Co-translational Interactions of Apoprotein B with the Ribosome and Translocon during Lipoprotein Assembly or Targeting to the Proteasome*

Hepatic lipoprotein assembly and secretion can be regulated by proteasomal degradation of newly synthesized apoB, especially if lipid synthesis or lipid transfer is low. Our previous studies in HepG2 cells showed that, under these conditions, newly synthesized apoB remains stably associated with the endoplasmic reticulum (ER) membrane (Mitchell, D. M., Zhou, M., Pariyarath, R., Wang, H., Aitchison, J. D., Ginsberg, H. N., and Fisher, E. A. (1998) Proc. Natl. Acad. Sci. U. S. A. 95, 14733–14738). We now show that independent of lipid synthesis, apoB chains that appear full-length are, in fact, incompletely translated polypeptides still engaged by the ribosome and associated with the ER translocon. In the presence of active lipid synthesis and transfer, translation and lipoprotein assembly are completed, and the complexes exit the ER. Upon omitting fatty acids from, or adding a microsomal triglyceride transfer protein inhibitor to, culture media to reduce lipid synthesis or transfer, respectively, apoB was degraded while it remained associated with the ER and complexed with cytosolic hsp70 and proteasomes. Thus, unlike other ER substrates of the proteasome, such as major histocompatibility complex class I molecules, apoB does not fully retrotranslocate to the cytosol before entering the ubiquitin-proteasome pathway. Although, upon immunofluorescence, apoB in proteasome-inhibited cells accumulated in punctate structures similar in appearance to aggresomes (cytosolic structures containing molecules irreversibly lost from the secretory pathway), these apoB molecules could be secreted when lipid synthesis was stimulated. The results suggest a model in which 1) apoB translation does not complete until lipoprotein assembly terminates, and 2) assembly with lipids or entry into the ubiquitin-proteasome pathway occurs while apoB polypeptides remain associated with the translocon and attached to the ribosome.

Apoprotein B100, the major structural protein of atherogenic very low density and low density lipoprotein particles, is an unusually large secretory hepatic protein with a molecular mass ϳ550 kDa. The regulation of the assembly and secretion of the apoB100-containing lipoproteins has been studied in various hepatic cell models, including the human hepatocarcinoma cell line HepG2. In human intestine, enterocytes edit the transcript of the apoprotein B gene so that a shorter protein, apoB48, is formed. Because HepG2 cells do not have this editing activity (2), throughout this report, the abbreviation apoB will refer exclusively to apoB100 and its incompletely translated polypeptides.
One of the early steps of apoB-lipoprotein biogenesis in the endoplasmic reticulum (ER) 1 is the association of translocated domains of apoB with its "lipid ligands" in a process mediated by the lipid transfer activity of microsomal triglyceride transfer protein (MTP) (1). After the initial co-translational lipidation of apoB, the remainder of the assembly process is thought to occur post-translationally (3). If either the synthesis or transfer of lipid ligands is limited, apoB is rapidly degraded, as illustrated not only by numerous studies in vitro (e.g. Refs. 4 and 5), but also by the human genetic disease abetalipoproteinemia, which results from mutations of MTP (6). We and others have shown in HepG2 cells that most of this metabolically regulated intracellular degradation of apoB occurs through the ubiquitin-proteasome pathway (7)(8)(9) in a process that appears to involve the cytosolic chaperone hsp70 (8,10).
Studies in yeast and mammalian systems have identified the proteasome as the principal means of disposal of a growing list of ER-associated proteins that presumably become misfolded because of either a structural mutation or a failure to assemble properly into a multimeric complex (11,12). This disposal process has been called ER-associated degradation (ERAD). Because the components of the ubiquitin-proteasome machinery are located in the cytosol, degradation of secretory and integral membrane proteins by this pathway requires that domains are, or become, accessible to the cytosol. Studies on the degradation of major histocompatibility complex class I heavy chains (13) and the cystic fibrosis transmembrane conductance regulator (14), among other examples, have shown that multi-ubiquitinylated substrates accumulated in the cytosol when proteasome activity was inhibited, supporting a model in which complete retrotranslocation of these proteins from the ER to the cytosol occurs (14). At least for the cystic fibrosis transmembrane conductance regulator, as undegraded protein accumulated in the cytosol, large aggregates ("aggresomes"), irreversibly lost from the secretory pathway, formed and were visible by fluorescence microscopy (14,15).
In contrast to these other ERAD substrates, we and others have recently shown that, in HepG2 cells, apoB destined for degradation does not accumulate in the cytosol when the proteasome is inhibited (16), but remains associated with the ER in close proximity to the translocon protein Sec61p (1,17). Thus, complete retrotranslocation may not be a universal feature of ERAD. Rather, only a subset of domains of the substrate may become exposed to and engaged by cytosolic components. This would be consistent with the localization in yeast of the ubiquitin-conjugating enzymes Ubc6p and Ubc7p to the cytosolic face of the ER membrane (11,18) and the demonstration of the 26 S proteasome itself on the cytosolic surfaces of the nuclear envelope and the ER membrane in fission yeast (19) and mammalian cells (20).
The focus of this study is the relationship between apoB and the translational and translocational machinery under metabolic conditions favoring either lipoprotein assembly or proteasomal degradation. From a variety of approaches, we have obtained data to support a model in which apoB translation in HepG2 cells is not completed until lipoprotein assembly has concluded. If conditions are not favorable to assemble a lipoprotein, apoB polypeptides destined for degradation form a complex with the proteasome and the cytosolic chaperone hsp70, leading to degradation. Overall, then, for apoB in HepG2 cells, both lipoprotein assembly and ERAD appear to be co-translational rather than post-translational processes.

EXPERIMENTAL PROCEDURES
General Cell Culture Methods and Immunological Reagents-HepG2 cells were maintained in minimal essential medium (Life Technologies, Inc.) containing 10% fetal bovine serum, 200 M L-glutamine, and 200 units/ml penicillin/streptomycin and were studied after achieving ϳ80% confluency. In all but the immunofluorescence experiments, cells were pretreated for 60 min with 10 M proteasomal inhibitor (lactacystin), which was then included in all subsequent labeling and chase media. In addition, unless otherwise noted under "Results," all pulsechase experiments were performed either with oleic acid (OA; to stimulate lipid synthesis and lipoprotein assembly) complexed to BSA (0.8 mM final concentration of OA in a 5:1 molar ratio with BSA) or with 0.16 mM BSA added to the chase medium.
Rabbit antiserum against the 26 S proteasome and the fluorescent substrate for in vitro proteasome activity assay (benzyloxycarbonyl-Gly-Gly-Phe-p-aminobenzamide) were generous gifts from Drs. M. Orlowski and C. Cardozo (Department of Pharmacology, Mount Sinai School of Medicine, New York). Rabbit antiserum to Sec61␣ was prepared by Research Genetics (Huntsville, AL) and used as described previously (21). Goat anti-human apoB antiserum and monoclonal antibody for immunoprecipitation and immunofluorescence, respectively, and antisera to human apoA-I and calnexin were purchased from Calbiochem. Monoclonal antibody to hsp70 was purchased from Stressgen Biotech Corp. (Victoria, Canada). Antiserum to human ␣ 2 -macroglobulin (␣ 2 M) was purchased from BIODESIGN International (Saco, ME).
In some apoB immunoprecipitation experiments, we used epitopespecific monoclonal antibodies, Bsol 14 (specific for the N terminus) and Bsol 22 (specific for the C terminus), which were purchased from the Lipoprotein Research Facility of the Ottawa Heart Institute (Ottawa, Canada). Anti-puromycin antiserum was provided by one of us (W. J. W.) and is described in Ref. 22.
Interaction between Sec61␣ and ApoB-Proteasome-inhibited HepG2 cells were incubated for 15 min in the presence of 100 Ci/ml Express 35 S-protein labeling mixture (labeled methionine and cysteine; PerkinElmer Life Sciences). Isotope-containing medium was then removed, and the cells were incubated in chase medium (minimal essen-tial medium containing 5 mM methionine and 2 mM cysteine) for 20 or 60 min. At the end of the chase period, cells were harvested in crosslinking buffer (1ϫ PBS, 5 mM EDTA, and 0.4% (w/v) digitonin). The homobifunctional cross-linking agent DSP (Pierce) was added to a final concentration of 0.25 mM, and the cells were incubated on ice for 40 min. The cross-linking reaction was quenched by adding ice-cold Tris (pH 8) to a final concentration of 50 mM and incubating on ice for 15 min. Cells were then lysed in denaturing lysis buffer (150 mM NaCl, 50 mM Tris (pH 8), 5 mM EDTA, 0.5% (w/v) deoxycholate, 1% (v/v) Triton X-100, and 0.1% (w/v) SDS containing 1ϫ protease inhibitor mixture (Roche Molecular Biochemicals)) with shaking overnight at 4°C. The cell lysates were clarified by centrifugation at 1500 ϫ g for 3 min, and the supernatants containing equivalent trichloroacetic acid-precipitable counts were then used for sequential immunoprecipitations. The sample was first exposed to antiserum against Sec61␣ as described (21), and the resulting immunocomplexes were captured by protein A-Sepharose and released from the beads by boiling in elution buffer (10% (v/v) ␤-mercaptoethanol, 4% (w/v) SDS, and 20% (v/v) glycerol). The eluate was diluted in 1ϫ NET, and the second immunoprecipitation was performed with anti-apoB antiserum as described previously (8). The eluate from the second immunoprecipitation was analyzed by 3-17% gradient SDS-PAGE and fluorography. The signals were quantified using a densitometer (Bio-Rad GS 700) or a PhosphorImager (Molecular Dynamics, Inc.).
Immunoprecipitation of ApoB by Epitope-specific Antibodies-Proteasome-inhibited HepG2 cells were incubated for 15 min in the presence of 35 S-protein labeling mixture and chased in isotope-free chase medium. At 0, 20, 40, and 60 min, microsomes were prepared (see below) and lysed in denaturing lysis buffer, and equal volume aliquots were separately immunoprecipitated with monoclonal antibodies Bsol 14 and Bsol 22 as described (8). The immunoprecipitates were analyzed by SDS-PAGE and fluorography as described above. In pilot experiments, labeled apoB from conditioned medium was similarly studied, and comparable recoveries were obtained with both antibodies. In some experiments, the topology of apoB was probed by treating microsomes with trypsin and recovering the remaining apoB with Bsol 14 or Bsol 22. Briefly, microsomes were resuspended in trypsin digestion buffer (25 mM KCl, 0.25 M sucrose, 5 mM HEPES, and 5 mM EDTA), and trypsin was added to a final concentration of 75 g/ml. The samples were incubated on ice for 45 min, and the reaction was arrested by adding 75 g/ml soybean trypsin inhibitor and 1ϫ protease inhibitor mixture. The samples were then divided and processed for denaturing immunoprecipitation with epitope-specific Bsol 14 and Bsol 22 antibodies. Under these conditions, there was quantitative cleavage of the cytosolic domain of the transmembrane ER protein calnexin and full protection of the ER luminal protein ␣ 2 M (data not shown).
Puromycin Incorporation into ApoB-Proteasome-inhibited HepG2 cells were incubated for 5 min in the presence of 35 S-protein labeling mixture and then incubated in chase medium. Puromycin (10 M) was added to the medium either at the start of the chase period (0-min time point) or after 70 min. In either case, incubation was continued for an additional 5 min. ApoB and ␣ 2 M (control protein) polypeptides that had incorporated puromycin were isolated by sequential immunoprecipitation using anti-apoB or anti-␣ 2 M antibodies, respectively, in the first step, followed by anti-puromycin antiserum (22) in the second step. The proteins in the final immunoprecipitate were resolved by SDS-PAGE and detected by fluorography as described above. To verify the specificity of anti-puromycin antiserum, 100 M puromycin was added to an aliquot of a control sample before incubation with anti-puromycin antiserum (see Fig. 3).
Removal of Ribosomes from the Endoplasmic Reticulum-Proteasome-inhibited HepG2 cells were incubated for 15 min in the presence of 35 S-protein labeling mixture and chased for 0, 20, or 60 min. Cells were harvested in 1ϫ PBS containing 0.4% (w/v) digitonin to semipermeabilize the cells. Equal volume aliquots were treated with or without 1 mM puromycin and 0.5 M potassium acetate and then incubated on ice for 15 min and at 37°C for 15 min to remove ribosomes from the ER membrane (23). Cell pellets were collected by centrifugation at 1500 ϫ g for 10 min and resuspended in DSP-containing buffer, and apoB cross-linked to Sec61␣ was isolated by sequential immunoprecipitation.
Interaction between Microsomal ApoB and the Proteasome-Proteasome-inhibited HepG2 cells were incubated for 15 min in the presence of 35 S-protein labeling mixture and chased for 20 or 60 min. At each time point, plates of cells were placed on ice and sequentially washed with ice-cold PBS, buffer containing 0.25 M sucrose and 5 mM EDTA, and SH buffer (0.25 M sucrose and 5 mM HEPES (pH 8.0)). Cells were scraped in SH buffer containing 1ϫ protease inhibitor mixture and sonicated twice on ice at 50% duty cycle for 30 pulses using a Vibra Cell TM sonicator (Sonics & Materials, Danbury, CT). The cell suspension was centrifuged at 10,000 ϫ g at 4°C for 10 min to remove the nuclei and unbroken cells. Microsomes were separated by ultracentrifugation at 100,000 ϫ g for 1 h at 4°C in a Beckman TLA 100.4 rotor. The pellets were resuspended in 1 ml of 0.5 M sucrose and 5 mM HEPES (pH 8.0) and centrifuged again for 1 h at 100,000 ϫ g at 4°C.
Microsomal pellets were resuspended in nondenaturing lysis buffer (0.5% (w/v) deoxycholate, 150 mM NaCl, and 50 mM Tris-HCl (pH 8)) overnight with rocking at 4°C. The suspension was clarified by centrifugation for 5 min at 1500 ϫ g. The supernatant containing the microsomal membranes was diluted in nondenaturing immunoprecipitation buffer (150 mM NaCl, 50 mM Tris-HCl (pH 8), and 5 mM EDTA) and incubated with anti-proteasome antiserum for 3 h with rocking at 4°C. Immunocomplexes captured by protein A-Sepharose were eluted by boiling in buffer containing 2% (v/v) ␤-mercaptoethanol. To isolate proteasome-associated apoB, the eluates were diluted, and a second immunoprecipitation was performed using antiserum to apoB. The final immunoprecipitate was analyzed by SDS-PAGE and fluorography as described above.
Interactions among Microsomal ApoB, hsp70, and the Proteasome-Proteasome-inhibited HepG2 cells were labeled to steady state by incubation for 2 h in the presence of 35 S-protein labeling mixture. The cells were then permeabilized (4°C for 15 min) with buffer containing 1ϫ PBS, 0.4% (w/v) digitonin, and 1ϫ protease inhibitor mixture with or without 50 units/ml apyrase (Sigma). Microsomal membranes were prepared as described above and subjected to immunoprecipitation with the anti-hsp70 monoclonal antibody. One-half of the immunoprecipitate was used for Western blotting (Renaissance kit, PerkinElmer Life Sciences) with anti-proteasome antiserum (to detect hsp70-proteasome interactions), and the other half was used for a second immunoprecipitation with anti-apoB antiserum (to detect hsp70-apoB interactions). The immunoprecipitate was then analyzed by SDS-PAGE and fluorography as described above.
Immunofluorescence Studies-HepG2 cells grown on collagen-coated coverslips were preincubated for 2 h with 10 M lactacystin in minimal essential medium containing 1% fetal bovine serum. Cells were then washed, and the medium was replaced with one containing either 0.16 mM BSA or 0.8 mM OA (as a complex with BSA) to stimulate lipid synthesis. The cells were fixed 0, 30, or 90 min after OA addition using 3% paraformaldehyde in 1ϫ PBS. Fixed cells were permeabilized with 0.1% Triton X-100 and incubated with anti-apoB antibody (primary antibody) and then with Texas Red-conjugated anti-rabbit IgG (Jackson ImmunoResearch Laboratories, Inc.) as the secondary antibody as described (1). In control experiments, the same protocol was performed using anti-human apoA-I antiserum as the source of primary antibody.

Kinetics of the Close Association of ApoB with the Translocon
Protein Sec61␣-We have previously reported that pulse-labeled apoB that is protected from degradation by proteasomal inhibitors stays associated with the ER translocon protein Sec61␤ for at least 1 h after the start of the chase period (1). This prolonged association with Sec61␤ was disrupted after 40 min when triglyceride synthesis, and therefore lipoprotein assembly, was stimulated by including OA in the chase medium. Based on the time estimated for hepatic lipoprotein assembly and ER exit (ϳ37 min) (24), this result suggested that throughout the biogenesis of an apoB-lipoprotein complex in the HepG2 cell, apoB remained in close proximity to the translocon.
The interaction of translocating chains with Sec61␣ and Sec61␤ can vary depending on the length of the polypeptide or its transmembrane topology (e.g. Ref. 25). Thus, to test whether the previous results for Sec61␤ represented a special relationship of that protein with apoB or a more general association of apoB with the translocon, the pattern of interaction of pulselabeled apoB with Sec61␣ was determined.
HepG2 cells pretreated with lactacystin were incubated in the presence of [ 35 S]methionine/cysteine for 15 min and chased for up to 60 min in isotope-free medium containing excess methionine/cysteine (chase medium). Cells were harvested at 0-, 20-, or 60-min time points, and apoB adjacent to Sec61␣ was detected by DSP-based cross-linking and sequential immunoprecipitation (see "Experimental Procedures"). As shown in Fig. 1, similar to the published results for Sec61␤ (1), a substantial amount of Sec61␣-associated apoB (ϳ23% of the total labeled apoB pool) (Fig. 1B) could be detected as late as the 60-min time point when OA was omitted from the chase medium. In the presence of OA, the fraction of apoB adjacent to Sec61␣ was significantly reduced at 60 min (Fig. 1A, lane 5 versus lane 4). Note that the efficiency of chemical cross-linking of proteins with disuccinimide-based reagents typically averages no more than 10% (26), strongly suggesting that the relatively robust results for the apoB-Sec61␣ interactions summarized in Fig. 1 are representative of the behavior of the bulk of apoB polypeptides studied under these conditions.
To exclude the possibility that OA was affecting apoB-translocon interactions independent of its promotion of lipoprotein assembly, such as by disrupting the ER membrane, the above experiment was repeated in the presence of 0.1 M BMS-200150, an inhibitor of MTP lipid transfer activity (kindly provided by Drs. David Gordon and John Wetterau, Bristol-Myers Squibb Co.) (27). If the effect of OA were nonspecific, blocking MTP activity and lipoprotein assembly would not alter the interaction of apoB with the translocon. As shown in Fig. 1 (lane 6), however, the decrease in apoB-Sec61␣ cross-linking at 60 min could be reversed by adding BMS-200150 to the medium in the presence of OA. Thus, the decreased proximity of apoB to Sec61␣ in the presence of OA most likely reflects successful lipoprotein assembly and exit from the ER, and not a nonspecific effect of the fatty acid on apoB-Sec61␣ association.
Differential Recovery of ApoB by Epitope-specific Monoclonal Antibodies-The above and previous (1) results demonstrate that apoB remains adjacent to the translocon during lipoprotein assembly. We wondered whether the apoB bands in Fig. 1 included some incompletely translated apoB that was indistinguishable upon SDS-PAGE from bona fide full-length apoB100. If translation were incomplete, we would expect that in a pulsechase study, the appearance of an epitope at the extreme C terminus should be abnormally delayed relative to one at the N terminus. Therefore, we compared the relative recoveries of the N-and C-terminal epitopes from microsomes isolated at different time points during the 60-min chase. As in the experiments above, proteasome-inhibited HepG2 cells were pulse-labeled; and after 0, 20, 40, and 60 min of chase, cells were harvested, and microsomes were collected by ultracentrifugation (see "Experimental Procedures"). Equivalent aliquots of the resuspended microsomal pellets were immunoprecipitated with Bsol 14 or Bsol 22 monoclonal antibodies, which recognize the Nterminal (amino acids 405-539) and C-terminal (amino acids 4521-4536) regions of apoB, respectively. The signals on the resulting fluorograms were quantified by densitometry, and the abundance of labeled apoB recovered by the anti-C terminus antibody was divided by the abundance recovered by the anti-N terminus antibody. This fraction was multiplied by 100 to derive the parameter we have termed the "C/N ratio." The results are summarized in Fig. 2A. The C/N ratio was 100 for apoB in the conditioned medium (i.e. medium collected after 2 h of continuous labeling in the presence of OA), demonstrating equal recovery efficiencies of the antibodies when both epitopes were present on fully translated apoB. At the end of the 15-min labeling period (i.e. 0 min of chase), the C/N ratio was only 50, indicating that only 50% of the microsome-associated apoB molecules contained the C-terminal epitope. At 20 min of chase, there was still a relative deficiency of the Cterminal epitope. Based on the average elongation rate of hepatic proteins, apoB translation would be expected to be finished by ϳ11 min. Therefore, the data indicate a greater than expected lag in the appearance of the C-terminal epitope, re-flecting a delay in completing apoB translation. In the presence of OA, the C/N ratios at the 20-and 40-min time points were similar to those in Fig. 2A, but were lower at 60 min, most likely representing the exit from the ER of full-length apoB molecules assembled into lipoproteins (data not shown).
As shown in Fig. 2A, some apoB polypeptides were gradually elongated to include the C-terminal epitope, as evidenced by the C/N ratio approaching 100 at the 60-min time point. We were interested in the topology of this extreme C-terminal region, so protease protection studies using these epitope-specific antibodies were performed (see "Experimental Procedures"). Representative results are shown in Fig. 2B. Note that at the 60-min time point, relative to the N-terminal epitopes, the C-terminal epitopes were more sensitive to trypsin digestion under conditions of limited lipid synthesis (i.e. without OA). When OA was added to stimulate lipid synthesis and lipoprotein assembly, the relative protection of the C-terminal epitope was significantly increased.
In control experiments, if microsomes were pretreated with Triton X-100 to disrupt membranes, the concentration of trypsin used resulted in 100% disappearance of the signal for apoB. In addition, this concentration was not damaging to microsomal integrity, as it degraded only the cytosolic tail of calnexin, an integral ER membrane protein (data not shown). Thus, the results shown in Fig. 2B most likely reflect the topological relationship between apoB and the ER membrane and imply that when lipid synthesis is inadequate, there is incomplete translocation and cytosolic exposure of the extreme C-terminal domain of apoB. In contrast, when lipid synthesis is stimulated, this domain is shielded from the cytosol, most likely because it was fully translocated.
ApoB Remains Functionally Bound to the Ribosome throughout the Time Required for Lipoprotein Assembly-The prolonged proximity of apoB polypeptides to the translocon and the kinetics of the differential recoveries of the C-and Nterminal epitopes suggested that the nascent apoB polypeptides were still functionally bound to the ribosome as peptidyl-tRNAs. This possibility was tested by assessing whether puromycin could be incorporated into labeled apoB chains during the chase period, particularly at the later time points. For comparison purposes, puromycin incorporation into another secretory protein, ␣ 2 M, was also assessed.
HepG2 cells pretreated with lactacystin were incubated for 5 min in the presence of [ 35 S]methionine/cysteine and then incubated for up to 70 min in chase medium containing either OA/BSA or BSA alone. Puromycin (10 M) was added to the medium after 70 min of chase, and incubation was continued for another 5 min. Labeled apoB or ␣ 2 M chains that had incorporated puromycin were recovered by sequential immunoprecipitation with either anti-apoB or anti-␣ 2 M antibody in the first step and anti-puromycin antibody in the second step.
As shown in Fig. 3 (lanes 6 and 7), in the presence or absence of OA, there were no ␣ 2 M-puromycin conjugates detected. Surprisingly, there were apoB-puromycin conjugates (lanes 3 and  4), with the labeled apoB species appearing full-length (based on the migration of apoB recovered from the conditioned medium). The percent of apoB conjugated to puromycin was ϳ20% of total labeled apoB, consistent with the 23% recovery of apoB cross-linked to Sec61␣ late in the chase period when OA was absent (Fig. 1). The comparable quantitative recoveries under essentially the same experimental conditions strongly imply that the apoB polypeptides conjugated to puromycin or crosslinked to Sec61␣ belong to the same pool of molecules.
If OA was present in the chase medium, puromycin incorporation into apoB was significantly lower (Fig. 3, lane 4 versus  lane 3), consistent with the loss of association of apoB with the translocon at the time when lipoprotein assembly should be completed, resulting in exit from the ER (Fig. 1) (1, 24). In contrast to these results, if puromycin was added at the end of the labeling period, both labeled nascent apoB and ␣ 2 M chains incorporated puromycin, as expected, and a range of incomplete translation products from 45 to 550 kDa for apoB and from 70 to 180 kDa for ␣ 2 M was observed (Fig. 3, lanes 2 and 5). That the sequential immunoprecipitation was specific for apoB-puromycin conjugates was indicated by a large reduction in the recovery of apoB when excess puromycin was added to the reaction mixture just before adding the anti-puromycin antibody (lane 1).
These results imply that apoB translation either slows or pauses very close to its carboxyl terminus. Overall, the data in Figs. 1-3 suggest that when lipid synthesis is stimulated, apoB translation is completed, and assembled lipoproteins are transported out of the ER.
Association of ApoB with the Translocon Is Independent of the Presence of the Ribosome-Since nascent apoB is bound to the ribosome via the tRNA, and the ribosome is bound to the translocon, the prolonged proximity of apoB to the translocon (Fig. 1) may not be due to direct apoB-translocon interactions. To determine whether the release of the ribosome from the ER membrane would result in the loss of apoB-Sec61␣ cross-linking, HepG2 cells were pretreated, labeled, and chased for the Sec61␣ cross-linking experiment as described above. At the indicated time points (Fig. 4), cells were scraped into buffer, and the suspension was divided into two aliquots. One was treated with 1 mM puromycin in the presence of 0.4 M potassium acetate prior to cross-linking (to disrupt the ribosometranslocon interaction) (23), and the other processed was similarly, but without adding puromycin (see "Experimental Procedures"). Both samples were then treated with the DSP cross-linker, lysed, and subjected to sequential immunoprecipitation with anti-Sec61␣ and anti-apoB antisera (see "Experimental Procedures"). As shown in Fig. 4, the puromycin/high salt treatment, which releases ribosomes from translocons (28), did not affect the extent of cross-linking of apoB to Sec61␣ in samples taken at the different chase points. This suggests that along with apoB engagement by the ribosome (Fig. 3), there are Immunoprecipitates were resolved by SDS-PAGE; apoB signals were quantified by densitometry; and the ratio of the recovery attained with Bsol 22 to that attained with Bsol 14 (referred to as the C/N ratio) was calculated for each time point. The y axis is the (mean C/N ratio ϫ 100%) Ϯ S.E. (n ϭ 4). Note the relative deficiency of C-terminal epitopes at the 20-and 40-min time points and the equal recoveries of N-and C-terminal epitopes from the conditioned medium. B, the topology of the N-and C-terminal epitopes of microsomal apoB was analyzed by protease sensitivity studies using epitope-specific N-and C-terminal antibodies. Equivalent volume aliquots of microsomes isolated from proteasome-inhibited HepG2 cells that had been pulse-labeled and chased for 60 min were incubated on ice for 45 min with or without 75 g/ml trypsin. Microsomes were then lysed, immunoprecipitated (IP) with Bsol 14 (N-terminal specific antibody (Ab)) or Bsol 22 (C-terminal specific antibody), and analyzed on 3-17% SDS-polyacrylamide gel (left panel). ApoB signals were quantified, and the percentage protection was calculated for each epitope in the absence of OA (lipid deficiency) and in the presence of OA (lipid sufficiency). The protection of the C-terminal epitope is expressed as the mean fraction Ϯ S.E. (three separate experiments) relative to the average percentage protection of the N-terminal epitope in the absence and presence of OA (right panel). Note that the C-terminal epitope was highly susceptible to trypsin digestion in the absence of OA, whereas the majority of the C-terminal epitopes were resistant to trypsin digestion in the presence of OA. also interactions with ER proteins or lipids that serve to retain apoB in the translocon during lipoprotein assembly.
Interactions of Microsome-associated ApoB with the Proteasome and the Cytosolic Chaperone hsp70 -The results above and in our previous reports (1,8,16) imply that apoB can be directed to either lipoprotein assembly or degradation while still associated with the translocon and the ribosome. Thus, rather than interacting with cytosolic factors, such as hsp70 and proteasomes, only after being completely retrotranslocated, apoB may be engaged by these factors while still bound to the ER membrane in a "bitopic" state (i.e. with part of apoB accessible to the cytosol and part in the ER lumen). To test this possibility, HepG2 cells were labeled to steady state for 2 h in the presence of 10 M lactacystin and then semi-permeabilized with digitonin. To test for the ATP dependence of interactions among apoB, hsp70, and the proteasome, some samples were treated with 50 units/ml apyrase to deplete ATP. Microsomes were collected by ultracentrifugation from both apyrasetreated and buffer-treated lysates, and the pellets were resuspended in nondenaturing lysis buffer. hsp70-associated proteins were isolated by immunoprecipitation with anti-hsp70 monoclonal antibody, and one-half of this immunoprecipitate was resolved by SDS-PAGE and immunoblotted with antiproteasome antiserum (see "Experimental Procedures").
As shown in Fig. 5, 26 S proteasome subunits ranging from 25 to 35 kDa were detected. This interaction with hsp70 was specific as shown by its ATP dependence (Fig. 5, lane 3 versus  lane 4). The other half of the hsp70 immunoprecipitate was diluted, and a second immunoprecipitation was performed with anti-apoB antiserum. Labeled apoB was recovered (lanes 1 and 2), and the amount recovered increased after apyrase treatment (lane 1 versus lane 2), indicating an ATP dependence consistent with previous results for hsp70-apoB interactions (10).
To investigate the kinetics of the association of microsomal apoB with the proteasome, HepG2 cells were pretreated with lactacystin; pulse-labeled for 15 min; and then chased for 0, 20, or 60 min as before. Microsomes were isolated at these time points and subjected to sequential immunoprecipitation under nondenaturing conditions (see "Experimental Procedures") using anti-proteasome antiserum in the first step and anti-apoB antiserum in the second step. As shown in Fig. 6, proteasomes and labeled microsomal apoB co-immunoprecipitated at all time points, and the extent of this interaction was lower in the presence of OA. These results are consistent with the decreased degradation of apoB (29) and the decreased cytosolic exposure of the C-terminal domain when HepG2 cells are incubated with OA (Fig. 2B).
The interaction between apoB and the proteasome was specific as judged by the absence of the apoB band if nonimmune serum was used in the first immunoprecipitation step (Fig. 6,  lane 1). We also tested whether proteasomes co-immunoprecipitating with microsomal apoB were functional by performing assays for the chymotrypsin-like activity of the proteasome. By incubating the co-immunoprecipitated proteasomes with a syn- Puromycin was then added, and incubation was continued for 5 min. Cells were lysed, and labeled apoB and ␣ 2 M polypeptides that incorporated puromycin were isolated by immunoprecipitation with anti-apoB or anti-␣ 2 M antiserum, respectively, in the first step and anti-puromycin antiserum in the second step. Immunoprecipitates were resolved by SDS-PAGE, and labeled polypeptides were visualized by fluorography. A representative result is shown. Note that when puromycin was added at 0 min, labeled apoB and ␣ 2 M chains of various lengths that had incorporated puromycin were visible (lanes 2 and 5, respectively). In contrast, when puromycin was added after 70 min of chase, only apoB polypeptides were visible (lanes 3 and 4 versus lanes 6 and 7). Note that the ␣ 2 M lanes were deliberately overexposed to ensure that there was no residual signal detectable at the 70-min time point. In the presence of OA, apoB-puromycin conjugates were decreased (lane 3 versus lane 4). To control for the specificity of the anti-puromycin antiserum, excess puromycin was added to some samples before the second immunoprecipitation to compete against labeled apoB-puromycin conjugates; note the absence of signal (lane 1).

FIG. 4.
Effect of the removal of ribosomes on the proximity of apoB to Sec61␣. Proteasome-inhibited HepG2 cells were labeled with [ 35 S]methionine/cysteine for 15 min and chased for 0, 20, or 60 min. Cells were harvested in 1ϫ cross-linking buffer. One-half of the cells were treated with 1 mM puromycin (Puro) and 0.5 M potassium acetate and then incubated (15 min on ice, followed by 15 min at 37°C) to release the ribosomes from the nascent polypeptide chains. The other half were treated similarly, but without puromycin. After cross-linking with DSP (see "Experimental Procedures"), cell lysates were sequentially immunoprecipitated with anti-Sec61␣ antiserum in the first step and anti-apoB antiserum in the second step. Immunoprecipitates were resolved by SDS-PAGE, and a representative fluorogram is shown. 5. Interactions among microsomal apoB, proteasomes,  and hsp70 in HepG2 cells labeled to steady state. Proteasomeinhibited HepG2 cells were labeled to steady state by incubation with [ 35 S]methionine/cysteine for 2 h. The cells were treated with permeabilization buffer with or without 50 units/ml apyrase before microsomes were prepared as described under "Experimental Procedures" and subjected to nondenaturing immunoprecipitation (IP) with anti-hsp70 antibody. The immunoprecipitate was divided in two aliquots; one was processed for a second immunoprecipitation with anti-apoB antiserum, and the other was used for immunoblotting (Western blotting (WB)) with anti-proteasome antiserum. Both apoB (lanes 1 and 2) and 26 S proteasome subunits ranging from 25 to 35 kDa (lanes 3 and 4) were recovered from the hsp70 immunoprecipitate. Note that these signals increased in apyrase-treated samples, indicating that the interactions with hsp70 were ATP-dependent. thetic substrate (benzyloxycarbonyl-Gly-Gly-Phe-p-aminobenzamide) (30), lactacystin-inhibitable activity was readily detected, suggesting that the microsomal apoB-associated proteasomes were indeed functional (data not shown).
Overall, these results imply that the proteasome and hsp70, both of which we have previously shown to be involved in ERAD of apoB (8), interact with apoB while it is still associated with the ER membrane. In other words, complete retrotranslocation of apoB is not required for its interaction with cytosolic factors and its entry into the ubiquitin-proteasome pathway.
ApoB Accumulates in Non-aggresomal Structures when Proteasome Activity Is Inhibited-We have previously shown by indirect immunofluorescence in HepG2 cells that when proteasomes are inhibited, apoB accumulates and assumes a punctate appearance that co-localizes with ER markers (1,8). Biochemical studies suggested that apoB accumulating under these conditions can still be recruited to lipoprotein assembly and the secretory pathway when OA is added to stimulate lipid synthesis (1). This implies that the intracellular localization of proteasome-protected apoB, as detected by immunofluorescence, will be changed by OA.
To test this, HepG2 cells were treated with lactacystin for 2 h, and then the cells were chased in either BSA-or OA/BSAcontaining medium for 30 or 90 min. The cells were immunostained using either anti-apoB or anti-apoA-I antibody (control protein) as the primary antibody. As shown in Fig. 7 (left  panel), a punctate appearance of apoB, similar to that shown previously (1,8), was apparent at the start of the chase period. Note that in the presence of BSA, there were no obvious changes in the level or distribution of the immunofluorescent signals over time (upper center and right panels). In marked contrast, with OA, the signal became diffuse by 30 min and was significantly lower in intensity by 90 min, consistent with the mobilization and secretion of the accumulated apoB after lipid synthesis was stimulated. Supporting this interpretation is that the apoB signal at 30 min overlapped with a marker (ERGIC53) (31) for the intermediate compartment, which is enriched in secretory vesicles (data not shown). These results were not due to nonspecific effects of OA, based on the lack of changes in the intensity or distribution of the signal for apoA-I (data not shown).
Taken together with the published biochemical results (1,8), these data demonstrate that apoB accumulating in punctate structures after proteasomal inhibition can be recruited to lipoprotein assembly and secretion. As will be described further under "Discussion," the apoB-associated punctate structures are unlikely to represent aggresomes, which are also punctate in immunofluorescent appearance. Although they also consist of an ERAD substrate (the cystic fibrosis transmembrane conductance regulator) that accumulates when the proteasome is inhibited, they contain protein molecules irreversibly lost from the secretory pathway (14,15). DISCUSSION As polypeptides emerge from the translocon, they encounter ER chaperones, such as BiP and calnexin, and undergo a "quality control" process in which conformation-dependent sorting FIG. 6. Kinetics of the interaction between microsomal apoB and the proteasome. Microsomes were isolated from proteasomeinhibited HepG2 cells that had been labeled with [ 35 S]methionine/ cysteine for 15 min and chased for 0, 20, or 60 min in the presence or absence of OA. Proteasome-associated apoB was recovered by sequential immunoprecipitation using either anti-proteasome antiserum (lanes 2-6) or preimmune serum (lane 1; negative control) in the first step and anti-apoB antiserum in the second step. Immunoprecipitates were resolved by SDS-PAGE, and the signals corresponding to labeled apoB were displayed by fluorography. Note that proteasome-associated apoB was detected at all time points, but that the level was significantly reduced at 60 min in the presence of OA.

FIG. 7.
After the proteasome is inhibited, apoB accumulates in punctate structures that are reversed when lipid synthesis is stimulated. HepG2 cells were treated with 10 M lactacystin for 2 h to promote apoB accumulation. The medium was replaced with medium containing lactacystin and either 0.8 mM OA or 0.2% (w/v) BSA. Cells were fixed 0, 30, or 90 min later and immunostained with anti-apoB monoclonal antibody, followed by Texas Red-conjugated anti-mouse IgG. Note that the apoB immunofluorescent signal in the BSA-treated cells remained similar in intensity and distribution throughout the time course. In marked contrast, in the presence of OA, the punctate appearance of the apoB fluorescent signal changed by 30 min to a diffuse pattern; and by 90 min, only a residual Golgi region signal was present, consistent with the mobilization of the majority of accumulated apoB to lipoprotein assembly and secretion from the cells. results in either a native folded state that exits the ER or a malfolded state that is retained and subject to ERAD (see Ref. 32 for a recent review). We have previously shown that when HepG2 cells are deprived of fatty acids, the association of apoB with its lipid ligands is reduced, and ERAD mediated by the ubiquitin-proteasome pathway ensues (1,8). Complete assembly of apoB with lipids to form a hepatic lipoprotein particle culminating in ER exit has been estimated to take ϳ40 min based on the kinetic analysis of subcellular fractionation data (24). It has also been shown that the initiation of lipid association with nascent apoB polypeptides in HepG2 cells is cotranslational (33), consistent with the interaction of metabolically labeled apoB with either Sec61␤ (1) or Sec61␣ (Fig. 1A) (17). In the common models proposed for hepatic lipoprotein formation (e.g. Refs. 3, 34, and 35), however, it has been proposed that the initial lipidation is co-translational, but subsequent lipidation (referred to as a "second step") occurs posttranslationally to complete the assembly process.
The data in this report argue strongly that not only the initial, but the bulk, if not all, of the assembly process in HepG2 cells occurs co-translationally. A key line of evidence to support this view derives from the finding that apoB is adjacent to translocon proteins throughout the time period required for lipoprotein assembly. Furthermore, the disappearance of apoBtranslocon interactions appears to coincide with the expected exit of the completed lipoprotein particle from the ER. A simple explanation for the persistent interaction between apoB and the translocon is that apoB does not leave the translocon until translation has terminated. Yet, in many studies of HepG2 cells, including our own, labeled apoB appears to be full-length (i.e. apoB100, ϳ550 kDa) at early time points in the chase period. SDS-PAGE, however, does not have the resolution to reveal minor differences in size among apoB polypeptides that would occur if translation slowed or paused near the carboxyl terminus of the protein. Thus, we determined directly whether apoB polypeptides were still bound to functional ribosomes as peptidyl-tRNAs late in the chase period by their ability to react with puromycin, an aminoacyl-tRNA analog, and by the kinetics of C-terminal synthesis.
As shown in Figs. 2 and 3, the results independently support the notion that the assembly of a lipoprotein particle and the termination of apoB translation are coupled events. In proteasome-inhibited cells, puromycin could be incorporated into 35 Slabeled apoB polypeptides as long as 70 min following pulse labeling, indicating the functional engagement of nascent apoB chains with the ribosome well into the chase period. In the presence of OA, the decrease in incorporation of puromycin added after 70 min of chase (Fig. 3), a time sufficient for the complete assembly of a lipoprotein, most likely reflects the completion of translation and exit from the ER. Because OA is not totally effective in sorting apoB to lipoprotein assembly (29), the residual apoB at 70 min in ϩOA samples most likely represents nascent apoB that is not associated with a full complement of lipid ligands.
Another indication of a delay in completing apoB translation is the significant lag in the appearance of the C-terminal epitope of apoB relative to that of the N-terminal epitope (Fig.  2). This lag can only be explained by a slower rate of, or pause in, the translation of the region of apoB mRNA encoding the extreme carboxyl-terminal portion of the full-length protein.
Interestingly, the location of the C-terminal epitope is just distal to one of the strongest pause-transfer sequences detected by Lingappa and co-workers (36) in their survey of the entire length of apoB. In a cell-free assay system, they have shown that sequences dispersed throughout the length of apoB can mediate pauses in translocation (37,38). In addition, we (39) and others (40) have observed pauses in translation of apoB. Both types of pauses have been hypothesized to allow time for the assembly of nascent apoB chains with lipids (3). Making an analogy to the coordinated efforts of the ribosome and translocon during the insertion of a transmembrane domain into the ER membrane bilayer (41), the results (puromycin incorporation, delayed C-terminal epitope appearance, and prolonged interaction with Sec61p) imply that the translational and translocational apparatuses cooperate to stabilize a topographical state of apoB necessary to achieve the final stage of lipoprotein assembly in HepG2 cells. Relevant to this point is a proposed "stop-translation" sequence (in the bacterial -receptor), which is thought to halt translation to allow sufficient time for important interactions between the nascent polypeptide and membrane lipids and proteins that are required for secretion (42). The fact that a considerable length of apoB polypeptide is translated prior to a final assembly stage is consistent with recent studies showing that glycosylation sites in apoB as distal as apoB68 are modified in HepG2 cells (43) in the absence of OA.
The prolonged binding of the nascent apoB chain to the ribosome via the tRNA and the affinity of the ribosome for the translocon lead, not surprisingly, to the positioning of the apoB within or near the translocon late in the assembly process. Yet, when ribosomes were stripped off the ER membrane, apoB was still adjacent to Sec61␣ (Fig. 4). Thus, the cross-linking of apoB to Sec61␣ results not only from their spatial proximity in the ribosome-translocon complex, but also from protein-protein interactions. These interactions with the nascent chain may be mediated by translocon components or by translocon-associated proteins, as suggested by studies on ␤-sheet and pausetransfer sequences contained in apoB (44,45). Also consistent with the data, apoB may remain in association with the ER membrane by interacting with ER proteins in the vicinity of the translocon (as suggested by studies of apoB and ER chaperones) (46,47) or with lipid components (because of the many hydrophobic domains of apoB). This last possibility is compatible with the inability of urea to effectively extract apoB from microsomes. 2 Delayed translation (and hence termination of translation) then may not reflect the need of the ribosome to be a structural anchor of apoB polypeptides, but may provide the length of polypeptide and time required to position nascent apoB properly so that it interacts with a variety of protein and lipid factors that help achieve or maintain the conformation needed to complete lipoprotein assembly. This would not be surprising, given that many transmembrane and secretory proteins, albeit more typical than apoB, undergo chaperone-mediated conformational maturation to the native folded state before ER exit is allowed (32). That the maturation process is prolonged for apoB is consistent with its previously reported tardiness (relative to other secretory proteins) in exiting the ER of HepG2 and rat hepatic cells (24,48).
Our results also speak to the fate of nascent apoB chains not successfully assembled into a lipoprotein particle. In contrast to the paradigm for other substrates of ERAD, such as major histocompatibility complex class I molecules in cytomegalovirus-infected cells and mutant carboxypeptidase Y in yeast (see Ref. 49 for a recent review), we have previously shown in proteasome-inhibited HepG2 cells maintained in OA-deficient medium that apoB does not appear to be completely retrotranslocated or "dislocated" to the cytosol prior to degradation (1). The proteasome and cytosolic factors involved in apoB degradation therefore appear to interact with ER-associated apoB polypeptides, as supported by the finding of ATP-dependent interactions among microsome-bound apoB, proteasomes, and cytosolic hsp70 (Figs. 5 and 6). It should be noted that the pretreatment of fatty acid-deficient HepG2 cells with lactacystin allowed the study of apoB intermediates of greater length than would have been ordinarily present in this metabolic state, based on data showing that targeting of apoB polypeptides to the ubiquitin-proteasome pathway can occur as early as midway through translation (16,50). Nonetheless, many features of the proteasomal targeting of apoB appear to be independent of polypeptide length, namely, stimulation by a deficiency in lipid synthesis or transfer, the increased exposure of the C-terminal region relative to the N terminus when lipoprotein assembly is not successful, and an increased interaction with hsp70 (16,51). We have previously shown that increased hsp70 expression promotes apoB ubiquitinylation (16) and proteasomal degradation (8), and we presume that the hsp70 bound to microsomal apoB (Fig. 5) is fulfilling functional roles, consistent with the decrease in apoB-proteasome interaction when OA is present (Fig. 6).
The lack of cytosolic accumulation of apoB in proteasomeinhibited cells (1) indicates that, unlike other ERAD substrates, such as the cystic fibrosis transmembrane conductance regulator (14,15) and influenza virus antigen (52), apoB does not form aggresomes, cytosolic structures containing proteins that are irreversibly lost to the secretory pathway. Consistent with this was the finding that the stimulation of lipid synthesis and lipoprotein assembly by OA resulted in the disappearance of the apoB-containing punctate immunofluorescent structures that formed after proteasome inhibition (Fig. 7). That this disappearance represented the rapid recruitment of the accumulated apoB to lipoprotein assembly and secretion is supported by our two previous studies (1,8) in which there was quantitative recovery of metabolically labeled apoB from the conditioned medium of HepG2 cells treated similarly to the cells in Fig. 7. Because these apoB-containing punctate structures co-localized with ER markers in previous immunofluorescence studies (1,8), it is plausible that they are part of the recently described "ER-associated bodies" formed at exit sites in which another proteasomal ERAD substrate, a mutant form of a yeast ABC transporter Ste6p, has been shown to accumulate (53). It is therefore intriguing to speculate that the diffuse apoB immunofluorescent signal observed 30 min after OA addition represents a plethora of secretory vesicles containing apoB-lipoprotein complexes that budded off the ER after lipid synthesis and assembly were stimulated.
In summary, we have used proteasome-inhibited HepG2 cells to study the molecular events in the two possible paths of apoB metabolism: lipoprotein assembly or ERAD. Both paths appear to be co-translational. The assembly process, which occurs when there is sufficient transfer of lipid ligands, functions while apoB polypeptides are incompletely translated and are interacting with the ribosome, translocon, ER chaperones, and, perhaps, the lipid bilayer. Lipoprotein assembly, apoB translocation, and apoB mRNA translation appear to be coordinated events that serve to maintain apoB in an appropriate functional and topographical state. If lipid ligands are insufficient, then ER-associated apoB polypeptides interact with cytosolic factors that target it to the ubiquitin-proteasome pathway. Given the potential of the proteasome (54) or chaperones (55,56) to act as molecular motors by virtue of their intrinsic ATPase activities, apoB is most likely extracted from the ER membrane rather than becoming fully retrotranslocated into the cytosol prior to its degradation.
A schematic summarizing the essential features of the proposed model is given in Fig. 8. Basically, lipoprotein assembly and apoB degradation resemble the quality control mechanisms governing the molecular sorting of other secretory and transmembrane proteins in the ER, in that exit and retention (and ultimate degradation) represent the successful separation of conformational variants. The differences between apoB and other ERAD substrates most likely reflect special requirements of lipoprotein assembly, unusual structural features of the apoB protein, and the regulation of the quality control process by the lipid metabolic state.