled permeabilized cells with cow liver cytosol supplemented with GST-ubiquitin. Compared with control cow liver cytosol, the appearance of deglycosylated heavy chains in the cytosol was reduced (Fig. 5A, lanes 16 -20 versus  lanes 11-15). In addition, a high molecular mass species appeared during the chase period, which was not seen in the absence of GST-ubiquitin. The high molecular mass species were GST-ubiquitinated heavy chains as demonstrated by reimmunoprecipitation with either ubiquitin antibodies (bottom panel, lanes 41-45) or GST antibodies (Fig. 5B, lanes 13-16). In most experiments, a fraction of 15-20% of the total population of heavy chains present at the beginning of the chase period were poly-GST-ubiquitinated, but occasionally up to 55% was modified. This is a much higher percentage of modification than seen with wild-type ubiquitin (see lanes 11-15 and 36 -40), which may suggest that deubiquitination is impaired by GST-ubiquitin (lane 44). This is probably due to chains containing endogenous ubiquitin in addition to GST-ubiquitin. Indeed, when ubiquitin was depleted from the cytosol (8) before the addition of GSTubiquitin, the appearance of the deglycosylated species was completely blocked, and essentially all modified chains remained in the membrane (lanes 21-25). The maximum amount of GST-ubiquitinated heavy chains was reached after 20 -30 min of incubation (Fig. 5B), and these adducts were stable for at least 2 h. Similar results were obtained with both the wildtype and K-R mutant heavy chains (data not shown). The GST-ubiquitin modified heavy chains were sensitive to protease treatment (data not shown), as expected from the fact that ubiquitination is a cytosolic event. These results show that the attachment of multiple GST-ubiquitin molecules to heavy chains is not sufficient for their release into the cytosol. They suggest that poly-GST-ubiquitin prevents retrotranslocation by disrupting a recognition signal for a downstream component. Apparently, the GST moiety also prevents recognition by the deubiquitinating enzymes.

Poly-GST-Ubiquitin Is Not Recognized by a Ubiquitin-binding Domain in the Ufd1-Npl4
Cofactor of the ATPase p97-One of the candidates for a downstream component is the ATPase p97, which has been shown to function together with its cofactor Ufd1-Npl4 in retrotranslocation (17)(18)(19)(20)(21). We first tested whether poly-GST-ubiquitination of MHC class I heavy chains would abrogate substrate recognition by p97, which would provide an explanation why these chains can no longer be moved into the cytosol. To determine substrate binding by p97, astrocytoma cells expressing US11 and HA-tagged wild-type heavy chains were treated with proteasome inhibitor, labeled, and permeabilized. The endogenous cytosol was replaced with cow liver cytosol that was either depleted of ubiquitin as described (8) or first depleted and then replenished with ubiquitin or GST-ubiquitin (Fig. 6A). Where indicated, His-tagged recombinant p97 protein was added, and the samples were chaseincubated for 30 min and then separated into pellet and supernatant fractions. Immunoprecipitation of the heavy chains showed that in the presence of ubiquitin, the deglycosylated species appeared in the supernatant, regardless of whether p97 was added or not (upper panel, lanes 4 and 8). In the absence of ubiquitin, retrotranslocation was blocked (lanes 9 -12). In the presence of GST-ubiquitin, retrotranslocation was also reduced, and the characteristic high molecular adducts appeared in the membrane fraction (lanes 13-16). To detect p97-associated heavy chains, the samples were first immunoprecipitated with His antibodies and then with HA antibodies (lanes 17-32). Binding of heavy chains was seen regardless of whether or not ubiquitin was present during the incubation (lanes 21-24 versus lanes [25][26][27][28], demonstrating that p97 can interact with nonmodified substrate. In the absence of His-p97, no material was immunoprecipitated (lanes [17][18][19][20]. p97 also interacted with poly-GST-ubiquitinated heavy chains (lanes 29 -32). Thus, the fact that poly-GST-ubiquitinated heavy chains cannot be released into the cytosol cannot be explained by disruption of substrate recognition by p97.
Because previous experiments implicated the cofactor Ufd1-Npl4 in the recognition of the polyubiquitin chains, 2 we next tested whether poly-GST-ubiquitin may not be recognized by this cofactor, providing a possible explanation for why this modification disrupts retrotranslocation. Polyubiquitin chains were synthesized in vitro using purified recombinant proteins. The assay contained ubiquitin, the E1, the E2, and a GST fusion to a cytosolic fragment of the E3. We employed Ubc7 and gp78 because they are known to be involved in the ER degradation of at least some proteins (29 -32). When all of the components were incubated together in the presence of ATP, polyubiquitin chains were synthesized (Fig. 6B, upper panel, lane  1). These chains were tested for interaction with the Ufd1-Npl4 (U/N) complex. The mammalian cofactor complex has two polyubiquitin-binding domains: one in Ufd1 and one in a zinc finger domain of Npl4. Because the Npl4 ubiquitin-binding domain is lacking in yeast (33) and the complex likely functions analogously to the mammalian one, the binding site in Ufd1 appears to be most important in retrotranslocation. We therefore tested the binding of both wild-type Ufd1-Npl4 and of a mutant complex that lacked the zinc finger domain in Npl4 (U/N⌬ZF). Polyubiquitin chains were incubated with these complexes, and binding was assessed by immunoprecipitation with Ufd1 antibodies followed by immunoblotting with ubiquitin antibodies. With the wild-type complex, a significant fraction of the polyubiquitin chains was bound (Fig. 6B, upper panel, lane 3; the lower panel demonstrates coprecipitation of Ufd1 and Npl4). As expected, the mutant complex (U/N⌬ZF), containing only the Ufd1 binding site for polyubiquitin, had reduced but still significant binding activity (lane 4). Next, we performed binding experiments with poly-GST-ubiquitin instead of wild-type polyubiquitin. The polymerization reaction with GST-ubiquitin was less efficient, but high molecular mass adducts were generated (lane 5). The poly-GST-ubiquitin chains bound to U/N but not to U/N⌬ZF as demonstrated with both ubiquitin and GST blots (lanes 7 and 8 versus lanes 3 and  4 in upper and lower panels). The generated poly-GST-ubiquitin chains were linked through lysine 48 of ubiquitin, as demonstrated by the fact that poly-GST-ubiquitin chains could not generated with GST-K48R or GST-K0 ubiquitin (which lacks all 7 lysines in ubiquitin) (Fig. 6C, compare lane 4 with  lanes 5 and 6). Taken together, these results show that poly-GST-ubiquitin chains are not recognized by the Ufd1 ubiquitinbinding domain of the cofactor Ufd1-Npl4, even though the chains have the correct ubiquitin linkage. On the other hand, intact Ufd1-Npl4 can interact with poly-GST-ubiquitin, likely through the zinc finger domain in Npl4. Our data are consistent with the hypothesis that poly-GST-ubiquitinated heavy chains may not be recognized by the cofactors of p97 during retrotranslocation. DISCUSSION Our previous results showed that polyubiquitination is required for retrotranslocation of MHC class I heavy chains from the ER membrane into the cytosol (8). Now, we demonstrate that polyubiquitination is insufficient for retrotranslocation. When the modification reaction was performed in the presence of AMP-PNP or when GST-ubiquitin instead of wild-type ubiquitin was employed in the reaction, polyubiquitinated heavy chains stayed on the membrane and were not moved into the cytosol. In addition, membrane-associated heavy chains modified with wild-type polyubiquitin could be released into the cytosol by the addition of ATP. Together, these data show that there is an ATP requirement in at least two steps of retrotranslocation: one for polyubiquitination, which can be satisfied by AMP-PNP, and the other for a downstream component that requires hydrolysis of the ␥-phosphate of ATP. Because polyubiquitination alone was insufficient to move a polypeptide into the cytosol, we conclude that the driving force for retrotranslocation is not provided by a simple ratcheting mechanism. This is in contrast to posttranslational translocation in the forward direction, in which a secretory protein can be moved completely into the ER lumen by a ratcheting mechanism (16). Although polyubiquitination does not suffice to move a polypeptide chain completely into the cytosol, it may still prevent movements back into the ER lumen. The initial steps of retrotranslocation must be independent of polyubiquitina-  1-25, and B, lanes 1-8), followed by a second round of immunoprecipitation with ubiquitin antibodies (A, lanes 26 -45, and B, lanes 9 -12) or GST antibodies (B, lanes [13][14][15][16]. HC Ϫ CHO and HC ϩ CHO indicate heavy chains without or with carbohydrate chain, respectively. Molecular mass marker (M) bands are 220, 97, 66, 45, and 30 kDa. The band migrating around 43 kDa is possibly heavy chain deubiquitinated during sample preparation (*).
tion of the substrate, because the modification occurs on segments that were originally in the ER lumen. It may not be too surprising that polyubiquitination does not provide a simple ratcheting mechanism, because the distances between consecutive lysine residues, the usual attachment sites in a polypep-tide chain, would make the ratchet in general inefficient. In addition, some proteins contain few or even no lysines.
Our data not only indicate the existence of an ATP-dependent step following polyubiquitination but also suggest that polyubiquitin serves as a recognition signal for a downstream component, The samples were then incubated at 37°C for 0 or 30 min and separated into supernatant (S) and membrane pellet (P) fractions. One portion was subjected directly to immunoprecipitation (IP) with ␣HA (lanes 1-16), and another was first immunoprecipitated with His antibodies for p97, followed by a second round of immunoprecipitation with ␣HA (lanes 17-32). All of the samples were analyzed by SDS-PAGE and autoradiography. After immunoprecipitation with His antibodies, part of the sample was analyzed by Coomassie Blue staining to show amounts of precipitated His-tagged p97 (lanes 33-48). B, polyubiquitin chains were synthesized in vitro using ubiquitin or GST-ubiquitin as substrate (see "Experimental Procedures"). The samples were incubated with buffer (Ϫ) or with the complex of purified Ufd1 and either Npl4 (U/N) or Npl4 lacking the zinc finger domain (U/N⌬ZF). Bound polyubiquitin chains were analyzed after immunoprecipitation with Ufd1 antibodies by blotting with ubiquitin antibodies. The lower panel shows an immunoblot with antibodies to Ufd1, Npl4, and GST. 30% of the material used for the binding experiments was subjected directly to SDS-PAGE (input). C, polyubiquitin chains were synthesized in vitro using either wild-type ubiquitin, a GST fusion to wild-type ubiquitin, or GST-fusions to K48R or K0 ubiquitin (in the latter all lysines are changed to arginines). In lane 3, the reaction contained wild-type ubiquitin and GST. The chains were separated by SDS-PAGE and analyzed by immunoblotting with ubiquitin antibodies. because a substrate that is poly-GST-ubiquitinated could not be moved into the cytosol. This idea of polyubiquitin as a recognition signal is strengthened by the fact that a specific lysine 48 linkage within the polyubiquitin chain is required for retrotranslocation, the same linkage that is required for the proteasome that acts subsequent to retrotranslocation in substrate degradation.
The simplest model is that a downstream component recognizes polyubiquitin and functions in an ATP-dependent process. Previously, the ATPase p97 and its cofactor Ufd1-Npl4 have been implicated in retrotranslocation. Our data suggest that this complex may be involved in the ATP-and polyubiquitindependent step. A requirement for ATP hydrolysis by p97 would explain why AMP-PNP did not allow retrotranslocation. Furthermore, a p97 mutant defective in ATP hydrolysis leads to a similar phenotype as AMP-PNP, i.e. the accumulation of polyubiquitinated substrate on the ER membrane. 2 The binding of polyubiquitin to the p97 complex is also consistent with it being the immediate downstream component. The binding occurs to an N-terminal domain of Ufd1 in the cofactor complex. 2 Although the zinc finger domain in Npl4 can also bind polyubiquitin (25), this interaction is likely not important for retrotranslocation. It does not discriminate between lysine 48 and lysine 63 linkages, 2 and the yeast homolog of Npl4 lacks the zinc finger domain (33) altogether but likely functions analogously to the mammalian protein in retrotranslocation. The fact that this domain can even interact with mono-ubiquitin (34) may explain the interaction we observe with poly-GST-ubiquitin. In contrast, the ubiquitin-binding domain in Ufd1 is specific for lysine 48 linkages in polyubiquitin chains and does not recognize lysine 63 linkages, 2 consistent with our finding that lysine 48 is required for retrotranslocation. We have found that Ufd1-Npl4 lacking the zinc finger domain in Npl4 does not bind poly-GST-ubiquitin, even though the chains have the correct lysine 48 linkages. The lack of interaction of poly-GST-ubiquitin with the ubiquitin-binding domain in Ufd1 could explain why this modification blocks the movement of MHC class I heavy chains into the cytosol. However, it is possible that other polyubiquitin-binding proteins, such as the proteasome or deubiquitinating enzymes, are also prevented from binding and cause the retrotranslocation defect. For example, deubiquitinating enzymes may be important to remove the polyubiquitin chains from polypeptides so that they can be moved through the relatively narrow pore in the double-barrel structure of the p97 ATPase (35). This function would be similar to the role suggested for deubiquitinating enzymes in the degradation of proteins by the proteasome. The inhibitory effect of AMP-PNP on deubiquitination seen in our experiments could be explained by an enzyme similar to that found on the proteasome (26,36).
Interestingly, p97 itself can bind to both nonubiquitinated and polyubiquitinated substrate molecules. This suggests that there are two consecutive interactions of the p97-Ufd1-Npl4 complex with retrotranslocating heavy chains. It first binds to nonubiquitinated segments of the substrate and then, following ubiquitination, to the polyubiquitin chain. How exactly the polypeptide substrate would subsequently be moved into the cytosol remains unclear, and our data also do not exclude that a component other than the p97-Ufd1-Npl4 complex could be involved. However, one of the possibilities is that the interaction of the cofactor with the polyubiquitin chain activates the ATPase p97 to "pull" the polypeptide chain out of the membrane.