The assembly and secretion of apolipoprotein B-48-containing very low density lipoproteins in McA-RH7777 cells.

We have used an extraction procedure, which released membrane-bound apoB-100, to study the assembly of apoB-48 VLDL (very low density lipoproteins). This procedure released apoB-48, but not integral membrane proteins, from microsomes of McA-RH7777 cells. Upon gradient ultracentrifugation, the extracted apoB-48 migrated in the same position as the dense apoB-48-containing lipoprotein (apoB-48 HDL (high density lipoprotein)) secreted into the medium. Labeling studies with [(3)H]glycerol demonstrated that the HDL-like particle extracted from the microsomes contains both triglycerides and phosphatidylcholine. The estimated molar ratio between triglyceride and phosphatidylcholine was 0.70 +/- 0.09, supporting the possibility that the particle has a neutral lipid core. Pulse-chase experiments indicated that microsomal apoB-48 HDL can either be secreted as apoB-48 HDL or converted to apoB-48 VLDL. These results support the two-step model of VLDL assembly. To determine the size of apoB required to assemble HDL and VLDL, we produced apoB polypeptides of various lengths and followed their ability to assemble VLDL. Small amounts of apoB-40 were associated with VLDL, but most of the nascent chains associated with VLDL ranged from apoB-48 to apoB-100. Thus, efficient VLDL assembly requires apoB chains of at least apoB-48 size. Nascent polypeptides as small as apoB-20 were associated with particles in the HDL density range. Thus, the structural requirements of apoB to form HDL-like first-step particles differ from those to form second-step VLDL. Analysis of proteins in the d < 1.006 g/ml fraction after ultracentrifugation of the luminal content of the cells identified five chaperone proteins: binding protein, protein disulfide isomerase, calcium-binding protein 2, calreticulin, and glucose regulatory protein 94. Thus, intracellular VLDL is associated with a network of chaperones involved in protein folding. Pulse-chase and subcellular fractionation studies showed that apoB-48 VLDL did not accumulate in the rough endoplasmic reticulum. This finding indicates either that the two steps of apoB lipoprotein assembly occur in different compartment or that the assembled VLDL is transferred rapidly out of the rough endoplasmic reticulum.

Immunoelectron microscopy studies have shown that apolipoprotein (apo) 1 B is present in the rough endoplasmic reticulum (ER), but very low density lipoprotein (VLDL)-sized particles are not (1). VLDL particles with immunoreactive apoB first appeared in the smooth termini of the rough ER; the smooth ER contained VLDL-sized particles without immunoreactive apoB (1). Based on these results, a two-step model for the assembly of VLDL was proposed. Dynamic evidence for this model was obtained by pulse-chase studies of apoB-100 and apoB-48 (2,3). The first step occurs during the translation of apoB and gives rise to a partially lipidated form of apoB (2,4). In the case of apoB-100, this partially lipidated particle appeared to be loosely associated with the ER membrane (5). In the case of apoB-48, a particle resembling high density lipoprotein (HDL) has been identified. The secretion of this dense, apoB-48-containing, HDL-like lipoprotein varied inversely with that of VLDL (2). Therefore, we hypothesized that this particle is a precursor of apoB-48 VLDL (2). A second VLDL precursor was identified as an apoB-free "lipid droplet" in the smooth ER (1,6). The assembly of both precursors is dependent on the microsomal triglyceride transfer protein (5,7,8).
The mechanism for the second step, fusion of the two precursors (1), is less well understood. We have demonstrated that brefeldin A inhibits the major lipidation of apoB (9). However the exact localization of the brefeldin A-sensitive mechanism in the assembly pathway remains to be elucidated.
Cotranslational or early post-translational degradation of apoB is important in regulating the amount of apoB that passes through the first step (10,11). This degradation involves ubiquitination and proteasomes (11). Recent results indicate that apoB is completely translocated to the lumen of the ER (12), suggesting that the early post-translational degradation follows the pathway described for misfolded proteins (i.e. the protein is retracted through the translocation channel) (13). However, there is strong evidence that the degradation involves nascent as well as full-length apoB chains (14,15), suggesting that this proteasomal degradation may also involve other pathways (for reviews, see Refs. 16 and 17).
Studies of VLDL assembly have been hampered by the fact that much of the apoB present in the cell remains associated with the microsomal membrane after carbonate extraction of the luminal proteins and therefore cannot be analyzed. Recently, we developed a procedure that extracts virtually all of the apoB-100 from the microsomal membranes without releasing integral membrane proteins (5). In this study, we used this new extraction procedure in a series of experiments to analyze the assembly of apoB-48 VLDL.

EXPERIMENTAL PROCEDURES
Materials-Eagle's minimum essential medium, nonessential amino acids, glutamine, penicillin, and streptomycin were obtained from ICN Biomedicals (Costa Mesa, CA). Fetal calf serum was from Biochrom KG (Berlin, Germany) and brefeldin A from Epicenter Technologies (Madison, WI). Methionine, fatty acid-free bovine serum albumin, sodium pyruvate, disodium carbonate, sodium hydrogen carbonate, phenylmethylsulfonyl fluoride, pepstatin A, and leupeptin were from Sigma. Rabbit immunoglobulin was from Dako (Glostrup, Denmark), and rabbit anti-rat transferrin IgG was from Organon Teknika (West Chester, PA). Trasylol (aprotinin) was from Bayer (Leverkusen, Germany). Immunoprecipitin and Eagle's minimum essential medium without methionine were from Life Technologies, Inc. N-Acetyl-Leu-Leu-norleucinal as well as enzymatic assays for the determination of phospholipids or triglycerides were from Boehringer Mannheim. Amplify, [ 35 S]methionine/cysteine mix, Rainbow protein molecular weight markers, and the ECL Western blotting analysis system were from Amersham Pharmacia Biotech, and Ready-Safe was from Beckman (Fullerton, CA). All chemicals used for SDS-polyacrylamide gel electrophoresis (SDS-PAGE) and alkaline phosphatase-conjugated goat anti-rabbit and rabbit anti-mouse were from Bio-Rad (Hercules, CA). Blue-stabilized substrate for alkaline phosphatase and trypsin (sequencing grade) were from Promega (Milwaukee, WI). Cyanogen bromide-activated Sepharose 4B was from Amersham Pharmacia Biotech, and ␣-cyano-4-OH cinnamic acid was from Aldrich (Milwaukee, WI). Antibodies to chaperones (binding protein, protein disulfide isomerase, glucose regulatory protein 94, and calreticulin) were purchased from Affinity BioReagents (Golden, CO).
Metabolic Labeling-The cells were pulse labeled and chased as described (2). Cells and the microsomal fraction were isolated as described (18). The luminal content of the vesicles was separated from the vesicle membranes by the sodium carbonate method (19), as modified (2). The following protease inhibitors were used: 0.1 mM leupeptin, 1 mM phenylmethylsulfonyl fluoride, 1 mM pepstatin A, 5 mM N-acetyl-Leu-Leu-norleucinal, and aprotinin (100 kallekrein-inhibitory units/ml). In some experiments, the luminal content of the microsomal vesicles was extracted with the deoxycholate/carbonate procedure described recently by Rustaeus et al. (5).
Subcellular Fractionation-The total microsomal fraction was obtained (18) and fractionated on a sucrose gradient. A 3.8-ml linear sucrose gradient (32.5-40%, w/v) was layered on a cushion of 0.5 ml of 65% sucrose (w/v), and the sample was layered on top of the gradient. All solutions contained 3 mM imidazole, pH 7.4, with the same protease inhibitors as used for metabolic labeling. The gradients were centrifuged in a Beckman Vti-65.2 vertical rotor at 50,000 rpm for 3 h at 12°C. The gradient was unloaded from the bottom into 22 fractions. NADPH cytochrome c reductase and galactosyl transferase were used as marker enzymes for the ER and the Golgi apparatus, respectively (20). Gradient fractions were also assayed by Western blot for calnexin, a marker for the rough ER.
Sucrose Gradient Ultracentrifugation-Lipoproteins in the microsomal lumen or in the medium were separated by sucrose gradient ultracentrifugation (2). The gradient was formed by layering, from the bottom of the tube, 2 ml of 49% sucrose, 2 ml of 25% sucrose, 5 ml of the sample in 12.5% sucrose (the sucrose was in phosphate-buffered saline), and 3 ml of phosphate-buffered saline. All solutions contained 0.1 mM leupeptin, 1 mM phenylmethylsulfonyl fluoride, 1 mM pepstatin A, 5 mM N-acetyl-Leu-Leu-norleucinal, aprotinin (100 kallekrein-inhibitory units/ml), and 0.5 mM EDTA. The gradients were centrifuged in a Beckman SW40 rotor at 35,000 rpm for 65 h at 12°C and unloaded from the bottom of the tube into 12-13 fractions.
Immunoprecipitation and Immunoaffinity Chromatography of ApoB and Electrophoresis-ApoB was immunoprecipitated from the cells, medium, and sucrose gradient fractions as described (2,21). Immunoaffinity chromatography of apoB-containing fractions extracted from the microsomes or present in the culture medium was carried out as described (18). SDS-PAGE, autoradiography, and determination of the radioactivity in proteins separated in the gels were performed as detailed elsewhere (21).
Lipid Determination-Lipids from McA-RH7777 cells were extracted as described by Olegård and Svennerholm (22) with slight modifications (23). Phosphatidylcholine and triglycerides were separated by thin layer chromatography with chloroform:methanol:water (65:25:4 v/v) followed by petroleumether:diethylether:acetic acid (80:20:1 v/v). The spots were visualized by iodine, scraped off, and extracted with 1 ml of chloroform:methanol 1:2 (phosphatidylcholine) or chloroform (triglycerides). The extracted lipids were dried under nitrogen in a conical tube, solubilized in 20 l of ethanol, and analyzed by enzymatic assays. As standards, pure phosphatidylcholine and triglycerides were chromatographed and processed in parallel with the sample. The recovery of triglycerides and phosphatidylcholine during the chromatography and extraction steps was 94 Ϯ 16% (mean Ϯ S.D.; n ϭ 5) and 65 Ϯ 4% (n ϭ 5). The recovery of triglycerides was also tested with radioactive tracer added to the cell homogenate; this experiment showed a recovery of 94 Ϯ 8% (n ϭ 4). The intra-assay variation was 4.5% for triglycerides and 8.1% for phosphatidylcholine.
Phosphatidylcholine and triglycerides were radiolabeled by incubating the cells for various periods with [ 3 H]glycerol (0.6 Ci/ml of culture medium). Cells were extracted, phosphatidylcholine and triglycerides were separated as described above, and specific radioactivity was determined (dpm/mg). In some experiments, apoB-containing lipoproteins were isolated by immunoaffinity chromatography from the luminal content or the medium. During the extraction of the lipids from these fractions, unlabeled phosphatidylcholine and triglycerides were added as carriers. The lipids were separated as described above, and the spots corresponding to triglycerides and phosphatidylcholine were scraped into scintillation vials; 1 ml of cyclohexane was added, and the radioactivity was determined in the presence of Ready-Safe scintillation mixture.
Isolation and Characterization of Proteins-Proteins associated with microsomal lipoproteins were isolated and identified as follows. Rat liver microsomes were isolated (23), and the luminal content was extracted with sodium carbonate (19). The extract (6 ml) was overlayered with 29 ml of phosphate-buffered saline (8 mM disodium hydrogen phosphate, 1.5 mM potassium dihydrogen phosphate, 137 mM sodium chloride, and 2.7 mM potassium chloride, pH 7.4, d ϭ 1.006 g/ml). After centrifugation for 22 h at 40,000 rpm in a Beckman Ti-60 rotor at 4°C, the gradients were fractionated from the top. The upper one-third of the tube (d Ͻ 1.006 g/ml) was collected. Pooled fractions of this supernatant (corresponding to three to five rat livers) were loaded onto a Mono Q column equilibrated with 50 mM Tris-HCl, pH 7.8, with 300 mM sucrose, 1 mM EDTA, 2 mM deoxycholate, 0.5% Triton X-100, 6 M urea, and 40 mM sodium carbonate. The column was eluted with a linear gradient of sodium chloride (0 -250 mM) at a flow rate of 0.5 ml/min. Fractions (0.5 ml) were collected, and the proteins in each fraction were separated by SDS-PAGE on 10% gels. Gels were stained with silver.
Fractions containing the same protein patterns were combined, concentrated, and subjected to SDS-PAGE on 3-15% gradient gels. The gels were stained with Coomassie Brilliant Blue, and the bands were cut out, destained with 50 l of a mixture of 50% ammonium bicarbonate (25 mM) and 50% acetonitrile, dried, and digested for 15 min with 0.1-0.2 mg of trypsin in 20 l of 50% ammonium bicarbonate (25 mM) and 50% acetonitrile. Ammonium bicarbonate (25 mM, pH 8) was added to cover the gels, and incubation was continued for 12 h at 37°C. Fragments were extracted with 10 -50 l of a mixture of 75% acetonitrile and 5% trifluoroacetic acid (in water).
Mass spectra were obtained on a TofSpec-E time-of-flight mass spectrometer (Micromass; Manchester, UK) equipped with a time-lagged focusing unit; TOF2UI version 3.4 was used for data collection and OPUS version 3.4 for data analysis. ␣-Cyano-4-OH cinnamic acid (10 mg/ml in water/acetonitrile, 50/50, v/v) was used as matrix without further purification. The ␣-cyano-4-OH cinnamic acid solution (0.5 ml) was mixed with the gel extract (0.5 ml) on the target and allowed to dry at room temperature. Spectra were collected in reflectron mode at an accelerating voltage of 20 kV with a 600-ns delayed extraction and a pulse of approximately 2.4 kV. Approximately 200 nitrogen laser pulses (3 ns, 337 nm) were carried out on each sample. For external calibration, a mixture of ACTH and angiotensin II was used (protonated 2465.2 and 1046.5, respectively). The peptide mass fingerprinting software program MS-Fit was run over the Internet. Monoisotopic masses were used for the searches; mass tolerance was Ϯ 200 ppm.
Generation of a Monoclonal Antibody against Rat Riboforin-To determine the effects of deoxycholate/carbonate extraction (5) on integral microsomal proteins, we first generated a monoclonal antibody against riboforin. BALB/c mice were immunized with solubilized microsomal membrane proteins from rat liver (23). Positive hybridomas were identified by enzyme-linked immunosorbent assay with the antigen (solubilized rat liver microsomes) and analyzed by Western blot of solubilized rat liver microsomes. One hybridoma reacted with a 60-kDa protein and was recloned to monoclonality. The immunoglobulins were isolated from the hybridoma culture medium and coupled to cyanogen bromide-activated Sepharose 4B as recommended by the manufacturer (Amersham Pharmacia Biotech) and used for immunoadsorption experiments (23). Using this immunoadsorbent, we recovered the 60-kDa protein that reacted with the monoclonal antibody. This protein was cut out of the Coomassie-stained gel, digested with trypsin, and analyzed by mass spectrometry as described above. A data search identified the protein as rat riboforin.
Upon gradient ultracentrifugation, the major amount of apoB-48 extracted by the deoxycholate/carbonate method migrated in the HDL density range (Fig. 1B). Using a modified gradient, we compared the densities of secreted apoB-48 and deoxycholate/carbonate-extracted apoB-48. The major amount of the secreted apoB-48 was present in the VLDL and the HDL density regions, as described previously (2). ApoB-48 that banded in the HDL density region (apoB-48 HDL) migrated in the same position as the apoB-48 extracted from the microsomes (Fig. 1C); we will refer to this form of apoB-48 as intracellular apoB-48 HDL. Thus, intracellular apoB-48 HDL and secreted apoB-48 HDL have very similar buoyant densities, and each migrated in the gradient in the expected position for a lipoprotein (in comparison with a nonlipidated protein of similar molecular weight) (Fig. 1C, III). Thus, the membraneassociated apoB-48, like apoB-100 (5), was extracted from the microsomes as a tentative lipoprotein.
To determine if intracellular apoB (both B-100 and B-48)containing HDL is associated with lipids and contains a lipid core, we began by estimating the incubation time needed to obtain steady-state labeling of the phosphatidylcholine and triglyceride pools of the cell. The cells were incubated with [ 3 H]glycerol (0.6 Ci/ml of culture medium) for 0, 1, 2, 5, 8, and 22 h. After each incubation, phosphatidylcholine and triglycerides were isolated, and the specific radioactivity (dpm/mg) was determined. Triglycerides reached a plateau after 8 h; phosphatidylcholine plateaued between 1 and 8 h, after which the specific radioactivity decreased (data not shown). We therefore incubated the cells for 8 h with [ 3 H]glycerol (0.6 Ci/ml culture of medium). The specific radioactivities of the total phosphatidylcholine (4,025 dpm/g of lipid) and triglyceride (2,044 dpm/g of lipid) pools of the cell were determined. The apoBcontaining lipoproteins in the HDL density region of the deoxycholate/carbonate extract of the microsomes were isolated by immunoaffinity chromatography, and the radioactivity in phosphatidylcholine and triglycerides was determined. Assuming that the specific radioactivity of the glycerolipids in this apoB fraction was the same as that of the total cell, we estimated the weight ratio between triglycerides and phosphatidylcholine in intracellular apoB HDL to be 0.85 Ϯ 0.15 (n ϭ 3; molar ratio, 0.70 Ϯ 0.09), indicating less triglyceride than phospholipid. Thus, intracellular apoB HDL has an immature lipid core.
The labeled cells were also chased for 2 h, and apoB HDL in the medium was isolated by gradient ultracentrifugation followed by immunoaffinity chromatography. Analyzed as described above, the weight ratio between triglycerides and phosphatidylcholine was 0.27 Ϯ 0.10 (n ϭ 3; molar ratio 0.22 Ϯ 0.08), indicating that this particle has a lipid core.
Next, we performed pulse-chase experiments to follow the turnover of intracellular apoB-48 HDL and correlated the find-ings with VLDL assembly and the appearance of apoB-48 HDL in the medium. The cells were pulse labeled with [ 35 S]methionine/cysteine for 10 min and chased for 0 -120 min. Radioactive apoB-48 was first seen in the microsomal intracellular apoB-48 HDL (Fig. 2). Not until maximal apoB-48 radioactivity was reached in this fraction did any significant amount of apoB-48 radioactivity appear in the VLDL fraction. In fact, the decrease in the apoB-48 radioactivity which followed this maximum accounted for the increased radioactivity in apoB-48 VLDL and in secreted apoB-48 HDL.
VLDL Assembly and the Size of Nascent ApoB Polypeptides-To determine the length of apoB required for VLDL assembly, we performed pulse-chase studies of nascent apoB chains. To obtain a continuous series of apoB polypeptides of different lengths which could be tested in the assembly process, we truncated apoB with cycloheximide, detached the nascent FIG. 1. Deoxycholate/carbonate extraction of riboforin and apoB-48 from the microsomal membrane. Panel A, Western blots, performed with antibodies to riboforin, of proteins in the extract (I) and membrane pellet (II) after extraction with carbonate or deoxycholate/ carbonate. The microsomes were extracted with sodium carbonate alone or together with 0.025% deoxycholate and 1.2 M potassium chloride. The extract and the membrane pellet were subjected to SDS-PAGE with 10% gels and blotted against antibodies to riboforin. Panel B, distribution of the extracted apoB-48 after gradient ultracentrifugation. Cells were labeled with [ 35 S]methionine/cysteine for 2 h and homogenized, and the total microsomal fraction was recovered. The microsomes were extracted with sodium carbonate alone or together with 0.025% deoxycholate and 1.2 M potassium chloride. The extracts were fractionated on sucrose gradients, and apoB-48 from each fraction was recovered by immunoprecipitation and analyzed by SDS-PAGE and autoradiography. (Only the bands corresponding to apoB-48 are shown in panels B and C.) Panel C, comparison of apoB-48 recovered by deoxycholate/carbonate extraction and the dense apoB-48-containing lipoproteins secreted into the medium (apoB-48 HDL). The cells were labeled as in panel B. The samples were separated by sucrose gradient ultracentrifugation, and apoB-48 was recovered as described in panel B. I, apoB-48 extracted from microsomes by the deoxycholate/carbonate method (microsomal intracellular apoB-48 HDL). II, the dense apoB-48 lipoproteins secreted into the medium (apoB-48 HDL). III, nonlipidized protein, possibly unprocessed complement factor C3 (33), similar in size to apoB-48; this protein was precipitated with immunoglobulins from nonimmune rabbits.
polypeptides from the ribosomes with puromycin, and chased them through the secretory pathway of the cell into the medium (4). The cells were pulse labeled for 10 min and chased for 0 -30 min. After each chase period, the cells were treated with cycloheximide and puromycin and then chased for another 180 min in the presence of cycloheximide and puromycin to allow the nascent chains to form lipoproteins and be secreted into the medium. The medium (containing full-length apoB-100/48 as well as the nascent apoB polypeptides that were released into the secretory pathway and secreted during the 180-min chase) was subjected to gradient ultracentrifugation. ApoB was recovered from each fraction by immunoprecipitation and analyzed by SDS-PAGE.
The major amount of apoB radioactivity was recovered with lipoproteins in the VLDL density range (Fig. 3A). After a 0-min chase, apoB nascent polypeptides corresponding to approximately apoB-40 and longer were associated with the VLDL fraction. However, an only small amount of nascent apoB chains shorter than apoB-48 was incorporated into VLDL. The antiserum used to immunoprecipitate apoB recognized nascent apoB chains in the medium at least as short as apoB-20 (Fig.  3B). Nascent polypeptides of the size of apoB-20 to apoB-40 banded in the HDL density region and were incorporated into apoB-48 HDL (Fig. 3, A and B). These results demonstrate that apoB-20 and longer nascent chains can assemble apoB-48 HDL and that apoB-40 is the approximate minimal length for VLDL assembly. However, efficient VLDL assembly requires apoB chains about the size of apoB-48.
Proteins Associated with Microsomal VLDL-To determine if proteins other than known apolipoproteins were associated with luminal lipoprotein particles (i.e. could be involved in the second step of VLDL assembly), we isolated the luminal content of rat livers and characterized proteins in the d Ͻ 1.006 g/ml fraction. Three fractions (I-III in Fig. 4) isolated from this supernatant by ion-exchange chromatography were characterized by major proteins with molecular masses around 70 kDa (I), 60 kDa (II), and 100, 70, and 50 kDa (III). After separation by SDS-PAGE and Coomassie staining, the stained bands (1-5 in Fig. 4) were cut out of the gel and digested with trypsin; the tryptic fragments were identified by the matrix-assisted laser desorption ionization-time of flight technique (Table I). All isolated proteins were identified by data-base search as chaperones: band 1 was identified as binding protein, band 2 as protein disulfide isomerase, band 4 as calcium-binding protein 2, and band 5 as calreticulin. Band 3 did not correspond to any protein in the rat data base, but a search of the mouse data base identified it as glucose regulatory protein 94.
The identities of these proteins were confirmed by immunoblotting of the VLDL fraction with antibodies against binding protein, protein disulfide isomerase, glucose regulatory protein 94, and calreticulin (Fig. 5, lane 1). In these experiments, the supernatant (containing VLDL) was dialyzed and lyophilized before it was subjected to electrophoresis and immunoblotting. The predominant amount of chaperones was recovered in the infranatant after this centrifugation (not shown). This infranatant (depleted of VLDL by the ultracentrifugation) was subesequently recentrifuged under the same conditions at pH 7.4 or 11, and the supernatant was recovered and immunoblotted as described above. No or very small amount of immunoreactive chaperones were detected in this supernatant (Fig. 5, lane 2 (pH 7.4) and lane 3 (pH 11). These results indicate that the recovery of chaperones in the supernatant depends on the presence of VLDL and does not reflect the fact that chaperones are abundant microsomal proteins. The experiments were designed so the two supernatants (Fig. 5, lanes 2 and 3) could be compared directly with the VLDL fraction (Fig. 5, lane 1). An After each chase period, the microsomal fraction and the culture medium were recovered, and the microsomes were extracted with deoxycholate/carbonate. This extract and the cultured medium were subjected to gradient ultracentrifugation, and apoB-48 was recovered by immunoprecipitation followed by SDS-PAGE. ApoB-48 radioactivity (dpm) is shown as a function of time. aliquot of the infranatant was also analyzed and was shown to contain the chaperones; the results obtained at pH 11 are shown in Fig. 5, lane 4.
Next, the infranatant recovered after the first ultracentrifugation was incubated at pH 11 with native VLDL recovered from rat plasma (0.2 mg of VLDL protein/mg of protein in the infranatant). After 25 min (i.e. the time used for the extraction of the luminal content of the microsomes), VLDL were reisolated, and the amount of chaperones associated with the lipoproteins was estimated by immunoblotting under the conditions used above. No or very small amounts of chaperones were associated with VLDL under these conditions (Fig. 5, lane 5), but chaperones were recovered in the infranatant (Fig. 5, lane  6). These results indicate that the association between the chaperones and VLDL does not depend on a high pH during the extraction. The experiment was also carried out at pH 7.5 with the same results (not shown). To force the system to reveal any binding of chaperones to native VLDL, we deliberately chose a higher concentration of lipoproteins than was present in the microsomes.
The described immunoblot experiments were repeated twice with the same results. Together these immunoblot experiments support the possibility of a specific interaction between the chaperones and intracellular VLDL.
Activation of VLDL Assembly by Oleic Acid and Subcellular Localization of ApoB-48 during the Assembly Process-Next, we sought to identify the subcellular compartment in which apoB-48 VLDL accumulated during the assembly process. The basis for these experiments was the observation that the assembly of apoB-48 VLDL in McA-RH7777 cells requires incubation with oleic acid (2). First, we determined the length of the incubation required to initiate the assembly process. The cells were labeled for 30 min, chased for 120 min in the absence of oleic acid, and then chased for 0 -60 min in the presence of oleic acid. ApoB-48 VLDL were present in the microsomal lumen after a 15-min chase with oleic acid.
Next, pulse-chase studies were combined with subcellular fractionation. Microsomes from McA-RH7777 cells were fractionated by ultracentrifugation in a sucrose gradient (Fig. 6A). Microsomes derived from the rough ER (calnexin; see Fig. 6B) were present in the first 10 -12 fractions but could not be detected in fractions 14 -22 (compare Fig. 6, A and B). Most of the trans-Golgi marker (galactosyl transferase) was found in fractions 10 -20, with a maximum in fraction 15. The fractions recovered from the gradient were combined into four pools (I-IV in Fig. 6A). As judged by electron microscopy, pools I and II contained mainly granulated vesicles (not shown). Pool IV comigrated with the trans-Golgi marker and contained no detectable calnexin; however, some rough membranes were detected in this fraction (not shown).
To identify the subcellular fraction in which apoB-48 VLDL FIG. 4. Proteins associated with lipoproteins isolated from the luminal content of rat liver microsomes. The sodium carbonate extract from rat liver microsomes (6 ml) was overlayered with phosphate-buffered saline (29 ml, d ϭ 1.006 g/ml) and centrifuged at 40,000 rpm for 22 h in a Beckman Ti-60 rotor at 4°C. The upper one-third of the tube was collected and chromatographed on a Mono Q column equilibrated with 0.05 M Tris-HCl, pH 7.8, 300 mM sucrose, 1 mM EDTA, 2 mM deoxycholate, 0.5% Triton X-100, 6 M urea, and 40 mM sodium carbonate. Retained proteins were eluted with a linear gradient of sodium chloride (0 -250 mM) in the same buffer. The major proteins were eluted in three regions of the column. Fractions from each region were combined, concentrated, and subjected to SDS-PAGE on 3-15% gradient gels under reducing conditions. The Coomassie-stained bands (1)(2)(3)(4)(5) were excised, trypsinized, and analyzed by the matrix-assisted laser desorption ionization-time of flight technique (see Table I).

TABLE I Identification of proteins isolated from microsomal VLDL
Microsomes were isolated from rat livers and the luminal content extracted. The density fraction Ͻ 1.006 were recovered from the luminal content by ultracentrifugation and the protein content separated by ion-exchange chromatography and SDS-polyacrylamide gel electrophoresis (see Fig 4). Coomassie-stained proteins were cut out of the gels and digested with trypsin. The molecular mass of the obtained fragments was determined with the matrix-assisted laser desorption ionization-time of flight technique and submitted to data search. The molecular masses of the tryptic fragments that matched with the date base and their positions in the identified protein are given in the first accumulated, we pulse labeled the cells for 30 min and chased them for 120 min. The second step of VLDL assembly was induced by adding oleic acid, and a second chase of 0, 15, 30, or 60 min was performed. After sodium carbonate or deoxycholate/carbonate extraction, the appearance of apoB-48 VLDL in the luminal content of pools I-IV was followed. Radioactive apoB-48 VLDL were first detected in the luminal content of pool IV, when no radiolabeled apoB-48 VLDL could be detected in the other fractions (Fig. 6C), regardless of which extraction procedure was used. The experiment was repeated four times with the same results. The time needed for apoB-48 VLDL to appear in pool IV was 15-30 min.

DISCUSSION
This study shows that deoxycholate/carbonate extraction releases apoB-48 from the microsomal membrane without releasing integral membrane proteins. During sucrose gradient ultracentrifugation, the released intracellular apoB-48 HDL comigrated with the secreted apoB-48 HDL, and both displayed the characteristics of lipoproteins. The intracellular particle contained both phosphatidylcholine and triglyceride at a molar ratio of 0.7, indicating a lipid core. These analyses were based on the incorporation of [ 3 H]glycerol into the two glycerolipids and the assumption that the specific radioactivities of triglycerides and phosphatidylcholine in the lipoprotein are similar to those observed in the cell. The analyses were carried out under conditions in which very small changes in the specific radioactivities of the two lipids were observed in the cell, indicating a steady-state situation. We believe that these conditions increase the possibility that the specific radioactivities of phosphatidylcholine and triglycerides of the intracellular lipoproteins are similar to those in the cell. Interestingly, the secreted apoB HDL had a significantly lower molar ratio of phosphatidylcholine to triglyceride (0.22) than the intracellular lipoprotein. We suggest two possible explanations for this difference. The first explanation is that intracellular apoB HDL contains both apoB-48 and apoB-100, whereas the secreted apoB-48 HDL contains virtually only apoB-48 (2). It is possible that apoB-48 assembles with less neutral lipid during the cotranslational lipidation. This mechanism is consistent with what we observed during studies of the cotranslational lipidation of apoB in HepG2 cells (4). The second explanation is that the secreted lipoprotein is an underlipidated form that has failed to form VLDL. This mechanism is consistent with our previous observation of an inverse relationship between apoB-48 VLDL and the secreted apoB-48 HDL (2).
In turnover studies, radioactive apoB-48 first appeared in the intracellular apoB-48 HDL and was not observed in VLDL in significant amounts until after the radioactivity had reached a maximum in intracellular apoB-48 HDL. The subsequent decrease in the radioactivity in intracellular apoB-48 HDL accounted for all of the increase in radioactivity in apoB-48 VLDL and in the secreted apoB-48 HDL. These results show that apoB-48 HDL is produced in the first step of the assembly process and is the precursor for VLDL produced during the second step. Moreover, intracellular apoB-48 HDL appears to be the precursor for secreted apoB-48 HDL.
These findings, together with our previous demonstration of an inverse relationship between secretion of apoB-48 HDL and apoB-48 VLDL (2), provide the basis for a model of apoB-48 VLDL assembly (Fig. 7). The first step in the assembly process gives rise to a pre-VLDL (i.e. intracellular apoB-48 HDL), which can then either go through the second step to form VLDL or be secreted as apoB-48 HDL.
Structural Requirements for Assembly of ApoB-48 VLDL-This study also demonstrated that apoB must reach the size of apoB-20 for the HDL-like particle to be assembled and secreted and must reach the size of apoB-40 to undergo the second step of VLDL assembly. In these experiments, we used our previously described method (4) to generate a series of truncated forms of apoB. In contrast to transfection studies, this method uses endogenous apoB and therefore avoids artifacts that can occur from the overexpression of a foreign gene or during the selection of stably transfected cells. The observation that apoB-40 can be assembled into VLDL supports the results of studies with transfected forms of apoB truncated at the carboxyl terminus (24), which indicated that apoB-37 can be assembled into VLDL. Not until apoB has reached the size of apoB-48, however, is there a quantitatively important recruitment of precursor particles to the second step of the assembly process.
ApoB-48 (as well as apoB-40) contains the amino-terminal globular domain and the major amount of the amphipathic ␤-sheet domain (25). The amino-terminal globular domain is important for the assembly of apoB lipoproteins (26), whereas the amphipathic ␤-sheet domain seems to be important for the strong and irreversible interaction between apoB and the lipid core. However, the same type of structure is present in the region between apoB-20 and apoB-40 (i.e. on the nascent chains The infranatant recovered during the first ultracentrifugation was recentrifuged under the same conditions at pH 7.4 or 11. The supernatant was recovered, dialyzed, lyophilized, and blotted as described above. Lanes 2 and 3 show blots of the supernatants obtained at pH 7.4 and 11, respectively. The amount of infranatant used for this recentrifugation was adjusted so that the intensity of the immunoblots shown in lanes 2 and 3 could be compared directly with that obtained with the VLDL fraction (lane 1). Lane 4 shows the blot of a portion (one-tenth) of the infranatant recovered after the second centrifugation (only the results obtained at pH 11 are shown). Lanes 5 and 6, the supernatant (lane 5) and infranatant (lane 6) obtained after ultracentrifugation of the VLDL-depleted luminal content that had been incubated with VLDL from rat plasma. A portion of the infranatant obtained after the first centrifugation was incubated for 25 min with rat plasma VLDL (0.2 mg of VLDL protein/mg of protein in the infranatant) was at pH 11. After the incubation, VLDL were reisolated, and the VLDL supernatant was dialyzed, lyophilized, and blotted as described above. Lane 5 shows the immunoblot of this VLDL supernatant against the different chaperones. The amount of infranatant used for these experiments was adjusted so that the intensity of the immunoblot shown in lane 5 could be compared directly with those shown in lanes 1-3. Lane 6 shows the corresponding blot of a portion of the infranatant. that are secreted with the HDL-like particle but not with VLDL). Thus, our results define more specifically the structural requirements for apoB to participate in the second step of VLDL assembly. Interestingly, although some nascent apoB chains as short as apoB-40 were assembled into VLDL, the efficiency of VLDL assembly increased rapidly when apoB reached the size of apoB-48. This could indicate that the carboxyl-terminal 17% of apoB-48 (i.e., apoB-40 -apoB-48), predicted to be amphipathic ␣-helices (25), has a central role in the interaction between apoB-48 HDL and the second step.
The results obtained in McA-RH7777 cells, which produce bona fide VLDL, differ completely from those obtained with the same method in HepG2 cells, which do not produce significant amounts of VLDL (4). In HepG2 cells, the length of nascent apoB-100 chains is inversely related to the density of the lipoprotein particles they assemble, ranging from HDL to low density lipoprotein (4). Thus, in a cell that does not produce VLDL, the lipidation is proportional to the length of the translated polypeptide. These observations reflect a fundamental difference between the cotranslational assembly process in HepG2 cells (4) and the two-step process for VLDL assembly in McA-RH7777 cells (2).
Chaperone Proteins in VLDL Assembly-Another major observation in this study is that VLDL particles isolated from the secretory pathway were associated with several chaperone proteins. Recently, Linnik and Herscovitz (27) independently demonstrated a similar network of chaperones associated with apoB, independent of lipidation, in HepG2 cells. HepG2 cells may be regarded as a model system for the first (cotranslational) step of VLDL assembly (4,28). Thus, these important results demonstrate that the first step in the assembly process is guided by chaperones (27). Our observations demonstrate that chaperones are also involved in the second step, when the particles have the density of VLDL. Thus, the whole assembly process seems to be guided by chaperones. A similar network of chaperones has been identified in association with other immature proteins in the secretory pathway. For example, influenza hemagglutinin is associated with binding protein, glucose regulatory protein 94, calreticulin, and calnexin before the homotrimer is formed (29,30). We did not identify calnexin associated with the intracellular lipoproteins, which is logical because it is an integral membrane protein, and the experiment was designed to investigate lipoprotein particles in the luminal content. One explanation is that these lipoproteins  7. Two-step model of VLDL assembly. In the first step of VLDL assembly, apoB is cotranslationally translocated to the lumen of the endoplasmic reticulum to form a pre-VLDL particle. The pre-VLDL can acquire the major amount of lipids in the second step, which either occurs immediately before the VLDL particle is transferred out of the rough ER or occurs in a compartment other than that in the first step. A network of chaperones is associated with the VLDL particle that is under assembly. BiP, binding protein; PDI, protein disulfide isomerase; CaBP2, calcium-binding protein 2; GRP 94, glucose regulatory protein 94. already have lost their contact with calnexin, as the interaction between apoB and calnexin is important for early events in the assembly process (31). Alternatively, the interaction between this chaperone and the lipoprotein may be disrupted by the extraction procedure.
Localization of ApoB-48 during the Assembly Process-The final observation made in this paper is that apoB-48 VLDL accumulated in a compartment separate from the rough ER; no apoB-48 VLDL was detected in the rough ER. One possible explanation for this observation is that the second step of VLDL assembly occurs not in the rough ER but in a smooth membrane compartment (pool IV). Alternatively, apoB-48 VLDL may accumulate, but not be assembled, in pool IV. Three observations argue against the latter possibility. First, no apoB-48 VLDL was detected in any of the rough ER fractions, even though they contained considerable amounts of radioactive apoB-48. Second, the results were the same irrespective of which extraction procedure (carbonate or deoxycholate/carbonate) was used to extract apoB-48 from the microsomes. Third, the appearance of apoB-48 VLDL in pool IV corresponded to the start of the second step during the oleic acid incubation. However, we could not exclude completely the possibility that VLDL is assembled in the rough ER. Thus the very small pool of VLDL in the rough ER may be the result of a rapid transferred of the assembled VLDL out of this compartment. Such a mechanism could be compatible with our failure to detect VLDL in the rough ER. Kinetics studies in rat indicate that the transfer out of the ER is the rate-limiting step in secretion (32). This rate-limiting step could be the conversion of pre-VLDL to VLDL. The observation that VLDL are associated with a network of chaperones, which contains ER retention signals, could suggest that the lipoprotein is retained in a post-rough ER, pre-Golgi smooth membrane compartment until the chaperones have left the particle.
In conclusion, the results presented here support the twostep model for the assembly of apoB-48 VLDL (Fig. 7). The process starts with the translation of the protein, forming a full-length, partially lipidated, pre-VLDL intracellular apoB-48 HDL. The pre-VLDL acquire the major amount of lipids in the second step, forming bona fide VLDL. The second step occurs either in a smooth membrane compartment or immediately before the particle is transferred out of the rough ER. The entire assembly process involves a series of chaperone proteins.