Apolipoprotein B-Containing Lipoprotein Particle Assembly: Lipid Capacity of the Nascent Lipoprotein Particle

formation of a nascent lipoprotein particle, and iii) the “lipid pocket” created by the first 1000 amino acid residues of apoB-100 is PL-rich, suggesting a small bilayer type organization and has a maximum capacity on the order of 70 molecules of lipid. This model is supported by the all-atom molecular model of the βα 1 lipid pocket presented in the accompanying paper.


Introduction
between apoB:931 to apoB:1000, the nature of particles is changed from a lipid-poor to a lipidated particle in the HDL 3 -like density range. The lack of a 1:1 molar ratio of apoB to MTP observed in this study supports a model in which the first 1000 amino acid residues of apoB are competent to complete the "lipid pocket" without a structural requirement for MTP. This nascent lipoprotein intermediate has a relatively constant Stokes diameter of 112 Å, a mean density of 1.21 g/ml, and has a maximum capacity on the order of 70 molecules of lipid per particle, primarily phospholipids. The surface:core lipid ratio of approximately 4:1 supports a bilayer type arrangement that is not responsive to the presence of oleate in the culture medium.

Amino Acids 931-1000 of ApoB-100 are Critical for the Formation of a Lipoprotein Particle.
To determine the effect of the 69 amino acid residues between apoB:931 and apoB:1000 on the relative flotation density of apoB:931-and apoB:1000-containing particles, cells were incubated overnight in serum-free Dulbecco's modified Eagle's medium (DMEM). Lipoproteins (d < 1.23 g/ml) and infranatant (d > 1.23 g/ml), isolated from the conditioned media, were analyzed by nondenaturing gradient gel electrophoresis (NDGGE) and immunobloted with monospecific polyclonal antibody to apoB-100. As shown in Fig. 1 A, lane 1, apoB:931 expressing cells secreted a major particle with Stokes diameter (S d ) of 110 Å and a minor particle with S d of 96 Å. Only a small fraction of the larger apoB:931-containing particle was recovered in the d < 1.23 g/ml fraction (compare lane 2 to lane 3), indicating that these particles are predominantly lipidpoor. The apoB:1000 expressing cells secreted a major particle with S d of 112 Å and trace amount of smaller particle with S d of 95 Å ( Fig. 1 B, lane 1). In contrast to apoB:931-containing particles, a larger fraction of apoB:1000 was recovered in d < 1.23 g/ml ( Fig. 1 B, lane 2) and a smaller fraction was recovered in d > 1.23 g/ml ( Fig. 1 B, lane 3), suggesting that the apoB:1000-containing particles are relatively lipid-rich. As shown in Fig. 2 B, lane 2, only the larger, apparently monodisperse, apoB:1000-containing particle floated.
Cells were incubated with 0.4 mM [ 14 C]-labeled oleic acid bound to 0.75% bovine serum albumin (BSA) and conditioned medium was subjected to NDGGE followed by autoradiography. As shown in Fig. 2  antibody to human apoB-100. In contrast, and consistent with results shown in Fig. 1 A, apoB:931-expressing cells secreted two forms of apoB-containing particles with S d of 110 Å and 96 Å (Fig. 2 A, lane 4) and apoB:1000-expressing cells secreted a major monodisperse particle with S d of 112 Å (Fig. 2 A, lane 5). Autoradiography of a duplicate gel demonstrated that the apoB:931-containing particles contained little, if any, radioactivity (Fig. 2 B, lane 4). In contrast, the band corresponding to apoB:1000-containing particles demonstrated approximately a 10-fold greater level of radioactivity ( Fig. 2 B, lane 5) than that associated with the apoB:931 band ( Fig.   2 B, lane 4). The bands on top of the gel might be aggregated proteins that remained in the large pore region of the gel or endogenous rat lipoproteins labeled in the lipid moiety. Similar results were obtained when studies were carried out using [ 3 H]-labeled glycerol (data not shown).
Cells were then labeled with [ 3 H]glycerol or [ 14 C]oleic acid and secreted apoB-containing lipoproteins were isolated by immunoprecipitation or NDGGE and analyzed for lipids as described in Materials and Methods. The results shown in Table I demonstrate that the apoB:1000-containing particles contain at least 4 times as much labeled lipid as the apoB:931containing particles. The above results clearly show that the 69 amino acid residues between apoB:931 and apoB:1000 are necessary for the formation of a lipidated particle. and TAG-59 (containing two fatty acids of 18 carbons and one fatty acid of 20 carbons) account for 29%, 34%, 35%, and 2%, respectively, of the total TAG subspecies in the control medium (without oleic acid) and 8%, 18%, 71%, and 3%, respectively, in the oleate-supplemented medium. This variation in the two experimental conditions, together with uncertainty of the exact number of 18-carbon chain fatty acids that might be labeled, could potentially introduce error in calculations of the number of lipid molecules, especially TAG, per apoB.

Analysis of [ 3 H]-Labeled Lipids Associated With
We used immunoprecipitation as the first step in determining the lipid composition of the secreted apoB-containing particles. Since apoB:931 was secreted mostly as a lipid-poor particle with peak density of 1.25 g/ml or greater 24 and contained very low levels of radiolabeled lipids ( Table 1 and Fig. 1 B), the number of lipid molecules per apoB:931 particle could not be accurately calculated by equations used in this study. Therefore, we have not shown any results on the lipid composition of apoB:931 that consisted of only a few PL molecules per particle.
Results of the calculation of the lipid composition of metabolically labeled apoB:1000containing particles isolated by immunoprecipitation with monospecific polyclonal anti-human apoB-100 are shown in Table II. ApoB:1000-containing particles secreted by control cells (incubated without oleic acid) contained 57% PL and 33% TAG and those secreted by the oleatesupplemented cells contained a higher content of TAG (58%) and a lower content of PL (39%).
Based on these results, the calculted stoichiometries of PL and TAG molecules per apoB:1000 were 36 and 19, respectively, in the absence of oleate and 22 and 31, respectively, in the presence of oleate.

Analysis of ApoB:1000-Containing Particles by NDGGE Demonstrates the Formation of
Stable, Monodisperse Lipididated Particle. To circumvent the potential nonspecific precipitation and/or adsorption of rat TAG-rich particles by immunoprecipitation, we isolated apoB:1000containing particles by NDGGE. Cells were metabolically labeled with [ 3 H]glycerol in the presence and absence of 0.4 mM oleic acid bound to 0.75% BSA and the labeled conditioned medium was concentrated and applied to NDGGE. To determine the lipid composition of apoB:1000-containing particles that floated, d < 1.23 g/ml fraction was isolated from the conditioned medium and was also subjected to NDGGE. Bands corresponding to apoB:1000containing particles with S d of 112 Å in total medium (Fig. 3 B, lanes 1-3), identified by immunoblotting of a duplicate gel (Fig. 3 A, lane 1), and d < 1.23 g/ml lipoproteins (Fig. 3 B,   lanes 4-6), identified by immunoblotting of a duplicate gel (Fig. 3 A, lane 2), were excised, digested and extracted for lipids as described in Material and Methods. In the absence of oleic acid, the isolated particles, from both the total medium and d < 1.23 g/ml fraction, contained 71-74% PL and 18-21% TAG and this composition was not altered by incubation of cells with oleate ( Table III). The calculated stoichiometries of PL and TAG per apoB:1000 in both the total conditioned medium and d < 1.23 g/ml fraction were 50 and 12, respectively, and were not responsive to the presence of oleate in the culture medium (Table III). Thus, the surface to core lipid ratio of apoB:1000-containing particles measured in this way was approximately 4:1 and was not affected by the addition of oleic acid to the culture medium.

Mass Analysis of Lipids Aassociated with ApoB:1000-Containing Particles Isolated by
Immunoaffinity Chromatography Supports the Stoichiometry of the "Lipid Pocket" Determined by Metabolic Labeling. To obtain independent confirmation of the results obtained from metabolic labeling and NDGGE studies, apoB:1000-containing particles were isolated by immunoaffinity chromatography of unlabled conditioned medium on an immunosorber with affinity purified monospecific polyclonal antibody to human apoB-100. Aliquots of the retained fraction, before and after 10-fold concentration, were analyzed for purity by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE) and NDGGE in conjunction with immunoblotting. A single band with the expected apoB:1000 molecular weight of 112 kDa was detected on SDS-PAGE by both Colloidal Blue staining (Fig. 4 A, lanes 1 and 2) and immunoblotting with anti-human apoB-100 ( Fig. 4 B, lanes 1 and 2). Analysis of the retained fraction on NDGGE showed a single band with the predicted S d of 112 Å by both Colloidal Blue staining ( Fig. 4 C, lane 2) and immunoblotting (Fig. 4 D, lane 2). The isolated apoB:1000containing particles secreted by control cells contained 87% PL, 7% TAG and 7% CE and particles secreted by oleate-supplemented cells contained 80% PL, 13% TAG and 7% CE ( Table   IV). The calculated numbers of PL, TAG, and CE molecules per particle were 64, 5, and 5, respectively, in the control cells and 56, 8, and 7, respectively, in oleate-supplemented cells ( Table IV). The surface to core lipid ratio in the presence of oleic acid was approximately 4:1 and thus supports the results obtained with metabolic labeling and NDGGE (Table III). These results indicate that the "lipid pocket" created by the first 1000 amino acid residues of apoB-100 has a maximum capacity of approximately 70 molecules of lipid.
As shown in Tables III and IV, the lipid composition and the lipid capacity of the apoB:1000-containing particles isolated by NDGGE and immunoaffinity chromatography were not responsive to oleate supplementation of the cells. In contrast, the lipid composition and hence the surface to core lipid ratio of apoB:1000-containing particles isolated by immunoprecipitation were altered by oleate supplementation of the cells (Table II) and were also different from those isolated by either NDGGE or immunoaffinity chromatography. Particles isolated by immunoprecipitation contained considerably more TAG molecules per particle and this was further increased by the addition of oleate to the culture medium, i.e., the surface:core lipid ratio of particles secreted by control and oleate-treated cells were 2.3 and 0.8, respectively. We suggest that isolation of truncated apoB-containing particles by immunoprecipitation might introduce an artifact in the lipid composition of secreted particles, especially in hepatic cell lines, perhaps by nonspecific immunoprecipitation and/or adsorption of endogenous rat TAG-rich particles to Protein G.  Based on these results we suggested that MTP might be an integral structural component of the lipid pocket. This would imply that one molecule of MTP should bind to one molecule of apoB:1000. To test this potential association between MTP and apoB:1000, cells were incubated in serum-free DMEM for 24 h and concentrated conditioned medium was subjected to NDGGE.
The gel was stained with Colloidal Blue and the band corresponding to apoB:1000 ( Fig. 3 B) was excised and analyzed by mass spectroscopy. In separate experiments, the concentrated conditioned medium was subjected to NDGGE and proteins were transferred onto PVDF membranes and stained with Coomassie Blue. The band corresponding to apoB:1000, identified by immunoblotting of a duplicate gel, was analyzed by amino acid sequencing. We were unsuccessful in dectecting any measurable MTP in apoB:1000-containing particles by either mass spectroscopy or amino acid sequencing (data not shown). These results suggest strongly that MTP is not a structural component of the "lipid pocket" but do not rule out its role in the lipidation of the apoB:1000-containing particles.

Discussion
In our previous studies 19,20 based on sequence homology between the βα 1 domain of apoB-100, i.e., the first 1000 amino acid residues of the mature protein, and LV, we proposed that formation of a LV-like lipid pocket is necessary for lipid transfer to apoB-containing lipoprotein particles. We suggested that initiation of particle assembly occurs when βα 1 domain folds into a three-sided LV-like lipid binding cavity, or alternatively, the lipid pocket is formed by association of the region of the βα 1 domain homologous to the βA and βB sheets of LV with βDlike amphipathic β sheet from MTP. 19,20 In the present paper, we have shown that apoB:1000 forms stable particles without the 1:1 molar ratio of MTP to apoB required if MTP was necessary as an integral structural component of the lipid pocket. In the accompanying paper (Richardson et al., submitted), we propose an allatom model for the formation of the "lipid pocket" by the first 1000 amino acid residues of apoB-100 that is completely consistent with the results of the present study. In this model, we describe a hairpin-bridge mechanism for lipid pocket completion in which a portion of a nonhomologous loop from one of the two amphipathic β sheets of the lipid pocket folds as an amphipathic helical hairpin to bridge the distance between the two sheets. This creates a third side to the lipid pocket without a structural requirement for MTP.
Based upon the depth of the pocket (approximately 40 Å), similar to the thickness of the hydrophobic core of a phospholipid bilayer, we have suggested it probable that lipids form an asymmetric bilayer assembly, containing a neutral lipid lens as described previously, 20 in the nascent lipid pocket. A minimum of 44 POPC molecules were manually docked into the lipid pocket of the model for the βα 1 domain of apoB, a number quite close to the experimental number of approximately 50 phospholipids per particle found in this study.
We previously demonstrated 24 that apoB:931 formed a particle with a constant diameter of 110 Å across a wide range of densities and a mean density of 1.25 g/ml or greater, denser than the classical HDL density range of 1.063-1.21 g/ml. The major particle formed by apoB:1000 had a diameter that remained constant at approximately 112 Å across a wide range of densities and a mean density of 1.208 g/ml, within the HDL 3 density range of 1.125-1.21 g/ml. In that same study, the larger construct, apoB:1200, containing a significant number of the amphipathic β strands located in the β 1 domain, formed a large particle with a diameter that increased with decreasing density, ranging from 118 Å to 127 Å, and a mean density of 1.197 g/ml, within the HDL 3 density range. This suggests that apoB:1200-containing particles possess additional, but varying, numbers of lipids compared to apoB:1000-containing particles.
In the present study we have experimentally confirmed that over a short stretch of 69 amino acids from apoB:931 to apoB:1000, the nature of the secreted particle is changed from an essentially lipid-free particle well outside the HDL density range to a lipidated particle within the HDL 3 density range. As shown in Table I and Fig. 2, the incorporation of [ 3 H]glycerol and [ 14 C]oleate into the lipid moieties of apoB:1000-containing particles was 4-10-fold higher than that in apoB:931-containing particles.
The apoB:1000-containing particles contained, on average, 50 molecules of PL, 12 molecules of TAG, and 6 molecules of cholesteryl ester per particle, for a surface:core lipid ratio of approximately 3:1. The X-ray crystal structure of lamprey LV 25 suggests a "lipid-pocket" containing approximately 27 molecules of PL and 11 molecules of TAG per LV monomer, a surface:core lipid ratio of approximately 2:1. To put these ratios in perspective, the surface:core lipid ratio of spheroidal HDL 3 particles is 1.5:1, 26 supporting the concept that the lipid in the apoB:1000-containing particles is in the form of a bilayer assembly, rather than in the form of a mixed micelle assembly like HDL 3 .
Negative stain electron microscopy of isolated apoB:1000-containing particles showed that these particles differ from typical HDL in two ways: 1) they are asymmetric in shape and 2) their edges are fuzzy in appearance. The majority of the particles display an elongated shape, with Biochemical studies have suggested that assembly of apoB-100 into a lipoprotein particle occurs co-translationally 8,9 and requires the activity of MTP. 9,16 However, little was known about the exact mechanisms by which apoB is assembled into a TAG-rich lipoprotein, i.e., the minimum structural requirement for initiation of lipoprotein assembly, the kinetics of lipid recruitment, specifically bulk lipid addition, the composition and number of lipids in the primordial lipidated particle, and the role of MTP in these processes. One often quoted mechanism for the physical assembly of lipid particles containing apoB is the budding oil droplet. 7 In this model, the N-terminal portion of apoB is embedded in the inner monolayer of the endoplasmic reticulum (ER) membrane, where it nucleates an oil droplet from the supersaturated rough ER membranes. Upon completion of apoB synthesis this oil-droplet is detached from the bilayer to form the nascent lipoprotein. However, thermodynamic considerations make it unlikely that lipoproteins assemble through the wholesale remodeling or dismantling of membrane bilayers.
An alternate model for the initiation of apoB assembly has been suggested by Small and colleagues. 27 In this model, the formation of the "primordial" lipoprotein particle by a multistep process involves the initial recruitment of PL by the N-terminal region followed by incorporation of core lipids directed by the presence of β-sheets at and beyond apoB-29. 27 They proposed that the N-terminal 20% of apoB interacts with the internal PL leaflet of the ER by amphipathic structures, mainly β-sheets, present between apoB-13 and apoB-20. They suggested that the recruitment of TAG is due to its association with the region of the 28 predicted amphipathic βstrands extending from apoB-29 to apoB-41. 27 Our results demonstrating that apoB:931 does not contain the structural elements to form a lapidated particle are at odds with recent studies by Shelness et al. 28 showing that apoB-20.5 (amino acids 1-931) secreted by transfected COS cells contained sufficient structural and functional domains to form a secretion-competent lipoprotein particle. Contrary to our results demonstrating that apoB:931 particles have a mean density of 1.25 g/ml and are lipid-poor, the apoB20.5 secreted by COS cells had a peak density of 1.20 g/ml and contained 34 molecules of surface lipids and 49 molecules of core lipids, predominately TAG. 28 These investigators 28 showed that truncated apoB as small as apoB-19.5 (amino acids 1-884) and apoB-20.1 (amino acids 912) formed small HDL 3 -like particles that contained 23-25 molecules of surface lipids and 28-36 molecules of core lipids. They suggested 28 that apoB-containing lipoproteins are initially formed as small, dense emulsion particles, with a surface:core lipid ratio of ≤ 1:1, where apoB inserts itself into a saturated membrane surface and desorbs lipid as a preformed core-containing lipoprotein, a model consistent with the budding oil droplet mechanism. 7 We are not sure of the reasons for the observed differences in densities and lipid composition of apoB:931 in our study and that reported by Shelness et al. 28 We speculate that this discrepancy may, in part, be due to the fact that we used hepatic-derived cells and Shelness et al. 28 Although we do not know the exact mechanism by which apoB:1000 acquires PL, the data presented here indicates that gradual lipid transfer into an apoB-containing particle during biosynthesis requires translocation of a critical length of apoB sequence, i.e., the βα 1 domain (residues 1-1000), necessary for creation of a competent lipid pocket. In support of this hypothesis, several studies 31-33 have suggested that acquisition of lipid occurs stepwise along the secretory pathway. Our model is also consistent with the hypothesis that the initial step in the assembly of apoB-containing lipoproteins involves the recruitment by the N-terminal domain of apoB of surface lipids, primarily PL, and a small quantity of core lipids during and immediately after translation. 34 It has been suggested 34 that in the next step, cytosolic TAG droplets are hydrolyzed and TAG is re-synthesized on the ER for assembly with apoB and translation/translocation continues. As more TAG, PL and cholesterol are added to the particle, its size increases and a fully lipidated VLDL particle is eventually formed. 34 MTP might be involved in translocation of apoB, assembly of PL with apoB, and/or transfer of TAG to the core of the nascent VLDL particles. 34 Based on the results presented in this paper and the hairpin-bridge mechanism for the formation of the lipid pocket described in the accompanying paper (Richardson et al. submitted), we suggest the following model for the initiation of lipoprotein particle assembly. The lipid pocket formed by apoB:1000 begins to recruit PL and some TAG co-translationally to form a primordial lipoprotein particle. MTP serves as a shuttle to deliver lipids into the lipid pocket upon translation of apoB to and beyond residue 1000. Translation of the amphipathic β strands in the β 1 domain provides a mechanism for stabilization of the particle beyond the nascent apoB:1000 particle. When the particle reaches a critical size, the salt bridges holding the hairpinbridge in place break and apoB undergoes conformational change. As the hairpin-bridge is unlocked, βA and βB sheets separate; lipids are added to the flexible basal opening of the lipid pocket and the hydrophobic helixes of helix-turn-helix associate with the growing particle. This open conformation of apoB would allow its association with a much larger surface than the protein itself. We suggest that this is the conformation that apoB assumes in LDL, IDL, and VLDL.
In summary, we have demonstrated that amino acids 931-1000 of apoB-100 are critical for the initiation of apoB lipoprotein assembly. Based on experimentally derived results and molecular modeling, we propose that initiation of particle assembly occurs when βα 1 domain (amino acid residues 1-1000) folds into a three-sided LV-like lipid binding cavity to form the "lipid pocket" without the structural requirement of MTP. This hydrophobic cavity is subsequently loaded with PL and forms a particle with lipid composition and surface:core lipid ratio consistent with LV "lipid pocket". Lipid composition, total number of lipid molecules per particle, and the particle peak density of apoB:1000-containing particles are not responsive to oleic acid supplementation indicating that the "lipid pocket" formed by the N-terminal 1000 residues of apoB-100 has a fixed lipid capacity on the order of 50 PL for a total stoichiometry of 70 lipid molecules. Negative stain EM of isolated apoB:1000-containing particles revealed a structure consistent with that predicted for the apoB:1000 "lipid pocket" by molecular modeling.
nondenaturing conditions 37-39 using monospecific polyclonal antibody to human apoB-100 coupled to Protein G-Sepaharose CL-4B as previously described. 37,38 The beads were washed six to seven times until background count was obtained in the wash and then extracted for lipids as described below.
Nondenaturing Polyacrylamide Gradient Gel Electrophoresis (NDGGE). In unlabeled experiments, aliquots of concentrated total medium, lipoprotein fraction (d < 1.23 g/ml), and infranatant fraction (d > 1.23 g/ml) were subjected to 4-20% NDGGE for 48 h at 4°C in buffer containing 24 mM Tris-HCl, pH 8.3 and 192 mM glycine. Proteins were transferred onto PVDF membrane and immunoblotted with anti-human apoB-100 as described below. In metabolic labeling studies, aliquots of total medium and d < 1.23 g/ml lipoprotein fraction were run on 4-20% NDGGE; gels were stained and the bands corresponding to apoB:931 and apoB:1000, identified by their Stokes diameter and immunoblotting of a duplicate gel, were excised and analyzed for lipids described below. The incorporation of [ 3 H]-labeled glycerol or [ 14 C]-labeled oleic acid into total lipids of intact apoB-containing lipoproteins was also determined by NDGGE of the labeled conditioned medium and autoradiography of the amplified and dried gels.

Isolation of Truncated ApoB-Containing Particles by Immunoaffinity Chromatography.
Affinity purified monospecific polyclonal antibody to human apoB-100 was coupled to the cross-linked agarose activated with N-hydroxysuccinimide (Affi-Gel 10). After exhaustive washing of the Affi-Gel with cold, double-distilled water, the antibody solution was added to the gel slurry (2 mg protein/ml of gel) and the mixture was gently shaken for 2-3 h at room temprature. The supernatant was removed and the remaining active sites were blocked; gel was  27 The number of lipid molecules per apoB particle was calculated from percent composition of PL, DAG, and TAG determined by metabolic labeling and TLC analysis or by mass determination of PL, TAG, and CE in retained fraction from anti-apoB immunosorber. The equation for calculating lipid:apoB molar ratios of apoB:1000-containing particles was derived by using the calculated molecular weight of 111,375 kDa for nonglycosylated apoB:1000, based on amino acid sequence using DNAMAN program, and particle density of 1.208 g/ml, determined by density gradient ultracentrifugation followed by NDGGE and immunoblotting as previously described. 24 Electron Microscopy. For electron microscopy, concentrated retained fraction from anti-apoB immunosorber was dialyzed against ammmonium acetate buffer (2.6 mM, pH 7.4) and nagatively stained with 2% sodium phosphotungstate as previously described. 47