The Pathway of Us11-Dependent Degradation of Mhc Class I Heavy Chains Involves a Ubiquitin-Conjugated Intermediate

Control and US11-expressing U373-MG astrocytoma cells (Jones et al. 1995) were cultured as described previously (Wiertz et al. 1996a).

Cells were detached from tissue culture flasks with trypsin and incubated in suspension in methionine- and cysteine-free DME for 1 h at 37°C. Cells at 1 × 10⁷/ml were pulse-labeled for 10 min at 37°C in 290 μCi/ml [³⁵S]methionine and cysteine (³⁵S-Protein Express Labeling Mix; New England Nuclear). At the beginning of the chase period, 5 mM nonradioactive methionine and 1 mM cysteine were added. Samples were taken at various timepoints and lysates were made as described below. In all pulse–chase experiments, whether on intact or permeabilized cells (see below), when present, the proteasome inhibitor ZL₃VS (Bogyo et al. 1997) was at 50 μM throughout.

Cells were labeled as described above, but for only 3 min, placed on ice, and washed once with PBS. They were then resuspended at 1.6 × 10⁷ cells/ml in PB (25 mM Hepes 7.3, 115 mM potassium acetate, 5 mM sodium acetate, 2.5 mM MgCl₂, 0.5 mM EGTA) containing 0.025% digitonin (Merck, purified as described in Gorlich and Rapoport 1993), an ATP-regenerating system (Feldman et al. 1997), and protease inhibitors (10 μg/ml leupeptin, 5 μg/ml chymostatin, 3 μg/ml elastatinal, and 1 μg/ml pepstatin). After a chase period at 37°C, lysates were made from samples taken at various time points and immunoprecipitations were carried out as described below. 5 min into the chase period, >95% of cells were permeable to Trypan blue and >90% of their lactate dehydrogenase activity was released (Tan et al. 1992; data not shown).

Permeabilized cells were resuspended at 6 × 10⁶ cell equivalents/ml in homogenization buffer (PB with 250 mM sucrose, PMSF, aprotinin, and leupeptin). Homogenization was carried out using a ball bearing device (Balch and Rothman 1985). The homogenates were fractionated by centrifuging sequentially at 1,000 g for 10 min, 10,000 g for 30 min, and 100,000 g for 1 h, and the resulting pellets were resuspended in homogenization buffer. The resuspended pellets and the 100,000 g supernatant were diluted with NP-40 lysis buffer and immunoprecipitations were carried out as described below. In the experiments shown in Fig. 12, the intact cells were resuspended at a concentration of 1.2 × 10⁷ cells/ml in homogenization buffer that contained an ATP regenerating system (Feldman et al. 1997).

Soluble, cytosolic proteins were squeezed out of permeabilized cells by centrifugation. At the indicated chase times, two samples were taken from each permeabilization reaction. Both were centrifuged in a microfuge at 14,000 rpm at 4°C for 10 min. The supernatant and pellet of one sample were remixed to represent the total starting material. The supernatant of the other sample was removed and saved. The pellet fraction was resuspended in PB containing digitonin and the ATP regenerating system. Lysates of each fraction (total, supernatant, and pellet) were made and immunoprecipitations were carried out, as described below.

Samples from the permeabilization reactions were added to ice-cold trypsin in PB, such that the final trypsin concentration in each reaction was as indicated (see Fig. 4). After 30 min on ice, trypsin digestion was stopped by the addition of the protease inhibitors PMSF and Nα-tosyl-lys chloromethyl ketone, hydrochloride (TLCK; Calbiochem-Novabiochem). Denaturing SDS lysates were made and heavy chains were immunoprecipitated as described below.

US11 astrocytomas were harvested, washed once in DME containing calf serum and glucose at 0.45 g/liter, and resuspended at 2 × 10⁶ cells/ml in the same media containing 1 mCi/ml ³H-mannose (New England Nuclear) with or without 50 μM ZL₃VS. After labeling at 37°C for 1 h with frequent agitation, denaturing SDS lysates were made and sequential immunoprecipitations with αHC and αUb serum were carried out as described below and in the legend to Fig. 6 B.

As noted in the figure legends, lysates were made in three different ways, depending on the experiment. In all cases, 1–2 × 10⁶ cells (or cell equivalents) were used to make 1–1.5 ml of lysate for each immunoprecipitation. The type of lysate did not affect the overall outcome of any of the immunoprecipitation experiments (data not shown), although the yield of immunoprecipitated heavy chain was greater when SDS was present. Nondenaturing NP-40 lysates were made by resuspending cell pellets, permeabilized cells, or cell fractions so that the final buffer was 0.5% NP-40 (or Igepal CA-630; Sigma Chemical Co.), 50 mM Tris, pH 8, and 10 mM MgCl₂. Samples were agitated for 20 min at 4°C and then clarified by centrifuging in a microfuge at full speed for 10 min. The resulting supernatant was used for immunoprecipitation. Denaturing SDS lysates were made by resuspending cell pellets, permeabilized cells, or cell fractions in 100–150 μl of 1% SDS and 2 mM DTT. Samples were heated to 95°C for 5 min, cooled to room temperature, agitated vigorously, and diluted into NP-40 buffer so that the final lysate used for immunoprecipitation was 0.1% SDS, 0.2 mM DTT, 0.5% Igepal, 50 mM Tris, pH 8, and 10 mM MgCl₂. In some experiments (Fig. 7, Fig. 9C and Fig. D, Fig. 11, and Fig. 12), nondenaturing NP-40 lysates were made and then SDS and DTT were added after the clarifying spin to final concentrations of 0.1% and 0.2 mM, respectively. All immune complexes were recovered by precipitation with fixed Staphylococcus aureus bacteria (Staph A).

Anti-heavy chain serum (αHC) was made by injecting rabbits with the luminal domains (amino acids 1-275) of both HLA-A2 and HLA-B27 that had been expressed in bacteria (kindly provided by D. Garboczi, Harvard University). This serum recognizes free MHC class I heavy chains, as did a similar antibody which was described previously (Neefjes et al. 1986). Antibodies were raised against purified bovine ubiquitin (Baboshina and Haas 1996) exactly as described (Haas and Bright 1985). The quality of the αUb serum that was obtained was evaluated in Western blots by its ability to bind Cdc34p-linked multi-ubiquitin conjugates (Banerjee et al. 1993; kindly provided by Seth Sadis, Harvard Medical School). US11 antiserum was raised in rabbits against a mixture of three US11 peptides (MPELSLTLFDEPPPLVETE, ESLVAKRYWLRDYRVPQRT, and FWGLYVKGWLHRHFPWMF) that were coupled to keyhole limpet hemocyanin. The monoclonal antibody 12CA5, which recognizes the influenza hemagglutinin (HA) epitope, was purified from tissue culture cell supernatants by standard methods (Harlow and Lane 1988). The monoclonal antibody 66Ig10 was used to immunoprecipitate transferrin receptor (TfR; van de Rijn et al. 1983).

Immunoprecipitated heavy chains were eluted from Staph A by heating to 95°C in 2% SDS and then diluted into Con A precipitation buffer (final conditions: 0.5 M NaCl, 50 mM Tris, pH 7.5, 1 mM CaCl₂, 1 mM MgCl₂, 1 mM MnCl₂, 0.5% Igepal, 0.1% SDS, and 0.1% BSA). Con A–Sepharose (Pharmacia Biotech) was added and samples were mixed gently for 15 h at 4°C. The precipitates were washed four times in Con A precipitation buffer before being analyzed by SDS PAGE.

One-dimensional isoelectric focusing (IEF), and fluorography were carried out as described previously (Ploegh 1995). Quantitation of proteins labeled with ³⁵S and ¹²⁵I was done with a Fuji PhosphorImager BAS1000. Densitometry on gels of ³H-mannose–labeled proteins was done with a BioRad Fluor-S MultiImager.

HLA-A*0201 heavy chain (HLA-A2), with the mouse H-2 K^b signal sequence replacing its own (Story et al. 1999), was cloned into the pcDNA3.1 expression vector (Invitrogen Corp.). An HA epitope tag (Surdej and Jacobs-Lorena 1994) was inserted in the α2 domain at amino acid 127, where amino acid 1 is the first amino acid of the protein after the signal sequence is removed. For the K→R HA/A2 construct, site-directed mutagenesis and standard cloning techniques were used to change the three lysine residues in the cytosolic tail (positions 311, 316, and 340) to arginines. The HA/A2 constructs were transfected into US11 and control astrocytoma cells using Superfect (Qiagen). Cell lines stably expressing wild-type (wt) HA/A2 or K→R HA/A2 astrocytomas were selected with G418-sulfate (GIBCO BRL).

Bovine ubiquitin (Sigma Chemical Co.) was iodinated using chloramine T (Ciechanover et al. 1980). Each 310-μl iodination reaction contained 2.6 mg/ml ubiquitin, 3.5 mCi/ml Na¹²⁵I (NEN Life Science Products), and 0.33 mg/ml chloramineT in 0.32 M potassium phosphate, pH 7.6. Iodination was allowed to proceed for 1 min and was stopped by addition of sodium metabisulfite and nonradioactive NaI. Iodinated ubiquitin was separated from unincorporated Na¹²⁵I by running Sephadex G-25 (Pharmacia Biotech) spin columns that had been equilibrated with PB. The final ubiquitin concentration was estimated to be 200 μM, labeled to 10,000–15,000 cpm/pmol. ¹²⁵I-ubiquitin was added to US11 cells in permeabilization reactions at a final concentration of 60 μM (see Fig. 6 A) or 15 μM (see Fig. 9 A).

To begin to dissect the mechanism of US11-dependent dislocation and degradation of MHC class I heavy chains, a permeabilized cell system was developed. We devised permeabilization conditions, using the mild detergent digitonin, such that cells appear microscopically intact, but are permeable to Trypan blue and ATP as well as to proteins such as hexokinase, trypsin, and lactate dehydrogenase (see below and Materials and Methods). In a typical experiment, human astrocytoma cells expressing US11 or control astrocytoma cells were pulse-labeled briefly at 37°C with [³⁵S]methionine to load the endoplasmic reticulum (ER) with radioactive MHC class I heavy chains. Labeled cells were permeabilized by incubation in a buffer containing a low concentration of digitonin and an ATP-regenerating system, and returned to 37°C for a chase period. Lysates were made from samples taken at various timepoints and class I heavy chains were recovered by immunoprecipitation with rabbit anti-heavy chain serum (αHC).

As in intact cells (Fig. 1 A, lanes 1–3 and 7–9), in the permeabilized cell system, heavy chains were degraded only when US11 was present (Fig. 1 B compare lanes 1–6 with lanes 7–12). The half-life of MHC class I heavy chain in permeabilized US11 cells was ∼10 min (Fig. 1 C), somewhat longer than its 2–3-min half-life in intact US11 cells (Wiertz et al. 1996a). In the presence of the proteasome inhibitors ZL₃VS (Fig. 1 D) or lactacystin (data not shown), degradation was largely prevented and a lower molecular mass heavy chain species accumulated in US11 cells. This species is endoglycosidase H (Endo H) resistant (data not shown) and its molecular mass by SDS-PAGE corresponds to that of the deglycosylated heavy chain species that accumulates in the cytosol of intact US11 cells treated with proteasome inhibitors (Fig. 1 A lanes 4–6). As was observed in intact cells, the change in molecular mass of the heavy chain in the permeabilized cells is accompanied by a change in isoelectric point upon hydrolysis of the glycoamide bond. This is best seen by comparison with bacterial N-glycanase (PNGase F) treated samples (Fig. 2 compare lane 3 with lanes 2 and 6). Thus, deglycosylated heavy chains also accumulate in permeabilized US11 cells in the presence of proteasome inhibitor.

As in intact cells, the deglycosylated heavy chains in the permeabilized cells were soluble and cytosolic (Fig. 3 A). The glycosylated heavy chains from both permeabilized US11 and control cells fractionated mostly with the particulate fractions, as did the control membrane proteins transferrin receptor (TfR) and US11. The light chain β₂m, a soluble secretory protein, fractionated with membrane pellets in permeabilized control cells, as expected from its tight association with the MHC class I heavy chain. In US11 cells, because there is little heavy chain, most of the β₂m is soluble in the ER lumen (Wiertz et al. 1996a). We found β₂m in both the 10-K pellet and the 100-K supernatant fractions in permeabilized US11 cells (Fig. 3 A), indicating that a portion of the ER content was released during homogenization.

Because the mechanical homogenization was a rather harsh procedure, resulting in some disruption of vesicles, a squeeze-out fractionation technique was also applied. At each timepoint after permeabilization, samples were taken, subjected to centrifugation in a microfuge, and separated into pellet and supernatant fractions. Under these conditions, β₂m fractionated identically in US11 and control cells, >90% of the β₂m was found in the pellet fractions, confirming that little disruption of vesicles occurred. US11 was found only in pellet fractions (data not shown). The deglycosylated heavy chain was found mostly in the squeezed-out, soluble fractions (Fig. 3 B, lanes 6 and 9) whereas the glycosylated heavy chain was found in the membrane pellet fractions.

The cytosolic localization of the deglycosylated heavy chain in permeabilized cells was confirmed by proteolysis protection experiments. Samples of permeabilized cells were treated with increasing amounts of trypsin. Trypsin at 100 μg/ml degraded nearly all of the deglycosylated heavy chain present, while the glycosylated, membrane-bound heavy chain was largely protected (Fig. 4 A). Further experiments showed that the faster-migrating heavy chain species, seen at 100 μg/ml and 200 μg/ml trypsin, corresponds to glycosylated heavy chain lacking the 30–amino acid cytosolic tail. It was not immunoprecipitable with antibodies specific for the heavy chain cytosolic tail (data not shown) and it was sensitive to treatment with endo H (Fig. 4 B). Thus, deglycosylated heavy chain in permeabilized US11 cells accumulates in the cytosol and not in a membrane-bound compartment.

In intact astrocytomas expressing US11 or US2, accumulation of deglycosylated heavy chain requires ATP (Wiertz et al. 1996b; Tortorella, D., unpublished observation). To test whether this also applies to the permeabilized cell system, we depleted ATP from permeabilization reactions carried out in the presence of proteasome inhibitor. Simply omitting the ATP-regenerating system from the permeabilization reactions significantly reduced the amount of deglycosylated heavy chain that appeared (Fig. 5, lanes 5–8). When remaining ATP was depleted by the addition of hexokinase and glucose, no deglycosylated heavy chain was detectable (Fig. 5, lanes 9–12). Moreover, ATP could not be substituted with the nonhydrolyzable ATP analogue AMPPNP (Fig. 5, lanes 13–16). Thus, as in intact cells, the US11-dependent accumulation of deglycosylated heavy chain in permeabilized cells requires ATP. Taken together, these data demonstrate that the permeabilized cell system faithfully reproduces the US11-dependent degradation of MHC class I heavy chain seen in intact cells.

Although ubiquitin conjugates of ER proteins degraded in the course of quality control have been detected in both mammalian cells and in yeast (Ward et al. 1995; Hiller et al. 1996; Hampton and Bhakta 1997; Mayer et al. 1998), ubiquitinated MHC class I heavy chains in cells infected with HCMV or in cells expressing US11 or US2 have not been reported. We used a direct and sensitive method to test for the presence of ubiquitinated heavy chain intermediates in permeabilized US11 cells. Instead of metabolically labeling cells before permeabilization, we permeabilized the cells in the presence of ¹²⁵I-ubiquitin. After a 10-min incubation at 37°C, lysates were made and heavy chains were isolated by immunoprecipitation. Iodinated, high molecular mass heavy chain species accumulated in permeabilized US11 cells, but only when proteasome inhibitor was present (Fig. 6 A, lanes 2 and 3). Much less of this material was seen in control cells, despite the presence of proteasome inhibitors (lane 5). The high molecular mass of the iodinated products is consistent with their being polyubiquitin-conjugated MHC class I heavy chains. They are not coimmunoprecipitating ubiquitinated proteins, because they are also immunoprecipitable with αHC antibody from permeabilized cells lysed under denaturing conditions (data not shown). Nor are they conjugates formed after lysis, because the appearance of ¹²⁵I-ubiquitin heavy chains requires incubation with permeabilized cells (lane 4).

To establish in intact cells that similar ubiquitin conjugates occur, anti-ubiquitin antibodies were raised and used in immunoprecipitation experiments. US11 and control cells were pulse-labeled with [³⁵S]methionine for 10 min and chased at 37°C in the absence or presence of proteasome inhibitor. MHC class I heavy chains were immunoprecipitated and the material bound to the αHC antibodies was dissociated and denatured with SDS. One aliquot was analyzed directly (Fig. 6 B, lanes 1–16), and the rest was diluted into immunoprecipitation buffer and subjected to a second round of immunoprecipitation with anti-ubiquitin serum (αUb). High molecular mass, ubiquitin-conjugated heavy chains were detected in US11 cells but not in control cells (Fig. 6 B, lanes 17–24 vs. lanes 25–32). Some nonubiquitinated heavy chains were also recovered, probably due to nonspecific interactions (lanes 17, 21, 25, and 29). Thus, MHC class I heavy chain is ubiquitinated in intact US11 cells just as it is in permeabilized cells.

In the absence of proteasome inhibitor, a small amount of heavy chain was found ubiquitinated immediately after labeling (Fig. 6 B, lane 18), but subsequently disappeared during the chase period, presumably because it was degraded. In the presence of proteasome inhibitor, the amount of ubiquitinated heavy chain observed remained essentially constant throughout the chase period (lanes 22–24). This probably reflects a balance in the action of ubiquitinating and deubiquitinating enzymes in the absence of appreciable degradation of the heavy chain.

The majority of heavy chain in US11 cells is not ubiquitinated. Given the short labeling time in these experiments and the stability of ubiquitin in cells, we believe that most of the radioactivity in the ubiquitinated heavy chains is incorporated into the heavy chains themselves and not the ubiquitin molecules. With this assumption, we estimate that, at most, 10–20% of the MHC class I heavy chains are ubiquitinated in intact US11 cells treated with proteasome inhibitor. In the absence of proteasome inhibitor, at most 5% of the heavy chains were ubiquitinated.

Consistent with the observation that dislocation is slower in permeabilized cells than in intact cells, we found that accumulation of ubiquitinated heavy chain is also slower under those conditions (Fig. 7 A). Interestingly, both deglycosylated heavy chains and ubiquitinated heavy chains accumulate at approximately the same rate (Fig. 7 B), suggesting that a common, preceding step is rate-limiting.

Having demonstrated the occurrence of ubiquitinated heavy chains, we wished to know whether ubiquitination precedes deglycosylation or vice versa. Two independent approaches were taken to determine the fraction of ubiquitinated heavy chains that are glycosylated. First, intact US11 cells were treated with proteasome inhibitors and labeled with ³H-mannose. Very little ³H-mannose-labeled, ubiquitinated heavy chain was detected. On very long (2 mo) exposures of gels from three independent experiments, the amount of ³H-mannose–labeled heavy chain precipitated with αUb serum was barely above background levels (data not shown). Furthermore, in αHC immunoprecipitations, ³H-mannose–labeled, nonubiquitinated heavy chains were efficiently recovered, but very little mannose label was detectable in the area of the high molecular mass, ubiquitinated HC species (Fig. 8 A, lanes 3 and 4). In contrast, such species were readily detectable in ³⁵S-labeled cells (Fig. 8 A, lanes 1 and 2). Ubiquitinated glycosylated heavy chains account, at most, for only 1–2% of all glycosylated heavy chains in these experiments. Thus, it appears that the majority of the ubiquitinated heavy chains in US11 cells are deglycosylated.

This finding is supported by a second set of experiments, in which glycosylated heavy chains were isolated by their ability to bind the lectin Con A. ³⁵S-labeled ubiquitinated heavy chains were isolated by immunoprecipitation with αHC serum, followed by reimmunoprecipitation with αUb serum. Heavy chains were eluted from the αUb antibodies with SDS and each sample was divided in two. Half was analyzed directly by SDS-PAGE and the other half was precipitated with Con A–Sepharose. Only a very small amount of ubiquitinated HC precipitated with con A (Fig. 8 B, compare lanes 1 and 2 with lanes 3 and 4). The precipitation was specific for glycosylated material because, in control experiments carried out in parallel, very little deglycosylated, nonubiquitinated HC was precipitated with Con A (Fig. 8 C, lanes 1 and 2). No heavy chain was recovered in samples incubated with Con A in the presence of the competitor methyl α-d-mannopyranoside (Fig. 8 B, lanes 5 and 6, Fig. 8 C, lane 3) or in samples incubated only with Sepharose beads (data not shown). We calculated that only ∼10% of the ubiquitinated heavy chains bound to Con A. Taken together, these data suggest that either deglycosylation precedes ubiquitination or that ubiquitinated heavy chains are rapidly deglycosylated.

Does ubiquitination occur while heavy chains are still associated with membranes? We first examined permeabilized US11 cells that had been treated with proteasome inhibitors because ubiquitinated heavy chains are most abundant under those conditions. In fractionation experiments, after mechanical homogenization of permeabilized cells that had been incubated in the presence of ¹²⁵I-ubiquitin, 50% of the ubiquitinated heavy chain was found in the 100-K supernatant and 30% in the 100-K pellet (Fig. 9 A). In squeeze-out fractionation experiments on permeabilized US11 cells labeled with ¹²⁵I-ubiquitin in the same way, ∼75% of ¹²⁵I-ubiquitinated heavy chains fractionated with the cytosol (data not shown). A similar result was obtained in ³⁵S-labeled permeabilized US11 cells, where ∼80% of ubiquitinated heavy chains fractionated with the cytosol after 20 min of chase (Fig. 9 B, lanes 10–12). Thus, most of the ubiquitinated heavy chains in US11 cells treated with proteasome inhibitors are released from the ER, although a small percentage seems to be membrane-associated.

We next examined the localization of ubiquitinated heavy chain in cells not treated with proteasome inhibitors. To follow the small amount of ubiquitinated heavy chain that accumulates in intact cells under these conditions, we permeabilized the cells after the chase period and fractionated their contents using the squeeze-out technique (see Materials and Methods). We found that almost all of the ubiquitinated heavy chain fractionated with the membrane pellets in US11 cells not treated with proteasome inhibitor (Fig. 9 C, lanes 5–7). This result was not simply due to the failure to achieve permeabilization because, under the same conditions, proteasomes and other cytosolic proteins can be squeezed out of these cells (data not shown). Moreover, in an experiment carried out in parallel using cells treated with proteasome inhibitor, deglycosylated heavy chain (Fig. 9 D, lanes 2 and 5) and ubiquitinated heavy chains (lanes 8 and 11) were squeezed into the cytosolic supernatant. Note that in the presence of proteasome inhibitor, the amount of soluble, ubiquitinated heavy chain increases with time (Fig. 9 D, compare lanes 8 and 9 with lanes 11 and 12). Thus, these results are consistent with those shown in Fig. 9A and Fig. B. They suggest that, both in the presence and absence of proteasome inhibitor, heavy chain is ubiquitinated while it is still associated with the ER membrane. Over time, in the absence of proteasome inhibitor, the ubiquitinated heavy chain is degraded, while, in the presence of proteasome inhibitor, it is released into the cytosol.

With heavy chain ubiquitination occurring at the ER membrane, it seemed possible that ubiquitination might be the initiating signal for heavy chain dislocation. To determine whether ubiquitination of the heavy chain is required for its dislocation, we prevented ubiquitination of the heavy chain cytosolic tail by removing all lysines from the tail domain. Two different HA epitope-tagged heavy chain constructs, based on the HLA allele A2, were made, one with a wild-type (wt) cytosolic tail and one with the three cytosolic tail lysines mutated to arginine (K→R). Cell lines stably expressing the HA-tagged heavy chains (HA/A2) were selected. We found that K→R HA/A2 is degraded at approximately the same rate as wt HA/A2 in US11 cell lines, while both HA/A2 proteins are stable in control cells (Fig. 10A and Fig. B, and data not shown). When pulse–chase experiments were conducted in the presence of proteasome inhibitor, deglycosylated HA-tagged wt and K→R heavy chain intermediates accumulated at roughly the same rate (Fig. 10 C). In all cases, the HA/A2 heavy chains were dislocated and degraded more slowly than endogenous heavy chain in cells not expressing HA/A2 constructs. We attribute this to the overexpression of the HA/A2 degradation substrates, which may saturate the degradation machinery (Story et al. 1999). These results were reproducible in multiple, independently derived cell lines. Ubiquitination of the heavy chain cytosolic tail, therefore, is not required to initiate US11-dependent destruction of the protein.

Next, we asked whether the K→R heavy chain is ubiquitinated. We isolated ubiquitinated HA/A2 from [³⁵S]methionine-labeled cell lines by sequential immunoprecipitation with αHA and αUb antibodies. Ubiquitin-conjugated K→R HA/A2 was detected in US11 cells both in the absence (Fig. 11 A, lane 13) or presence (Fig. 11 B, lanes 13 and 16) of proteasome inhibitor. In fact, K→R HA/A2 was ubiquitinated to approximately the same extent as wt HA/A2. The finding that dislocated heavy chains are not ubiquitinated exclusively on residues in the cytosolic tail implies that dislocation must start before ubiquitination of the heavy chain.

In squeeze-out fractionation experiments, the majority of ubiquitinated K→R HA/A2 and wt HA/A2 fractionated with membrane pellets in the absence of proteasome inhibitors (Fig. 11 A, lanes 13–21). In the presence of proteasome inhibitors, slightly more than half of the ubiquitinated HA/A2 in each cell line fractionated with the cytosol (Fig. 11 B, lanes 13–24). These results were confirmed in experiments in which the cell lines expressing the two different HA/A2 alleles were homogenized mechanically and fractionated by differential centrifugation. Both ubiquitinated HA/A2 species showed identical fractionation behavior. In the absence of proteasome inhibitors, they were clearly membrane-associated, with most found in the 1-K pellet (Fig. 12 A). In the presence of proteasome inhibitors, the majority of the ubiquitinated HA/A2 was in the 100-K pellet and supernatant (Fig. 12 B). Although we have not characterized the heavy chain in the 100-K pellet fraction further, we have found proteasome subunits in this fraction as well (data not shown), raising the possibility that some ubiquitinated heavy chain is tightly associated with proteasomes in the presence of proteasome inhibitor. Taken together, these data support the idea that the ER luminal domain of heavy chain is ubiquitinated while the protein is still associated with the membrane.

We have developed a permeabilized cell system that recapitulates many of the important aspects of US11-dependent dislocation and degradation of MHC class I heavy chains that have been observed in intact cells. In the presence of proteasome inhibitors, deglycosylated heavy chain accumulates in the cytosolic fraction of permeabilized cells. As in intact cells, appearance of this deglycosylated species is ATP-dependent.

We have used this permeabilized cell system, in combination with experiments carried out in intact cells, to identify and order intermediates in US11-dependent dislocation. In so doing, we have provided strong evidence that US11-dependent degradation of class I heavy chain follows a sequence of events similar to that described for other proteins that exit the ER for degradation in the cytosol. First, using two independent methods, we have shown that ubiquitinated MHC class I heavy chains accumulate in US11 cells. Experiments involving the addition of ¹²⁵I-ubiquitin to permeabilized cells followed by immunoprecipitation with αHC serum provide direct evidence that heavy chain is ubiquitinated. Double immunoprecipitation experiments on lysates of ³⁵S-labeled cells, using antibodies against MHC class I heavy chain and against ubiquitin, confirmed the identity of the ubiquitinated heavy chain. Importantly, ubiquitinated heavy chains were detected in US11 cells regardless of whether they had been treated with proteasome inhibitors, indicating that they are not an artifact of proteasome inhibitor addition.

The majority of the ubiquitinated heavy chain that accumulates in US11 cells treated with proteasome inhibitor is not glycosylated, suggesting that deglycosylation may precede ubiquitin conjugation. However, the presence of a small population of glycosylated, ubiquitinated heavy chains and the observation that, at least in permeabilized US11 cells, ubiquitinated heavy chains seem to accumulate simultaneously with deglycosylated heavy chains (Fig. 7), suggest that there might be no obligatory order to the deglycosylation and ubiquitination steps.

We addressed the role of heavy chain ubiquitination in US11-dependent export from the ER by mutating all of the lysine residues in the heavy chain cytosolic tail to arginine, creating a K→R heavy chain mutant that lacks potential ubiquitination sites in its cytosolic domain. This mutant is still dislocated and degraded at approximately the same rate as wt heavy chain. It is also ubiquitinated. Thus, ubiquitination of the cytosolic tail of MHC class I heavy chain is not required to initiate heavy chain dislocation and degradation.

Our data lead to the following model for US11-dependent degradation of MHC class I heavy chain (Fig. 13). The first step is dislocation, in which all or part of the luminal domain of heavy chain enters the cytosol. We do not know the source of the force that pulls or pushes the heavy chain from the ER, nor do we know which segment of the protein exits the ER first. However, once the luminal domain of the heavy chain has been dislocated, the protein seems to undergo ubiquitination while still associated with the ER membrane. We find ubiquitinated heavy chains, including the K→R heavy chain mutants, in cell membrane fractions. Thus, at this stage, the ubiquitinated heavy chains are either completely dislocated from the ER but still tightly associated with the membrane on the cytosolic side (Fig. 13, arrow I), or they are partially dislocated from the ER, with part of the protein integrated in the ER membrane or held in the dislocation channel. Partially dislocated heavy chain could be in two different orientations: with a portion of the luminal domain situated in the ER membrane (Fig. 13, arrow II), or, more likely for energetic reasons, with the transmembrane domain in the ER membrane and with both NH₂ and COOH termini in the cytosol (Fig. 13, arrow III). ER membrane-associated ubiquitin-conjugating enzymes have been identified in yeast and they are required for ER-associated degradation (Sommer and Jentsch 1993; Biederer, 1996; Hiller et al. 1996; Biederer et al. 1997). Our results suggest that similar enzymes may play a role in US11-dependent heavy chain degradation.

Ultimately, the heavy chain is degraded. It is not clear how the heavy chain is brought into contact with the proteasome, nor where degradation takes place. If we assume that the proteasome inhibitor simply causes a backup of intermediates of the normal degradation pathway, then both soluble deglycosylated and soluble ubiquitinated heavy chains must also be present in the absence of inhibitor, but as very short-lived species. Thus, degradation would occur on cytosolic proteasomes. Alternatively, it is possible that the proteasome inhibitor actually alters the degradation pathway and that the soluble heavy chain species are an artifactual consequence of proteasome inhibition. If this is the case, then heavy chain degradation would occur at the ER membrane, carried out by a population of proteasomes that localizes there (Rivett 1998).

The work presented here suggests that US11 accelerates a process that normally occurs with misfolded proteins in uninfected cells. By inference, US2 probably operates similarly. In fact, MHC class I heavy chains are also ubiquitinated in cells expressing US2 (Shamu, C.E., unpublished observation). Thus, understanding the exact mechanism by which US11 and US2 induce rapid MHC class I heavy chain degradation will be important not only for understanding viral immune evasion, but also for understanding ER quality control. The speed and specificity with which heavy chain is degraded in the presence of US11 will be particularly useful for further characterization of the dislocation pathway using our permeabilized system. It should be possible to dissect the cytosolic requirements of the pathway and to carry out more detailed mechanistic studies.

We thank Domenic Tortorella, Seth Sadis, Michael Glickman, Margo Furman, Johannes Huppa, Alexei Kisselev, and Maurits Kleijnen for reagents, protocols, and helpful discussions. We thank Dan Finley, Melissa Rolls, and Peter Sorger for comments on the manuscript.

This work was supported by the National Institutes of Health (grant 5R37-A133456 to H.L. Ploegh) and Boehringer Ingelheim Inc. (Ridgefield, CT). C.E. Shamu was supported by the Jane Coffin Childs Memorial Fund for Medical Research and by a Special Fellowship from the Leukemia Society of America. C.M. Story was supported by the Cancer Research Institute of New York. T.A. Rapoport is an Investigator of the Howard Hughes Medical Institute.