The amino terminus of apolipoprotein B is necessary but not sufficient for microsomal triglyceride transfer protein responsiveness.

Human apolipoprotein (apo) B mediates the formation of neutral lipid-containing lipoproteins in the liver and intestine. The association of apoB with lipid is thought to be promoted by the microsomal triglyceride transfer protein complex. We have reconstituted lipoprotein assembly in an insect cell line that normally does not support this process. Expression of human microsomal triglyceride transfer protein (MTP) and apolipoprotein B48 (apoB48) together enabled Sf-21 insect cells to secrete 60-fold more lipoprotein-associated triacylglycerol than control cells. This dramatic effect demonstrates that effective partitioning of triacylglycerol into the secretory pathway requires an endoplasmic reticulum-associated neutral lipid transporter (provided by MTP) and an apolipoprotein to shuttle the lipid through the pathway. Expression of the human apoB48 gene in insect cells resulted in secretion of the protein product. Including both MTP subunits with apoB48 and oleic acid specifically increased apoB48 secretion 8-fold over individual subunits alone. To assess whether specific regions of apoB are necessary for MTP responsiveness, nine apoB segments were expressed. These included NH-terminal segments as well as internal and COOH-terminal regions of apoB fused with a heterologous signal sequence. ApoB segments containing the NH-terminal 17% of the protein were secreted and responded to MTP activity; however, a segment containing only the NH-terminal 17% of the protein was not significantly responsive to MTP. Segments lacking the NH terminus were not MTP-responsive, and five of six of these proteins were trapped intracellularly but, in certain cases, could be rescued by fusion to apoB17. These results suggest that the NH terminus of apoB is necessary but not sufficient for MTP responsiveness.

The transport of water-insoluble lipids through the circulation of all mammals is mediated by lipoprotein particles. Apo-lipoprotein B (apoB) 1 is an unusually large secretory protein (514 kDa) that is required for the assembly and secretion of triacylglycerol-rich lipoproteins from the liver and the intestine (1). ApoB production rates are highly variable in human populations. Clinically important lipoprotein disorders such as familial combined hyperlipidemia are often associated with apoB overproduction (2).
The microsomal triglyceride transfer protein (MTP) plays an essential role in the assembly of apoB-containing lipoproteins. MTP is a protein complex found in the endoplasmic reticulum (ER) lumen of liver and intestinal cells (3). This complex primarily transfers neutral lipids between membranes in vitro (4). MTP has two subunits: protein disulfide isomerase (PDI) and a 97-kDa subunit that possesses lipid transfer activity in vitro when coupled with PDI (5). Mutations in the 97-kDa subunit of MTP have been found in patients with abetalipoproteinemia (6 -8). This condition leads to only trace amounts of apoBcontaining lipoproteins in the plasma and a substantial decrease in plasma neutral lipid content.
Recent studies have shown that introduction of the 97-kDa subunit of MTP into cells that are nonhepatic and nonenteric in origin, enables these cells to secrete segments of apoB (9,10). ApoB secretion in these cases is increased when the cells are supplemented with oleic acid, a substrate for neutral lipid biosynthesis. The tissue-specific production of apoB-containing lipoproteins correlates with the tissue distribution of the 97-kDa subunit of MTP. These studies, coupled with our understanding of abetalipoproteinemia, reinforce the hypothesis that MTP promotes lipoprotein assembly and secretion by facilitating the coupling of apoB with lipid. The secretion of apoB and lipid appear to be interdependent in that apoB requires adequate neutral lipid for secretion and neutral lipid secretion requires apoB.
Studies of invertebrates (primarily insects) indicate that they have evolved an alternative mechanism for the release of neutral lipids from cells (11). While adult insects have large triacylglycerol stores in their lipoprotein-producing fat body tissue, insects cannot target triacylglycerol into the secretory pathway. Insects fail to assemble lipoproteins intracellularly and instead secrete the protein component of their lipoprotein in a lipid poor form (12). Mobilization of stored triacylglycerol occurs only after it is hydrolyzed to diacylglycerol. The diacylglycerol is released from the cell to the extracellular fluid, the hemolymph, where lipoprotein assembly takes place (13). This assembly process requires an extracellular lipid transfer particle (14) which, like MTP, has been shown to possess a lipid transfer activity in vitro (15).
No molecular explanation exists for the contrasting mechanisms of triacylglycerol mobilization that have evolved within the animal kingdom. We hypothesized that the ability of vertebrates to partition triacylglycerol into the secretory pathway is dependent upon ER retention of their neutral lipid transfer activity. If this hypothesis is correct, then providing invertebrate cells with an intracellular neutral lipid transfer activity should enable them to directly secrete triacylglycerol from their intracellular stores.
Sf-21 cells (from the fall armyworm Spodoptera frugiperda) are invertebrate cells that do not produce lipoprotein particles. We have utilized these cells to assess the requirements for the secretion of triacylglycerol and apoB. Although they accumulate high levels of intracellular triacylglycerol, Sf-21 cells release little triacylglycerol into the media. Here we show that expression of human MTP and the intestinal form of human apoB (apoB48) in Sf-21 cells is sufficient to confer upon them the ability to efficiently partition triacylglycerol into the secretory pathway.
The role of specific apoB sequences in the responsiveness to MTP has not been elucidated. COOH-terminal apoB truncations result in a decreased capacity to bind lipid, but do not eliminate apoB's ability to form lipoproteins (16 -18). However, the potential role of the NH 2 terminus of apoB in mediating MTP responsiveness is undefined. A segment of the NH 2 terminus of apoB is thought to undergo pausing during translocation across the ER membrane (19 -21). Thus, this region might be essential in the initial combination of protein with triacylglycerol that occurs during lipoprotein assembly. We therefore investigated the possibility that the NH 2 -terminal region is essential for apoB's responsiveness to MTP. We report that the NH 2 terminus of apoB is required for MTP responsiveness. However, by itself the NH 2 terminus is not MTP-responsive, indicating that it is necessary but not sufficient for MTP responsiveness. The current work further demonstrates that the targeting of triacylglycerol into the secretory pathway is dependent upon expression of an ER-associated neutral lipid transporter. The work also suggests that the evolutionary differences in animal lipid secretion are related to the secretion or retention of the organism's neutral lipid transfer activity.

EXPERIMENTAL PROCEDURES
Production of Recombinant Baculoviruses-The production of recombinant baculoviruses encoding human apoB17, human protein disulfide isomerase, and the 97-kDa subunit of human MTP are described elsewhere (22,23).
Generation of additional recombinant baculoviruses encoding regions of apoB represented in Fig. 11, was carried out as follows. Initially, a plasmid was constructed to contain an entire apoB100 minigene. Two oligonucleotides, 5Ј-GAT CCG CGG CCG CAT AGG CCA CTA GTG-3Ј and 5Ј-AAT TCA CTA GTG GCC TAT GCG GCC GCG-3Ј, were synthesized and annealed to generate the "Bam-Eco" polylinker containing (from the 5Ј end) BamHI, SacII, NotI, Sfi I, SpeI, and EcoRI sites. The polylinker was phosphorylated with polynucleotide kinase (U.S. Biochemical Corp.) prior to use. The Bam-Eco linker was ligated with a 2.5-kb EcoRI-BamHI fragment of apoB cDNA from pB18 (24) and pCMV5 that had been digested with BglII. This resulted in pB18LII. pB18LII was digested with BstEII (ϩ1359 of the apoB cDNA) and KpnI (in the linker region of pCMV5). This product was ligated to a BstEII-KpnI fragment from pB100 (24) that was generated by a complete digestion with KpnI and a partial digestion (at ϩ1359 of apoB cDNA) with BstEII. The resulting plasmid is pB100LII.
An 18-kb NotI-SmaI DNA fragment encoding the human apoB100 minigene was excised from pB100LII and inserted into a NotI-SmaIdigested baculovirus transfer vector pVL1392 (Invitrogen) to yield pAcB100. pB48LII was generated by engineering a translational stop signal at codon 2153 (as described (25)) of apoB53 in pB53L-L. A 7.2-kb NotI-SmaI fragment encoding human apoB48 was removed from pB48LII and inserted into NotI-SmaI digested pVL1392 to form pAcB48.
The baculovirus transfer vector pVTBac (26) (a gift from T. Vernet) containing the honeybee melittin signal peptide was used for the production of internal and COOH-terminal apoB constructs. Following cloning, all junctions were in frame with the honeybee melittin signal sequence as confirmed by sequencing at the Columbia University Cancer Center (CUCC) Sequencing Facility. Initially, an XbaI oligonucleotide linker with translational stop sequences in all three reading frames (CTAGTCTAGACTAG) was inserted into the SmaI site of pVTBac, yielding pVTBac*. A 6.5-kb BamHI fragment spanning sequences from apoB33 to apoB80 was removed from pB100LII and inserted into the BamHI site of pVTBac* yielding pAcB33-80. A 1.8-kb BamHI-BclI fragment encoding apoB33-46 was inserted into the BamHI site of pVTBac* to produce pAcB33-46. A 4.4-kb BclI-BamHI fragment encoding apoB48 -80 was inserted into the BamHI site of pVTBac* to produce pAcB48 -80.
To produce apoB69 -79, pVTBac was first modified by cutting with KpnI and filling in with the large fragment of DNA polymerase (Klenow) to generate blunt ends. A 63-bp synthetic DNA fragment (CGA ATC GAA GGT CGT AAA GAA ACC GCT GCT GCT AAA TTC GAA CGC CAG CAC ATG AAC AGC TAA) encoding the S peptide of RNase A (as in Ref. 27), a factor Xa cleavage site, and a translational stop codon were ligated to the blunt-ended pVTBac to generate pVTBac-S. pVTBac-S was digested with PstI and filled in with the Klenow to generate blunt ends. pAcB33-80 was digested with AccI and MscI to generate a 1.4-kb fragment encoding apoB69 -79. This fragment was treated with Klenow to generate blunt ends and was subsequently ligated to the linearized blunt-ended pVTBac-S to generate pAcB69 -79-S. This construct thus encoded apoB69 -79 fused in frame with the S peptide sequence as confirmed by sequencing at the CUCC Sequencing Facility.
To generate apoB78 -100, pB100LII was digested with BsiHKAI and was treated with T4 DNA polymerase to remove 3Ј-protruding sequences. Following digestion with MunI, a 4.5-kb fragment encoding apoB78 -100 was isolated. pVTBac was digested with EcoRI and SmaI, and the 4.5-kb apoB fragment was inserted to generate pAcB78 -100.
To produce apoB88 -100, two oligonucleotides were used to amplify the apoB exon 29 coding sequence from human genomic DNA via the polymerase chain reaction. The 5Ј primer contained 42 bases (5Ј-cgg gat cca cAG TCC TCT CCA GAT AAA AAA CTC ACC ATA TTC-3Ј), with the uppercase letters representing apoB sequence. The non-apoB sequence contains a BamHI site for cloning purposes, and the apoB sequence begins with the final two amino acids of exon 28. The 3Ј primer contained 27 bases (5Ј-tcc ccg GGC TGG CTC ACT GTA TGG TTT-3Ј), with the uppercase letters representing 3Ј apoB untranslated sequence.The lowercase (non-apoB) sequence contains a XmaI site for cloning purposes. Following amplification, the resulting 1.7-kb fragment was digested with BamHI and XmaI and inserted into pBluescript II SKϩ/Ϫ (Stratagene), which had also been digested with BamHI and XmaI. Following nucleotide sequencing, the same fragment was excised and inserted into pVTBac, which was previously digested with BamHI and XmaI. The construct was in frame with the honeybee melittin signal peptide.
To generate the apoB17-(69 -79) fusion, pAcB17-Bam was digested with BamHI and SnaBI. A 1.5-kb BamHI-SnaBI fragment encoding apoB69 -79 was removed from pAcB69 -79-S and was inserted into pAcB17-Bam. This construct was linearized with BamHI and was filled in with the Klenow fragment to place apoB69 -79 in frame with apoB17. The resulting product was pAcB17-(69 -79-S). All fusion junctions were confirmed by sequencing at the CUCC Sequencing Facility.
The resulting transfer vectors encoding regions of apoB were then used to produce recombinant baculoviruses using linearized viral DNA (Invitrogen) according to the manufacturer's suggestions. Recombinant viruses were identified, plaque-purified, amplified, and titered as described (28,29).
Analysis of Intracellular Lipids following Oleic Acid Treatment of Sf-21 Cells-Sf-21 cells were grown in suspension using TC 100 medium (Life Technologies, Inc.) with 10% fetal bovine serum (Hyclone). Cells were plated in 60-mm diameter tissue culture dishes at a density of 4 ϫ 10 6 cells/dish. Cells were infected for 1 h with wild-type baculovirus at a multiplicity of infection of 5. Following infection, cells were washed and fed SF 900 serum-free media (2 ml) (Life Technologies, Inc.). Twenty-seven hours postinfection, the above media were replaced with 2 ml of SF 900 media containing 0.5% bovine serum albumin (BSA) or 0.5% BSA complexed with 1 mM oleic acid. In experiments where an oleic acid tracer was used, [ 3 H] oleic acid was included at 6.25 Ci/ml. Seventeen hours later, cells were washed and scraped into 1 ml of phosphate-buffered saline (PBS), pH 6.2 (150 mM NaCl, 2.8 mM KCl, 1.5 mM KH 2 PO 4 , 6.5 mM NaHPO 4 , pH 6.2). Protein measurements were determined by a modified Lowry assay (30) using bovine serum albumin as a standard. Lipids were extracted (31) and subjected to TLC using a hexane:ether:acetic acid (80:20:2, v/v/v) solvent system. Mass measurements of various lipids were determined by charring with concentrated sulfuric acid (32) and compared with standard curves generated using triolein, diolein, cholesteryl oleate, and phosphatidylcholine. For measurements involving [ 3 H]oleic acid tracer, TLC was performed as above, and resolved lipids were identified using a Berthold automatic TLClinear analyzer and a Berthold CHROMA software package (version 6.23). Lipid species were scraped from the plate, and quantitation was performed via scintillation counting.
The BSA-oleate complexes were made as a 10 ϫ stock in PBS, pH 6.2, as follows. 10% (w/v) fatty acid-free BSA (Sigma) was dissolved in 10 ml of PBS, pH 6.2. Sixty-one mg of sodium oleate was dissolved in two ml of absolute ethanol plus 40 l of 4 N NaOH with gentle heating. The ethanol was evaporated with heating under a stream of nitrogen. In cases where an oleic acid tracer was used, 1.25 mCi of [ 3 H]oleic acid was added and then dried under nitrogen. The dried oleate was redissolved in 10 ml of PBS, pH 6.2, with heating. The oleate/PBS solution was equilibrated to room temperature while mixing with a stir bar. The BSA solution was then added, and mixing continued for 10 min. This 10 ϫ solution was filter-sterilized and stored at Ϫ20°C.
Analysis of Secreted Lipids from Oleic Acid-treated Cells-To assess the effect of MTP and apoB48 on lipid secretion from Sf-21 cells, the release of oleic acid-containing lipids was assessed using uninfected cells and cells infected with wild-type (control), apoB48 and MTP (23) viruses. Cells were infected as above with a total multiplicity of infection of 5 viruses/cell. In experiments where multiple viruses were used to infect cells, the following virus ratios were used: wild type:PDI, 3:2; wild type:PDI:97-kDa subunit, 3:1:1; apoB48:PDI, 3:2; apoB48:PDI:97-kDa subunit, 3:1:1. Media changes and harvesting were done as described above. After harvest, the media were centrifuged at 1500 ϫ g for 10 min. Media lipids were extracted and quantitated as described above. Expression of the individual proteins being tested was monitored via immunoblotting of cell lysates (for MTP) and media (for apoB48) (not shown).
Analysis of Lipoprotein-associated Lipids-In order to assess the lipid content of apoB48-containing lipoproteins, media samples (1.75 ml) were adjusted to a density of 1.24 g/ml with sodium bromide and were underlayered beneath 10.25 ml of a NaBr solution with a density of 1.20 g/ml. Following ultracentrifugation in an SW-41 rotor at 175,000 ϫ g for 27 h, the top 1 ml was removed. The lipids were extracted and analyzed as described above. To analyze phospholipids, samples were chromatographed using a chloroform, methanol, 40% methylamine, water (60:36:5:5, v/v/v/v) solvent system with phosphatidylinositol, phosphatidylserine, sphingomyelin, phosphatidylcholine, and phosphatidylethanolamine as standards. Quantitation was carried out as described above.
Analysis of Apolipoprotein B Segments Produced by Recombinant Baculoviruses-Sf-21 cells were plated in 60-mm diameter tissue culture dishes at a density of 4 ϫ 10 6 cells/dish. Cells were infected for 1 h with recombinant viruses (or wild-type control virus) at a multiplicity of infection of 5 viruses/cell in TC 100 media (Life Technologies, Inc.) plus 10% fetal bovine serum. Following infection, cells were washed twice with 2 ml of TC 100 media and were covered with 2 ml of SF 900 serum free media (Life Technologies, Inc.). Forty-two hours postinfection the media was removed, and the cells from each plate were lysed in 500 l of lysis buffer (2% sodium dodecyl sulfate, 0.05 M Tris, pH 9.0, 6 M urea, 0.1% EDTA, 2 mM benzamidine, 0.2 mM phenylmethylsulfonyl fluoride). Protein concentrations were determined (30) using BSA as a standard. Twenty-five g of total cell protein was subjected to SDS-polyacryl-amide gel electrophoresis (SDS-PAGE) (33). Immunoblotting (34) was carried out with anti-human apoB monoclonal antibodies (1D1, CC3.4, D7.2, MB47, MB43, Bsol 16, and Bsol 7) (35-37) and alkaline phosphatase-conjugated rabbit anti-mouse IgG (Sigma). All expected immunoreactive products were detected (data not shown). The apoB100 construct produced two discrete protein products, one the size of apoB100 and the other the size of apoB48. The mechanism by which this heterogeneity occurs is unknown, although mRNA editing (38,39) and premature polyadenylation of apoB mRNA (40) are known to result in apoB48 production in other systems. Specific proteolysis of apoB prior to secretion is another possible mechanism of formation of the shortened product.
Stimulation of ApoB48 Secretion by Oleic Acid-Cells were plated out and infected with wild-type control virus or virus encoding apoB48 with an multiplicity of infection of 5 as described above. Twenty-seven hours postinfection, the 2 ml of SF 900 media were replaced with 2 ml of SF 900 media containing 0.5% BSA or 0.5% BSA complexed with 1 mM oleic acid. The media were harvested 16 h later. Following centrifugation for 10 min at 1500 ϫ g the media received phenylmethylsulfonyl fluoride (0.2 mM) and benzamidine (2 mM). One ml of medium was precipitated with cabosil (41), and the resulting pellet was resuspended in 200 l of cabosil resuspension buffer (2% SDS, 0.05 M Tris, pH 9.0, 6 M urea, 0.1% EDTA, 0.2 mM phenylmethylsulfonyl fluoride, 2 mM benzamidine). This has been demonstrated to be an effective method for concentrating apoB protein products (17,41). Aliquots of this sample were then subjected to SDS-PAGE under reducing conditions and immunoblot analysis. A rabbit anti-pig apoB polyclonal antibody was utilized, followed by a goat anti-rabbit horseradish peroxidase-conjugated second antibody. Enhanced chemiluminescence (Amersham Corp.) and exposure to Kodak XAR-5 x-ray film was used to detect bound antibody. Immunoreactive apoB48 was quantitated using a Molecular Dynamics (Personal Densitometer SI) scanning laser densitometer. Band volumes were integrated using the MD ImageQuant software, version 4.1. To ensure that analysis was in the linear range of detection, several sample volumes (2-20 l) were analyzed, and several timed exposures were made. To monitor levels of total apoB protein expression, cell lysates were prepared as described above, and immunoblotting was carried out (using 25 g of total cell protein) with anti-apoB antibodies.
Analysis of the Effect of MTP on ApoB48 Secretion-Cells were infected as described above with a total viral multiplicity of infection of 5 viruses/cell for each condition tested. The viral ratios used were as follows: for cells expressing apoB48 and PDI (apoB48:PDI ϭ 3:2), for cells expressing apoB48 and the 97-kDa subunit (apoB48:97-kDa subunit ϭ 3:2), for cells expressing apoB48 and both MTP subunits (apoB48:PDI:97-kDa subunit ϭ 3:1:1). Twenty-seven hours postinfection, media were changed as in the oleic acid experiments (above) and were harvested 16 h later. Media harvesting, Cabosil precipitation, and quantitative immunoblotting were carried out as described above. To monitor protein expression, cell lysates were prepared as described above, and immunoblotting was carried out (using 25 g of total cell protein) with anti-apoB, anti-PDI, and anti-97-kDa subunit antibodies.
Analysis of the Specificity of the Oleic Acid and MTP Effects-Cells were infected and media changes were carried out as described above.
To assess the specificity of the oleic acid effect on apoB48 secretion, cells were infected with either apoB17-producing virus or apoB48-producing virus (multiplicity of infection of 5). The media introduced at 27 h postinfection contained 0.5% BSA with or without 1 mM oleic acid. The media also contained 100 Ci/ml [ 35 S]methionine (DuPont Express Protein Labeling Mix). To assess the specificity of the MTP effect on apoB48 secretion, cells were infected at a total multiplicity of infection of 5 with the following viruses at the indicated viral ratios: wild-type virus:PDI, 3:2; wild-type virus:PDI:97-kDa subunit, 3:1:1; apoB48:PDI, 3:2; apoB48:PDI:97-kDa subunit, 3:1:1). The media change at 27 h postinfection included 0.5% BSA complexed with 1 mM oleic acid and 100 Ci/ml [ 35 S]methionine. Eight hours after the media changes, media samples were harvested, and 5 l of total secreted protein was analyzed by SDS-PAGE under reducing conditions in 4 -15% gradient gels. Following electrophoresis, the gels were treated in Enhance (ICN) according to the manufacturer's instructions and dried. Fluorography of the dried gels was carried out at Ϫ70°C.
Pulse-chase Analysis of the Intracellular Degradation of ApoB48 -Cells were infected as described above with apoB48, PDI, and 97-kDa subunit-encoding viruses. Nineteen hours after infection, the media were replaced with 2 ml of fresh SF900 containing 0.5% BSA with or without 1 mM oleate. Sixteen hours later, the media were removed, and cells were washed in PBS (pH 6.2) and then incubated in SF900 medium free of L-cysteine or methionine for 1 h at 27°C, in order to deplete the intracellular pool of the two amino acids. The cells were pulselabeled with 1 ml of 0.1 mCi/ml [ 35 S]methionine/cysteine mix containing medium and chased in SF900 medium with an excess amount of cold methionine and cysteine. At each time point, the medium was collected after spinning down the cell debris, and the cells were lysed by sonication. Cell extracts and media samples were immunoprecipitated by incubating with polyclonal antibodies against apoB followed by protein G-agarose. After extensive washing, the precipitated proteins were solublized and subjected to SDS-PAGE fractionation. The bands corresponding to apoB 48 were cut out from the gels and counted for radioactivity. To ensure quantitative recovery of apoB, a second round of immunoprecipitation was performed. As a control for nonspecific interactions with the antibody and/or the protein G beads, cells infected with PDI or MTP viruses alone were also carried through the whole procedure.
Analysis of the Buoyant Density of ApoB48 -Cells were infected at a total multiplicity of infection of 5 as described above. The medium change at 27 h postinfection contained 0.5% BSA complexed with 1 mM oleic acid and 50 Ci/ml [ 35 S]methionine. Eight hours later the media were harvested as above, and samples (1.75 ml) were subjected to NaBr density gradient ultracentrifugation (42) in an SW-41 rotor at 175,000 ϫ g for 32 h. The resulting gradient range was 1.016 -1.247 g/ml. Twelve 1-ml fractions were collected from the top of each tube. Each fraction was precipitated with Cabosil and resuspended in 30 l of Cabosil resuspension buffer (see above). Twenty l from each sample was electrophoresed in a 4 -15% SDS-PAGE gel under reducing conditions, and fluorography was carried out as described above.
Analysis of ApoB Segments and ApoB17 Fusions for Secretion and MTP Responsiveness-NH 2 -terminal, internal, and COOH-terminal apoB constructs as well as apoB17 fusion proteins were tested for their secretion and responsiveness to MTP. Cells were infected as described above with individual apoB viruses along with one (PDI) or both subunits of MTP. Media were changed at 27 h postinfection, and new media contained 0.5% BSA complexed with 1 mM oleic acid. Media were harvested 17 h later (as described above), and 1 ml was precipitated with Cabosil. Pellets were resuspended as above in 200 l of resuspension buffer, and aliquots ranging from 5 to 60 l were analyzed by SDS-PAGE and quantitative immunoblotting (see above). All immunoblotting was carried out with a rabbit anti-pig apoB polyclonal antibody except for apoB78 -100 and apoB88 -100 (which required a monoclonal antibody-Bsol 16) and apoB69 -79 (for which MB47 was used). To monitor protein expression, cell lysates were prepared as described above, and immunoblotting was carried out (using 25 g of total cell protein) with anti-apoB, anti-PDI, and anti-97-kDa subunit antibodies.

Sf-21 Cells Actively Synthesize Triacylglycerol and Store It
Intracellularly-To stimulate the production of neutral lipids in Sf-21 cells, virally infected cells were exposed to 1 mM oleic acid using BSA as a carrier. The intracellular levels of lipid species known to participate in either vertebrate or invertebrate lipoprotein assembly were quantitated. The 17-h oleic acid treatment resulted in an 11-fold increase in intracellular triacylglycerol content (Fig. 1). No significant change in the levels of phospholipids or diacylglycerols was observed. This increase in triacylglycerol content transformed the phenotype of the Sf-21 cells so that they more closely approximated that of both vertebrate and invertebrate lipoprotein-producing cells.
Introduction of a [ 3 H] oleic acid tracer, along with the 1 mM oleic acid, resulted in effective incorporation of the tracer in all lipid pools analyzed except cholesteryl ester (Fig. 2). The lack of detectable cholesterol ester is consistent with the observation that Sf-21 cells lack the enzyme required for its formation (43). The triacylglycerol pool contained two-thirds of the intracellular tracer, supporting the mass analysis and indicating that synthesis of triacylglycerol was a major element of the lipogenic response. Infections with recombinant viruses did not significantly alter the intracellular lipid mass levels or tracer distributions when compared with wild-type virus infection (not shown).
Sf-21 Cells Secrete Little Triacylglycerol from Their Intracellular Stores-In order to characterize the neutral lipids released from oleic acid-stimulated Sf-21 cells, measurements were made of 3 H-media lipids following exposure of the cells to oleic acid for 17 h. Use of a [ 3 H]oleic acid tracer ensured that the species measured were products of cellular lipid biosynthesis and secretion and not media lipid components. Despite a large elevation in intracellular triacylglycerol levels in response to oleic acid, very little was released into the culture media (Fig. 3). This inability of Sf-21 cells to effectively secrete triacylglycerol, despite high intracellular levels, is consistent with invertebrate physiology; invertebrates release diacylglycerol rather than triacylglycerol.
Infection of Sf-21 cells with wild-type baculovirus did little to affect the media neutral lipid composition (Fig. 3). This indicates that any change, upon infection with recombinant viruses, would likely reflect the effect of heterologous proteins expressed by these viruses.
Human MTP and Apolipoprotein B Stimulate Triacylglycerol Secretion from Sf-21 Cells-Following introduction of the apoB48 and MTP genes (either separately or together), triacylglycerol secretion was measured. While expression of both MTP subunits in Sf-21 cells results in neutral lipid transfer activity (23), this was insufficient to promote triacylglycerol secretion (Fig. 3). Expression of apoB48 in Sf-21 cells had a modest effect on triacylglycerol release, consistent with apoB48 having a limited ability to bind neutral lipids and carry them through the secretion pathway without the assistance of MTP.
In contrast, co-expression of apoB48 with MTP resulted in a dramatic increase in triacylglycerol secretion by the invertebrate cells. The total media triacylglycerol level under these conditions was 10-fold higher than in media from uninfected cells. Thus, apoB48 and MTP are sufficient to efficiently partition cellular triacylglycerol into the secretory pathway. No effect on media diacylglycerol levels was observed in the presence of apoB48 and MTP (not shown), demonstrating that the effect is specific for triacylglycerol.

MTP-mobilized Triacylglycerol Is Assembled into Lipoprotein Particles-
The previous experiments suggested that MTP redistributed triacylglycerol into the secretory pathway of Sf-21 cells and combined it with apoB48. To quantitate the amount of triacylglycerol secreted in association with lipoprotein particles, the media was subjected to ultracentrifugation in sodium bromide at a density of 1.20 g/ml. Sixty times more triacylglycerol was present in the d Ͻ 1.20 g/ml fraction from cells expressing both apoB48 and MTP when compared with control cells (Fig. 4). This large difference clearly demonstrated that invertebrate expression of these vertebrate gene products dramatically stimulated triacylglycerol secretion in the form of nascent lipoprotein particles. By contrast, little difference was detected in the diacylglycerol found at this density, demonstrating a triacylglycerol-specific effect.
In addition to neutral lipid and protein, lipoproteins contain a surface layer of phospholipid. We therefore assessed the phospholipid content of the Sf-21 cell-produced apoB48 lipoproteins. Analysis of the phospholipids that floated with the apoB48 and triacylglycerol indicated that phosphatidylcholine and phosphatidylethanolamine were both present in the lipoprotein particles (data not shown).
ApoB48 Secretion by Sf-21 Cells Is Specifically Stimulated by Oleic Acid and MTP-In mammalian liver, apoB secretion is regulated post-translationally, primarily by the amount of triacylglycerol synthesized in the cells. To determine if insect cells expressing apoB are also capable of responding to the regulatory effects of triglyceride, apoB secretion was measured in the absence and presence of 1 mM oleate. Under these conditions, the triacylglycerol content of the Sf-21 cells increases 11-fold (Fig. 2). The level of apoB48 secretion was increased 6-fold in the presence of exogenous oleic acid (Fig. 5). No detectable change in the level of intracellular apoB48 was observed in the presence of oleic acid (data not shown).
To investigate whether the effect of oleic acid treatment was specific for apoB48 secretion or if it was influencing general protein secretion, analysis of total cell protein secretion was carried out. Total 35 S-labeled secreted proteins were analyzed in the presence and absence of exogenous oleic acid (Fig. 6). In cells expressing apoB48, oleic acid stimulated the secretion of apoB48, while the levels of other secreted proteins were unaffected. The secretion of apoB17 from control cells was unaffected by the presence of oleic acid, further demonstrating that the longer apoB48 protein was specifically influenced by the presence of oleic acid.
In order to assess the effect of MTP on apoB secretion, cells expressing apoB48 were co-infected with either the individual MTP subunits or with both subunits together. Expression of both MTP subunits with apoB48 in the presence of oleic acid dramatically increased apoB48 secretion (Fig. 7). This increase was 8-fold more than that observed in the presence of either individual subunit alone. This pronounced stimulation of apoB48 secretion by MTP is not seen in the absence of exogenous oleic acid (not shown), demonstrating that MTP's activity is dependent on triacylglycerol availability. By quantitative Western blot analysis of oleate-treated cells, we estimate that the mass of apoB accumulating in the tissue culture medium in the presence of MTP after 17 h was 2-10% of the intracellular apoB mass (data not shown).
To assess whether or not the effect of MTP on apoB48 secretion was specific, total protein secretion was analyzed in cells expressing apoB48 and one or both subunits of MTP (Fig. 8). Induction of apoB48 secretion was clearly seen in the presence of both MTP subunits. No other secreted proteins were stimulated by MTP, demonstrating its specificity for apoB48. An additional media protein (97 kDa) is detectable as a result of MTP expression. We have determined immunochemically that this is the large subunit of MTP (some of which is secreted during overexpression) and not an MTP-stimulated protein (not shown).
The MTP Induction of ApoB Secretion Is Not a Consequence of Rescue from Early Intracellular Degradation-Previous studies in HepG2 cells have shown that oleate can increase apoB secretion by diminishing the proportion of apoB subject to early post-translational degradation (44). ApoB proteolysis is very efficient, resulting in the degradation of 80% of intracellular apoB within 40 min (44). We assessed apoB48 degradation in Sf-21 cells over a period of 4 h. The cells were subjected to a 30-min pulse with [ 35 S]methionine and [ 35 S]cysteine followed by a chase in tracer-free medium for times up to 4 h. The experiments showed that the rate of tracer incorporation into apoB is not affected by oleate or by co-expression of MTP, indicating that oleate and MTP do not increase apoB secretion by increasing apoB biosynthesis (data not shown). In contrast to primary hepatocytes or HepG2 cells, Sf-21 cells do not rapidly degrade significant proportions of newly synthesized apoB (Fig. 9). ApoB radioactivity was not detectable in the culture media until after 4 h, explaining why there was no decline in intracellular apoB radioactivity during this chase period. At longer times, however, oleate or MTP exerted a major effect on the amount of apoB secreted (Figs. 5 and 7).
ApoB48 Secreted by Sf-21 Cells Has the Buoyant Density of a High Density Lipoprotein-While oleic acid and MTP were able to stimulate the secretion of apoB48, density gradient ultracentrifugation was required to assess whether the secreted apoB48 was in the form of a lipoprotein particle. Analysis of FIG. 5. Analysis of apoB48 secretion from cells treated with oleic acid. The relative amounts of apoB48 secreted by Sf-21 cells were analyzed by quantitative immunoblotting, utilizing an anti-apoB polyclonal antibody. ApoB48 was quantitated in the media of control cells (incubated with 0.5% BSA and no exogenous oleic acid) and from cells treated with 0.5% BSA and 1 mM oleic acid. ApoB48 levels from cells without oleate (Ϫoleate) were normalized to 1, and levels from cells with oleate (ϩoleate) were expressed relative to that. Values represent the mean of four determinations, and error bars represent standard deviation. A representative immunoblot is shown. The increase in reactivity was not paralleled by an increase of intracellular reactivity, and no immunoreactivity occurred when cells infected with wild-type (control) virus were tested (not shown). Relative media levels of apoB48 were analyzed by quantitative immunoblotting following co-expression with individual MTP subunits and both subunits together. MTP-mediated stimulation of apoB48 secretion required both MTP subunits. ApoB48 levels from PDI-expressing cells were normalized to 1, and levels from other cells were expressed relative to that. Values represent the mean of three determinations, and error bars represent standard deviation. Representative immunoblots of apoB48 (secreted) and the MTP subunits (intracellular) are shown. No increase in intracellular apoB48 content was detected in the presence of MTP, and no apoB48 was detectable using cells lacking apoB48-encoding viruses (not shown).

FIG. 8. Specificity of the stimulation of apoB48 secretion by MTP.
To assess the specificity of MTP stimulation of apoB48 secretion, oleate-treated cells were treated with [ 35 S]methionine, and total secreted proteins were analyzed by fluorography. MTP induction of apoB48 secretion is clearly seen, while no other proteins secreted by the cells appear to respond to MTP expression. An additional media protein (97 kDa) is detected from cells expressing both MTP subunits. We have determined immunochemically that this is the large subunit of MTP (some of which is secreted during overexpression) and not an MTPstimulated protein (not shown).
apoB48 buoyant density showed that oleic acid and MTP stimulation of apoB48 secretion correlated with the formation of apoB48 containing high density lipoprotein particles (Fig. 10). The density distribution of these particles is comparable with that of particles from apoB48-producing cells of hepatic origin (16 -18). MTP expression resulted in a higher level of apoB48 secretion but did not significantly alter the buoyant density of the secreted lipoprotein particles, indicating an increase in the number but not in the size of the particles. (The apparent difference in extent of the MTP effect between Figs. 7 and 10 is due to the fact that additional viruses were included in the experiment shown in Fig. 7 to control for the total multiplicity of infection. This was not done in the experiments depicted in Fig. 10.) Secretion and MTP Sensitivity of ApoB Segments Correlates with the Presence of the NH 2 Terminus of ApoB-To assess whether specific regions of apoB are required for secretion of the protein and MTP responsiveness, nine apoB segments (Fig. 11) were expressed in oleic acid-treated Sf-21 cells, in the presence and absence of MTP. All segments containing the NH 2 -terminal 17% of apoB were secreted from the cells. MTP stimulated the secretion of all the segments containing this NH 2 -terminal apoB segment (Fig. 12). In contrast, none of the apoB segments that lacked the NH 2 terminus were responsive to MTP. In fact, five of these gene products were only detectable inside the Sf-21 cells and were not secreted, even in the presence of MTP. The most COOH-terminal protein product of apoB (B88 -100) was secreted by the cells but was unresponsive to MTP.
The NH 2 Terminus of ApoB Confers Secretion Competence and MTP Responsiveness to ApoB Internal Segments-To determine if the NH 2 terminus of apoB could confer secretion competence and MTP responsiveness to internal apoB segments, the NH 2 -terminal 17% of apoB (B17) was fused with three internal apoB fragments. Cells expressing these three segments (B33-46, B69 -79, and B48 -80) were unable to se-crete them (Fig. 12). Fusion of these internal segments to apoB17 enabled all three of these apoB segments to be secreted by Sf-21 cells (Fig. 13). Co-expression of the fusion proteins with MTP resulted in a 3.6-fold induction in the secretion of the longest construct (B17-(48 -80)) while the shorter fusion proteins were insensitive to MTP. This is consistent with the size correlation observed in Fig. 12 but does not rule out a role for specific sequences in lipid binding.

DISCUSSION
To understand the mechanism of triacylglycerol targeting to the secretory pathway of cells, we introduced mammalian gene products into invertebrate host cells. Invertebrate cells are unable to effectively secrete triacylglycerol, although they actively synthesize triacylglycerol when given media supplemented with free fatty acid.
Individually, expression of the mammalian genes for apoB or MTP did not promote substantial triacylglycerol secretion. However, in combination, apoB and MTP promoted a striking increase in the levels of triacylglycerol secreted from the cells. This observation suggests that the productive partitioning of triacylglycerol into the secretory pathway requires an ER-associated neutral lipid transfer activity (provided by MTP) and a vehicle by which the lipid can be shuttled through the pathway and out of the cell (provided by apoB).
The lipoproteins secreted by Sf-21 cells expressing apoB and MTP contained phospholipid and triacylglycerol. Unlike mammalian cells, cultured Sf-21 insect cells do not synthesize cho- lesterol esters (43), thus none were detected in the secreted lipoproteins. This suggests that cholesteryl ester is not a required substrate for the formation of apoB-containing lipoproteins.
Invertebrate fat body tissue produces apolipoproteins that are capable of binding and transporting neutral lipids. However, their neutral lipid transfer activity (lipid transfer particle) is localized extracellularly, in the hemolymph. It is therefore likely that the differences in lipoprotein assembly that have arisen within the animal kingdom (substrate usage and site of assembly) are due to differences in the secretion or retention of the respective lipid transfer activities.
The location of the animal's lipid transfer activity is likely to influence the neutral lipid species (triacylglycerol or diacylglycerol) that is released from the cell. Vertebrates and invertebrates both appear to mobilize cytosolic triacylglycerol stores by first hydrolyzing them into more soluble diacylglycerol (11,45). In vertebrates, it appears that the diacylglycerol is reesterified to triacylglycerol at the cytoplasmic face of the ER (45). The ER localization of MTP places it in close proximity to the re-esterification reaction, making it accessible to triacylglycerol. In addition MTP transports nonpolar lipids more actively than amphipathic lipids (46). MTP's subcellular location and substrate specificity are therefore likely to contribute to the partitioning of triacylglycerol rather than diacylglycerol into the secretory pathway where it is complexed with apoB. In contrast, the invertebrate neutral lipid transfer activity (lipid transfer particle) is extracellular. The lack of triacylglycerol re-esterification activity at the plasma membrane would likely limit the availability of this lipid for extracellular assembly. However, diacylglycerol is more soluble and membrane-permeable than triacylglycerol (47,48), making it a better candidate for transfer across the plasma membrane for extracellular lipoprotein assembly.
The ability of Sf-21 cells to secrete apoB48 enabled us to study the effect of MTP on apoB secretion and determine whether specific regions of apoB are important for its secretion and MTP-responsiveness. These experiments demonstrate a clear role for both subunits in apoB secretion. In Sf-21 cells, co-expression of apoB48 with the 97-kDa MTP subunit alone had little or no effect on apoB48 secretion (Fig. 7). This suggests that the 97-kDa subunit alone is incapable of stimulating apoB secretion and that this subunit is also incapable of utilizing endogenous insect PDI as a productive subunit. Co-expression of both subunits of MTP in Sf-21 cells has been shown to result in detectable levels of neutral lipid transfer activity (23). To determine if specific regions of apoB are required for secretion and MTP responsiveness, nine apoB segments (Fig. 11) were expressed with or without both MTP subunits in oleic acid-treated cells. B-"48" represents the apoB48-like protein secreted by cells infected with the apoB100-encoding virus. B-100H (B-100 heavy) represents the full-length protein secreted by the same cells. Quantitative immunoblotting was used to assess levels of induction. For each construct, secretion levels in the presence of PDI alone were normalized to a value of 1, and levels in the presence of both subunits are expressed relative to that. Five of the segments were retained intracellularly and were not detected in the media (ND), even in the presence of MTP. Error bars represent the standard deviation from three independent determinations. Independent t-test analysis for apoB17 resulted in p ϭ 0.06005 and for apoB88 -100 p ϭ 0.32946. Representative immunoblots are shown for secreted and intracellular apoB. Odd-numbered lanes represent samples from cells expressing the apoB segment and PDI. Even-numbered lanes represent samples expressing apoB and both MTP subunits.
In the present study, expression of apoB48 in the presence of both MTP subunits increased apoB48 secretion 8-fold over either individual subunit alone. This demonstrates a requirement for both MTP subunits and active lipid transfer for the stimulation of apoB secretion. Analysis of total protein secretion demonstrated specificity of the MTP effect for apoB48 (Fig. 8).
ApoB produced in hepatoma cells is subject to rapid intracellular proteolysis under conditions that do not favor apoB secretion. Its secretion is enhanced by oleate through rescue from proteolysis (44,49). In primary rat hepatocytes, apoB is rapidly degraded, but its secretion rate is unaffected by oleate (50). Pulse-chase experiments demonstrated that apoB is not rapidly degraded in Sf-21 cells (Fig. 9). Accordingly, the protease inhibitor, N-acetyl-Leu-Leu-norleucinal, which inhibits apoB degradation in mammalian cells (51,52), seems to have little if any effect on Sf-21 apoB levels. 2 The fact that we see a basal level of apoB secretion in the Sf-21cells in the absence of MTP might be attributable to the lack of an intracellular apoB degradation system. Therefore, the ability of oleate and MTP to stimulate apoB secretion in this system might not occur as a result of rescue of apoB from proteolysis. Our data also suggest that apoB synthesis is not changed by oleate or MTP. Our results are consistent with intracellular transport rather than degradation being rate-limiting for apoB secretion, as recently suggested by Bonnardel and Davis (53). Therefore, this system might be useful for identifying the primary impediment to apoB secretion and the mechanism by which MTP exerts its effect.
Many studies have been carried out to define regions of apoB that may participate in lipoprotein formation. Limited trypsin proteolysis has demonstrated that some regions of apoB100, when trypsinized, readily dissociate from LDL, while other regions remain tightly associated with the lipoprotein particle (54). Complementary studies where proteolytic fragments of apoB100 were incubated with lipid microemulsions demonstrated that specific regions of apoB100 are more lipophilic than others (55,56). Although these studies suggest that certain regions of apoB may readily bind lipid, they provide little information about which regions are essential for lipoprotein assembly.
Sequence based predictions suggest that apoB100 contains several amphipathic motifs throughout its sequence. A recent analysis suggests that apoB has a pentapartite structure of three amphipathic ␣-helical stretches alternating with two am-phipathic ␤-sheet stretches (57). The ␤-sheet regions are predicted to exist as irreversible lipid-associating domains that encompass amino acids 827-1961 and 2611-3867.
Systematic COOH-terminal truncation of apoB100 suggests that there are no specialized regions within the COOH-terminal 70% of the protein that are essential for lipoprotein assembly (16 -18). These studies suggest that the lipid binding capacity of apoB is a function of its total length. While these studies thoroughly examined the effect of apoB COOH-terminal deletions, all constructs retained the NH 2 terminus of apoB.
The present study tested whether apoB contains specific regions that are required for protein secretion and MTP responsiveness. Nine apoB segments spanning the entire length of the protein were expressed with and without MTP. These segments included NH 2 -terminal, COOH-terminal, and internal apoB fragments (Fig. 11) and varied in length from 10% of apoB through the full-length apoB100.
Of the initial constructs tested, only those containing the NH 2 terminus of apoB were secreted by the cells and were stimulated by MTP (Figs. 7 and 12). In contrast, none of the constructs that lacked the NH 2 terminus of apoB were stimulated by MTP. The majority of these apoB segments (5 out of 6) were trapped inside the cells and remained so, even in the presence of MTP. The trapped segments consisted mainly of internal regions of apoB. All of the trapped segments overlapped with apoB sequences predicted to contain irreversible lipid-associating amphipathic ␤-sheets (57). Secretion-negative apoB segments obtained from tunicamycin-treated cells showed greater electrophoretic mobility than those from untreated cells, indicating that the trapped internal segments were at least targeted to the ER. 3 This suggests that the heterologous signal sequence functioned properly.
The one segment that lacked the NH 2 terminus of apoB but was secreted was apoB88 -100. This segment corresponds almost exactly with the final exon of apoB (exon 29). The apoB88 -100 region does not overlap with a predicted irreversible lipid-associating region of apoB, but it contains sequences predicted to form reversible lipid-binding ␣-helixes much like non-apoB apolipoproteins. Although apoB88 -100 was capable of being secreted, it was not responsive to MTP. This is also the case for non-apoB apolipoproteins.
Analysis of the nine apoB segments tested suggested that the NH 2 terminus may be important for secretion of internal regions of apoB and for MTP responsiveness. To further explore this idea, three fusion proteins were generated. These contained three internal fragments that were secretion-defective and nonresponsive to MTP. Each was fused in frame with the NH 2 -terminal 17% of apoB (apoB17). Following fusion with apoB17, all of the internal fragments could be detected in the media of expressing cells (Fig. 13), suggesting that apoB17 had conferred secretion competence on them. Only the longest of the three was responsive to MTP. This construct, apoB17-(48 -80), represents 49% of the total apoB sequence, and its secretion was stimulated 3.6-fold by MTP.
The reason for the importance of the NH 2 terminus of apoB in directing apoB secretion and enabling MTP responsiveness is currently unclear. This region may interact directly with MTP and may mediate MTP sensitivity by providing a nucleation point for lipid acquisition by apoB. Previous work suggests this region must be completely translated before nascent apoB can be lipidated (58). Other studies have suggested that the NH 2 terminus of apoB undergoes novel translocational pausing, resulting in transient transmembrane intermediates (19,20) and may integrate into the inner leaflet of the mem- FIG. 13. Analysis of apoB17 fusion proteins. To assess any requirement for the NH 2 terminus of apoB for secretion and MTP sensitivity, apoB17 was fused in frame with three internal segments of the protein. The length of the fusions (relative to apoB100) is indicated. Quantitative immunoblotting (as for Fig. 12) demonstrated that all three fusions were detectable in the media of infected cells and that secretion of the longest fusion was stimulated by MTP. ND indicates no detectable induction. The induction value for apoB17-(48 -80) is a mean of three determinations Ϯ S.D. brane (59). (An alternative model is that apoB pauses during translation rather than translocation (59).) Additional studies suggest that the NH 2 terminus eventually translocates into the ER lumen (21). It has been suggested that translocation then pauses so that apoB exists as a transient transmembrane protein with the NH 2 terminus extending into the lumen of the ER, while the remainder of the protein remains cytosolic. Without the impetus to resume translocation (i.e. binding of lipid presented by MTP within the ER lumen), the arrested apoB is susceptible to cytoplasmic degradation and subsequent secretion of the already lumenal NH 2 terminus (21).
The NH 2 terminus of apoB is very rich in cysteine residues. Twelve of the protein's 25 cysteine residues are found in the NH 2 -terminal 11% of the protein (60,61). All 12 of these cysteines are involved in disulfide bond formation, while only 4 of the remaining 13 cysteines are found in disulfide bonds (62). Without ER lumenal proteins to hold them in place and promote their forward translocation, secretory proteins are free to retrotranslocate and free themselves from the ER membrane (63). A protein such as apoB that is capable of periods of translocation arrest may require additional mechanisms to prevent retrotranslocation. A compact, lumenal NH 2 terminus (held together with disulfide bonds) may form a disulfide knot, thus preventing retrotranslocation and keeping apoB accessible to lipids presented by MTP. Subsequent lipid acquisition is likely to stimulate the forward translocation of the rest of the protein, ultimately leading to lipoprotein maturation and the subsequent secretion of the lipid and apoB complex.
In summary, we have created a system with which to study the process of apoB and MTP-mediated lipoprotein assembly. Our ability to promote lipoprotein formation in insect cells (utilizing several gene products encoded by separate viruses) has allowed us to probe the importance of each individual protein. A detailed analysis of apoB suggests that the NH 2 terminus of this protein is essential but not sufficient for MTPmediated lipoprotein formation. Future studies should allow a better understanding of the mechanism by which the NH 2 terminus has its effect and of which NH 2 -terminal elements are essential. Similar studies may help define the functional elements of MTP and should aid in the analysis of other gene products that might participate in lipoprotein formation.