Apolipoprotein B sequence requirements for hepatic very low density lipoprotein assembly. Evidence that hydrophobic sequences within apolipoprotein B48 mediate lipid recruitment.

We studied the structural requirements of apolipoprotein (apo) B for assembly of very low density lipoproteins (VLDL) using rat hepatoma McA-RH7777 cells expressing human apoB (h-apoB). Recombinant h-apoB48, like endogenous rat apoB48 (r-apoB48), was secreted as VLDL in addition to high density lipoproteins (HDL) by transfected cells, indicating that the N-terminal 48% of apoB contains sequences sufficient for VLDL assembly. Truncation of the C terminus of h-apo-B48 to -B42 or -B37 had little effect on the ability of apoB to assemble VLDL, whereas truncation to -B34 or -B29 markedly diminished or abolished VLDL formation. None of the truncations affected the integration of apoB into HDL. To determine whether the ability to assemble VLDL is governed by apoB length or by sequences beyond apoB29, we created chimeric proteins that contained human apoA-I and a segment derived from between the C-terminal 29 and 34%, 34 and 37%, or 37 and 42% of apoB100. The resulting chimeras, namely AI/B29-34, AI/B34-37, and AI/B37-42, were secreted by the transfected cells as lipoproteins with buoyant density (d < 1.006 g/ml), electrophoretic mobility (pre-β), and size characteristics of human plasma VLDL. The chimeras could assemble discrete VLDL particles devoid of endogenous r-apoB100, and could actively recruit triglycerides and phospholipids into the lipoproteins. However, these chimeras were secreted inefficiently. Pulse-chase analysis showed that less than 5% of the newly synthesized AI/B proteins were secreted, and more than 70% was degraded intracellularly. Degradation of the chimeras could be blocked by the cysteine protease inhibitor N-acetyl-leucyl-leucyl-norleucinal, but the treatment did not enhance their secretion. Protease protection analysis of microsomes isolated from transfected cells indicated that >65% of AI/B chimeras (compared with <25% of r-apoB100) were inaccessible to exogenous trypsin. These data suggest that the recruitment of large quantities of triglycerides during VLDL formation is not governed simply by apoB length, but is mediated by short hydrophobic sequences ranging from 152 to 237 amino acids (3-5%) of apoB. The existence of multiple such hydrophobic sequences within apoB48 may facilitate efficient assembly of hepatic VLDL particles.

tities of triglycerides during VLDL formation is not governed simply by apoB length, but is mediated by short hydrophobic sequences ranging from 152 to 237 amino acids (3-5%) of apoB. The existence of multiple such hydrophobic sequences within apoB48 may facilitate efficient assembly of hepatic VLDL particles.
Human apolipoprotein (apo) 1 B is a large and amphipathic protein which serves as the structural backbone for the assembly of triglyceride-rich lipoproteins. Two forms of apoB are found in human plasma: the full-length apoB100 consisting of 4536 amino acids, and apoB48 representing the N-terminal 2152 amino acids (ϳ48%) of the protein (1). The enormous size and lipophilic properties of these two polypeptides have been suggested to underlie their ability to recruit triglycerides into hepatic very low density lipoproteins (VLDL) and intestinal chylomicrons (2). Early studies of C-terminally truncated apoB variants found in human familial hypobetalipoproteinemia (3) have provided in vivo evidence that the ability of apoB to recruit lipids is compromised by the truncation. Subsequently, the impact of apoB length on the extent of lipid recruitment during apoB-containing lipoprotein (LpB) formation has been demonstrated by several laboratories using transfected cell lines expressing the truncated forms of human apoB (h-apoB) (4 -9). The in vivo observations and cell culture studies combined have unambiguously shown that the length of apoB polypeptide plays an important role in the formation of LpB. Despite this evidence, several other in vivo observations have indicated that the ability of apoB to form VLDL may not be solely determined by the length of the polypeptide. The most compelling evidence is that apoB48, a protein only half the size of apoB100, can mediate the assembly and secretion of both chylomicrons (in the intestine) and VLDL (in the liver of rats and mice). Moreover, some of the C-terminally truncated forms of apoB (e.g. apoB37) have been detected in plasma VLDL of subjects with hypobetalipoproteinemia (3). Therefore, the ability to assemble triglyceride-rich lipoproteins can not be simply a function of apoB length as suggested by the cell culture studies. Rather, amino acid sequences that reside within the N-terminal 48%, or even within 37%, of apoB100 may play important roles in the assembly of triglyceride-rich LpB.
Assembly of VLDL in the liver is a complex process perhaps occurring in multiple steps. Experimental evidence accumulated so far suggests that assembly of VLDL, at least the apoB48-containing VLDL in rat liver, may proceed through two steps. The initial step occurs as the apoB polypeptide translocates from the cytosolic to the lumenal side of the endoplasmic reticulum (ER) membrane (8), a process presumably facilitated by the microsomal triglyceride transfer protein (MTP) (10) in the presence of sufficient lipid supply. Considerable progress has been made in understanding the apoB sequences involved in the translocation process, showing that the rate of apoB translocation across the ER membrane could regulate the level of apoB secretion. It has been observed that translocation of apoB may pause (11,12) or even stop (13), producing several transmembrane intermediates of apoB. Indeed, immunoreactive apoB has been detected on the cytosolic side of hepatic microsomal membranes (14 -17). It has been suggested by some (18), but disputed by others (19), that translocation pausing is essential for lipidation of the apoB polypeptides. Translocation arrested apoB polypeptides, resulting from insufficient lipid supply, probably do not participate in LpB assembly and are ultimately degraded. Little is known about the mechanism that is responsible for apoB degradation, nor is it clear which specific apoB sequences are involved in lipid recruitment during the initial step of VLDL assembly. Biochemical evidence for the second step of hepatic apoB48-VLDL assembly has recently been presented (20). It has been suggested that following an initial co-translational assembly step which generates an apoB48-containing species resembling high density lipoprotein (HDL), a so-called "second step" takes place in which a large amount of neutral lipid is recruited to convert this primordial apoB48-HDL to apoB48-VLDL. In rat hepatoma McA-RH7777 cells, the second step was dependent upon oleate supplementation of the culture medium (20). This two-step model is in accord with previous observations (21) and is supported by recent studies (22) using primary rat hepatocytes. The current study was intended to search for sequences within h-apoB48 that mediate the second step lipid recruitment during VLDL synthesis.
The in vivo data obtained from studies of human hypobetalipoproteinemia also suggest that C-terminal truncation of apoB100 may impair apoB's ability to recruit lipid during VLDL assembly. While the relatively large truncated apoB forms (e.g. ՆapoB37) were invariably found in VLDL and low density lipoproteins (LDL) in addition to HDL, the short forms (e.g. apo-B31 and -B32.5) were found only in HDL and the d Ͼ 1.21 g/ml fractions (3). These data imply that apo-B31 and -B32.5 may not contain the sequences necessary for VLDL formation, whereas apoB37 does. A recently proposed pentapartite structural model of human apoB100 (h-apoB100) suggests that the protein is composed of three amphipathic ␣ helix domains and two amphipathic ␤ strand domains (23). One of the ␤ strand domains is located within apoB48 (between apo-B18 and -B43) which may represent the functional lipid-binding domain engaged in the second step lipidation to form VLDL. Alternatively, the in vivo data from hypobetalipoproteinemia may suggest that the requirement for the second step lipidation of apoB could be determined by the length of lipidbinding sequences, since hydrophobic structures that potentially mediate apoB binding to lipids are distributed quite evenly throughout the h-apoB100 molecule (2). In this study, we used five C-terminally truncated h-apoB forms and three chimeras containing h-apoB segments that were expressed in stably transfected McA-RH7777 cells to determine whether the sequence requirement for hepatic VLDL assembly is deter-mined by apoB length or by domains within the protein. Our results suggest that the amphipathic ␤ strands between the C termini of apo-B29 and -B42 (B29 -42) play important roles in VLDL assembly, and segments (as short as 152 amino acids) derived from B29 -42 are sufficient to mediate VLDL formation.

EXPERIMENTAL PROCEDURES
Materials-DNA restriction and modification enzymes, and oligonucleotide linkers were purchased from New England Biolabs or from Life Technologies, Inc. Sequenase (Version 2.0) was obtained from U. S. Biochemical Corp. All reagents for cell culture were purchased from Life Technologies, Inc., and reagents for polyacrylamide gel electrophoresis (PAGE) from Bio-Rad Laboratories. Immunoblotting reagents, [ 35 S]methionine/cysteine (ProMix TM , 1000 Ci/mmol) for cell labeling and [ 35 S]methionine (1000 Ci/mmol) for in vitro translation were obtained from Amersham Corp. [2-3 H]Glycerol (9.6 Ci/mmol) was obtained from ICN Pharmaceuticals Canada Ltd. Silica gel plates (20 cm ϫ 20 cm, Silica Gel 60) for thin layer chromatography were from EM Science (Gibbstown, NJ). Sheep polyclonal antibodies, which recognized human or rat apolipoproteins, and the cysteine protease inhibitor Nacetyl-leucyl-leucyl-norleucinal (ALLN) were purchased from Boehringer Mannheim. Protein A-Sepharose CL6B and cyanogen bromide activated Sepharose were obtained from Pharmacia Biotech Inc. (Montreal, Quebec). Monoclonal anti-human apoB antibody 1D1 (24) and anti-human apoA-I (h-apoA-I) antibody 5F6 (25) were provided by R. Milne and Y. Marcel (Ottawa Heart Institute). Polyclonal (26) and monoclonal (27) antibodies to rat apoB (r-apoB) were gifts from R. Davis (San Diego State University) and L. Wong (Louisiana State University), respectively. Rabbit antibody to bovine protein disulfide isomerase was a gift from M. Michalak (University of Alberta).
Construction of Truncated ApoB Expression Plasmids-All expression plasmids were prepared using the vector pCMV5 (28) from which polylinker sequence between the HindIII and XbaI restriction sites had been removed. Truncated apoB cDNAs were prepared using pB53L-L (29) as starting material. The plasmid pB42 (nucleotides 20 -5849 of the apoB100 cDNA) was generated by excision of the EcoRI to ClaI fragment from pB53L-L and ligation into pCMV5 that had been digested with the same enzymes. The plasmid pB37 (nucleotides 20 -5290 of the apoB cDNA) was created by removal of the SalI-SalI fragment from pB53L-L and recircularization of the remaining fragment with T4 DNA ligase. To create the plasmid pB34 (nucleotides 20 -4838 of the apoB cDNA), pB37 was digested with DraI, ligated with an MluI linker, and digested with EcoRI and MluI to obtain the apoB insert which was then ligated into pCMV5 that had been digested with EcoRI and MluI. The plasmid pB29 (nucleotides 20 -4124 of the apoB cDNA) was generated by XhoI digestion of pB34, end-filling with Klenow polymerase, ligation with MluI linker, digestion with EcoRI and MluI to release the B29 fragment and ligation of the fragment into pCMV5 that had been digested with the same enzymes. The non-apoB amino acids introduced by cloning at the C termini of the apoB proteins were: KL in apoB42, RL in apoB37, FDAYR in apoB34, and DAYR in apoB29.
Construction of Expression Plasmids for ApoA-I/B Fusion Proteins-An EcoRI-StuI fragment encoding the N-terminal 217 amino acids of h-apoA-I was excised from the apoA-I cDNA (30). The blunt StuI end was ligated with a ClaI linker, and the resulting 800-base pair fragment was ligated into pCMV5 that had been digested with EcoRI and ClaI to create pAI/StuI. The expression plasmids, pAI/B29 -34, pAI/B34 -37 and pAI/B37-42 were assembled using the 800-base pair EcoRI-StuI fragment and corresponding apoB sequences. Appropriate linker oligonucleotides (NcoI linker for pAI/B29 -34 and pAI/B37-42, MluI linker for pAI/B34 -37) were ligated to the 3Ј end of the EcoRI/StuI fragment to accept the apoB sequences. To obtain the B29 -34 segment (nucleotides 4124 -4838 of the apoB cDNA), pB34 was digested with XhoI (nucleotide 4124 of the apoB cDNA), end-filled, ligated with NcoI linkers, and excised by digestion with NcoI and MluI (located in the pCMV5 polylinker). The B34 -37 segment (nucleotides 4838 -5290 of the apoB cDNA) was obtained from pB37 by DraI digestion (nucleotide 4838 of the apoB cDNA), ligated with MluI linker, and excised by MluI and XbaI digestion. For the B37-42 segment (nucleotides 5290 -5848 of the apoB cDNA), a 550-base pair SalI-SalI fragment (from nucleotide 5290 of the apoB cDNA to pCMV5 polylinker) was excised from pB42, end-filled, ligated with NcoI linker, and digested with NcoI and ClaI (located in the pCMV5 polylinker). To assemble the pAI/B plasmids, the pCMV5 vector was digested with EcoRI and MluI (for pAI/B29 -34), XbaI (for pAI/B34 -37), or ClaI (for pAI/B37-42). Compatibly linkered apoA-I and apoB fragments were inserted into appropriately digested pCMV5 by a three-piece ligation reaction. Extraneous amino acids, introduced at the junctions between the apoA-I and apoB sequences by the ligation procedures, were: PWL in pAI/B29 -34, DAS in pAI/B34 -37, and PW in pAI/B37-42. For in vitro transcription and translation studies, the AI/B chimera coding sequences were excised from the pCMV expression constructs by EcoRI and MluI digestion and inserted into pSPT19 (Pharmacia). Expression of AI/B chimeras was driven by the SP6 promoter. All expression plasmids, purified twice by CsCl gradient centrifugation, were analyzed by restriction enzyme digestion, and by DNA sequencing to verify the pAI/B junctions and the Cterminal amino acid codons, respectively.
Cell Culture-McA-RH7777 cells (CRL-1601) were obtained from the American Type Culture Collection (Rockville, MD). Cell culture conditions, transfection protocols, and selection of stable transformants were performed as described previously (31) using the truncated apoB and AI/B chimera plasmids generated above.
Characterization of Medium Lipoproteins-Transfected cells (70 -80% confluent in 100-mm dishes) were incubated for up to 24 h in Dulbecco's modified Eagle's medium (DMEM) containing 20% serum (standard medium), or in oleate-supplemented medium (OS medium, standard medium supplemented with 0.4 mM oleate conjugated with bovine serum albumin). The conditioned media (5 ml) were fractionated by ultracentrifugation in a sucrose density gradient (20) to separate lipoproteins between 1.006 and 1.17 g/ml (47% (w/v) sucrose) or 1.25 g/ml (54% (w/v) sucrose). After centrifugation (35,000 rpm, 65 h in SW41 Ti rotor), 12 fractions (ϳ1 ml each) were collected from the top of the tube. Lipoproteins in each fraction were concentrated by immunoprecipitation with polyclonal antibodies to apoB (for truncated h-apoB variants) or to apoA-I (for chimeras), resolved by SDS-PAGE, and visualized by immunoblot analysis as described previously (8).
Agarose Gel Electrophoresis-Ultracentrifugally isolated VLDL were fractionated on thin agarose films (Lipogel, Beckman Instruments) according to the manufacturer's recommendations. After completion of electrophoresis, the lipoproteins were transferred to nitrocellulose, and the presence of r-apoB, h-apoB or h-apoAI was revealed by immunoblot analysis using the appropriate antibodies as described previously (32).
Metabolic Labeling Studies-Metabolic labeling and pulse-chase analysis of the recombinant proteins using [ 35 S]methionine/cysteine were performed as described previously (9). In some experiments, serum, oleate and the cysteine protease inhibitor ALLN were included in the labeling or chase media as indicated in the figure legends. In experiments where both lipids and apolipoproteins were labeled, the cells (60-mm dishes) were incubated with 15 Ci [ 3 H]glycerol and 200 Ci of [ 35 S]methionine/cysteine for up to 4 h in 2 ml of cysteine-and methionine-free DMEM containing 20% serum. Since the cell density has been shown to have an important effect on lipid and apolipoprotein synthesis and secretion in McA-RH7777 cells (33), we performed labeling experiments at similar cell densities (cell protein variation within 15%) with the transfected cells. Medium VLDL were isolated by ultracentrifugation, and AI/B chimera-containing species were purified by immunoadsorption (see below). Lipids were extracted from the immunoadsorbant beads with chloroform:methanol (1:1, v/v) and resolved by thin layer chromatography as described previously (34). The 3 H-labeled triglyceride and phosphatidylcholine were quantified by liquid scintillation counting. Apolipoproteins were released from the delipidated immunoadsorbant beads using SDS-PAGE sample buffer, resolved by SDS-PAGE (3-15% gels), and 35 S-labeled proteins were visualized by autoradiography as described previously (9).
Immunoadsorption of ApoA-I/B Chimeras-A h-apoAI-specific immunoadsorbant was prepared by covalently coupling monoclonal antibody 5F6 to cyanogen bromide activated agarose. Ultracentrifugally isolated VLDL were incubated at 4°C with the immunoadsorbant beads under non-denaturing conditions for 16 -24 h in phosphate-buffered saline. The beads were then pelleted in a microcentrifuge and washed four times with phosphate-buffered saline, and the bound fraction was extracted for analysis of lipids or protein as described above.
Microsome Isolation and Protease Digestion-All isolation procedures were performed at 0 -4°C. Transfected cells from two 100-mm dishes were combined and disrupted by 20 passes through a ball bearing homogenizer (H & Y Enterprise, Redwood City, CA) in buffer containing 10 mM Tris-HCl, pH 7.4, 250 mM sucrose, 0.1 mM leupeptin, 0.1 mM phenylmethylsulfonyl fluoride, aprotinin (100 kallikrein inhibitor units/ml), and ALLN (40 g/ml). Intact cells and nuclei were removed by centrifugation for 10 min at 10,000 ϫ g, and microsomes were isolated from the postnuclear supernatant as a pellet following centrifugation for 15 min at 400,000 ϫ g in a Beckman TLA-100.3 rotor. The microsomes were resuspended in buffer containing 10 mM Tris-HCl, pH 7.4, and 250 mM sucrose. To release the lumenal content fraction, the microsomes were treated with 0.1 M sodium carbonate, pH 11.3, on ice for 30 min as described elsewhere (20). For protease digestion experiments, protease inhibitors were removed from the microsome preparations by ultracentrifugation in buffer without protease inhibitors, and proteolysis was performed immediately (35).
In Vitro Transcription and Translation of AI/B Chimeras-The AI/B expression plasmids (prepared in pSPT19) were linearized by PvuII digestion and transcribed for 60 min using SP6 polymerase (Riboprobe Core System, Promega). The transcripts were purified and subsequently translated in rabbit reticulocyte lysate (Promega) containing [ 35 S]methionine in the presence of rat liver microsomes (16). Aliquots of the translation mixture were treated with trypsin in the presence or absence of Triton X-100 as described above.
Other Methods-Cell protein was quantified according to Lowry et al. (36) using bovine serum albumin as a standard.

C-terminally Truncated ApoB Proteins Retain the Ability to
Assemble VLDL-It was reported that endogenous r-apoB48 of McA-RH7777 cells could assemble VLDL through two steps in an oleate-dependent manner (20). In this study we first determined the effect of sequential C-terminal truncation of human apoB48 on its ability to assemble VLDL. Four truncated proteins: h-apo-B29, -B34, -B37, and -B42, in addition to h-apoB48 (33) were expressed in McA-RH7777 cells (Fig. 1A). The h-apoB of the appropriate molecular weight and immunoreactivity was detected in the medium of each stably transfected cell line by immunoblot analysis (Fig. 1B). Under standard cell culture conditions (i.e. DMEM containing 20% serum), all of the secreted h-apoBs were associated with HDL (d ϭ 1.08 -1.18 g/ml) species (data not shown), consistent with previous results using the same expression system (4,8). When the culture medium was supplemented with oleate (0.4 mM), a portion of the secreted h-apoB48 was recovered in VLDL (Fig. 1C, fractions 1 and 2) in addition to HDL (fractions 8 -10). This result suggests that recombinant h-apoB48, like r-apoB48, can also assemble VLDL in an oleate-dependent manner in the transfected cells. The oleate-dependent VLDL formation was also observed for the C-terminally truncated h-apoBs such as h-apo-B42, -B37, and -B34, indicating that the ability to assemble VLDL was not severely affected by the truncation (although VLDL formation by h-apoB34 was compromised) (Fig. 1C). In contrast, formation of VLDL was entirely abolished by truncation to h-apoB29. None of the truncations affected the secretion of apoBs as HDL (Fig. 1C). Sequential flotation of the secreted lipoproteins confirmed that h-apo-B48, -B42, -B37 and -B34 indeed formed VLDL (d Ͻ 1.006 g/ml) (data not shown). These data are reminiscent of observations in human hypobetalipoproteinemia indicating that apoB37 forms VLDL but apo-B31 and -B32.5 do not (3). Thus, these results provide in vitro evidence that the lipid-recruiting sequences for VLDL formation most likely reside beyond the N-terminal 30% of apoB100.
A-I/B Chimeras Synthesized by McA-RH7777 Cells Can Assemble VLDL-The failure of apoB species containing less than 30% of the N-terminal sequences (i.e. ՅapoB30) to form VLDL might also be explained by the lack of sufficient polypeptide length. To determine whether the ability of apoB to recruit lipid was determined by the protein length or by unique properties of the apoB sequences, we designed chimeras that contained segments of apoB derived from between the C termini of apo-B29 and -B42, a region enriched with amphipathic ␤ strands (23). The B29 -42 region was dissected into three segments, B29 -34, B34 -37, and B37-42 (between the C-terminal 29 and 34%, 34 and 37%, and 37 and 42%, respectively, of apoB100), using convenient restriction enzyme sites, and each segment was ligated in-frame to the 3Ј end of the h-apoA-I cDNA ( Fig. 2A). The h-apoA-I sequence is composed almost entirely of amphipathic ␣ helices (37) and is devoid of cysteine residues and N-glycosylation sites. The number of amino acids in the h-apoB segments varied from 237 in B29 -34 (containing two cysteines and three predicted N-glycosylation sites) (38), 152 in B34 -37 (containing one cysteine), to 187 in B37-42 ( Fig.  2A). The corresponding chimeras, namely AI/B29 -34, AI/B34 -37, and AI/B37-42, were stably expressed in McA-RH7777 cells. Fig. 2B shows that proteins with the predicted molecular mass and reactivity with antibodies to either h-apoA-I (left panel) or h-apoB (right panel) were expressed by the transfected cells. No degradation products of the chimeras were detected within the cells or in the medium by immunoblotting, nor was there any indication of proteolytic cleavage at the junction between apoA-I and apoB sequences of the chimeras.  Fig. 1C, except that 54% (w/v) sucrose was used in the bottom layer to achieve the higher density limit (1.24 g/ml), and antibody 5F6 was used to detect AI/B chimeras. D, secretion of 35 S-labeled r-apoB100 and AI/B29 -34 on lipoproteins. The AI/B29 -34-producing cells were labeled with [ 35 S]methionine/cysteine in standard medium (f) or OS medium (å). At the indicated time, VLDL (d Ͻ 1.006 g/ml) and HDL (d Ͼ 1.006 g/ml) were isolated from the medium by ultracentrifugation. 35 S-Labeled r-apoB100 (top panel) and AI/B29 -34 (bottom panel) were purified from the lipoproteins by immunoprecipitation, resolved by SDS-PAGE, and quantified by liquid scintillation counting.
Density gradient ultracentrifugation of the medium lipoproteins (Fig. 2C) showed that each chimera was secreted as VLDL when cells were cultured in oleate-supplemented media (the top three panels). The ability to associate with VLDL might be correlated with the h-apoB segment length within the chimeras, since the longest chimera (AI/B29 -34) was more predominant on VLDL than either AI/B34 -37 or AI/B37-42. The control protein AI/StuI, however, was secreted exclusively as HDL (fractions 7-9 of the panel labeled as AI/StuI). Unlike the C-terminally truncated apoBs (Fig. 1C), only small amounts of the AI/B chimeras were secreted as HDL (Fig. 2C). Metabolic labeling studies using AI/B29 -34 as representative (Fig. 2D) showed that secretion of 35 S-AI/B29 -34 VLDL (left panels, d Ͻ 1.006 g/ml), like endogenous r-apoB100 VLDL, increased 5-fold with oleate supplementation. Secretion of the radiolabeled AI/ B29 -34 and r-apoB100 in dense lipoproteins (right panels, d Ͼ 1.006 g/ml) was less affected by oleate supplementation. These data were the first indication that a small segment of h-apoB in the chimera, either B29 -34, B34 -37, or B37-42, was sufficient to mediate lipid recruitment to form VLDL.
Characterization of A-I/B Chimera-containing VLDL-It has been shown that formation of apoB48 VLDL in McA-RH7777 cells differs from that of apoB100 VLDL by its oleate dependence (20). We examined the oleate dependence of AI/B chimera VLDL formation and found that it resembled that of apoB48 VLDL. Fig. 3A shows data obtained using AI/B29 -34producing cells as representative. In the absence of oleate, both AI/B29 -34 (Fig. 3A, fraction 7-9, without oleate, AI/B29 -34) and r-apoB48 (fractions 8 -9, without oleate, B48) were secreted exclusively in fractions of d Ͼ 1.02 g/ml, whereas r-apoB100 was found primarily in fractions of d Ͻ 1.02 g/ml (fractions 1-3, without oleate, B100). Upon oleate supplementation, AI/B29 -34, like endogenous r-apoB48, was secreted predominantly as VLDL in addition to small amounts as HDL (Fig. 3A, with oleate, second and third panels). Formation of VLDL in this oleate-dependent fashion was also observed with chimeras AI/ B34 -37 and AI/B37-42 produced by the respective cell lines (data not shown). Secretion of r-apoB100 as VLDL was stimulated by oleate (Fig. 3A, compare without oleate to with oleate). In experiments where the medium lipoproteins were fractionated by sequential flotation at d ϭ 1.006 g/ml (Fig. 3B), we confirmed that the oleate-induced VLDL indeed had buoyant density less than 1.006 g/ml (compare lanes labeled as ϪOA and ϩOA in vivo of the left panels). Secretion of the chimera as VLDL (d Ͻ 1.006 g/ml) was not a result of nonspecific association with preexisting VLDL, since mixing of the conditioned medium postsecretionally with 0.4 mM oleate (ϩOA in vitro) did not result in formation of VLDL containing the chimeras. As was the case for density gradient ultracentrifugation in Fig.  3A, sequential flotation (Fig. 3B, right panels) also demonstrated that secretion of AI/B chimeras in the d Ͼ 1.006 g/ml fraction was not affected by oleate. In a separate experiment where triglyceride-rich VLDL secreted by McA-RH7777 cells were mixed with the conditioned medium from the chimeraproducing cells, the chimeras also failed to transfer post-secretionally to the lipid-enriched VLDL (data not shown). As shown below, the AI/B chimera VLDL could be separated from r-apoB100 VLDL by immunoaffinity techniques. Thus, these results indicate that the characteristic nonexchangeability of apoB is retained in the AI/B chimeras, and they can, like The conditioned media were fractionated as described in Fig. 1C, and endogenous r-apoB100 (top panels), r-apoB48 (middle panels), or AI/B29 -34 (bottom panels) were visualized by immunoblotting with monoclonal antibody LRB220 (to r-apoBs) or 5F6 (to AI/B29 -34). The exposure times for r-apoB48 was 100 s, 10-fold longer than that for r-apoB100, to visualize low level expression of endogenous r-apoB48. The AI/B29 -34 panel was also overexposed to visualize the protein in HDL region. B, sequential flotation of secreted AI/B chimeras. The AI/B chimera or AI/StuI-producing cells were cultured in standard (ϪOA) or OS medium (ϩOA in vivo) for 16 h. For each chimera, an aliquot of conditioned standard medium was supplemented with 0.4 mM oleate and incubated at 37°C for 24 h (ϩOA in vitro). Each sample was separated into d Ͻ 1.006 and dϾ1.006 g/ml fractions, resolved by SDS-PAGE and visualized by immunoblotting. C, Superose 6 column chromatography of VLDL containing AI/B chimeras. Ultracentrifugally isolated VLDL (d Ͻ 1.02 g/ml) were chromatographed on Superose 6 columns (1.6 ϫ 30 cm) with PBS running buffer at a flow rate of 0.75 ml/min. Apolipoproteins in each fraction (2 ml) were concentrated by adsorption onto fumed silica (34), resolved by SDS-PAGE, and visualized by immunoblotting. The quantity of individual proteins, estimated by densitometric scanning of the resulting fluorograms, is expressed as arbitrary absorbance units (A.U.). D, agarose gel electrophoresis of VLDL containing AI/B chimeras. Ultracentrifugally isolated VLDL (d Ͻ 1.020 g/ml) were resolved by agarose gel electrophoresis, and the presence of lipoprotein species containing r-apoB, h-apoB, or AI/B chimeras were determined by immunoblotting with the indicated antibody. Human lipoprotein standards (hVLDL and hLDL) were used to define pre-␤ and ␤ mobility, respectively. endogenous r-apoB100, independently assemble VLDL particles.
Additional physicochemical analysis provided further evidence that the AI/B chimera VLDL had the characteristics of VLDL. Gel filtration chromatography using Superose 6 columns (Fig. 3C) revealed that each chimera VLDL eluted in the void volume (V 0 ), earlier than human LDL (hLDL). Agarose gel electrophoresis (Fig. 3D) showed that the chimeras (e.g. AI/ B29 -34 and AI/B37-42) formed VLDL with pre-␤ mobility, and there was no indication of aggregation of these lipoproteins. The same gel filtration characteristics (i.e. elution as VLDL) and electrophoretic mobility (i.e. pre-␤) were also observed for VLDL containing the C-terminally truncated h-apoBs such as apo-B48 and -B34 (data not shown).
Overexpression of AI/B Chimeras Stimulates Lipid Secretion-The following experiments were performed to demonstrate that AI/B chimeras were able to actively recruit lipids for VLDL synthesis. Using AI/B29 -34 as a representative and AI/StuI as control, we metabolically labeled the respective cell lines with both [ 3 H]glycerol and [ 35 S]methionine/cysteine. The VLDL (d Ͻ 1.02 g/ml) were isolated from the medium by ultracentrifugation and the AI/B-containing VLDL was purified by immunoadsorption. In a preliminary experiment which is shown in Fig. 4A, we found that the purified VLDL contained only AI/B29 -34 and lacked r-apoB100 (right lane, 5F6 bound). Unfortunately, intact AI/B containing VLDL particles could not be eluted from the immunoaffinity adsorbant for further analysis. However, using the immunoaffinity technique, we could purify the radiolabeled AI/B chimera VLDL and analyze their lipid and protein composition. The autoradiograms shown in Fig. 4B demonstrated that the separation of 35 S-labeled r-apoB100 VLDL and AI/B29 -34 VLDL secreted by AI/B29 -34-producing cells was essentially complete, and the immunoaffinity recovery of the AI/B29 -34 VLDL was 85% as determined by counting of the radiolabeled VLDL. The 35 S-labeled AI/B29 -34 specifically bound to the anti-h-apoAI affinity beads (left panel, 5F6 bound), whereas 35 S-labeled r-apoB100 bound predominantly to the anti-apoB immunoadsorbant (right panel, ␣B-VLDL). Less than 5% of the recovered 35 S-labeled r-apoB100 bound nonspecifically to the anti-apoAI affinity beads as determined by scanning densitometry. As shown in the immunoblots of Fig. 2C, AI/StuI-producing cells secreted none of the AI/StuI protein in the VLDL fraction. In control dual labeling experiments using the AI/StuI-producing cells, 35 S-labeled r-apoB100 was recovered only in VLDL fractions treated with the anti-apoB immunoadsorbant (Fig. 4C, right panel, ␣B-VLDL). Notably, VLDL that contained apoE and/or apoA-I and essentially no r-apoBs was secreted by AI/StuIproducing cells (Fig. 4C, 5F6 bound). The level of synthesis (data not shown) and secretion (Fig. 4, compare B and C) of endogenous r-apoB100 was similar between AI/StuI and AI/ B29 -34-producing cells. The McA-RH7777 cells produced very little endogenous r-apoB48 as compared with r-apoB100 (9, 34) (also see below). Endogenous rat apoE and rat apoA-I were co-precipitated with VLDL (d Ͻ 1.02 g/ml) from AI/B29 -34producing cells with both immunoadsorbants (Fig. 4B). Association of these exchangeable apolipoproteins with VLDL may reflect their physical preference for large particles (39). Lipid composition of the affinity purified VLDL samples was also determined. Fig. 5A shows that the amount of [ 3 H]triglyceride and [ 3 H]phosphatidylcholine (left panels) secreted as AI/B VLDL (5F6 bound) by AI/B29 -34-producing cells was 2-fold higher than by the control AI/StuI-producing cells. Notably, the amount of [ 3 H]triglyceride and [ 3 H]phosphatidylcholine associated with AI/B chimera VLDL (5F6 bound) was comparable to that associated with r-apoB100 VLDL (␣B bound). The 5F6bound VLDL species secreted by AI/StuI-producing cells that contained apoE and/or apoA-I but little apoB also contained triglyceride and phospholipid (Fig. 5A, 5F6 bound). The origin of these VLDL-like particles (Figs. 4C and Fig. 5A, 5F6 bound) is unclear and remains to be determined. The amount of [ 3 H]triglyceride and [ 3 H]phosphatidylcholine (right panels) associated with r-apoB100 VLDL (␣B bound) did not differ between the two cell lines. The increase in triglyceride concentration in the AI/B29 -34 VLDL was also evident by high performance thin-layer chromatography (data not shown). The initial rate of incorporation of [ 3 H]glycerol into triglyceride and phosphatidylcholine was similar in the two cell lines (Fig. 5B). These data demonstrate that the chimeras can actively recruit lipids (both neutral lipid and phospholipid) to form VLDL.
Secretion of AI/B Chimera-containing VLDL Is Inefficient and the Majority of Newly Synthesized Chimera Is Degraded by The ultracentrifugally isolated d Ͻ 1.02 g/ml fraction (VLDL) was separated into VLDL containing chimera (5F6-VLDL) or r-apoB (␣B-VLDL) by selective immunoadsorption. Proteins and lipids were extracted from the immunoadsorbant and resolved by SDS-PAGE and thin layer chromatography, respectively. C, metabolic labeling of AI/StuI-producing cells. The experiments were performed the same as B except the AI/StuI-producing cells (1.5 mg of cell protein/60-mm dish) were used. The autoradiograms in B and C were exposed for 2 days to visualize r-apoB100 (upper panels) or for 5 days for AI/B29 -34 (lower panels).
an ALLN-sensitive Protease-Although oleate supplementation stimulated chimera secretion as VLDL, the secretion efficiency of these proteins was low. Pulse-chase analysis after 4-h pretreatment with oleate (using AI/B29 -34 as representative), revealed that less than 2% of the newly synthesized chimeras were secreted as VLDL (d Ͻ 1.02 g/ml) at the end of 3-h chase (Fig. 6, A and B). Under the same conditions, the secretion efficiency of endogenous r-apoB100 was over 40% (Fig. 6, A and  B). However, since the number of 35 S-labeled AI/B29 -34 molecules synthesized was nearly 20-fold higher than that of r-apoB100 (the number of methionine and cysteine in the chimera was 20 times fewer than in apoB100, and the amount of label incorporated was about the same at time 0 in Fig. 6B), the number of chimera molecules secreted was comparable to that of r-apoB100. This calculation assumes that the cysteine/methionine content of r-apoB100 is the same as h-apoB100. Both AI/B29 -34 and endogenous r-apoB100 were secreted as VLDL (Fig. 6A, left panels) but not as HDL (right panels) by cells cultured with oleate-supplemented medium. The AI/B29 -34 chimera, like endogenous r-apoB100, displayed rapid posttranslational degradation within the cells (Fig. 6B). At the end of the chase, less than 20% of the initial labeled proteins were left in the cells. The McA-RH7777 cells synthesized (Fig. 6B) and secreted (Fig. 6A) little endogenous r-apoB48 as compared with r-apoB100.
The effects of lipid supplementation and protease inhibitor on the secretion of AI/B chimeras were further examined and compared with endogenous r-apoB100 (Table I). In the absence of lipid supplement, virtually no chimeras were secreted, nor did 20% serum alone stimulate their secretion. In contrast, low levels of secretion (ϳ12% of total) were observed for endogenous r-apoB100 under serum-free conditions. Secretion of the chimeras was stimulated (2-4% of total) only when 0.4 mM oleate, together with serum, was added to the medium. Secretion of endogenous r-apoB100, however, was stimulated (ϳ24% of total) even by serum alone. Oleate supplementation of the serum-containing media had no further stimulatory effect on the secretion of r-apoB100. The cysteine protease inhibitor ALLN blocked the intracellular degradation of endogenous r-apoB100 and the chimeras. However, ALLN did not enhance their secretion. Thus, even under conditions most favorable for VLDL synthesis and secretion, the secretion efficiency of the AI/B chimeras was significantly lower than that of endogenous r-apoB100. The protection of AI/B chimeras from intracellular degradation by ALLN suggests that sequences susceptible to the ALLN-sensitive proteolysis may colocalize with the lipidbinding regions of apoB.
The low secretion of AI/B chimeras might result from inefficient translocation of the proteins across the ER membrane. We tested this possibility by assessing the topology of these  5. Secretion of lipids associated with VLDL containing the AI/B chimeras or r-apoB100. The experiments were performed as described in Fig. 4, B and C. A, the 3 H-labeled triglyceride ( 3 H-TG) and phosphatidylcholine ( 3 H-PC) associated with the purified VLDL. B, synthesis of triglyceride and phosphatidylcholine in the cells. Lipids were separated by thin layer chromatography and quantified by liquid scintillation counting: f, AI/StuI-producing cells; å, AI/B29 -34-producing cells. TG, triglyceride; PC, phosphatidylcholine.
proteins with respect to the microsomal membranes using protease protection assay. In the intact microsomes isolated from the transfected cells, a significant portion of AI/B29 -34 (67%), AI/B34 -37 (68%), and the control AI/StuI (60%) were protected from exogenous trypsin as determined by scanning densitometry of the immunoblots (Fig. 7A, compare lane 1 with lane 3 of the left panels). Prolonged exposure of the immunoblots (lane 4 of the left panels) revealed several smaller protected fragments derived from trypsin digested AI/B29 -34 and AI/B34 -37, but not the control protein AI/StuI, suggesting that only a small portion of the AI/B chimeras are found in a transmembrane topology. There was a striking difference in the trypsin accessibility of apo-B100 and -B48. While only 10 -26% of r-apoB100 was trypsin resistant (as determined in two independent experiments), apoB48 was almost totally resistant to trypsin; 91% of recombinant h-apoB48 expressed in transfected cells, and 100% of endogenous r-apoB48 were trypsin-resistant (Fig.  7A, compare lane 1 with lane 3 of the right panels). Protected fragments with M r ϳ250,000 and Ͻ200,000 (indicated by arrowheads) derived from trypsin digestion of r-apoB100 were also observed, indicating a significant portion of r-apoB100 may exist as transmembrane intermediates. As expected, the ER resident protein disulfide isomerase was inaccessible to exogenous trypsin and was fully protected (Fig. 7A). These data suggest that the translocation efficiency is markedly different between apo-B100 and -B48. In separate experiments, AI/B chimeras were translated and translocated in vitro using microsomes isolated from rat liver, and the resulting samples were subjected to trypsin digestion to assess the topology of the AI/B chimeras (Fig. 7B). One predominant protein species was observed when chimera RNA was translated in the absence of microsomal membranes (lane 1). In the presence of microsomal membranes (lane 2) an additional processed AI/B29 -34 polypeptide of larger apparent molecular weight was detected (downward arrowhead) reflecting N-glycosylation of the primary translation product, as predicted from the primary sequence of the apoB segment (see Fig. 2A). Addition of microsomal membranes to AI/B34 -37, AI/StuI (lane 2, middle and lower panels) and AI/B37-42 (not shown) translation mixtures produced polypeptides of smaller apparent molecular weight than the primary translation product (upward arrowheads), reflecting signal peptide cleavage. In all instances the unprocessed species were totally degraded by exogenous trypsin (lane 3). In contrast, the majority of the processed species of AI/B

Effect of lipid supplementation and ALLN on secretion of AI/B chimeras and r-apoB100
Transfected cells (60-mm dishes) were incubated for 2 h prior to pulse-chase analysis in the following medium: 1) serum-free DMEM, 2) DMEM ϩ 20% serum, 3) medium 2 ϩ 0.4 mM oleate, and 4) medium 3 ϩ ALLN (40 g/ml ALLN, from a 40 mg/ml stock solution in dimethyl sulfoxide). The pulse-chase experiments were performed as described in Fig. 6, A and B, except that cells were pulse-labeled (30 min) and chased (2 h) with medium containing the indicated supplements. Radioactivity associated with r-apoB100 or AI/B chimera in the cells and media at the initiation and at the end of the chase were quantified. Results are expressed as percent of the initial radiolabeled protein recovered at the end of the 2-h chase. Data for r-apoB100 are the mean Ϯ S.D. of three cell lines (i.e. AI/B29 -34, AI/B34 -37, and AI/B37-42  7. Trypsin digestion of microsome-associated AI/B chimeras and r-apoB. A, Microsomes prepared from cell lines expressing AI/B29 -34, AI/B34 -37, AI/StuI (left three panels), or recombinant h-apoB48 (middle panel on the right) were subjected to limited proteolysis with trypsin (50 g/ml) in the presence or absence of 0.5% Triton X-100. Individual proteins were visualized by immunoblotting with 10-s exposure of the ECL reaction. Lanes 4 on the left were over exposed (60 s) in order to visualize the smaller protected fragments generated by trypsin digestion of AI/B chimeras. Endogenous protein disulfide isomerase (PDI) was used as a lumenal protein marker (bottom panel on the right). Data for endogenous r-apoB in AI/B chimera-producing cells are also shown (top panel on the right). B, protease protection analysis of AI/B chimeras generated by in vitro translation in the presence of rat liver microsomes. Transcripts were translated in rabbit reticulocyte lysate containing [ 35 S]methionine with or without microsomes, as indicated. After 2-h translation, protease protection analysis was performed as described in A and the products were visualized by autoradiography. Upward and downward arrowheads indicate the products of microsome-mediated signal peptide cleavage and N-glycosylation, respectively. chimeras were inaccessible to exogenous trypsin: 76.8 Ϯ 0.8% of AI/B29 -34, 87.7 Ϯ 11.6% of AI/B34 -37, and 91.8 Ϯ 12.0% of AI/B37-42 (mean Ϯ S.D., n ϭ 3). These data together suggest that, unlike endogenous r-apoB100, AI/B chimeras are efficiently translocated across the ER membranes.
Finally, we determined the distribution of AI/B chimeras and endogenous r-apoB100 between lumen and membrane of the microsomes by sodium carbonate treatment. The microsomes were isolated from transfected cells that had been labeled with [ 35 S]methionine/cysteine for 4 h. Only a minor portion (Ͻ15%) of the 35 S-labeled AI/B chimeras and endogenous r-apoB (B100 and B48) was recovered in the lumen irrespective of oleate supplementation. Under the same conditions, the majority (Ͼ80%) of the lumenal marker PDI was released from the membrane fraction by the sodium carbonate treatment (data not shown). These results suggest that the majority of the chimera and apoB are associated with the microsomal membranes.

DISCUSSION
Little is known about how lipids are recruited by apoB during VLDL assembly. The current study has attempted to determine if the ability to assemble apoB48-VLDL is governed by length or by unique sequences of the protein, and provided evidence that VLDL assembly is probably mediated by hydrophobic, lipid-binding domains within apoB48. Results obtained from experiments with both the truncated h-apoBs (ranging from apo-B29 to -B48) and the AI/B chimeras clearly demonstrate that the full-length of apoB48 is not a prerequisite for VLDL assembly. Rather, short h-apoB sequences (152-237 amino acids), most likely the amphipathic ␤ strands present as multiple copies within B29 -42, can mediate the recruitment of a large quantity of lipids in a manner similar to the proposed "two-step" assembly for apoB48-VLDL. Thus, the apoB sequences required for VLDL assembly are much shorter than was previously expected.
Structural Requirement of Human ApoB for Hepatic VLDL Assembly-The finding that AI/B chimeras could form VLDL was surprising and prompted us to analyze extensively the physicochemical characteristics of these VLDL particles. The experimental data obtained from density ultracentrifugation (Figs. 2C and 3A and 3B), gel permeation chromatography (Fig.  3C), and agarose gel electrophoresis (Fig. 3D), combined with the apparent absence of endogenous r-apoB by immunoaffinity purification (Fig. 4), supported the suggestion that the AI/B chimeras (containing 3-5% of apoB sequences) can form discrete, secretion competent VLDL. The VLDL containing AI/B chimeras are not formed by nonspecific exchange of the chimeras onto pre-existing VLDL (Fig. 3B). Rather, they are formed by active recruitment of triglyceride and phosphatidylcholine into the lipoproteins (Fig. 5A). To date, it has not been possible to ascertain the stoichiometry of AI/B chimeras per VLDL particle. Nor could the morphology of the AI/B chimera VLDL be examined because elution of intact particles from the immunoaffinity adsorbants was unsuccessful. However, even considering the possibility that more than one AI/B chimera could be present per VLDL particle, one might still wonder how a protein containing only 3-5% of the apoB sequence can mediate VLDL assembly? We have searched for structural features within the three apoB segments (i.e. B29 -34, B34 -37, and B37-42) that could potentially be responsible for the ability of AI/B chimeras to recruit lipids. Since N-glycosylation sites are present in some (e.g. B29 -34) but not the other segments (e.g. B34 -37 and B37-42) ( Fig. 2A), we have eliminated N-glycosylation as a requirement for VLDL assembly. Likewise, since cysteines (two in B29 -34, one in B34 -37, and zero in B37-42), proline-rich clusters (2) (one in B29 -34, and zero in B34 -37 or B37-42), or the postulated "pause-transfer" motifs (11,12) (four in B37-42, one in B29 -34, and zero in B34 -37) are not uniformly distributed among the three segments, we have concluded that none of these structural features is essential for the AI/B VLDL assembly. The only common feature among the three apoB segments is their hydrophobicity. The hydropathy values, as determined by the Kyte and Doolittle algorithm (PCGene™, Intelligenetics, Inc., 15-residue window) revealed that each apoB segment generated a chimera which was more hydrophobic (grand average hydrophobicity ϭ Ϫ6.59 for AI/ B29 -34, Ϫ6.20 for AI/B34 -37 and Ϫ5.67 for AI/B37-42) than AI/StuI (grand average hydrophobicity ϭ Ϫ8.06). Furthermore, according to the algorithm of Segrest et al. (23), each apoB segment was predicted to consist predominantly of amphipathic ␤ strands. It is conceivable that these uniquely enriched ␤ strands within B29 -42 are the critical determinants of VLDL assembly. However, although the presumed role of the amphipathic ␤ strands in VLDL has been demonstrated by the current study, the role of other structural elements, such as amphipathic ␣ helices, in VLDL assembly has not been excluded. The present study, nevertheless, suggests that the structural determinants for VLDL could be relatively short (3-5% of apoB100) and present in multiple copies. The presence of multiple segments enriched in hydrophobic sequences within apoB48 would certainly enhance the efficiency of lipid recruitment during VLDL synthesis. It should be noted that according to Segrest et al. (23) amphipathic ␤ strands also occur within apoB29 (between apo-B18 to -B29) but the protein does not assemble VLDL (Fig. 1C). Thus, the lipid recruiting function for VLDL assembly may not be an indiscriminant property of all the amphipathic sequences.
Topological Considerations for AI/B Chimera-containing VLDL Assembly-When the microsomal membrane topology of AI/B chimeras was compared with that of endogenous r-apoB100, similarities and differences between the two species were observed. In the microsomal fractionation studies, both r-apoB100 and AI/B chimera were similarly resistant to the sodium carbonate treatment and associated with the microsomal membranes (Ͼ85% of total). The avid membrane association of AI/B chimeras is most likely attributable to the presence of the hydrophobic apoB segments within these chimeras. Although such a high proportion of apoB on the microsomal membranes was previously reported using rat liver (40) and was unlikely an artifact of the sodium carbonate treatment, other researchers (22) have found less apoB associated with the membranes (as low as 20 -40%). It is possible that the high proportion of membrane associated apoB observed in the present study was attributable to the addition of ALLN during microsome preparation, preventing degradation of a labile pool of membrane-bound apoB. Although the AI/B chimera and r-apoB100 showed similar membrane affinity, their topologies were different, as demonstrated by protease protection analysis. The former was predominantly resistant (ϳ70% protection in microsomes isolated from transfected cells (Fig. 7) and 80 -90% protection by in vitro translation assay using rat liver microsomes), whereas the latter was extremely sensitive (ϳ20% protection) to proteolysis. These data could be interpreted as indicating that the AI/B chimeras are predominantly on the lumenal side of the membrane while r-apoB100 is to a large extent in a transmembrane configuration. In this regard, the AI/B chimeras were similar to apoB48 which was also fully resistant to exogenous trypsin. At present, the reason for the striking differences in trypsin sensitivity between apoB100 and apoB48 is not clear. If the accessibility to trypsin at steady state indeed reflects apoB translocation efficiency, then the transmembrane topology of apoB100 is most likely determined by sequences within the C-terminal 50% of the protein. The translocation efficiency of apoB48 and apoB100 needs to be further examined. Furthermore, although it is reasonable to speculate that the membrane associated AI/B chimeras (on the lumenal leaflet of the ER membrane) may serve as precursor of the secreted VLDL, whether or not the formation of AI/B chimera VLDL indeed occurs on the ER membrane also remains to be determined experimentally. Recently, it was reported that the ER membrane associated apoB100, and probably apoB48 as well, was the precursor for the formation of VLDL in McA-RH7777 cells (41). Results of the AI/B chimera studies suggest strongly that sequences beyond the N-terminal 30% of apoB100 probably play important roles in VLDL assembly through their unique interaction with the ER membrane. Further studies using the AI/B chimera approach may allow us to define the structural features of h-apoB involved in VLDL assembly.
AI/B Chimeras Are Less Able than Truncated ApoBs to Form Secretion Competent HDL-There were two unique features associated with the secretion kinetics of the AI/B chimeras. First, secretion of the chimeras under standard conditions (i.e. DMEM containing 20% serum) was extremely poor, and the vast majority of them was degraded within the cells (Table I). This is in sharp contrast to the truncated h-apoBs such as apoB37 which could be efficiently secreted as HDL without significant intracellular degradation (4). Second, under the oleate-supplemented conditions, the chimeras were secreted predominantly as VLDL (Fig. 2C), unlike the truncated h-apoBs (e.g. apo-B42 and -B37) which could be secreted as both HDL and VLDL (Fig. 1C). Although currently there is no satisfactory explanation for the two features, several possible mechanisms have been considered. Since degradation of AI/B chimeras could be specifically inhibited by ALLN, it is unlikely that their degradation is the result of nonspecific proteolysis. Rather, degradation of the chimeras is mediated by an ALLNsensitive mechanism characteristic of apoB degradation. Likewise, since no proteolytic cleavage at the junction between apoA-I and apoB sequences was observed in the transfected cells, it is unlikely that degradation of the chimeras results from instability of the protein per se. The inability to recover substantial quantities of AI/B chimera HDL from the medium may suggest that the chimeras are less able than truncated apoBs to form HDL or that HDL particles can be formed but are unstable. The reason for the impaired ability of chimeras to form stable HDL particles is not immediately clear. Although it is plausible that the impairment is attributable to the lack of sufficient polypeptide length of the chimeras, more likely it is because they do not contain the unique N-terminal apoB domain. Distinct from the rest of the molecule, the N-terminal portion of apoB (ϳ15%) contains several intrachain disulfide linkages and probably assumes a globular conformation (2,23). Recently, several groups (41)(42)(43) working with different systems have reported that the N-terminal cysteine-rich domain of apoB plays an important role in the initial cotranslational assembly of LpB. In the case of AI/B chimeras, the apoA-I portion of the proteins may not be sufficient to initiate lipidation to form a primordial HDL particle. Thus, the insufficient lipidation of AI/B chimeras on the ER membrane may underlie their susceptibility to intracellular degradation. However, in the presence of abundant lipid supply, these membrane associated AI/B chimeras are able to recruit lipid through the h-apoB portion of the proteins to form VLDL.
We have thus provided evidence that massive lipid recruitment during VLDL synthesis is governed by discrete hydrophobic sequences within apoB48. In mammals, apoB48 is synthesized from the same apoB gene that encodes apoB100 through the apoB mRNA editing (44). Although both apo-B100 and -B48 are essential for triglyceride transport, it has been shown that in the intestine (45) and rat liver (46) where apoB48 is synthesized, the turnover rate of apoB48 in response to the lipid is higher than that of apoB100, suggesting that the N-terminal portion of apoB is more efficient than the full-length protein in mobilizing lipid. The ability of apoB48 to assemble VLDL makes the C-terminal portion of apoB100 superfluous for triglyceride transport, this might be the rationale for the apoB mRNA editing mechanism to have evolved in the intestine to preclude futile translation. The physiological advantage for the evolution of apoB48 is also evident from the inverse relationship between apoB48 concentration and atherogenic lipoproteins (VLDL and low density lipoproteins) in the plasma (47). An improved understanding of the mechanisms responsible for apoB48 VLDL synthesis may help find means to lower plasma cholesterol and triglyceride levels.