Folding of the amino-terminal domain of apolipoprotein B initiates microsomal triglyceride transfer protein-dependent lipid transfer to nascent very low density lipoprotein.

The initial assembly of apolipoprotein B100 (apoB) into lipoprotein particles occurs cotranslationally. To examine steps required to initiate this process, the intracellular folding and assembly of the amino-terminal 28% of apoB (apoB28) was examined using several criteria including nonreducing gel electrophoresis, sensitivity to dithiothreitol (DTT)-mediated reduction, and buoyant density gradient centrifugation. In hepatoma cells, after a 1-min pulse with radiolabeled amino acids, labeled apoB28 migrated during gel electrophoresis in the folded position and was resistant to reduction in vivo with 2 mM DTT. A similar rate and extent of folding was observed in Chinese hamster ovary cells, a microsomal triglyceride transfer protein (MTP)-negative cell line that can neither lipidate nor efficiently secrete apoB28. Amino-terminal folding of apoB28 was essential for its subsequent intracellular lipidation as apoB28 synthesized in hepatoma cells under reducing conditions remained lipid poor (d > 1.25 g/ml) and was retained intracellularly. Upon DTT removal, reduced apoB28 underwent a process of rapid (t1/2 approximately 2 min) post-translational folding followed by a slower process of MTP-dependent lipidation. As with the cotranslational assembly pathway, post-translational lipidation of apoB28 displayed a strict dependence upon amino-terminal folding. We conclude that: 1) folding of the amino-terminal disulfide bonded domain of apoB is achieved prior to the completion of translation and is independent of MTP and events associated with buoyant lipoprotein formation and 2) domain-specific folding of apoBs amino-terminal region is required to initiate MTP-dependent lipid transfer to nascent apoB in the hepatic endoplasmic reticulum.

In addition to the protein and lipid biosynthetic capacities present in most cell types, the biogenesis of very low density lipoprotein requires at least two dedicated gene products: apoB, the major protein component of very low density lipoprotein (1)(2)(3)(4)(5)(6), and the microsomal triglyceride transfer protein (MTP), 1 a soluble lipid transfer protein localized to the ER of hepato-cytes and intestinal epithelial cells (7). As the initial assembly of apoB with lipids occurs cotranslationally (8), it was predicted that the amino-terminal domain may play a critical role in initiating the process of lipoprotein particle assembly (7,9). Indeed, the amino-terminal ϳ18% of apoB (apoB18) is atypical relative to the rest of the protein: (i) it is globular and highly disulfide bonded (6 of the 8 disulfide bonds in apoB100 are positioned within the amino-terminal 11% of the protein (10)), (ii) it demonstrates a relatively lower affinity for plasma low density lipoprotein particles than internal and carboxyl-terminal domains (2,(11)(12)(13), and (iii) in transfected cells, apoB18 lacks the capacity to recruit a significant amount of lipid and its secretion can be achieved independently of MTP (14 -17).
While the amino-terminal domain of apoB is incapable of forming core-containing lipoproteins on its own, several lines of evidence suggest that it nonetheless plays an essential role in lipoprotein secretion. When disulfide bond formation was disrupted in the amino-terminal domain of apoB by preincubating HepG2 cells with 2 mM DTT, the ability of apoB to form a secretable lipoprotein was irreversibly blocked (9). In addition, internal domains of apoB, which are otherwise incapable of undergoing secretion, could be rendered secretion competent by appending them to the carboxyl-terminal end of apoB17 (18). While proper folding of the amino-terminal domain of apoB appears essential to promote its secretion, due to apoB's large size (4,536 amino acid residues) and hydrophobicity, it has not yet been experimentally established which stage of the lipoprotein assembly and secretion process is dependent upon this amino-terminal folding event.
In the current report, this and other mechanistic aspects of the assembly and secretion of apoB-containing lipoproteins were explored by focusing on the biogenesis of the aminoterminal 28% of apoB (apoB28). This form of apoB has been shown previously to be of sufficient size to undergo secretion in the form of a buoyant, lipid core-containing lipoprotein particle (8,14,15,19). Furthermore, the secretion of apoB28 is MTPdependent, a hallmark of apoB biogenesis (7, 20 -22). While capable of MTP-dependent lipoprotein assembly, the apoB28 polypeptide chain is sufficiently small (1270 amino acid residues) to enable resolution of reduced and folded forms by nonreducing SDS-PAGE. This allowed us, using techniques that are not applicable to full-length apoB48 or apoB100, to directly monitor the relationship between intracellular folding of apoB and its capacity to undergo MTP-dependent lipidation in the ER.
The studies described here indicate that folding of the aminoterminal domain of apoB28 occurs either during or shortly after translation and can be achieved independently of MTP. Furthermore, the data indicate that folding of this domain is essential for apoB28's capacity to undergo subsequent MTP-dependent assembly with lipids. The requirement for amino-terminal folding was observed for both the normal cotranslational assembly pathway as well as during the posttranslational folding and lipidation that occurs when apoB28 was artificially reduced in the ER and then allowed to refold. These studies provide compelling evidence that the initiation of lipoprotein formation in the hepatic ER proceeds via two distinct stages: the amino-terminal globular domain of apoB undergoes a cotranslational folding event that occurs autonomously and prior to appreciable lipid recruitment. Once completed, this folding event facilitates the capacity of apoB's downstream sequences to undergo MTP-dependent assembly with lipids. The requirements for the production of small, dense apoB28-containing lipoprotein particles described here, may also be necessary to initiate the formation of native lipoproteins containing apoB48 or -100.

EXPERIMENTAL PROCEDURES
Materials-Restriction and DNA modification enzymes were from New England Biolabs. Tissue culture media and supplements were obtained from ICN (Irvine, CA) or MediaTech (Washington, D.C.). Cycloheximide, DTT, and IAA were from Sigma. Tran 35 S-label (an ϳ5:1 mixture of [ 35 S]Met and Cys) was from ICN. Anti-FLAG M2 monoclonal antibody was from Eastman Kodak Scientific Imaging Systems (New Haven, CT). Polyclonal antibodies to human apoB100 and albumin were obtained from Boehringer Mannheim. Centricon-10 and Centriprep-10 centrifugal concentrators were from Amicon (Beverly, MA).
Construction of an ApoB28-containing Expression Plasmid-A form of apoB28 containing an 8-amino acid carboxyl-terminal epitope tag and stop codon was constructed. The SacII/XhoI fragment from the apoB53containing plasmid pB53LII (obtained from Dr. Zemin Yao, University of Ottawa) was ligated to SacII and XhoI-digested pBluescript IISKϩ (Stratagene, La Jolla CA) to form plasmid SK28. A double-stranded oligonucleotide encoding the 8-amino acid epitope tag (DYKDDDDK) (23,24), stop codon, and flanking restriction ends was ligated to XhoI and KpnI digested SK28 to form SK28F. The fusion protein was excised from SK28F using the EcoRI and KpnI restriction enzyme sites flanking the apoB coding region and subcloned into EcoRI and KpnI-digested pCMV5 (25) to form plasmid apoB28F.
Tissue Culture and Transfection Conditions-COS-1 and HepG2 cells were cultured as described previously (9,26). McA-RH7777 cells were maintained in low glucose Dulbecco's modified Eagle's medium containing 10% fetal bovine serum. CHO cells were maintained in a 50:50 mixture of low glucose Dulbecco's modified Eagle's medium and Ham's F-12 containing 10% fetal bovine serum. Transient transfections of COS-1 cells by the DEAE dextran method and HepG2 cells by the calcium phosphate method were performed as described (27,28). To control for possible plate to plate variability in transient transfection efficiencies, HepG2 cells were trypsinized 24 h after transfection, mixed, and replated. Experiments were performed 16 -24 h after replating. For stable transfections, McA-RH7777 or CHO cells were cotransfected with apoB28F and pSV2neo (29) at a mass ratio of 20:1 as described (28). Stably transfected CHO cells were selected in 700 g/ml G418 and McA-RH7777 cells in 400 g/ml G418. Individual clones were expanded and screened for apoB28F expression and secretion by Western blot analysis. For routine propagation of transfected McA-RH7777 and CHO cell lines, G418 concentration was reduced to 350 g/ml.
Metabolic Labeling of Cells-Cells were washed with phosphatebuffered saline and preincubated in serum free, Met/Cys-deficient high glucose Dulbecco's modified Eagle's medium (COS-1 cells) or minimal essential medium (McA-RH7777, CHO, and HepG2 cells) for 20 min prior to labeling. For CHO cells, the Cys/Met-deficient minimal essential medium was supplemented with 35 g/ml L-proline. Labeling was initiated by addition of 100 Ci/ml [ 35 S]Met/Cys (total volume of labeling media was 0.4 ml/60-mm dish, 2 ml/100-mm dish, and 5 ml/150-mm dish). Chase was initiated by diluting labeling media with 2.5 volumes of chase media (minimal essential medium containing 2.5 mM additional Met and 1 mM additional Cys). Media was removed and replaced with 2 ml (100-mm dishes) or 5 ml (150-mm dishes) fresh chase media. In cases where post-translational lipidation and secretion were monitored (Figs. 6 and 7), cycloheximide was included in the chase to block new protein synthesis. After the pulse in the presence of DTT, chase media containing 100 M cycloheximide was added and incubation was continued for an additional 5 min. Media was then replaced with chase media containing 100 M cycloheximide but lacking DTT. Control experiments (not shown) revealed that a 5-min preincubation with 100 M cycloheximide reduced protein synthesis to less than 0.5% of control values. Immunoprecipitation of apoB28F from cells and media was performed as described (26).
Generation of Reduced ApoB28F-In instances where cells were labeled in the presence of DTT (Figs. 4, 6, and 7), DTT was added 1.5 min prior to addition of label. Under these conditions, approximately 80% of labeled apoB28F was in reduced form, primarily because the DTT present during the preincubation blocks the disulfide bond formation that normally occurs during translation. Because DTT is an inhibitor of protein synthesis (9, 30 -32), which in HepG2 cells occurs predominantly at the level of initiation (9), DTT preincubation reduces incorporation of radiolabeled amino acids into total protein (9,30,31). In Fig.  4, where a comparison is made between apoB28F labeled in the presence and absence of DTT, twice as many cells were employed when labeling was performed in the presence DTT.
Gradient Fractionation of Intracellular ApoB28F-Post-nuclear membranes were prepared by Dounce homogenization as described previously (26). Membranes were suspended in 300 l of membrane buffer (26) and adjusted to 0.1 M sodium carbonate, pH 11.5, by addition of a 1 M stock. Samples were diluted to 2 ml with 100 mM sodium carbonate, pH 11.5, and incubated on ice for 60 min with occasional low speed vortexing. Membranes were pelleted by centrifugation at 100,000 rpm (412,000 ϫ g) for 15 min at 4°C in a Beckman TL100 tabletop ultracentrifuge using the TLA 100.3 rotor. Supernatant fractions were transferred to clear polystyrene tubes and adjusted to 0.005% phenol red, 200 mM NaCl, 50 mM Tris-HCl, pH 7.5, 2.5 mg/ml bovine serum albumin and sufficient dilute HCl (ϳ108 l of a 1:5 dilution of concentrated HCl) to bring the pH to 7.5. The neutralized sample was adjusted to 1.25 g/ml with solid KBr and a final volume of 3 ml. Samples were transferred to polyallomer quickseal tubes, and centrifuged in the TLA 100.3 rotor at 100,000 rpm for 12-15 h at 15°C. The top 1-ml (d Ͻ 1.25 g/ml) and bottom 2-ml (d Ͼ 1.25 g/ml) fractions were recovered with a tube slicer (Beckman Instruments). Fractions were concentrated to ϳ50 l using Centricon 10 centrifugal concentrators (Amicon) and diluted to 2 ml with phosphate-buffered saline. After another round of centrifugation/concentration, samples were diluted into 1 ml of lysis buffer (1% Triton X-100, 25 mM Tris-HCl, pH 7.5, 300 mM NaCl, 2.5 mg/ml bovine serum albumin, 1 mM phenylmethylsulfonyl fluoride, 10 g/ml leupeptin, 10 g/ml pepstatin) and subjected to immunoprecipitation with anti-FLAG M2 monoclonal antibody as described (26). Pellets from carbonate extraction were washed 2 ϫ with 1 ml of TBS and resuspended in 200 l of 1% SDS, 25 mM Tris, pH 7.4, and boiled for 10 min. Samples were diluted with 800 l of lysis buffer and subjected to immunoprecipitation as described (26).
Density Gradient Centrifugation of Secretory ApoB28F-Eight ml of 1.25 g/ml KBr in phosphate-buffered saline was added to an SW41 polyallomer centrifuge tube (Beckman Instruments) and overlaid with 1 ml of conditioned media that had been adjusted to 1.20 g/ml with solid KBr. The 1-ml sample was then overlaid with 3 ml of 1.05 g/ml KBr. Tubes were centrifuged at 37,000 rpm (160,000 ϫ g) for 48 h at 20°C in an SW41 rotor. In Fig. 7, 9 ml of conditioned media were first concentrated to 1 ml using a Centricon Centriprep-10 concentrator prior to adjusting to 1.20 g/ml. Tubes were fractionated from the top using an Autodensiflow Gradient Fractionator (Labconco Industries) into 1-ml samples. The density of each fraction was determined gravitometrically. For immunoprecipitation, gradient fractions were exchanged into phosphate-buffered saline using Centricon-10 centrifugal concentrators as described above. After diluting into lysis buffer, fractions were immunoprecipitated with anti-FLAG M2 monoclonal antibody.

ApoB28F: A Model System for Examination of Early Stages of Hepatic Very Low Density Lipoprotein
Assembly-Because of the cotranslational kinetics of lipoprotein assembly, steps required to initiate small particles formed by the amino-terminal ϳ25% of apoB may be the same as those required for apoB48 and apoB100-containing lipoproteins. We therefore investigated whether an epitope-tagged form of apoB28 could serve as an appropriate model system for analyzing the relationship between apoB folding and subsequent steps required for the initiation of lipoprotein particle assembly.
A plasmid encoding the amino-terminal 28% of apoB with the 8-amino acid FLAG epitope (23, 24) and termination codon appended to its carboxyl terminus (apoB28F; "Experimental Procedures") was transiently transfected into HepG2 cells (a human hepatoma cell line). Cells were pulse-labeled for 10 min with [ 35 S]Met/Cys, and chased for 90 min in media containing an excess of unlabeled amino acids. Chase media was collected and subjected to density gradient centrifugation as described under "Experimental Procedures." As shown in Fig. 1A, apoB28F was secreted as a buoyant lipoprotein particle with a peak density of ϳ1.175 g/ml, a value similar to that reported by others for apoB28 (14,19). A similar peak density was also observed for apoB28F secreted from stably transfected rat hepatoma cells (McA-RH7777 cells; not shown). Furthermore, the secretion of apoB28F from the kidney fibroblast cell line, COS-1, was induced by ϳ5-fold upon coexpression with the large subunit of MTP (Fig. 1B, compare amount of apoB28F in media (M) in the presence (ϩ) and absence (Ϫ) of MTP expression). These results indicate that apoB28F undergoes MTP-dependent secretion as a buoyant lipoprotein particle. We next examined whether the folding state of apoB28F could be monitored by nonreducing SDS-PAGE, a common means of examining disulfide bond formation in intact cells (33). Labeled apoB28F secreted from McA-RH7777 cells was immunoprecipitated and boiled in the absence or presence of 50 mM DTT. After alkylation of free sulfhydryls with IAA, samples were analyzed by nonreducing SDS-PAGE. As observed in Fig. 1C, a subtle but reproducible mobility shift was observed upon DTT-mediated reduction of secretory apoB28F (Fig. 1C, compare lanes 1 and 2, 3 and 4).
ApoB28F Undergoes Rapid Folding into a DTT Resistant Form-Utilizing both mobility changes during nonreducing SDS-PAGE as well as onset of DTT resistance as criteria, the folding kinetics of apoB28F were examined in transfected McA-RH7777 cells. The underlying basis for the ability of proteins to undergo a transition from DTT sensitivity to DTT resistance in the ER is not fully understood. The transition may reflect conformational changes in the protein and/or different extents or stages of chaperone association (34,35). Nonetheless, for several model proteins previously examined, the rate of onset of DTT resistance in the ER closely parallels the time required to achieve a native conformation competent for anterograde transport in the secretory pathway (34,36).
Stably transfected McA-RH7777 cells were pulse-labeled for 1 min with [ 35 S]Met/Cys and chased for various times prior to adjusting the chase media to 2 mM DTT (30). After incubating with DTT for 5 min (a condition that allows elongation of nascent polypeptide chains (37)), cells were placed on ice and treated with an excess of IAA (30). ApoB28F immunoprecipitated from cell lysates was resolved by nonreducing SDS-PAGE. As observed in Fig. 2A, lanes 1-7, the predominant form of apoB28F migrated with the folded standard (lane R) and at no time after the pulse did addition of DTT cause appreciable reduction. Since proteins labeled during the 1-min pulse were, by definition, nascent polypeptides at the time that [ 35 S]Met/ Cys was added to the cells, and it is estimated that ϳ3-4 min is required to synthesize apoB28F (ϳ300 -400 amino acids per minute (37)), these results indicate that the disulfide bonds in apoB28F progressed to the DTT resistant form either before or shortly after (within 1 min) translation was completed. Although the anti-FLAG antibody allowed us to monitor only full-length protein, use of a polyclonal antibody with the potential to recognize amino-terminal epitopes of apoB28F (38) failed to detect an appreciable spectrum of nascent polypeptide chains under our labeling conditions (not shown). Furthermore, it may be difficult to detect and interpret SDS-PAGE mobility shifts based on differences in disulfide bond formation within a spectrum of nascent polypeptide chains.
The rapid onset of DTT resistance observed for apoB28F is in marked contrast to all other disulfide bonded proteins examined to date which undergo a time-dependent post-translational transition from DTT sensitivity to DTT resistance in the ER (9,34,36,39,40). As an example, the onset of DTT resistance of albumin in HepG2 cells was examined. Although 11 of the 17 disulfide bonds in albumin form cotranslationally, with the remainder forming almost immediately after translation (41,42), the disulfide bonds in albumin remained sensitive to DTT-mediated reduction well after completion of synthesis ( Fig. 2A, lane 1). Approximately 3 min was required for half of the labeled albumin to achieve DTT resistance after a 1-min pulse (Fig. 2B, lane 4). Other disulfide bonded secretory proteins examined in this way often display considerably longer periods of DTT sensitivity in the ER (34,36,40).
Rapid Folding of the Amino-terminal Domain of ApoB Occurs Independently of MTP and Buoyant Lipoprotein Formation-We investigated whether the rapid folding of apoB28F, as measured by its almost immediate onset of DTT resistance, was an autonomous property of the apoB protein or whether folding was dependent upon the presence of liver specific factors such as MTP. For this purpose, apoB28F was stably transfected into CHO cells, an MTP negative cell line that lacks the capacity to form secretable lipoproteins (17). Although CHO cells cannot efficiently secrete apoB28F (not shown), we also explored whether or not CHO cells were capable of lipidating apoB28F by the MTP-independent pathway that may exist in some nonhepatic cell lines (43). McA-RH7777 and CHO cells, stably transfected with apoB28F, were labeled for 2 h with [ 35 S]Met/Cys. Post-nuclear membranes were prepared and subjected to extraction with sodium carbonate, pH 11.5, to release lumenal contents (44,45). After centrifugation, pellet fractions were solubilized and immunoprecipitated with anti-FLAG M2 monoclonal antibody. Supernatant fractions were adjusted to 1.25 g/ml with KBr and subjected to density gradient centrifugation as described under "Experimental Procedures." The top 1-ml fractions (d Ͻ 1.25 g/ml) and bottom 2-ml fractions (d Ͼ 1.25 g/ml) were immunoprecipitated and analyzed along with the carbonate pellets by SDS-PAGE and fluorography. As shown in Fig. 3A, in McA-RH7777 cells, the carbonate extractable apoB28F was distributed in an ϳ2:1 ratio between the top (T) and bottom (B) gradient fractions. In contrast, CHO cells were incapable of converting apoB28F into a form with sufficient lipid to float in the d Ͻ 1.25 g/ml fraction.
To examine the contribution of MTP and events associated with buoyant lipoprotein formation on apoB28F folding, the experiment in Fig. 2 was repeated in CHO cells. When 2 mM DTT was added to transfected CHO cells immediately after the 1-min pulse, the folded form of apoB28F represented ϳ70% of total (Fig. 3B, lane 2). This result was somewhat different from that obtained in McA-RH7777 cells where ϳ90% of apoB28F resisted DTT immediately after the 1-min pulse ( Fig. 2A, lane  1). Both of these results, however, were distinct from the behavior of albumin (Fig. 2B) and other proteins whose disulfide bonds remain completely sensitive to DTT-mediated reduction for a characteristic period of time after completion of translation. That the majority of apoB28F in CHO cells achieved DTT resistance immediately after the pulse indicated that its capacity to fold rapidly occurred independently of MTP, and independently of events necessary for buoyant lipoprotein assembly. This conclusion was further supported by the finding that, when apoB28F was reduced and then allowed to refold, the rate and extent of post-translational folding in CHO and McA-RH7777 cells was also similar (see below).
Reduced ApoB28F Remains Lipid Poor in the Hepatic ER-To examine the intracellular consequences of blocking amino-terminal folding, the buoyant density of apoB28F extracted from cells that were labeled under control and reducing conditions was examined. For these experiments transiently transfected HepG2 cells were used because, unlike McA-RH7777 cells, they express detectable amounts of serum albumin which is useful as an endogenous disulfide bonded control protein (not shown). Forty-eight hours after transfection, HepG2 cells were pulse-labeled for 10 min in the absence or presence of 2 mM DTT. Where DTT was used, it was added 1.5 min prior to addition of label. Under these conditions, ϳ80% of the labeled pool of apoB28F is reduced ("Experimental Procedures"). After metabolic labeling, post-nuclear membrane fractions were prepared from control and DTT-treated cells and extracted with sodium carbonate, pH 11.5, to release their lumenal contents. Pellet and supernatant fractions were analyzed as described for Fig. 3A. As shown in Fig. 4, when cells were labeled in the absence of DTT, the majority of apoB28F was recovered in the carbonate supernatant (S) which in turn was distributed in an ϳ2:1 ratio between the top and bottom density gradient fractions. In DTT-treated cells, a greater proportion of apoB28F remained associated with the membrane pellet (ϳ50%). This material copelleted with membranes due either to membrane binding and/or because of DTT-induced aggregation (31,46). Of the carbonate releasable material, however, greater than 90% was recovered in the bottom (d Ͼ 1.25 g/ml) gradient fraction (compare lanes 5 and 6). A portion of each gradient fraction was also analyzed for albumin content. Under both reducing and nonreducing conditions, all of the albumin was recovered in the bottom fraction and was

FIG. 4. Reduced apoB28F is unable to form of a buoyant lipoprotein.
Transiently transfected HepG2 cells were pulse-labeled for 10 min in the absence or presence of 2 mM DTT as indicated. Where DTT was included, it was added 1.5 min prior to addition of label. For cells incubated without DTT, three 100-mm dishs of cells were employed. For cells incubated with DTT, six 100-mm dishes were used ("Experimental Procedures"). A post-nuclear membrane fraction was prepared and extracted with sodium carbonate, pH 11.5. Supernatant fractions (S) were subjected to density gradient centrifugation as described for Fig. 3A. ApoB28F contained in the top 1-ml (T) and bottom 2-ml (B) gradient fractions as well as carbonate pellets (P) were immunoprecipitated with anti-FLAG M2 monoclonal antibody. ApoB28F was analyzed by 6% nonreducing SDS-PAGE and fluorography. The position of reduced and folded forms of apoB28F are indicated. either fully reduced or folded, depending upon the labeling condition (not shown). These results indicate that folding of the amino-terminal domain is an essential prerequisite for apoB28F's ability to recruit lipid in the hepatic ER.
Post-translational Folding of ApoB28F-While disulfide bond formation is in many cases cotranslational, proteins artificially reduced in the ER by DTT can fold post-translationally and undergo efficient secretion (30,39,47). It was previously shown that reduced apoB100 was incapable of achieving secretion competence post-translationally (9). However, because of its large size it was impossible to assess whether this was due to an inability to achieve post-translational disulfide bond formation or whether subsequent steps required for its assembly and secretion were blocked. To examine this question, transiently transfected HepG2 cells were pulse-labeled for 10 min in the presence of DTT to produce reduced apoB28F. After the pulse, the cells were chased for various periods of time in the absence of DTT. After each chase period, cells were placed on ice, treated with IAA, and prepared for immunoprecipitation. As observed in Fig. 5A, upon removal of DTT, reduced albumin underwent a process of post-translational folding as evidenced by its almost immediate (within 0.5 min) conversion to a diffuse series of electrophoretic forms representing intermediates in folding. Within 4 min, the diffuse grouping of bands was predominantly converted to a more rapidly migrating species corresponding to the folded form. This form has been shown previously to undergo efficient secretion (9,30). When apoB28F was examined in the same extracts, it also underwent a timedependent conversion from the reduced to the folded form which, based on nonreducing SDS-PAGE mobility, was complete within 4 min. In addition to its comigration with native folded apoB28F by nonreducing SDS-PAGE, ϳ70% of the posttranslationally folded apoB28F resisted reduction with DTT within 4 min of DTT removal (not shown). This post-translational folding reaction was not facilitated by MTP or events associated with apoB28F lipidation as similar results were observed when the experiment was repeated in CHO cells (not shown). It appears, therefore, that the amino-terminal domain of apoB28F is capable of relatively rapid post-translational folding in the ER.
Post-translational Lipidation of ApoB28F Is Dependent Upon Folding of the Amino-terminal Domain-Since apoB28F can fold post-translationally, we examined whether post-translational folding was accompanied by a process of post-translational lipidation. Transiently transfected HepG2 cells were pulse-labeled in the presence of DTT for 10 min and then chased for 0, 5, or 15 min in the absence of DTT. At the end of the labeling period, post-nuclear membranes were prepared and treated with sodium carbonate, pH 11.5, to extract lumenal contents. After centrifugation, pellet and supernatant fractions were analyzed as described for Figs. 3A and 4.
As observed in Fig. 6A, lanes 2 and 3, cells labeled for 10 min in the presence of DTT contained predominantly lipid-poor apoB28F in the carbonate supernatants (i.e. most was recovered in the bottom fraction of the gradient). Although it was previously demonstrated that post-translational folding was essentially complete within about 4 min of DTT removal (Fig.  5), this folding was not accompanied by appreciable post-translational lipidation as the percentage of d Ͻ 1.25 g/ml apoB28F was not significantly changed after the 5-min chase (compare ratio of apoB28F in lanes 2 and 3 versus lanes 5 and 6). During this time frame, a greater proportion of apoB28F became carbonate extractable. Only ϳ35% of apoB28F was extracted after the 0-min chase whereas ϳ65% was extracted after the 5-min chase (compare ratio of apoB28F in supernatant and pellet fractions in lanes 1-3 and 4 -6). This result has been consistently observed and indicates that, as apoB28F undergoes posttranslational folding (Fig. 5B), some of it is converted from carbonate-resistant to carbonate-extractable form, perhaps reflecting disaggregation and/or release from the ER membrane. However, virtually none of the carbonate-resistant apoB28F that was released into the supernatant after 5 min of chase was lipidated. After 15 min of chase, the proportion of carbonateextractable apoB28F was unchanged. During this same time frame, however, the proportion of the apoB28F in the d Ͻ 1.25 g/ml fraction was increased by ϳ4-fold (compare lanes 2 and 3  with 8 and 9). This increase in the percentage of buoyant apoB28F formed post-translationally was dependent upon amino-terminal folding as it was not observed if DTT was present throughout the 15-min chase period (lanes 10 -12).
To quantitate the extent of post-translational lipidation of apoB28F following DTT removal, the experiment in Fig. 6A was performed in triplicate and mean Ϯ S.D. values were calculated. As shown in Fig. 6B, apoB28F consistently underwent a characteristic post-translational lipidation reaction that resulted in an ϳ4-fold increase in d Ͻ 1.25 g/ml apoB28F within 15 min of DTT removal. Taken together, the data in Fig.  5B and Fig. 6 indicate that the post-translational lipidation of apoB proceeds through two distinct stages. DTT removal from cells containing reduced apoB28F results in a process of posttranslational folding (Fig. 5B) and possible disaggregation which, however, is not accompanied by appreciable lipidation (Fig. 6A, lanes 5 and 6). After folding has been completed, apoB28F appears to undergo a relatively slow (as compared with the usual cotranslational pathway) process of lipidation that is easily detected within 15 min of DTT removal. The lipidation that occurs after the completion of folding is MTPdependent as it was not observed in CHO cells (not shown).
Characterization of ApoB28F-containing Lipoproteins Formed by the Post-translational Assembly Pathway-To determine if the apoB28F-containing lipoprotein particles formed post-translationally were secretion competent, transfected HepG2 cells were pulse-labeled with [ 35 S]Met/Cys in the presence of DTT to produce predominantly reduced apoB28F and then chased for either 0 or 60 min in the presence or absence of DTT. At the end of the 60-min chase in the presence of DTT, less than 2% of the apoB synthesized was recovered in the medium (Fig. 7A). When the chase was performed in the absence of DTT the percentage of apoB28F recovered in the medium increased by greater than 4-fold, confirming that the lipoproteins formed post-translationally underwent secretion.
To further characterize the nature of the apoB28F-containing lipoproteins formed post-translationally, their buoyant density gradient profile was examined. Transiently transfected HepG2 cells were pulse-labeled for 10 min in the presence of DTT and chased in the absence of DTT. After the chase, media was recovered, concentrated, and subjected to density gradient centrifugation as described for Fig. 1A. Although approximately 80% of the labeled apoB28F in media from such an experiment is due to post-translational folding and assembly (Fig. 7A), no evidence was obtained to indicate that the buoyant density profile of this lipoprotein population differed from that observed in control cells (Fig. 1A). This result, along with the secretion data (Fig. 7A), indicates that, although the usual pathway for apoB assembly occurs cotranslationally, apoB28F is capable of undergoing a post-translational folding and assembly process which gives rise to a native, secretion-competent lipoprotein particle. As is the case for normal cotranslational assembly (Fig. 4), the post-translational pathway also demonstrates that amino-terminal folding is a prerequisite for apoB's capacity to undergo MTP-dependent assembly with lipid.

DISCUSSION
Recent studies in hepatoma cells indicate that translation of the amino-terminal ϳ25% of apoB is sufficient to produce a small lipid core-containing lipoprotein particle (14,19). Once formed, this small dense particle appears to undergo further MTP-dependent enlargement and maturation concomitant with the ongoing translation of apoB (8). Additional lipid may also be added to nascent lipoprotein particles post-translationally (48 -50). While a conceptual framework exists for understanding how a small dense lipoprotein formed by the aminoterminal ϳ25% of apoB can be enlarged by both co-and posttranslational mechanisms, little is known about how the process of lipoprotein formation is initiated. Considering the cotranslational nature of lipoprotein assembly we assumed that steps required to initiate the formation of a small dense lipoprotein particle (e.g. that formed by apoB28) should be the same as those required for native forms of apoB48 and -100. We, therefore, examined the biogenesis of apoB28F with a particular focus on events that precede and, therefore, may be responsible for its capacity to function as an acceptor for MTPmediated lipid transfer in the hepatic ER.
Using both electrophoretic mobility changes during nonreducing SDS-PAGE and the onset of DTT resistance as criteria for protein folding in vivo, as used extensively by others (33-36, 39, 47), the rate and extent of folding of the amino-terminal domain of apoB28F was examined in both hepatoma and CHO cells. As with many disulfide bonded proteins, these studies revealed that apoB28F and, we presume by analogy apoB100, forms the bulk of its amino-terminal disulfide bonds during the process of translation. Unlike other proteins, however, apoB28F did not display a distinct temporal window in which its disulfide bonds were sensitive to DTT-mediated reduction in the ER (9,34,36,39,40). Interestingly, this uniquely rapid onset of DTT resistance was not dependent upon MTP or lipoprotein formation, as the rate and extent of folding was similar in CHO cells, a cell line that can neither lipidate nor secrete apoB28F (Fig. 3). Hence, the rapid folding of the aminoterminal domain is an autonomous property of the apoB protein and not a consequence of its extensive assembly with lipids.
The underlying basis for the rapid onset of DTT resistance of apoB28F is not fully understood. The ability of DTT to reduce newly disulfide bonded proteins in the ER is an ATP-requiring event suggesting the involvement of ER-localized chaperones (34,35). That the amino-terminal domain of apoB28F progresses so rapidly to a form that resists DTT may reflect its rapid rate of folding and/or its unusually short-lived interactions with ER chaperones. Irrespective of its underlying basis, it is clear that the disulfide bonds in the amino-terminal domain of apoB achieve DTT resistance cotranslationally. The amino terminus of apoB may, therefore, represent a modular domain whose folding occurs during its translation and whose function is essential for subsequent stages of lipoprotein assembly. This latter possibility was explored by examining the intracellular fate of apoB28F translated under conditions that blocked folding of its amino-terminal domain.
When folding of the amino-terminal domain was disrupted by preincubation of transfected HepG2 cells with 2 mM DTT, apoB28F was unable to recruit sufficient lipid to float at d Ͻ 1.25 g/ml. In contrast, ϳ50% of apoB28 was present in d Ͻ 1.25 g/ml fraction under control labeling conditions. This implies that, the strong lipid binding sequences in the form of predicted amphipathic ␤-sheet domains positioned between apoB20 and -28 (11) are not alone capable of sequestering appreciable lipid in vivo without the participation of the amino-terminal domain. This also indicates that the primary defect in apoB100 secretion caused by misfolding of the amino-terminal domain, which we observed in an earlier study (9), is most likely the result of an inability to recruit lipid and not due to production of a lipoprotein particle that is structurally aberrant and, therefore, incapable of anterograde sorting within the secretory pathway. Nevertheless, it is a formal possibility that, unlike apoB28F, longer forms of apoB with their more extensive lipid binding sequences may have the capacity to overcome the assembly block caused by misfolding of the amino terminus.
To further examine the relationship between amino-terminal folding and MTP-dependent lipid transfer to apoB, we explored the possibility that apoB28-containing lipoproteins could be assembled post-translationally. Although in many cases disulfide bond formation occurs cotranslationally, studies in DTTtreated cells have demonstrated that artificially reduced proteins readily undergo post-translational folding once the normal oxidizing conditions within the ER are re-established (30,39,40,47,51). When DTT was removed from transfected hepatoma cells, reduced apoB28F underwent a process of posttranslational folding, similar to that observed for albumin (Fig.  5) and other disulfide bonded proteins. Although the posttranslational folding of apoB28F into a DTT-resistant form was nearly complete within 5 min of DTT removal, very little posttranslational lipidation was observed within this time frame (Fig. 6). Hence, the folding of the amino-terminal domain of apoB into the DTT-resistant form is not, in and of itself, accompanied by extensive lipidation. However, after 15 min in the absence of DTT, a considerable proportion of apoB28F was chased from the d Ͼ 1.25 fraction to the d Ͻ 1.25 fraction. This result provided further compelling evidence that lipoprotein assembly cannot proceed until a critical function performed by the amino-terminal domain has been completed.
An alternative explanation for the inhibitory effects of DTT on apoB28F assembly is that, rather than inhibiting apoB folding, DTT may reversibly inhibit one or more factors necessary for its assembly with lipids. This possibility is unlikely for several reasons: the activity of the most likely such factor, MTP, is unaffected in vitro by 2 mM DTT (not shown). Furthermore, the secretion and buoyant density of apoB100 secreted from HepG2 cells is completely unaffected by DTT if it is added after the translation of the amino-terminal ϳ20% of the protein has been completed (9). Therefore, any cellular processes required for the cotranslational enlargement of apoB20-containing particles is unaffected by DTT. Most critically, however, preliminary results indicate that, within the amino-terminal 11% of apoB, substitution of single cysteine pairs involved in disulfide bond formation with serine residues, profoundly inhibits apoB28F secretion and intracellular processing. 2 In contrast to the apoB28F results reported here, previous studies in HepG2 cells indicated that once reduced, apoB100 was incapable of undergoing post-translational assembly into a secretion competent form. At least two possible reasons may be responsible for the difference in behavior between apoB100 and apoB28F. First, apoB28F was found to be quite stable in reduced form. After a 10-min pulse in the presence of DTT, followed by a 60-min chase in the absence of DTT, 75 Ϯ 7% of the apoB28F synthesized under reducing conditions was recovered from cell pellets and media. When DTT was maintained throughout the chase the recovery was 89 Ϯ 14% (both values are mean Ϯ S.D., n ϭ 3). This stability is in contrast to apoB100 where only 16% of apoB synthesized under reducing conditions was recovered in cells and media after a 60-min chase (9). Considering this dramatic difference in intracellular stability, it is possible that the rate of degradation for reduced, unlipidated apoB100 far exceeds its potential rate of post-translational folding and lipidation. The other feature that distinguishes apoB28F from apoB100 is the extent of their amphipathic lipophilic domains. It is possible that the capacity of the ER protein folding machinery to mediate the disaggregation and/or refolding of apoB's lipophilic domains, once translocated into the ER in unlipidated or underlipidated form, may be inversely related to the length of these domains and hence their degree of aggregation.
Although the present results demonstrate a distinct stage in apoB folding which precedes and appears to be essential for its MTP-dependent lipidation, the results do not directly demonstrate the specific role played by the amino-terminal domain of apoB. We had previously hypothesized that folding of the amino-terminal domain of apoB may be accompanied by an autonomous phospholipid recruitment step which would provide a lumenal lipid surface for MTP-mediated lipid transfer in the ER (9). This prediction is based, in part, on the fact that the amino-terminal domain of apoB can solubilize dimyristoylphosphatidylcholine vesicles in vitro (16) and that enrichment of hepatic membranes with the phosphatidylcholine/ethanolamine analog, phosphatidylmonomethylethanolamine, reduces the secretion of even small amino-terminal forms of apoB (52). In the present studies, folding of the amino-terminal domain of apoB was not accompanied by substantial lipid recruitment. However, depending upon the amount of lipid associated with amino-terminal folding, density gradient centrifugation may not be a sufficiently sensitive means of detecting such proteinlipid interactions. Recent results of Wu et al. (53) indicate that during the assembly of apoB100 in HepG2 cells, MTP and apoB100 form a complex which can be coimmunoprecipitated. It has since been shown that apoB-MTP interactions occur for forms of apoB containing as little as the amino-terminal 13% of the protein (54). These results indicate that a direct apoB-MTP interaction may be necessary for the selective transfer of lipid from the ER membrane or some other donor site to nascent apoB. The strict dependence of amino-terminal folding on the capacity of apoB to initiate lipid sequestration indicates that this domain may either directly (via a protein-protein interaction) or indirectly (via a protein-lipid interaction) mediate a critical apoB-MTP interaction necessary for lipoprotein formation.
In conclusion, it appears that while much of apoB is composed of domains with avid and in some cases irreversible lipid binding properties, these sequences alone may not be sufficient to drive the normal process of lipoprotein assembly. In addition to apoB's lipophilic sequences, a relatively soluble amino-terminal domain appears to play an essential role in the initiation and perhaps subsequent stages of lipoprotein particle formation. Future studies of the structure and function of this domain will likely reveal important mechanistic insights into the process of apoB-containing lipoprotein assembly. Finally, the development of the post-translational assay reported here, which uncouples the assembly of apoB with lipids from the process of translation, may provide an additional valuable system to further dissect steps required for apoB's unique form of assembly in the hepatic ER.