Nascent Lipidated Apolipoprotein B Is Transported to the Golgi as an Incompletely Folded Intermediate as Probed by Its Association with Network of Endoplasmic Reticulum Molecular Chaperones, GRP94, ERp72, BiP, Calreticulin, and Cyclophilin B*

We have previously demonstrated that endoplasmic reticulum (ER)-resident molecular chaperones interact with apolipoprotein B-100 (apoB) during its maturation. The initial stages of apoB folding occur while it is bound to the ER membrane, where it becomes partially lipidated to form a primordial intermediate. We determined whether this intermediate is dependent on the assistance of molecular chaperones for its subsequent folding steps. To that end, microsomes were prepared from HepG2 cells and luminal contents were subjected to KBr density gradient centrifugation. Immunoprecipitation of apoB followed by Western blotting showed that the luminal pool floated at a density of 1.12 g/ml and, like the membrane-bound pool, was associated with GRP94, ERp72, BiP, calreticulin, and cyclophilin B. Except for calreticulin, chaperone/apoB ratio in the lumen was severalfold higher than that in the membrane, suggesting a role for these chaperones both in facilitating the release of the primordial intermediate into the ER lumen and in providing stability. Subcellular fractionation on sucrose gradients showed that apoB in the Golgi was associated with the same array of chaperones as the pool of apoB recovered from heavy microsomes containing the ER, except that chaperone/apoB ratio was lower. KBr density gradient fractionation showed that the major pool of luminal apoB in the Golgi was recovered from 1.02 < d < 1.08 g/ml, whereas apoB in ER was recovered primarily from 1.08 < d < 1.2 g/ml. Both fractions were associated with the same spectrum of chaperones. Together with the finding that GRP94 was found associated with sialylated apoB, we conclude that correct folding of apoB is dependent on the assistance of molecular chaperone, which play multiple roles in its maturation throughout the secretory pathway including distal compartments such as the trans-Golgi network.

The correct folding of apolipoprotein B (apoB) 1 into its mature, secretion-competent form is a complex process that leads to the formation and secretion of triacylglycerol (TAG)rich lipoproteins, chylomicrons, and the atherogenic lipoproteins, very low density lipoproteins (VLDL) (1,2). A two-step model for the formation of VLDL was originally proposed on the basis of electron microscopic studies in rat liver (3). This model was further supported by latter studies in rat liver (4), in rat hepatoma cells (McA-RH7777; Refs. [5][6][7][8], and in transgenic mice lacking the gene for apoB in the intestine (9). According to this model, the first step involves partial lipidation of apoB to form a primordial intermediate with HDL/LDL-like density. In the second step this intermediate fuses with a large apoB-free lipid droplet composed primarily of TAG to form nascent VLDL or chylomicrons. In HepG2 cells, however, these distinct steps were not clearly demonstrated. Nonetheless, the initial steps of lipidation are thought to be similar to other systems in that they are mediated by microsomal triglyceride transfer protein (MTP) (reviewed in Refs. 10 -12) and occur co-translationally, while apoB is bound to the ER membrane (13). Partial lipidation of apoB leads to formation of an intermediate with HDL/ LDL-like densities. This intermediate is then released into the ER lumen. Subsequent recruitment of lipids by this intermediate seems to occur continuously (13) to ultimately form and secrete larger TAG-rich particles with primarily LDL-like density and, to a lesser extent, particles with VLDL and IDL-like densities.
In the absence of sufficient lipids, primarily TAG and cholesterol esters (1,14,15), or when MTP activity is diminished, either as a result of mutations (11,16) or as a result of specific inhibitors (7,17), apoB is unable to form lipoproteins and is targeted to proteasome-dependent degradation both co-and post-translationally (18,19) in hsp70-and hsp90-mediated mechanisms (20 -22).
In addition to MTP, the folding of apoB, like that of other secretory proteins, seems to require the assistance of ER-resident molecular chaperones including BiP, calnexin, calreticulin (CRT), ERp72, and GRP94 (18,(23)(24)(25)(26)(27)(28). Molecular chaperones bind transiently to nascent polypeptides to prevent their aggregation, and thereby maintain them in conformations competent for efficient folding. Once correct folding is attained, chaperones dissociate and folded proteins exit the ER. However, if nascent polypeptides fail to attain their native form, they are retained in the ER bound to chaperones, which ultimately target them for proteasome-dependent degradation (29) to prevent their transport to their target destination (30 -34).
Given the complexity of apoB folding, which involves lipidated membrane-bound and luminal intermediates, it was of interest to characterize in great detail the involvement of chaperones in various stages of its biogenesis.
We hypothesized that the membrane-bound pool of apoB will interact with molecular chaperones. However, it was unclear whether partially lipidated apoB that is translocated into the ER lumen as a primordial intermediate is still dependent on the assistance of molecular chaperones for its subsequent folding steps and, if so, whether these chaperones dissociate from apoB-containing particles before their transport to the Golgi. This study was designed to address these questions.
We demonstrate that the same array of molecular chaperones interact with both the membrane-bound and the luminal pools of apoB. However, the relative level of BiP, GRP94, ERp72, and cyclophilin B (CyPB) associated with the luminal pool of apoB is severalfold higher than that associated with the membrane-bound pool, suggesting a role for these chaperones in facilitating the release of apoB from the microsomal membrane. We further demonstrate that GRP94, ERp72, BiP, calreticulin (CRT), and CyPB remain bound to apoB that is transported to the Golgi, albeit the chaperone/apoB ratio was lower than that in the ER. This is consistent with the idea that these chaperones play multiple roles both in early and in late folding events occurring in the ER and the Golgi, respectively. Thus, apoB is recognized as an incompletely folded protein throughout most of the secretory pathway.

Materials
Antibodies-Goat polyclonal antibodies to apoB were from Biodesign and Calbiochem. Mouse monoclonal antibodies to apoB 2D8 were kindly provided by Dr. E. Marcel and Dr. Milne (University of Ottawa, Canada). B4 was a generous gift from Dr. J. Fruchard (Université de Lille, Lille, France) with permission from Dr. H. Bazin (University of Louvain, Louvain, Belgium). Polyclonal antibodies to GRP94 were kindly provided by Dr. S. Cala (Wayne State University, Detroit, MI). Rat monoclonal antibodies to GRP94, mouse monoclonal antibodies to BiP, rabbit antibodies to ERp72 and to calreticulin were from Stressgen Biotechnologies Corp. (Victoria, British Columbia, Canada); rabbit antibodies to CyPB and mouse monoclonal antibodies to Golgin 97 were from Molecular Probes. Monoclonal antibodies to GPP130 were kindly provided by Dr. H. Hauri (University of Basel, Basel, Switzerland). Mouse monoclonal antibodies to p58 and horseradish peroxidase (HRP)conjugated secondary antibodies were from Sigma.

Methods
Cell Cultures, Pulse-chase Experiments, and KBr Density Gradient Fractionation-HepG2 cells were pulse-labeled with [ 35 S]methionine/ cysteine and chased as described (26). Cell monolayers were subjected to cross-linking, then solubilized and immunoprecipitated with antibodies to apoB (26). Media were collected and spun at 800 ϫ g for 5 min, concentrated using microconcentrators, adjusted to 1.25 g/ml with solid KBr, overlaid by 1.006 g/ml solution, and spun in an SW41 rotor at 40,000 rpm for 20 -24 h. Density fractions were collected from the top and their density determined by an Abbe Refractometer (American Optical Corp.). Fractions were washed in microconcentrators, adjusted to 2% CHAPS, 50 mM HEPES, and subjected to immunoprecipitation.
Preparation and Extraction of Microsomes-HepG2 cells were incubated for 2 h in media supplemented with 0.8 mM oleic acid complexed to 0.2 mM albumin and 25 g/ml ALLN to increase lipid availability and inhibit degradation, respectively. Cells were either subjected to crosslinking and then scraped with a rubber policeman (26) or were scraped without prior cross-linking. Washed cells were suspended in homogenization buffer (0.25 M sucrose, 10 mM HEPES, pH 7.5, and the following protease inhibitors: 1 mM AEBSF, 5 g/ml ALLN, and 10 g/ml each aprotinin and leupeptin), then homogenized with a tight-fitting Dounce homogenizer with 20 strokes. In some experiments, homogenates were also passed through a 25-gauge needle. Homogenates were spun for 5 min at 6,000 ϫ g, and post-nuclear supernatants were spun for 40 min at 200,000 ϫ g at 3°C to pellet the microsomes.
Microsomes were extracted either with carbonate alone (0.1 M Na 2 CO 3 , pH 7.4) (35) or with carbonate supplemented with 0.025% sodium deoxycholate, with or without 1.2 M KCl (carbonate/DOC or carbonate/DOC/K, respectively) (36). Samples were then spun at 100,000 ϫ g for 30 min to separate the soluble contents (luminal) from the membranes. Soluble contents were neutralized, and salts removed by ultrafiltration. Membranes and luminal contents were then adjusted to 2% CHAPS, 50 mM HEPES containing protease inhibitors as described above, and apoB immunoprecipitated. In some experiments luminal contents and membranes were adjusted to 1% SDS in PBS, heated to 65°C for 15 min, diluted to 0.1% SDS with 1% Triton X-100 in PBS, and subjected to immunoprecipitation.
Subcellular Fractionation on Sucrose Gradients-HepG2 cells were incubated for 2 h in media supplemented with oleic acid and ALLN as described before. Cells were either subjected to cross-linking and then scraped with a rubber policeman (26) or were scraped without crosslinking as indicated in the figures. Washed cells were suspended in homogenization buffer, homogenized, and centrifuged as described above. Postnuclear supernatants were adjusted to 1.3 M sucrose and overlaid with 1.2, 1.15, 0.86, and 0.25 M sucrose, 10 mM HEPES (37). Gradients were spun for 18 h at 24,000 rpm at 4°C in SW41 rotor and eight fractions collected from the top as follows: top 1 ml, light Golgi Pellets were extracted either with carbonate or carbonate/DOC/K as described above. Fractions were then centrifuged for 30 min at 100,000 ϫ g to separate membranes from luminal contents. Membranes were washed in PBS and then solubilized in 2% CHAPS. Luminal contents were adjusted to pH 7.4, concentrated using microconcentrators, and adjusted either to PBS, pH 7.4, or to 2% CHAPS, 50 mM HEPES, pH 7.4, both containing protease inhibitors as described above. Membranes and luminal contents were then processed for immunoprecipitation. In some experiments, luminal contents were subjected to KBr density gradient centrifugation as described above and apoB in density fractions immunoprecipitated.
Immunoprecipitation and Western Blotting-Samples were precleared by incubation with gelatin-and protein G-Sepharose and nonimmune goat or rabbit serum, mouse, or rat IgGs for 1 h at 4°C. Beads were removed by centrifugation, and supernatants were incubated at 4°C for 18 -20 h with antibodies to apoB, GRP94, or ERp72 along with appropriate controls as indicated in the figures. Immunocomplexes were captured with protein G-Sepharose as described (26). For native immunoprecipitation, beads were washed in PBS followed by phosphate buffer containing 0.5 M NaCl and, finally, in PBS.
Immunocomplexes were solubilized by heating the beads to 95°C in Laemmli sample buffer (38) containing 10% ␤-mercaptoethanol and 6 M urea, then resolved by 4 -15% SDS-PAGE followed by electrotransfer onto PVDF membranes (26). ApoB, chaperones, ER, and Golgi marker proteins were probed with appropriate antibodies, visualized by the ECL system (NEN Life Science Products), and quantified by densitometry, using Duoscan T1200 Agfa Densitometer and Zero-Dscan image analysis system.
Lectin Blotting-Microsomes were prepared from un-cross-linked HepG2 cells as described above, then extracted with carbonate and luminal contents subjected to native immunoprecipitation with either rabbit antiserum to GRP94 or nonimmune rabbit serum. Immunocomplexes were resolved on 4 -15% SDS-PAGE and transferred onto PVDF membranes. Nonspecific sites were blocked by incubating the mem-branes in 5% albumin, PBS for 1 h. Membranes were then incubated in either buffer alone (phosphate buffer, pH 6.8, 1 mM CaCl 2 , 0.01% albumin) or buffer containing 20 milliunits/ml neuraminidase for 20 -24 h at 37°C. Membranes were then washed and incubated with biotinylated SNA-I, which was visualized using the ECL system following incubation with HRP-extravidin.

RESULTS AND DISCUSSION
Chaperones Interact Both with the Membrane-associated Pool of ApoB and with the Pool Released into the Microsomal Lumen-Unlike other secretory proteins, the newly translated pool of apoB remains associated with the ER membranes during early stages of folding, which involve MTP-dependent lipidation to form a primordial lipoprotein intermediate that is ultimately released into the ER lumen (39). We hypothesized that it is during these early stages of folding that molecular chaperones would be most critical for its successful folding and maturation. The partially lipidated intermediate that is translocated into the lumen is conceivably more stable than the membrane-bound pool and, therefore, may not require the assistance of molecular chaperones, so that it could proceed to acquire the bulk of lipids to form VLDL/IDL independently of chaperones. Alternatively, this intermediate may still depend on chaperone assistance for its subsequent folding steps. In that case, two scenarios may be envisioned: (a) it remains associated with the same array of molecular chaperone as those interacting with the membrane-bound pool, or (b) it remains associated with a subset of chaperones that may be necessary to keep it in a conformation competent for binding the bulk lipids. To test these possibilities, we determined first whether chaperones are associated with the luminal pool of apoB. To that end, crude microsomes were prepared from HepG2 cells that were pre-incubated with 0.8 mM oleate and the calpain inhibitor ALLN, which also effectively inhibits proteasomes (40) to increase folding efficiency. Microsomes were subjected to three extraction protocols to release the luminal contents: carbonate (0.1 M, pH 11.5) alone or carbonate supplemented either with deoxycholate (carbonate/DOC) or deoxycholate and KCl (carbonate/DOC/K). The latter two protocols were demonstrated to increase the efficiency of extraction of lipidated apoB from microsomal membranes prepared from McA-RH7777 cells (36). Immunoblotting showed that the inclusion of 0.025% DOC, which is well below its critical micellar concentration, did not solubilize the microsomal membranes as evidenced by the fact that the integral membrane protein, calnexin, was not extracted (data not shown). Therefore, these protocols can be applied to HepG2-derived microsomes. Quantification of apoB in membranes and lumen showed that the amount of apoB extracted by carbonate-supplemented DOC with or without KCl was similar (data not shown), and both protocols typically led to higher levels of extracted apoB compared with carbonate alone albeit the difference between these protocols was not statistically significant. Thus, an average of approximately 48 Ϯ 12 (range, 30 -60%) and 62 Ϯ 10% (range, 47-72%) (n ϭ 5) of the total pool of apoB was extracted by carbonate and carbonate/DOC, respectively. Therefore, either carbonate alone or carbonate/DOC with or without KCl can be used interchangeably with sufficient recovery of the luminal apoB pool for further characterization. Fig. 1A shows that apoB released either by carbonate or carbonate/DOC (lanes 1 and 3, respectively) was associated with the major ER-resident molecular chaperones including GRP94, ERp72, calreticulin, CyPB, and BiP (data not shown). Importantly, the level of co-immunoprecipitated chaperones was substantially higher than the corresponding controls (compare lanes 1 and 3 to lanes 2 and 4, respectively). Therefore, co-immunoprecipitated chaperones reflect apoB-chaperone interactions existing in living cells. Furthermore, following quantification of apoB and chaperone bands, we calculated the chaperone/apoB ratio and found that it was similar whether microsomes were extracted with carbonate or carbonate/DOC/K.
Next, it was of interest to determine the chaperone/apoB ratio in the lumen compared with the membrane-bound pool. In these series of experiments, however, we could not determine with certainty whether the spectrum and the level of molecular chaperones associated with these pools of apoB are comparable, because the efficiency of immunoprecipitated membrane-bound apoB was very low following solubilization of microsomal membranes by CHAPS. To substantially increase the efficiency of solubilization, membranes and luminal contents were heated to 65°C in the presence of 1% SDS prior to immunoprecipitation with antibodies to apoB. Fig. 1B shows that the same array of chaperones is associated both with the membrane-bound and the luminal pools of apoB (lanes 5 and 7, respectively), namely GRP94, ERp72, calreticulin, CyPB and BiP (data not shown). When antiserum to apoB was replaced with nonimmune serum, chaperone protein bands were diminished (lanes 6 and 8), strongly suggesting that the observed interactions of these chaperones with apoB exist in living cells. Notably, the presence of SDS in the immunoprecipitation buffer led to a substantial reduction in background, consistent with reduction of nonspecific ionic interactions between chaperone proteins and the protein G-Sepharose beads (lanes 6 and 8). Quantification of apoB and chaperone bands revealed that chaperone/apoB ratio in the lumen was at least 3-4-fold higher than that in the membranes. This was consistently observed for GRP94, ERp72, and CyPB, although the -fold increase varied between experiments. In contrast, calreticulin/apoB ratio in the lumen was essentially similar to that in the membranes. These findings therefore clearly demonstrate different roles for these chaperones in the folding and maturation of apoB. Thus, GRP94, ERp72, BiP, and CyPB presumably play a role not only in early folding events occurring cotranslationally while apoB is mem-  4). Luminal contents were separated from membranes, then adjusted to 2% CHAPS and subjected to immunoprecipitation with goat antiserum to human apoB (lanes 1 and 3) or nonimmune goat serum (lanes 2 and 4) followed by SDS-PAGE and Western blotting. B, cells were incubated as above and microsomes extracted with carbonate/DOC. Membranes (lanes 5 and 6) and luminal contents (lanes 7 and 8) were adjusted to 1% SDS in PBS and heated to 65°C for 15 min. The samples were then adjusted to 0.1% SDS with 1% Triton and subjected to immunoprecipitation either with antiserum to apoB (lanes 5 and 7) or goat nonimmune serum (lanes 6 and 8) followed by Western blotting. ApoB and chaperones were visualized by the ECL system, as described under "Experimental Procedures." M, membranes; L, lumen; NIS, nonimmune serum. brane-bound, but they also seem to facilitate the release of the lipidated intermediate from the membrane. One possibility is that they replace hydrophobic interactions between apoB and the ER membrane, thereby promoting its detachment from the membrane and providing stability to this intermediate once translocated into the lumen. Calreticulin on the other hand, being a lectin-like chaperone, may be involved in other aspects of apoB folding, so that no additional molecules are recruited to interact with the primordial intermediate prior to or after its release into the lumen. Nonetheless, calreticulin seems to play a role both during early and late folding events, as it also remains associated with the primordial intermediate after its translocation into the lumen.
What is the nature of the membrane-bound pool of apparently full-length apoB? Because these experiments were carried out under conditions that support optimal lipoprotein assembly (e.g. by providing oleic acid and inhibiting degradation by ALLN), this pool of apoB does not represent misfolded molecules. Rather, this pool is likely to represent the precursor of the primordial intermediate, which ultimately matures into secretion-competent lipoproteins with VLDL, IDL, and LDLlike densities. Being membrane-bound and having an apparent size of full-length apoB, this pool is presumably not fullytranslated, although very close to it (41). It follows that these molecules are almost fully lipidated as translation is closely coupled to lipidation in HepG2 cells (41). This pool, being in the process of translation, interacts both with the translocon and the ribosome (41). This could be partly responsible for a fraction of the pool to be nonextractable by either carbonate or carbonate/ DOC in protocols involving cross-linking of proteins. In addition, it is conceivable that domains enriched in ␤-strands predicted to form amphipathic ␤-sheets (42)(43)(44) interact with the luminal leaflet of the ER membrane until sufficient core lipids are recruited to form the primordial intermediate.
These experiments also demonstrated for the first time a direct interaction between CyPB and apoB. This suggests that CyPB is involved in the folding of apoB and may even be critical for its successful folding. This is based on studies showing that the immunosuppressant, cyclosporine A, prevents the secretion of apoB (45), concurrently with diminished interaction of CyPB with apoB. 2 Although cyclosporine A also binds to cyclophilin A, its cytosolic residence makes it an unlikely candidate chaperone to promote folding of apoB, which takes place in the ER.
In fact, it was demonstrated that, unlike the CFTR receptor (46), interaction of apoB with the cytosolic chaperones Hsp70 and Hsp90 leads to its targeting for degradation (20 -22). CyPB is both a folding catalyst belonging to peptidyl prolyl cis-transisomerases and a chaperone (47)(48)(49). Either activity or both may be important for the correct folding of apoB, because (a) apoB has a number of prolines that may be dependent on the peptidyl prolyl cis-trans-isomerase activity of CyPB for efficient folding, and (b) CyPB, like GRP94 and ERp72, seems to play a role in mediating the detachment of the primordial intermediate from the membrane. Further studies will be necessary to characterize the role of CyPB in apoB maturation in more detail.
Chaperones Interact with Lipidated ApoB with HDL-like Density Released into the Microsomal Lumen-The previous series of experiments (Fig. 1) demonstrated that molecular chaperones are associated with the luminal pool of apoB. However, these experiments did not provide direct evidence for the interaction between these chaperones and the lipidated pool of apoB. To directly demonstrate such interactions, microsomes prepared from cells following a 30-min pulse-labeling period were extracted with carbonate, and the soluble contents were subjected to KBr density gradient fractionation. This extraction protocol was chosen for two reasons: first, to allow for a direct comparison with previous studies involving careful characterization of apoB-100-containing particles in HepG2 cells (13); second, we determined that there was no qualitative difference in the density distribution of particles released by carbonate or carbonate/DOC (data not shown) consistent with other studies (36). Fig. 2A shows that approximately 80% of the newly synthesized pool of apoB was found in fractions having HDL-like density with a peak at 1.12 g/ml. This pool of apoB was associated with GRP94, ERp72 (panel C), as well as BiP, calreticulin, and CyPB (data not shown).
Because the association of chaperones with lipidated apoB is likely to increase the overall density of apoB-containing particles, it was necessary to ensure that the observed density distribution of apoB does not result from excessive cross-linking of proteins that are not otherwise associated with apoB in living cells. To that end, we performed another series of experiments in which cells were not subjected to cross-linking prior to microsomal preparation. These experiments showed that the density distribution of apoB present in the lumen was very similar to that obtained following cross-linking of cellular proteins (data not shown).  2. Lipidated apoB present in the microsomal lumen is associated with molecular chaperones and has a peak density of HDL-like lipoproteins, whereas secreted apoB fractionates with VLDL/IDL and LDLlike lipoproteins. HepG2 cells were incubated for 2 h with 0.8 mM OA, then pulse-labeled with [ 35 S]methionine/cysteine for 30 min and chased for 2h. Cells harvested after the pulse were subjected to cross-linking, and crude microsomes prepared and extracted with carbonate. Luminal contents and conditioned media collected after a 2-h chase were adjusted to 1.25 g/ml with solid KBr and subjected to density gradient centrifugation. ApoB in density fractions was immunoprecipitated with antibodies to apoB, resolved by SDS-PAGE, visualized by autoradiography, and quantified using Imagequant and expressed as percentage of total in lumen (A) and media (B). Chaperones associated with apoB in lumen were identified by Western blotting (C).
The size distribution of intracellular apoB-containing particles in HepG2 cells was analyzed by Bostrom et al. (13), who demonstrated that particles isolated from HDL-like densities (e.g. 1.065-1.170 g/ml) had a mean diameter of approximately 25 and 20 nm following pre-incubation of cells with or without oleate, respectively (13). This size is typical of plasma LDL particles, which contain apoB-100 as their only protein constituent. The fact that the mean density of these particles is in the HDL range is consistent with these particles having higher protein to lipid ratio compared with plasma LDL that is contributed by the associated chaperones. Indeed, these investigators suggested, among other possibilities, that perhaps the discrepancy between the size and the density may be due to the presence of proteins (unknown at the time), other than apoB on the lipoproteins isolated from the microsomal lumen (13). However, because molecular chaperones have a slow turn-over rate (e.g. several days), no additional proteins were detected following a short term radiolabeling protocol (13).
These series of experiments clearly demonstrated that apoB is released into the lumen as a lipidated intermediate with an HDL-like density that is still dependent on the assistance of molecular chaperones for its subsequent folding steps.
The density distribution of secreted apoB-containing particles is very different from that of the intracellular pool, as a significant fraction, approximately 44 and 34% is found in lower densities, e.g. d Ͻ 1.02 and 1.02 Ͻ d Յ 1.03 g/ml, respectively (Fig. 2B), clearly reflecting a higher lipid to protein ratio presumably because of both higher lipid content and lower protein content (caused by absence of chaperones). None is found in HDL-like densities, consistent with the finding that the particles are secreted free of chaperones (data not shown) and with the hypothesis that apoB-100 cannot attain a stable conformation when associated with particles having a lipidpoor core. This can be attributed to those domains starting at apoB-21 enriched in ␤-sheets (42,43), which presumably "prefer" to be bound to TAG because this is a more energetically favorable state (43,44). Therefore, it is conceivable that these domains are also most likely to be engaged in interaction with molecular chaperones, which stabilize them until enough lipids are recruited.
Chaperones Are Associated with Advanced Folding Intermediates of ApoB That Are Transported to the Golgi-The major folding steps of newly translated proteins that enter the secretory pathway take place in the ER, where resident chaperones and folding catalysts assist in their folding and assembly by direct physical interaction. Such interactions persist until proper folding is attained, whereby the affinity between chaperones and folding intermediates declines, ultimately leading to the release of chaperones (31,32). Correctly folded proteins are then transported to the Golgi for further modifications and for sorting to their target destination.
Because the folding of apoB is much more complex than most other secretory proteins and we have demonstrated that molecular chaperones remain associated with apoB following its translocation into the lumen as a lipidated intermediate (Fig.  2), it was of interest to determine at what stage of maturation chaperones dissociate from apoB. To that end, cell homogenates were fractionated on sucrose density gradients, apoB in subcellular fractions immunoprecipitated, and associated chaperones identified by Western blotting. Fig. 3A shows the distribution of marker proteins. Both Golgi markers, e.g. Golgin 97 (50), GPP130 (51), and p58 (52), and the ER marker calnexin (53) are shown. As evidenced from this figure, all Golgi markers are present in fraction 3. GPP130 demonstrated to reside in the cis/medial Golgi (51) is also present to a certain extent in fraction 2. On the basis of this analysis, fraction 3 represents cis/medial Golgi. This fraction also contains the trans-Golgi network (TGN) as it also contains the entire pool of sialylated apoB (data not shown).
The distribution of calnexin, an integral membrane protein of the ER, is quite different, as it is found predominantly in heavy microsomes. Thus, the ER fractionates with heavy microsomes. From here on, heavy microsomes and ER will be used interchangeably.
The distribution of luminal apoB is shown in panel B. As expected, a substantial fraction, approximately 40% is found in heavy microsomes, which include the ER. A smaller but sizable fraction, approximately 20%, was found in cis/medial Golgi (fraction 3). Such distribution suggests that, during its folding, apoB resides primarily in the ER demonstrated to be the site where its major folding steps take place (54). The presence of a sizable fraction of the intracellular pool of apoB in the Golgi is consistent with other reports demonstrating the presence of apoB in Golgi from HepG2 cells (55). Similarly, VLDL particles were also detected in Golgi from rat (56) and mouse liver (57). The presence of apoB in this compartment under steady-state conditions is consistent with modifications of apoB such as glycosylation and phosphorylation (58 -62), as well as possible remodeling by exchange of phospholipids between the Golgi membranes and the nascent particles (63,64).
Most interesting is the finding that apoB present in the Golgi compartment is associated with the same spectrum of molecular chaperones as those bound to apoB in the ER. These include GRP94, ERp72, calreticulin, BiP, and CyPB (panel C). These findings therefore demonstrate that apoB exits the ER in complexes containing the same chaperones assisting in its early folding stages. It is possible that other molecular chaperones such as protein disulfide isomerase (PDI) are also present in these complexes. Chaperone/apoB ratio within the subcellular compartments is shown in Fig. 3D. It appears that within the early secretory pathway (e.g. ER) the chaperone/apoB ratio is substantially higher. Thus, chaperones dissociate from the particles before transport to the Golgi. The relative amount is dependent on the specific chaperones. Thus, calreticulin/apoB ratio in Golgi is approximately half that in the ER, GRP94/, and ERp72/apoB ratios are approximately 30 and 20%, respectively, and CyPB/ and BiP/apoB decline to less than 10% (Fig.  3D). Thus, it appears that a substantial amount of chaperone molecules dissociate from apoB-containing lipoproteins prior to their arrival at the Golgi compartment consistent with apoB being in a more advanced folding stage. Nevertheless, the fact that a sizable fraction remains bound to apoB during and after its transport to the Golgi is consistent with different roles for these chaperones in late folding events occurring in post-ER compartments.
To ensure that the chaperones co-precipitated with the pool of apoB that is present in the Golgi interact with apoB in living cells, we performed two series of experiments.
In one series, cells were exposed to BMS-197636, a specific inhibitor of MTP that blocks the lipid transfer activity of MTP in vitro (65). As a result, apoB is unable to form lipoproteins and is rapidly degraded in a proteasome-dependent mechanism (17). Under these conditions it is expected that the major pool of apoB would be largely unfolded/misfolded and therefore primarily localized to the ER. Therefore, only very low levels of chaperones are expected to co-precipitate with antibodies to apoB in the Golgi. We found that this inhibitor reduced the amount of secreted apoB by greater than 80% (data not shown) concomitantly with a substantial reduction in the level of apoB in the Golgi compared with control levels (compare lane 1 with lane 3). Consistent with the diminished level of apoB in the Golgi, the level of GRP94 and ERp72 co-immunoprecipitated with apoB is also greatly reduced (compare lane 1 to lane 3). The reduced level of apoB in the ER lumen of cells treated with the inhibitor (compare lane 4 to lane 2) is consistent with diminished lipidation of apoB, thus preventing its translocation into the lumen.
Together, these findings demonstrate that co-precipitation of ER-resident molecular chaperones with apoB in the Golgi is dependent on the presence of apoB.
In the next series of experiments, cross-linking of cellular proteins was eliminated to minimize manipulations prior to fractionation and immunoprecipitation. To that end, un-crosslinked cells were homogenized and subjected to subcellular fractionation. Immunoprecipitation of luminal contents was carried out under native conditions, so that apoB-chaperone interactions were minimally disrupted. We used polyclonal antibodies to GRP94 and two monoclonal antibodies to apoB, 2D8 and B4, which bind to a domain encompassing residues 1438 -1480 (66) and 1854 -1878 (67), respectively. As shown in Fig. 5, antibodies to GRP94 co-immunoprecipitated apoB both in the Golgi (lane 9) and in the ER (lane 11), whereas control samples that were incubated with nonimmune serum co-precipitated substantially less apoB (lanes 10 and 12, respectively). Similarly, antibodies to apoB, B4, specifically co-immunoprecipitated GRP94 from Golgi and ER (lanes 5 and 7, respectively), whereas control samples precipitated only low levels of GRP94 (lanes 6 and 8, respectively). In a similar series of experiments, 2D8 and B4 antibodies co-immunoprecipitated calreticulin (data not shown), GRP94 and ERp72 from the Golgi (lanes 1 and 3, respectively) and the ER (lanes 2 and 4, respectively). Thus, the observed apoB-chaperone interactions following cross-linking of cellular proteins does not result from excessive cross-linking of cellular proteins, which are not otherwise in close contact with apoB, but rather reflect interactions that exist in living cells, and as long as the immunoprecipitation protocol does not require detergents, these interactions are not disrupted.
Thus, from the experiments shown in Figs. 4 and 5, we conclude that ER-resident molecular chaperones interact with apoB both in the ER and in the Golgi. We calculated the relative amount of apoB associated with GRP94 in the ER and Golgi from the amount of apoB immunoprecipitated with antibodies to GRP94 and the efficiency of the immunoprecipitation of GRP94. We found that 40 -60 and 25-35% of the total pool of apoB in the ER and the Golgi, respectively, is associated with GRP94 under steady-state conditions.
Interestingly, the pool of apoB that was immunoprecipitated from the Golgi (Fig. 5) (lanes 1 and 3) migrated slightly slower than its ER-derived counterpart (lanes 2 and 4) consistent with modifications of apoB occurring in a post-ER compartment, conceivably, the Golgi itself. These include modifications of the glycan residues, phosphorylation (62) or both. These differences between the pools of apoB were observed in a number of experiments and provide further support for efficient separation between the early secretory compartment, the ER, and the Golgi, and for the presence of distinct pools of apoB within these compartments. These findings, therefore, strongly suggest that apoB-molecular chaperone complexes exist within the Golgi compartment in living cells, excluding the possibility that such complexes are derived from vesicles originating in the ER.
Overall, these studies clearly demonstrate that apoB is transported to the Golgi as part of complexes containing ERresident molecular chaperones. The presence of different levels of chaperones compared with the ER suggests distinct roles for these chaperones in late folding events occurring in the Golgi.
The Major Pool of ApoB Present in the Golgi and in Heavy Microsomes Is Lipidated and Is Associated with the Same Array of Molecular Chaperones-The next series of experiments was designed to determine whether the chaperone-associated pool of apoB present in the Golgi is lipidated. Fig. 6A shows that the FIG. 4. MTP inhibitor changes the subcellular distribution of apoB and associated chaperones. HepG2 cells were incubated for 2 h with OA and ALLN without (lanes 1 and 2) or with 100 nM MTP inhibitor, BMS-197636 (lanes 3 and 4). Cells were then processed as in Fig. 3. ApoB and associated chaperones were identified by immunoblotting. G, Golgi. majority of the apoB pool in the ER is found in 1.08 Ͻ d Ͻ 1.2 g/ml, whereas that in the Golgi is found at lower densities, 1.02 Ͻ d Ͻ 1.08 g/ml. Nevertheless, apoB present in this lower density range is still associated with the same array of chaperones associated with the pool of apoB recovered from the ER, including GRP94, ERp72, BiP, calreticulin, and CyPB (Fig.  6B). The lower density of apoB in the Golgi reflects a higher lipid to protein ratio. This could be caused by either an increase in the relative content of lipid, decrease in relative content of protein without change in lipid content, or both. Increase in lipid content can occur by recruitment of neutral lipids in post ER pre-Golgi compartments, or in the Golgi itself. Such a model is consistent with continuous recruitment of lipids in these cells rather than a two-step model demonstrated in liver and McA-RH7777 cells (3)(4)(5)(6)(7)(8). The release of a substantial fraction of chaperones from apoB-containing particles prior to exit of particles from the ER (Fig. 3D) may partially account for the shift in the density of the particles in the Golgi with no substantial addition of lipid. To test this possibility, we estimated the content of lipid in the ER and Golgi.
The relative amount of lipid and protein was calculated from the density of the particle, which equals to the sum of protein and lipid density using densities of 1.33 and 0.94 g/ml for protein and lipid, respectively.
Using average densities of 1.16 and 1.05 g/ml for particles in the ER and Golgi, respectively, we calculated that the fraction of protein and lipid in particles in the ER is 0.56 and 0.44, and in the Golgi 0.28 and 0.72, respectively.
The molecular mass of protein is the sum of the mass of apoB and associated chaperones. To calculate that, we made two alternative assumptions. Particles in the Golgi have at least one molecule of each of the identified chaperones (e.g. GRP94 as a homodimer, ERp72, BiP, calreticulin, and CyPB) per particle (or per apoB molecule). Alternatively, there are two chaperone molecules per particle. According to the first assumption, the mass of protein in Golgi-derived particles is approximately 0.95 ϫ 10 6 daltons. The mass of protein in ER-derived particles where the chaperone/apoB ratio is severalfold higher, was calculated from the data presented in Fig. 3D and found to be approximately 2.52 ϫ 10 6 daltons. From the lipid to protein ratio, the molecular mass of lipid in particles derived from Golgi and ER is 2.44 and 1.98 ϫ 10 6 daltons, respectively. Thus, FIG. 6. Density distribution of apoB in Golgi and ER. Cells were incubated, and processed as in Fig. 3. Golgi and ER were extracted with carbonate/DOC/K and luminal contents adjusted to 1.25 g/ml with solid KBr and subjected to density gradient centrifugation. Twelve fractions were collected from the top and density measured as described under "Experimental Procedures." Fractions were combined as follows: fraction 1, d Յ 1.02 g/ml; fraction 2, 1.02 Ͻ d Ͻ 1.08 g/ml; fraction 3, 1.08 Ͻ d Ͻ 1.21 g/ml; and fraction 4, d Ͼ 1.21 g/ml. ApoB in combined fractions was immunoprecipitated and quantified by densitometry as described in Fig. 3. A, distribution of apoB in density fractions expressed as percentage of total apoB in Golgi (squares) and ER (circles). B, an immunoblot showing apoB and associated molecular chaperones in Golgi (left) and ER (right).

FIG. 5. Molecular chaperones co-immunoprecipitate with apoB in Golgi and ER without prior cross-linking of cellular proteins.
Cells were incubated and processed as in Fig. 3, except that cellular proteins were not cross-linked prior to subcellular fractionation. Luminal contents of Golgi (lanes 1, 3, 5, 6, 9, and 10) and ER (lanes 2, 4, 7, 8, 11, and 12) fractions obtained by carbonate extraction were neutralized and subjected to immunoprecipitation with monoclonal antibodies to apoB, 2D8 (lanes 1 and 2) or B4 (lanes 3-5 and 7), or with rabbit antiserum to GRP94 (lanes 9 and 11). Control samples were incubated with nonimmune mouse IgGs (lanes 6 and 8) or nonimmune rabbit serum (lanes 10 and 12) under native conditions. Immunocomplexes were resolved on SDS-PAGE and apoB and molecular chaperones identified by immunoblotting. Three representative experiments are shown: experiment 1 (lanes 1-4), experiment 2 (lanes 5-8), and experiment 3 (lanes 9 -12). In experiment 1, all lanes are from the same exposure time. In experiments 2 and 3, Golgi and ER should not be compared directly as they are from different exposure times. Samples immunoprecipitated with antibodies and the corresponding controls within the same compartment are from the same exposure time and can be compared with each other. G, Golgi; E, ER; NI, nonimmune (IgG or serum).
according to the first assumption, lipid content in Golgi-derived particles is approximately 23% higher than that in ER-derived particles.
If calculations are made assuming two chaperones per particle, the mass of protein in Golgi and ER-derived particles is 1.36 and 4.52 ϫ 10 6 daltons, respectively, and that of lipid is 3.50 and 3.55 ϫ 10 6 daltons, respectively. According to the second assumption, the content of lipid in Golgi-derived particles is similar to that of ER-derived particles. Thus, it is possible that changes in the amount of chaperones per particle can have a noticeable effect on the density of the particle. Although it is conceivable that the number of individual chaperone molecules per particle varies, the calculations shown above give a reasonable estimate of the actual number.
The size of the particles in the ER and the Golgi was calculated from the anhydrous volume (V) of the particle shown in Equation 2.
V particle ϭ 1 ϫ mass particle ϫ 10 24 /d ϫ N͑Avogadro's number͒ (Eq. 2) Assuming a spherical particle, the radius (r) can be calculated from the following equation. The calculated size of particles in the ER and Golgi is 230 and 210 Å diameter, respectively, assuming one chaperone molecule/apoB in Golgi or 280 and 244 Å, respectively, assuming two chaperone molecules/apoB in Golgi. Because these calculations are based on the volume of an anhydrous sphere, the actual size of these particles is approximately 20% larger.
Clearly, the size and the density of the particles reflect the fact that there are proteins other than apoB on these particles. According to these calculations, particles in the Golgi seem slightly smaller than those in the ER. This may be a result of the fact that the shape of the particles is not a perfect sphere; therefore, the calculations based on a spherical shape may deviate to a certain extent from the actual size. This is probably more pronounced for particles in the ER, which appear to contain severalfold more chaperone proteins compared with Golgi-derived particles. Nonetheless, these calculations give a good estimate of the true size of the hydrated particles, particularly as they are in agreement with those measured by negative-stain electron microscopy (13). Thus, under steady-state conditions, particles of at least LDL or IDL/VLDL size are present both in the ER and in the Golgi. This may indicate that in HepG2 cells the major recruitment of lipids occurs in the ER.
Notably, the density of apoB-containing particles in the Golgi is still substantially higher than that of the secreted particles, 40% of which are found in d Ͻ 1.02 g/ml (Fig. 2B). This is presumably because of associated chaperones in the Golgi. In the search for proteins that might be involved in the second step of VLDL assembly, Stillmark et al. (28) found that PDI, GRP94, ERp72, BiP, and calreticulin co-fractionated with apoB-48-VLDL from rat liver. However, the subcellular localization of these particles was not determined.
GRP94 Is Associated with ApoB That Had Been Transported to the TGN-The previous series of experiments clearly demonstrated that ER-resident molecular chaperones are associated with apoB that had been transported to the Golgi complex. However, it was unclear in which compartment within the Golgi these chaperones dissociate from apoB-containing particles. We were interested in determining whether molecular chaperones interact with apoB that had been transported to distal compartments of the secretory pathway, such as the TGN. To do that, we took advantage of the fact that the enzyme sialyltransferase resides in the TGN, and used lectin blotting to identify sialylated apoB. In the experiment depicted in Fig. 7, luminal contents were subjected to immunoprecipitation with antibodies to GRP94, followed by lectin and immunoblotting. Membranes were incubated either in buffer alone (lanes 1 and 2) or buffer containing neuraminidase (lanes 3 and 4). Membranes were then probed with the lectin SNA-I that preferentially binds to sialic acid attached to terminal galactose in ␣-2,6 linkage. Bound SNA-I was visualized following binding of HRP-avidin. A clear band of SNA-I was observed in the sample that was immunoprecipitated with antibodies to GRP94 and was incubated in buffer alone (lane 1), whereas only a faint band was seen in the control sample in which antibodies were replaced by nonimmune serum (lane 2). As expected, incubation of samples with neuraminidase prevented the binding of SNA-I (lane 3), consistent with removal of terminal sialic acid residues by this enzyme. Thus, the binding of SNA-I reflects the presence of sialic acids. Probing of these membranes with antibodies to apoB confirmed that the bands identified with SNA-I represent apoB-100. These findings, therefore, are consistent with the notion that at least a fraction of the pool of apoB that was modified by enzymes residing in the TGN was associated with GRP94. Similar results were obtained using antibodies to ERp72 (data not shown). From the relative amount of the sialylated pool of apoB that was co-precipitated with antibodies to GRP94 and from the efficiency of GRP94 immunoprecipitated with these antibodies, we estimated that approximately 15-25% of the sialylated pool of apoB is associated with GRP94. Thus, at least GRP94 and possibly ERp72 remain associated with apoB even after it had been transported to the TGN.
Given the role of ER-resident molecular chaperones in primary quality control in the ER (31,34), the question remains as to what is the role of their prolonged interaction with apoB well beyond the ER? It is generally thought that these chaperones assist in the folding of nascent polypeptides in the ER. The progression of the folding process is very closely monitored by these chaperones so that successful folding leads to the release of the properly folded nascent proteins, which then exit the ER and move to the Golgi free of chaperones. However, if correct folding is not achieved, misfolded proteins are retained in the ER bound to these chaperones, which prevent their transport to their target destination. This mechanism is very efficient in mammalian cells, so that only a very small fraction of misfolded proteins escapes this primary quality control (31,34). Backup mechanisms beyond the ER (provided by a different set of proteins; Ref. 31) ensure that those that escaped the ER are prevented from delivery to their destination. Consistent with such a tight FIG. 7. GRP94 is associated with apoB that has gone through TGN. Cells were incubated with OA and ALLN and then homogenized without prior cross-linking of proteins, and microsomes were prepared and extracted with carbonate. Luminal contents were then subjected to immunoprecipitation with rabbit antibodies to GRP94 (lanes 1 and 3) or nonimmune serum (lanes 2 and 4) and immunocomplexes in duplicates resolved by SDS-PAGE and electrotransferred onto a membrane. Membrane was blocked with albumin, and then one half was incubated with buffer alone (lanes 1 and 2) and the other half with buffer containing neuraminidase (lanes 3 and 4) for 20 h at 37°C. Both halves were then probed sequentially with the lectin SNA-I followed by antibodies to apoB. Bound lectin was visualized by extravidin-HRP. NIS, nonimmune serum. quality control mechanism, it was demonstrated that apoB can be targeted for degradation at different stages of its maturation beyond the ER (15,40,55,68,69). This, however, is an unlikely scenario occurring in HepG2 cells under the conditions used in this study, because the cells were incubated under conditions that support optimal folding of apoB resulting in a very efficient secretion. A more likely explanation for the prolonged interaction between ER-resident chaperones and apoB is that the maturation of apoB to form VLDL/IDL is a multistep process that begins in the ER and continues throughout the secretory pathway including distal compartments of the Golgi, as demonstrated by a number of investigators (56,57,63,64,70). These include modifications of apoB such as glycosylation and phosphorylation, remodeling of the phospholipid on the surface of the particles, and possibly continued lipidation (71,72). All of these may tag apoB as an incompletely folded (e.g. it is on a path for successful folding), rather than a misfolded protein, therefore requiring assistance of ER-resident chaperones for its stabilization. Molecular chaperones could also facilitate the transport of apoB-containing particles through the secretory pathway, consistent with a proposal that chaperones are also involved in protein trafficking, as demonstrated for Drosophila whereby the cyclophilin homologue, ninaA, appears to be required for rhodopsin Rh1 as it travels through the distal compartments of the secretory pathway (73), whereas CyPB and HSP47 were suggested to play a role in export of procol-lagen from the ER as they were found associated with procollagen in the intermediate compartment (74).
ApoB-associated chaperones could dissociate just before apoB VLDL/IDL particles are packaged into secretory vesicles, consistent with the finding that the pool of apoB recovered in the light Golgi fraction (Fig. 3, fraction 2) is not associated with chaperones. This fraction presumably contains secretory vesicles (37).
If, however, apoB fails to undergo complete maturation either in the early or in distal compartments of the Golgi, these chaperones may then target it for degradation.
In summary, we demonstrated the following. (a) CyPB directly interacts with apoB and therefore plays a role in folding and maturation of apoB. (b) all five major ER-resident molecular chaperones and folding catalysts (GRP94, ERp72, BiP, CRT, and CyPB) interact with the membrane-bound pool of apoB conceivably to keep it in a conformation competent for MTP-mediate lipidation. Except for CRT, these chaperones seem to facilitate the translocation of lipidated apoB into the microsomal lumen and to provide stability to the primordial intermediate in the lumen. (c) ApoB is transported to the Golgi in complexes containing the same molecular chaperones as those interacting with the luminal pool of apoB in the ER except that chaperone/apoB ratio is lower. Thus, apoB leaves the ER as an incompletely folded intermediate. (d) Because GRP94 interacts with apoB that has gone through the TGN it is conceivable that this chaperone (and perhaps others) plays a FIG. 8. A proposed model for chaperone-assisted folding of apoB to form lipoproteins. Molecular chaperones interact with apoB as it emerges through the translocon (shown in red) into the ER lumen. Following elongation, domains which are presumably enriched in amphipathic ␤-sheets interact with the membrane (43,44) cotranslationally, where initiation of MTP-mediated lipidation leads to the formation of a small lipid core (shown in yellow) (step I). Ongoing translation-coupled lipidation leads to elongation, enlargement of the core, and recruitment of additional molecular chaperones (step II). Recruitment of additional neutral lipid molecules leads to completion of translation (41), followed by dissociation of the ribosome (step III). At this stage enough lipids had been recruited and the lipidated intermediate begins to detach from the ER membrane. It appears, however, that for it to completely detach, additional chaperone molecules, GRP94, ERp72, and CyPB, but not CRT, are recruited to facilitate the release presumably by replacing the hydrophobic interactions with the membrane and to stabilize the primordial intermediate following its release into the lumen (step IV). This intermediate recruits more core lipids to form larger particles (step V). Prior to ER exit, a fraction of the bound chaperones dissociates (approximately 50 -80% of CRT, GRP94, ERp72, and calreticulin, and over 90% of BiP and CyPB), and the lipoprotein particles are transported to the Golgi (step VI). ER-resident chaperones remain associated with apoB-containing lipoproteins in the Golgi while additional modifications take place (step VII). Ultimately, all chaperones dissociate presumably before packaging into secretory vesicles (step VIII) and nascent VLDL is secreted. role in late folding events in the maturation of apoB occurring in distal compartments of the secretory pathway.
On the basis of the data presented in this report, we propose a model (Fig. 8) showing that the maturation of apoB is dependent not only on efficient MTP-mediated lipidation, but also on other ER-resident molecular chaperones that seem to play multiple roles in the biogenesis of apoB to form TAG-rich lipoproteins occurring throughout the secretory pathway, including the ER and distal Golgi compartments, contrary to current ideas assigning a role for molecular chaperones primarily in folding events occurring in the ER.