SEL1L nucleates a protein complex required for dislocation of misfolded glycoproteins

Abstract

Membrane and secretory proteins that fail to pass quality control in the endoplasmic reticulum are discharged into the cytosol and degraded by the proteasome. Many of the mammalian components involved in this process remain to be identified. We performed a biochemical search for proteins that interact with SEL1L, a protein that is part of the mammalian HRD1 ligase complex and involved in substrate recognition. SEL1L is crucial for dislocation of Class I major histocompatibility complex heavy chains by the human cytomegalovirus US11 protein. We identified AUP1, UBXD8, UBC6e, and OS9 as functionally important components of this degradation complex in mammalian cells, as confirmed by mutagenesis and dominant negative versions of these proteins.

Terminally misfolded membrane or secretory proteins that have entered the endoplasmic reticulum (ER) are typically transported back across the ER membrane into the cytosol, a process referred to as dislocation or retrotranslocation. Once in the cytosol, the proteasome degrades these misfolded proteins in a ubiquitin-dependent manner (1).

Although some components of the mammalian dislocation machinery show some sequence similarities to yeast proteins, their contribution to dislocation is not always clear. The presence of several mammalian orthologues for each yeast component of the dislocation machinery precludes a functional identification of the relevant mammalian components by homology.

Analysis of two viral proteins encoded by human cytomegalovirus, US2 and US11, has helped to define the composition of the dislocation protein complexes (2). Both US2 and US11 facilitate dislocation of newly synthesized Class I major histocompatibility complex (MHC) heavy chains (HCs), presumably to evade recognition by cytotoxic T cells (3). US2 and US11 are ER-resident type I transmembrane proteins that interact with Class I MHC HCs in the ER lumen and, from there, initiate their destruction (4, 5). US2 and US11 achieve this by recruiting different sets of proteins: US2 uses signal peptide peptidase (6) and other as yet unknown proteins, whereas US11 engages a pathway that includes Derlin-1 (7). Derlin-1 itself associates with the ubiquitin ligase HRD1 and gp78, both of which share sequence similarities with yeast Hrd1p (8). Whether HRD1 and gp78 are involved in the ubiquitination of Class I MHC HCs is an open question (9), although the human homologue of yeast Hrd3p, SEL1L, is involved in Class I MHC HC dislocation (10). Derlin-1, HRD1, and the transmembrane protein VIMP form a complex with p97 and its cofactors UFD1/NPL4 and might be involved in their recruitment to the ER (7, 8, 11).

How a luminal protein can cross the lipid bilayer is not known, and the existence of a proteinaceous pore, consisting of Hrd1p and/or Der1p has been suggested (7, 8, 11, 12), but alternative modes of extraction might exist (13). US11 hijacks a pathway that contributes to the degradation of aberrantly folded proteins independent of viral accessories, as shown by the degradative fate of α1-antitrypsin null Hong Kong, truncated ribophorin RI₃₃₂, and misfolded cystic fibrosis transmembrane conductance regulator ΔF508 (10, 14, 15). Here, we identified new components of the mammalian dislocation machinery essential for degradation, including the E2 ligase that cooperates with HRD1/SEL1L and two ER proteins that act downstream of the substrate selection process, and verified a recently reported SEl1L-interacting ER luminal protein important for substrate recognition, OS9 (16).

Results

Isolation and Identification of Proteins That Interact with SEL1L.

We conducted a large-scale immunopurification of SEL1L, using HA-tobacco etch virus (TEV)–tagged SEL1L transduced into HeLa cells [supporting information (SI) Fig. S1]. The HA-TEV tag was fused to the N terminus of SEL1L preceded by the signal sequence of H2-K^b. SEL1L was isolated by immunoprecipitation from digitonin extracts. Materials eluted with TEV protease were subjected to SDS/PAGE, and SEL1L-interacting polypeptides were identified by liquid chromatography tandem mass spectrometry (LC/MS/MS, Fig. S1). We recovered several proteins already known to be SEL1L interactors: HRD1, a ubiquitin E3-ligase involved in ER dislocation (8, 9); Derlin-2, a multispanning transmembrane protein required for exit of polyomavirus from the ER (17); the ATPase p97; and several other proteins involved in protein folding, such as PDI, BiP, and calnexin (Fig. S1b).

We identified two additional proteins not previously known to be part of the mammalian dislocation machinery: ancient ubiquitous protein 1 (AUP1), and UBXD8 (Fig. S1 a and b). We also identified UBC6e, an enzyme that serves as a ubiquitin-conjugating enzyme (E2) (18), and OS9, a protein involved in the degradation of mutant α1-antitrypsin (16). We propose that UBC6e is the E2-type activity that acts in concert with the ubiquitin ligase HRD1.

OS9 was identified as a protein amplified in osteosarcoma. OS9 is ubiquitously expressed and has alternative splice versions (19). The C terminus of OS9 interacts with HIF1α, a subunit of the protein hypoxia-inducible factor (HIF) 1. HIF1α is ubiquitinated and degraded, depending on O₂ levels in the cell. OS9 regulates HIF1α levels by increasing the rate of prolyl hydroxylation in HIF1α, thereby initiating ubiquitination (20).

The presence of an N-terminal signal sequence implies that OS9 is targeted to the ER lumen. OS9 has a glucosidase type II domain involved in binding to misfolded proteins. OS9, an ER-resident glycosylated protein, is part of the mammalian dislocation machinery through its interactions with SEL1L (16) (Fig. S1a). OS9 has been located to the cytosol (20). There might be a pool of OS-9 that is active in the cytosol or HIF1α might be regulated indirectly from within the ER. The yeast homologue Yos9p is a luminal ER protein that binds to the luminal domain of Hrd3p, the homologue of SEL1L. Yeast Yos9p targets terminally misfolded ER proteins to the dislocation machinery, which includes Hrd3p and Der1p (12, 21, 22).

Ancient ubiquitous protein 1 is proposed to interact with integrins (23), but its function is obscure. AUP1 has a CUE domain, involved in ubiquitin binding or in recruitment of ubiquitin-conjugating enzymes to the site of dislocation (24). AUP1 has an N-terminal membrane anchor, with the remainder predicted to be in the cytosol (23). AUP1 has not been implicated in any aspect of (glyco) protein quality control and lacks an obvious homologue in yeast.

UBXD8 (ETEA) is among the set of proteins up-regulated in T cells from atopic dermatitis patients (25). UBXD8 has a UBX domain, a UBA domain, a UAS domain, and a transmembrane domain (Fig. S1c). The UBA domain is found in many proteins of otherwise divergent structure and function and mediates binding to ubiquitin. The UBX domain is structurally similar to ubiquitin despite the lack of sequence homology. UBX domains may serve as adaptors for the multifunctional AAA ATPase p97 (26). The UAS domain is a domain of >100 aa of unknown function, which assumes a thioredoxin-type fold. The closest relative of UBXD8 in yeast cannot immediately be inferred because of the limited extent of sequence identity and lack of functional data.

UBC6e (UBE2J1) is an ortholog of yeast Ubc6p, an ER-anchored E2. In yeast, Ubc6p can function together with the ubiquitin ligase Doa10p and the cytosolic E2 Ubc7p (27, 28). There are two Ubc6p orthologs in mammalian cells, UBE2J1 and UBE2J2. UBE2J1 was termed UBC6e, and UBE2J2 is called UBC6 (18). UBC6e and UBC6 are both involved in the degradation of T cell receptor-α and CFTRΔF508 (14, 18). UBC6e forms a complex with Derlin-1 for CFTRΔF508 disposal (14). UBC6e displays less sequence identity (25%) to the yeast protein than does UBC6 (40%) (18). Unlike yeast Ubc6p, human UBC6e is a stable protein (29).

ER Localization of the SEL1L-Interacting Proteins.

Immunofluorescence with affinity purified anti-AUP1 and anti-UBXD8 antibodies shows the diagnostic reticular ER staining pattern and colocalization with the ER marker PDI (Fig. 1). Endogenous AUP1 and UBXD8 proteins thus reside in the ER, where dislocation occurs. UBXD8, AUP1, and UBC6e all readily cosediment with the microsomes in the absence of detergent and are largely resistant to extraction with alkaline sodium carbonate and urea (Fig. S2a). OS9, consistent with its predicted characterization as a soluble ER luminal protein, is readily extracted from the microsomes by alkaline sodium carbonate. OS9 is present as two splice variants (19), sensitive to digestion with endoglycosidase H (Fig. S2b), consistent with ER residency and the presence of the single predicted N-linked glycan. We verified the interaction of endogenous OS9, UBC6e, and AUP1 with SEL1L by immunoprecipitation followed by immunoblotting (Fig. S3).

Fig. 1.

AUP1 and UBXD8 are found in the ER. HeLa cells were fixed and incubated with anti-PDI antibody and affinity-purified antibodies against AUP1 (Upper) or UBXD8 (Lower) for immunofluorescence analysis. (Right) Merged images. (Scale bars: 10 μm.)

OS9 Overexpression Perturbs Dislocation of RI₃₃₂ but Not of Class I MHC via US11.

To examine a possible role for OS9 in dislocation, we designed an N-terminal GFP-tagged version and two mutant versions (R188A; E212D) of OS9 predicted to disrupt the mannose-6-phosphate receptor homology domain (MRH) or glucosidase II domain implicated in OS9-substrate interaction (30). Mutant or tagged versions of OS9 in US11-expressing cells only marginally affect Class I MHC HC dislocation (Fig. S4 and Fig. 4A, lanes 10–12), especially when compared with overexpression of UBC6e, AUP1-GFP, or UBXD8-GFP (Figs. 3 and 4). To explain the comparative dispensability of OS9 in US11-expressing cells, US11 might be directly responsible for substrate recognition instead of OS9 and target Class I MHC HC directly to the dislocation machinery.

Fig. 4.

AUP1 and UBXD8 are crucial for US11-mediated dislocation. (A) US11-expressing cells were transduced with either empty vector (pLHCX), AUP1-GFP, UBXD8-GFP, or GFP-OS9. The experiment was performed as in Fig. 3. (B) Quantitation of the amount of glycosylated Class I MHC HC to total HC counts. (C) US2-expressing cells were transduced with the same virus preparation used in a and a pulse–chase experiment was performed as for a. (D) Quantitation of the amount of glycosylated Class I MHC HC to total HC counts. (E) GFP-tag on UBXD8 hinders recruitment of p97. UBXD8-WT and UBXD8-GFP were expressed in 293T cells and immunoprecipitated with anti-UBXD8 antibodies from digitonin extracts. The eluates were separated on a 10% SDS/PAGE and immunoblotted with anti-p97 antibodies.

Fig. 3.

UBC6e is crucial for US11-mediated dislocation. (A) US11-expressing cells were transduced with either empty vector (pLHCX), UBC6e WT, or UBC6e C91S mutant. The three cell lines were treated with ZL₃VS pulse-labeled for 10 min with ³⁵S, chased for indicated time points, and lysed in 1% SDS; the lysate was then immunoprecipitated with anti-HC. The eluates were separated on 12% SDS/PAGE and visualized on film. (B) Quantitation of the amount of glycosylated HC to total HC counts. (C) US2-expressing cells were transduced with the same virus preparation used in a. The experiment was performed as in A. (D) Quantitation of the amount of glycosylated Class I MHC HC to total HC counts.

Does OS9 play a role in dislocation independent of viral proteins? We generated HeLa cells that overexpress wild-type OS9, GFP-OS9, or the mutant versions OS9 R188A and OS9 E212D. We then transiently transfected these cell lines with a truncated version of ribophorin, RI₃₃₂ (31), a protein dislocated in a SEL1L-dependent manner (10). For all constructs examined, we observe a delay in RI₃₃₂ degradation (Fig. 2). Endogenous ribophorin is stable and electrophoretically distinct, and it serves as a control for recovery. We conclude that OS9 is involved in the dislocation of the soluble glycoprotein RI₃₃₂. Overexpression of wild-type OS9 inhibits dislocation of RI₃₃₂, presumably because it interferes with the stoichiometry of the dislocation complex. We sought to verify our results with knockdown constructs against OS9, but none of the constructs yielded a level of reduction adequate to achieve inhibition of dislocation.

Fig. 2.

OS9 is crucial for dislocation of the terminally misfolded glycoprotein, RI₃₃₂. (A) HeLa cells were transduced with either empty vector (pLHCX), wild-type OS9, OS9 R188A, OS9 E212D, or GFP-OS9, using the same virus preparation as in Fig. S4. The five cell lines were then transfected with a construct that specifies truncated ribophorin, RI₃₃₂, pulse-labeled 36 h after transfection for 20 min with ³⁵S, chased for indicated time points, and then lysed in 1% SDS and immunoprecipitated with antibody raised against the luminal portion of ribophorin. The eluates were separated on 10% SDS/PAGE and visualized by autoradiography. # indicates a faster migrating version of RI₃₃₂. (B) Quantitation of the amount of RI₃₃₂ remaining at the indicated time points.

UBXD8, AUP1, and UBC6e Are Required for US11-Mediated Dislocation.

Because US11 and US2 both target Class I MHC molecules but apparently do so by initially recruiting different proteins, we used US2-mediated dislocation as a control for proper ER function (6, 7, 10). Manipulations that perturb ER function nonspecifically should affect dislocation via both the US2 and US11 pathways. Our criterion is thus to score as specific those manipulations that interfere with US11-mediated dislocation only.

Among the SEL1L-interacting proteins, UBC6e was the only protein known to act as an enzyme (E2) and whose catalytic center could be ascertained (18). We thus destroyed the catalytic activity of UBC6e by replacement of cysteine 91 with serine (29).

We installed a GFP tag onto the C terminus of AUP1 and UBXD8 and onto the N terminus of OS9. We reasoned that the GFP domain may interfere with, but not completely abolish, the function or recruitment capabilities of flanking domains, and thus yield inhibitory effects for the corresponding fusion proteins (7). Because UBXD8 has a UBX domain that might recruit p97 to the site of dislocation, the attachment of a globular GFP-sized domain in close proximity to the C-terminal UBX domain might interfere with this interaction. Similarly, the GFP-tagged version of Derlin-1 inhibits Class I MHC HC degradation in a US11-dependent fashion (7).

We observed strong inhibition of class I MHC HC degradation in US11 cells that overexpress UBC6e C91S, AUP1-GFP, and UBXD8-GFP (Figs. 3 and 4). In pLHCX vector control cells, most Class I MHC HCs have lost their N-linked glycan at the 30-min chase point because of peptide: N-glycanase (PNGase) activity (32). In the presence of proteasome inhibitor (ZL₃VS), the diagnostic deglycosylated dislocation intermediate accumulates (4). The overexpression of catalytically inactive UBC6e (C91S) or wild-type UBC6e strongly delays degradation of Class I MHC HC: >75% of HC remains in the ER (Fig. 3A, lanes 4–9). All three cell types express comparable levels of US11, which displays the typical delayed cleavage of its signal peptide (33) (Fig. S5a2, lanes 1–9). High levels of UBC6e C91S significantly slow the degradation of HCs in the absence of proteasome inhibitor (Fig. S6).

In US2 cells transduced with the same constructs, degradation continues unperturbed. Class I MHC HCs are dislocated at rates similar to those in control cells (pLHCX), compared with cells that overexpress UBC6e or UBC6e C91S (Fig. 3C, lanes 1–9). All three cell lines express similar levels of US2 (Fig. S5b2), with its usual mobility on SDS/PAGE: in addition to ER-inserted glycosylated US2, we detect a faster migrating US2 lacking its N-linked glycan, as US2 is inefficiently translocated into the ER (34). Both the US2 and the US11 cell lines were obtained by viral transduction of UBC6e C91S and WT UBC6e and show equivalent levels of expression of UBC6e (Fig. S5 a1 and b1, lanes 4–9). The ubiquitin-activating enzyme UBC6e is thus involved in Class I MHC HC dislocation in US11 cells but not in US2 cells. Because US2 cells remain capable of proper dislocation, ER function as such is not compromised.

We then examined the fate of Class I MHC HC when expressing the GFP-tagged versions of AUP1, UBXD8, and OS9 (Fig. 4). Cells that express AUP1-GFP showed inhibition of dislocation: 50% of Class I MHC HC remains in the ER after 30 min of chase (Fig. 4A, lanes 4–6 and Fig. 4B). In cells that express UBXD8-GFP, >75% of HCs fail to reach the cytosol (Fig. 4A, lanes 7–9 and Fig. 4B, compared with lanes 1–3). As mentioned above, GFP-OS9 did not significantly inhibit dislocation of Class I MHC HC. Again, US2 cells served as a control (Fig. 4 C and D). US2-dependent dislocation proceeded unperturbed in AUP1-GFP, UBXD8-GFP, and OS9-GFP cells (Fig. 4 C and D and Fig. S7).

We compared the ability of UBXD8 and UBXD8-GFP to recruit p97 into the dislocation complex. To this end, we overexpressed UBXD8 and UBXD8-GFP to the same levels in 293T cells and performed an immunoprecipitation with anti-UBXD8 antibodies from digitonin lysates. The recovered material was then analyzed by immunoblotting with anti-p97 antibodies (Fig. 4E). The amount of p97 recovered in immunoprecipitates from cells expressing UBXD8-GFP is much reduced compared with cells expressing wild-type UBXD8. We conclude that the GFP tag hinders recruitment of p97 to the ER membrane, and therefore impedes dislocation. The residual p97 recovered is attributable to the endogenous UBXD8 present in the cells.

Dominant Negative Constructs of UBC6e, AUP1, and UBXD8 Retain Class I MHC HC in the ER.

We used the monoclonal antibody W6/32, which recognizes only correctly assembled Class I MHC molecules in their fully native conformation (35), to explore whether inhibition of dislocation is accompanied by an increase in the amount of correctly folded Class I MHC molecules. We indeed found this to be the case (Fig. S8, cell lines used were those from Figs. 3 and 4) and conclude that the intermediates that accumulate when dislocation is inhibited retain their typical orientation within the ER. In pulse–chase experiments, the W6/32-reactive Class I MHC molecules do not undergo conversion of their high mannose to the complex-type glycans, as inferred by a lack of a shift in mobility assessed by SDS/PAGE. This observation is consistent with the ability of US11 to retain Class I MHC molecules in the ER, also when dislocation is blocked, as observed for the single point mutant in the transmembrane segment of US11 (36). This experiment also demonstrates that UBC6e C91S, UBC6e WT, AUP1-GFP, and UBXD8-GFP do not disrupt dislocation merely by preventing the association of US11 with Class I MHC HC: as with the empty vector control, US11 coimmunoprecipitates with W6/32-reactive Class I MHC HC in all the cell lines constructed (Fig. S8).

Discussion

We have identified three new components of the mammalian dislocation machinery. We used as the point of departure the isolation of SEL1L-interacting partners. We reasoned that via SEL1L, we should recover additional proteins involved in ER dislocation: both ER luminal components that may be involved in substrate recognition and, through its binding partner HRD1, additional cytosolic components that act downstream.

We chose a cell line that did not express US11 to isolate the SEL1L complex to avoid possible bias that might derive from remodeling of the dislocation machinery by US11 itself. The significance of the isolated proteins was verified by returning to our model dislocation cell lines, those expressing US11 or US2. US11 uses a pathway that is superficially similar to the Hrd1p/Hrd3p pathway in yeast (7, 8, 10).

From our analysis of the US11 pathway, the role of OS9 in dislocation is not immediately apparent. We thus turned to the ribophorin fragment RI₃₃₂ to assess a possible contribution of OS9 to dislocation, because RI₃₃₂ is destroyed in a SEL1L-dependent manner. It is imperative to keep in mind the time scale of dislocation of each of the substrates. Because US11-mediated dislocation proceeds rapidly (Class I MHC HC half-life is only 2–5 min), it is more sensitive to minor perturbations than dislocation of other longer-lived substrates such as RI₃₃₂. Because the effect of the OS9 mutants and GFP-OS9 on US11-mediated dislocation is less than that of AUP1-GFP and UBXD8-GFP, we consider the role of OS9 to be comparatively minor (Fig. 4). In contradistinction, the moderate effect of manipulating OS9 level on RI₃₃₂ degradation (Fig. 2) is sufficient to implicate OS9 in the quality control mechanism of RI₃₃₂, as is the case in SEL1L-dependent degradation (10).

Because overexpression of wild-type OS9 inhibits dislocation of RI₃₃₂, excess OS9 likely disrupts the architecture of the complex by titrating away components and rendering the dislocon incapable of efficiently processing substrates. We do not see such a difference in US11 cells that overexpress OS9 to similar levels. Why does a disruptive level of OS9 not affect the performance of the dislocon in US11-expressing cells? We attribute this discrepancy to the fact that US11 itself may stabilize the complex in a manner insensitive to excess OS9. Perhaps the rapidity of US11-mediated dislocation in itself also points to a stabilized dislocon and more efficient recognition inherent in the unique and specific interaction of US11 with its substrate.

Combined, these results are consistent with a model in which US11 serves the specific function of delivering Class I MHC HCs to the HRD1/SEL1L complex and accelerates their removal from the ER and degradation (Fig. 5). In HeLa cells, OS9 is an integral part of this complex, and contributes to substrate recognition. For neither mammalian OS9 nor its yeast homologue, Yos9, is it clear what (sets of) endogenous substrates each of them recognizes. The example of US11 shows that other proteins can assume a substrate recognition function in the context of the larger HRD1/SEL1L complex and deliver substrates to the ligase complex. OS9 is in fact essential for the degradation of mutant α1-antitrypsin (16).

Fig. 5.

Schematic of US11-mediated dislocation. Depicted is the complex that US11 uses to dislocate class I MHC HCs. Domains are italicized. Fn II, fibronectin type II domain; Glc II, glucosidase II domain.

We show that UBC6e is involved in the degradation of Class I MHC HCs in US11 cells. The identification of the responsible E2 has been an important goal, and one possible E2, the E2–25K protein, was uncovered by using a permeabilized cell system (37) to assay for its activity. However, this assay does not allow exchange or removal of membrane-bound molecules; thus, UBC6e, a membrane-anchored E2, would have escaped detection. From the in vivo data in intact cells presented here, we believe that UBC6e is the primary E2 enzyme that catalyzes the ubiquitination of Class I MHC HCs in US11 cells. Other E2s, especially if present in excess, might nonetheless be capable of performing the same reaction.

We also identified two UBX domain-containing proteins, UBXD2 and UBXD8, both of which associate with SEL1L. UBXD2 (Erasin), a mammalian UBX-containing protein linked to dislocation, participates in the degradation of CD3δ (38), but it does so through unknown mechanisms. Could UBXD8 be the possible homolog of Ubx2p, a protein that spans the ER membrane twice and is involved in recruiting p97 to the ER membrane (39)? We see strong inhibition of US11-mediated HC dislocation when overexpressing UBXD8-GFP. However, UBXD8 shares only 17% sequence identity with Ubx2p. Curiously, UBXD8 shares the same level of homology with Ubx3p (another cdc48p cofactor of unknown function). Ubx3p was not reported to be part of the dislocation complex in yeast (21, 22). UBXD8 and Ubx3p share similar organization, as reflected by the order of the distinct domains that are present: both are predicted to have a UAS and a UBX domain C-terminal to a single transmembrane domain. In contrast, Ubx2p has two transmembrane domains and lacks the UAS domain but does have a UBA domain at its N terminus. If UBXD8 were to be inserted as a type I or type II ER transmembrane protein, either the UBA or the UBX domain would reside within the ER lumen. Domains that specify involvement in the ubiquitination pathway are not usually found inside the ER. We see clear ER localization of UBXD8 in immunofluorescence and by sedimentation analysis of microsomes (Fig. 1 and Fig. S2); thus, we propose a similar mechanism of ER insertion as has been shown for Erasin or UBXD2. UBXD8 might be inserted in the ER membrane by dipping into the outer leaflet of the lipid bilayer (Fig. 5) with both tails exposed to the cytosol (38). UBXD8 and UBXD2 might both be involved in recruitment of p97 to the site of dislocation, together or separately, depending on the topology of the substrate. The GFP tag installed on UBXD8 hinders recruitment of p97, which might account for the slowed dislocation (Fig. 4E). We do not know how AUP1 acts as a dominant negative, but it is plausible that the GFP tag here also hinders the recruitment of a downstream, possibly unknown, component of the dislocation machinery.

It is now clear that UBC6e, AUP1, and UBXD8 are required for the exit of a type I ER membrane protein from the ER (Figs. 3–5). UBC6e and AUP1 each have one transmembrane segment, and UBXD8 may dip into the cytosolic face of the ER membrane, all of which may contribute to the formation of a proteinaceous channel. Each of these three proteins also contains conserved functional domains with cytoplasmic exposure (Fig. 5, and Fig. S1c). A schematic representation of the putative organization and composition of this complex is shown in Fig. 5. The initial step of the dislocation pathway involves recognition of the substrate. In the case of Class I MHC HC, this is primarily done by US11, but for RI₃₃₂, a glycosylated misfolded protein, OS9 is involved in the process. The other three proteins described here, AUP1, UBXD8, and UBC6e, also act before cytoplasmic disposition of the dislocation substrate (Fig. S8). UBC6e acts as an E2 ubiquitin ligase, and UBXD8 appears to play a role in the recruitment of the AAA ATPase p97. The role of AUP1 remains elusive, but its CUE domain may be involved in recruitment of another ubiquitin-conjugating enzyme. The identification of additional proteins that participate in these reactions, as reported here, is an important step toward a better understanding of the essential cellular process of dislocation.

Experimental Procedures

Antibodies, Cell Lines, Constructs.

Antibodies.

The cytosolic parts of the three proteins AUP1 (amino acid 62–411), UBC6e (amino acid 1–232), and UBXD8 (amino acid 361–445) were expressed as N-terminal His-tagged fusions in Escherichia. coli BL21 (DE3) Rosetta cells and purified. The recombinant His-tagged fusion proteins were sent to Covance Research Products to generate rabbit polyclonal antibodies. Antibodies against AUP1, UBC6e, and UBXD8 were affinity purified as described in ref. 7. Antibodies to Class I MHC HC, US2, and US11 have been described (34, 36). The anti-GFP, anti-PDI, and anti-OS9 antibodies were purchased from Abcam. Alexa Fluor 488-conjugated goat anti-mouse antibody and Alexa Fluor 568-conjugated goat anti-rabbit antibody were from Molecular Probes. Anti-ribophorin antibody and the RI₃₃₂ cDNA were a generous gift from N. Erwin Ivessa (Vienna Biocenter, Vienna, Austria).

Cell Lines.

U373, US2, and US11 cell lines have been described (10). HeLa and 293T cells were purchased from ATCC. Cells transduced with pLHCX-based vectors were selected and maintained in 125 μg/ml hygromycin B (Roche).

Protein Constructs.

The murine H2-K^b signal sequence was fused to the N-terminal HA-TEV tag of SEL1L to ensure proper ER localization. SEL1L was cloned from cDNA, using standard methods. The SEL1L sequence is unstable in bacteria, and several mutations occurred that were removed by single-point mutagenesis (Strategene). cDNA clones for UBXD8, OS9, UBC6e, and AUP1 were obtained from Open Biosystems, and the ORF was cloned into pcDNA3.1(+), pLHCX (Clontech), and pEGFP-N1 (Clontech). GFP-OS9 was cloned with the OS9 signal sequence replaced by the murine H2-K^b signal sequence followed by GFP.

Anti-HA Affinity Purification and MS/MS Analysis.

A total of 5·10⁸ HeLa cells were lysed for 30 min in 24 ml of ice-cold lysis buffer (2% digitonin, 25 mM Tris·HCl pH 7.4, 150 mM NaCl, 5 mM MgCl₂, complete protease inhibitor tablets [Roche], and 2.5 mM N-ethylmaleimide). The nuclei and cell debris were pelleted at 16,000 × g for 15 min, and the cleared lysate was incubated with 250 μl of anti-HA agarose beads (clone 3F10, Roche) for 3 h at 4°C with gentle agitation. The beads were washed with 50 ml of wash buffer (0.1% digitonin, 25 mM Tris·HCl pH 7.4, 150 mM NaCl, and 5 mM MgCl₂) and eluted with 100 units of TEV protease (AcTEV, Invitrogen) in 250 μl of wash buffer at 4°C overnight. The eluted material was collected, and the beads were washed with 500 μl of wash buffer. The washes and eluted materials were pooled and exchanged into 20 mM NH₄CO₃ pH 8.0, 0.1% SDS by using MicroSpin G-25 Columns (Amersham Biosciences). The eluate was concentrated in a speed-vac and separated by SDS/PAGE (10% acrylamide). Polypeptides were revealed by Coomassie blue staining, excised, and trypsinized as described (7). Peptides were sequenced by liquid chromatography.

Pulse–chase Experiments, Immunoblotting, SDS/PAGE.

Methods for pulse labeling, cell lysis, immunoprecipitation, pulse–chase regarding Class I MHC HCs in US11 and US2 cells, viral transduction of cells, transfection of cells with RI₃₃₂, SDS/PAGE, and fluorography have been described (10). All quantitation was performed on a phosphoimager.

Immunofluorescence, Microsomal Preparation.

Cells were seeded onto glass coverslips and allowed to attach overnight. Fixation was achieved with 4% paraformaldehyde for 20 min at room temperature. Cells were permeabilized with 0.1% Triton X-100 for 10 min at room temperature and incubated with the affinity-purified antibody as described in ref. 7. Imaging was performed on a spinning disk confocal microscope at ×100 magnification. Microsomes were prepared from U373 cells as described in ref. 40. Microsomes were incubated in the indicated buffer conditions for 30 min and centrifuged at 20,000 × g for 20 min. The pellet was resuspended directly in reducing sample buffer, and the supernatant was first trichloroacetic acid precipitated.