Scavenger Receptor Class B, Type I-mediated Uptake of Various Lipids into Cells

Scavenger receptor (SR)-BI is the first molecularly defined receptor for high density lipoprotein (HDL) and it can mediate the selective uptake of cholesteryl ester into cells. To elucidate the molecular mechanisms by which SR-BI facilitates lipid uptake, we examined the connection between lipid donor particle binding and lipid uptake using kidney COS-7 cells transiently transfected with SR-BI. We systematically compared the uptake of [ 3 H]cholesteryl oleoyl ether (CE) and [ 14 C]sphin-gomyelin (SM) from apolipoprotein (apo) A-I-containing reconstituted HDL (rHDL) particles and apo-free lipid donor particles. Although both types of lipid donor could bind to SR-BI, only apo-containing lipid donors exhibited preferential delivery of CE over SM ( i.e. non-stoichiometric lipid uptake). In contrast, apo-free lipid donor particles (phospholipid unilamellar vesicles, lipid emulsion particles) gave rise to stoichiometric lipid uptake due to interaction with SR-BI. This apparent whole particle uptake was not due to endocytosis, but rather fusion of the lipid components of the lipid donor with the cell plasma membrane; this process is perhaps mediated by

Scavenger receptor (SR)-BI 1 is a physiologically regulated receptor able to mediate the binding of HDL and the subsequent uptake of cholesteryl ester (CE) from the lipoprotein particle into cells (1)(2)(3). The uptake of CE from HDL is different from the endocytic uptake of cholesterol from low density lipoprotein (LDL), in that it does not involve the uptake of the whole lipoprotein particle. Instead, there is a preferential transfer of the CE from the HDL particle without concomitant uptake of the protein component of the lipoprotein (4). This selective CE uptake pathway had been studied extensively for many years prior to the discovery of SR-BI (5, 6), but its significance for lipoprotein metabolism remained uncertain until recently. The critical role for SR-BI in vivo has been demonstrated by gene transfer experiments showing that overexpression of this protein in the liver lowers plasma HDL cholesterol levels in mice (7,8) and decreases development of atherosclerosis (9,10). In contrast, a targeted null mutation in the SR-BI gene in mice increases plasma cholesterol levels and reduces delivery of HDL-CE to the adrenal gland (11).
SR-BI resides in the plasma membrane of cells and is composed of a glycosylated extracellular domain and two membrane-spanning domains near the N-and C-terminal regions of the molecule (11). The ϳ40-kDa extracellular portion of the receptor is critical for function because an antibody directed against this region inhibits the binding and selective CE uptake from HDL (12). Comparison of SR-BI to a closely related scavenger receptor, CD-36, shows that while both receptors can mediate binding of HDL, only SR-BI is efficient at promoting selective CE uptake and this function is attributable directly to the extracellular domain of SR-BI (13,14). Also, mutational analysis of this region of the receptor revealed 2 arginine residues that are critical for allowing the efficient uptake of CE from HDL (13). The binding of HDL to SR-BI and the selective uptake of CE from HDL are linked processes (14 -16). The first step in lipid uptake via SR-BI is the binding of the lipid donor particle to the receptor. The second step involves the movement of lipid from the SR-BI-bound lipid donor particle to the cell plasma membrane. Two models for the latter process have been suggested. Earlier studies from this laboratory led to the proposal that the receptor forms a nonpolar channel between a bound HDL particle and the cell plasma membrane through which hydrophobic molecules such as CE can diffuse down a concentration gradient created by the CE/phospholipid (PL) ratio of the HDL particle (16). Alternatively, it is thought that SR-BI might facilitate hemifusion between the phospholipid monolayer of an HDL particle and the external monolayer of the plasma membrane, thereby allowing lipid transfer to occur (17). It has been shown that selective lipid uptake from HDL occurs during HDL retroendocytosis (18); it is likely that the movement of CE molecules from HDL to the plasma membrane occurs both at the cell surface and in endocytic vesicles. In addition to facilitation of CE uptake, SR-BI mediates uptake of PL from reconstituted HDL particles (16,19). SR-BI can bind a variety of ligands (20,21) and exchangeable apolipoproteins are among them. Cross-linking studies have demonstrated that the class A amphipathic ␣-helix present in apolipoproteins is a recognition motif for the receptor (22), but multiple sites in apolipoproteins may be recognized by SR-BI (23). Furthermore, there are conflicting reports on the role of apoA-I and apoA-II in mediating HDL binding and selective CE uptake from HDL; Pilon et al. (24) have suggested that apoA-II is a better ligand for SR-BI than apoA-I but that apoA-II has an inhibitory effect on CE uptake. On the contrary, de Beer et al. (25) obtained results showing that apoA-II binds less well to SR-BI than apoA-I but is more efficient at promoting CE uptake than apoA-I. These results suggest that the nature of the ligand binding to SR-BI may influence the selective CE uptake process. Thus, a quantitative investigation of the role of exchangeable apolipoproteins in mediating binding to SR-BI and promoting SR-BI-mediated uptake of lipids is required for better understanding of the mechanism of action of SR-BI.
In this study, we examine the association between the binding of a lipid-donating particle and the subsequent lipid movement into the cell. The results suggest that binding of a lipid donor particle to SR-BI is insufficient to allow selective uptake of lipids by SR-BI; interaction of an apolipoprotein with SR-BI is essential for selective uptake of lipids from a donor particle.

Materials
Bovine brain phosphatidylserine (PS), liver phosphatidylinositol (PI), and 1-palmitoyl-2-oleoyl phosphatidylcholine (POPC) were purchased from Avanti Polar Lipids (Albaster, AL Apolipoproteins (apo) A-I and A-II were isolated from human HDL as described (26). Prior to use, the purified apolipoproteins were resolubilized in 6 M guanidine HCl and dialyzed against Tris-buffered saline (TBS) (10 mM Tris, 150 mM NaCl, 0.1 mM EDTA, pH 7.4). Human apoE-3 was expressed in Escherichia coli and purified according to the method of Morrow et al. (27) and solubilized in 6 M guanidine HCl and 1% (v/v) ␤-mercaptoethanol overnight and dialyzed extensively against 100 mM ammonium bicarbonate buffer prior to use (28). For some experiments, apolipoproteins were labeled with [ 125 I] using the iodine monochloride method (29); the specific activities of the labeled products were in the range of 300 to 1500 dpm/ng of protein.

Methods
Preparation of Small and Large Unilamellar Vesicles-POPC dissolved in chloroform:methanol (1:1, v/v) was dried under nitrogen onto the wall of a glass tube and then placed in a vacuum oven to completely remove any remaining solvent. The lipid was then rehydrated in TBS to form multilamellar vesicles (MLV). Using a sonicator (Branson Sonifier 350), this dispersion was sonicated on ice under nitrogen using a titanium tapered tip for 5 min followed by 1 min of cooling. This cycle was repeated 10 times until the initially cloudy lipid mixture became translucent. To remove the titanium debris that was produced during the sonication the samples were spun in a Beckman GPR centrifuge for 15 min at 3000 rpm. Then to separate any remaining MLV from the small unilamellar vesicle (SUV) layer, the sample was centrifuged in a Beckman 50 Ti rotor for 2 h at 4°C at 145,000 ϫ g (30). The top SUV layer was removed and used in competitive binding experiments. For lipid uptake experiments requiring incorporation of radioactive lipids, the above procedure was modified slightly. Radioactive lipid pairs ([ 3 H]COE or [ 3 H]FC with [ 14 C]SM) of interest in trace amounts (Ͻ1 mol % of each) were dried along with POPC, CO, FC (98:1:1 mol ratio) in a glass tube and processed as described above for unlabeled SUV. Negatively charged SUV containing either PS or PI were prepared in a similar manner and contained 48:48:1:1 mole ratios of PS or PI and POPC, FC, and CO. SUV used for measuring direct binding to cells at 4°C were prepared as described above except they contained only [ 14 C]SM as a marker. Preparation of SUV by the above method yielded vesicles 22 Ϯ 3 nm in diameter (30) and they were used within 10 days of preparation. For some experiments, apoA-I was complexed with SUV as described by Davidson et al. (31). In brief, dual-labeled SUV prepared as described above were incubated with apoA-I at an initial lipid to protein mole ratio of 170:1 at 24°C for 12 h. Thereafter, apoA-I not associated with SUV was removed using gel filtration chromatography (1 ϫ 30-cm Superdex 200 HR column, Amersham Pharmacia Biotech). The typical final lipid to protein mole ratio was 500:1; assuming that a typical SUV contained about 2000 PC molecules (32), 3 to 4 apoA-I molecules were associated with each vesicle.
Large unilamellar vesicles (LUV) were prepared with POPC:FC:CO (98:1:1 initial mol ratio) plus [ 3 H]COE and [ 14 C]SM in trace amounts (Ͻ1 mol %) using an extruder (Lipex Biomembranes, Vancouver, B.C., Canada). The appropriate amounts of lipids were dissolved in chloroform:methanol (1:1, v/v), mixed with radioactive lipids, and dried completely onto the wall of a glass tube under nitrogen. TBS was then used to hydrate and form a MLV suspension. To ensure complete recovery of all lipids from the wall of the glass tube, samples were subjected to 3 min of water bath sonication. The resulting MLV were passed 10 times through the extruder using two 200-m pore size filters at 100 pounds/ square inch of nitrogen. This procedure yielded ϳ90% LUV with a mean diameter of 185 Ϯ 15 nm (30). For experiments requiring the encapsulation of [ 3 H]inulin, the procedure was modified slightly. POPC, FC, CO, and [ 14 C]SM (5 Ci) in the appropriate amounts were dried onto the wall of a glass tube under nitrogen. Then, 500 Ci of [ 3 H]inulin (dissolved in phosphate-buffered saline) was added to the glass tube and the resulting MLV were passed through the extruder as described.
To separate the inulin trapped in the core of the LUV from unencapsulated material, BioGel A-1.5m (Bio-Rad) in a 1.6 ϫ 100-cm gel filtration column was used. Liquid scintillation measurements of the collected void volume showed that the LUV peak corresponded with all of the lipid label ([ 14 C]SM) and some of the [ 3 H]inulin. The column retained the overwhelming majority of the [ 3 H] counts (i.e. unencapsulated inulin). To ensure that [ 3 H]inulin associated with the LUV was a good marker for the aqueous compartment of the vesicle, a portion of the [ 3 H]inulin-containing LUV was disrupted by sonication. After gel filtration, liquid scintillation counting of the eluted LUV peak showed that at least 90% of the [ 3 H]inulin counts associated with the LUV were released. We attempted to encapsulate [ 3 H]inulin in SUV, but we were not able to achieve a sufficient level of incorporation for use in cell culture experiments.
Lipid Emulsion-Lipid emulsion particles composed of POPC and a core of triolein were prepared as described (33). In brief, POPC and triolein (1:1, w/w) dissolved in chloroform were dried onto the wall of a glass tube along with either [ 3 H]COE and [ 14 C]SM or [ 3 H]FC and [ 14 C]SM. The lipids were resolubilized in TBS containing 10% sucrose initially at 60°C. The lipid suspension was sonicated with a titanium flat tip (Branson Sonifier 350) for 5 min followed by a 5-min pause. This cycle was repeated 6 times after which the samples were centrifuged at a low speed for 30 min to remove titanium debris released from the tip. The sample was then placed in 12-kDa molecular mass cutoff dialysis tubing and dialyzed overnight to remove the sucrose. After this, the samples were centrifuged in a SW-40 rotor at 40,000 rpm for 20 min at 4°C. The middle milky emulsion layer was removed from the centrifuge tube, transferred to new tubes, and centrifuged for an additional 12 h under similar conditions. The top creamy layer was removed and used in lipid uptake experiments. A triolein to phospholipid ratio of 2:1 (w/w) was achieved and the lipid emulsion particle diameter was 35 Ϯ 5 nm (33).

Discoidal Reconstituted HDL (rHDL) Preparation and Characterization-rHDL containing [ 3 H]COE and human apolipoproteins A-I, A-II
, or E-3 were prepared by complexing the desired apolipoprotein with POPC, CO, and trace amounts (Ͻ0.01 mol %) of [ 3 H]COE using the cholate dispersion method according to Sparks et al. (34). An appropriate amount of POPC and CO was dried into a film first under a light stream of N 2, then under high vacuum. Subsequently, the lipids were hydrated in TBS, vortexed to generate MLV, and incubated with sodium cholate at 37°C for 1.5 h to generate detergent-mixed micelles. The solution was mixed every 15 min during the incubation and after 30 min was sonicated for 5-10 min in a bath sonicator to clarify the initially turbid solution. The mixture was then allowed to incubate for the remainder of the 1.5 h. Next, apolipoprotein freshly dialyzed from 6 M guanidine HCl into TBS was added to the detergent:lipid mixture at a POPC to protein ratio (mol:mol) of 100:1. The lipid-detergent-protein complexes were then aspirated into a syringe containing Bio-Beads (Bio-Rad) and the solution was rotated for 3 h at 4°C to facilitate the removal of cholate. Thereafter, the phospholipid-apoprotein discs formed were collected and separated from the Bio-Beads by ejecting the contents of the syringe through a 0.45-m sterile filter. Native polyacrylamide gradient gel electrophoresis (Phast System, Amersham Pharmacia Biotech) was used to confirm the formation of reconstituted HDL particles.
To obtain a homogenous population of rHDL particles each preparation was fractionated on a gel filtration column (1 ϫ 30-cm Superdex 200 HR column, Amersham Pharmacia Biotech). Identical size rHDL complexes (hydrodynamic diameters ϳ12 nm) were isolated and used within 10 days of preparation. The final mole ratio of apo:POPC:CO was typically 1:65:0.3. To measure uptake of various lipids in COS cells, rHDL were dual labeled with 14 C-and 3 H-labeled lipids of interest. These rHDL particles contained apoA-I, POPC, CO, and FC in mole ratios of 1:96:1:1 and trace amounts (Ͻ1 mol %) of two lipid labels of interest. All preparations of rHDL were made in parallel. Nondenaturing electrophoresis (Amersham Pharmacia Biotech) showed a homogenous band of equal size for all preparations. The combinations of radioactive lipid in each preparation were as follows: [ (16) and they were Ͼ99.9% pure.
Cell Culture Experiments-Kidney COS-7 cells were maintained and transiently transfected with the cDNA for SR-BI or the vector DNA as described by Connelly et al. (14). Concentration-dependent binding of [ 14 C]SM-labeled POPC-SUV and PC/PS-SUV was conducted at 4°C for 1 h in cells with and without SR-BI expression. After incubation, the cell monolayer was washed 3 times with 2 ml of ice-cold phosphatebuffered saline containing 0.1% bovine serum albumin plus one additional wash with 2 ml of phosphate-buffered saline alone. Then 0.1 N NaOH was used to solubilize the cells and an aliquot of the cell lysate was used to determine the cell associated radioactivity using liquid scintillation counting. Direct binding of ligand to COS-7 cells was analyzed using a nonlinear regression fitting to the equation, on Graph Pad Prism as described before (16). Binding of rHDL and selective CE uptake from rHDL were studied using COS-7 cells with and without SR-BI expression. rHDL containing [ 3 H]COE and 125 I-labeled apoA-I, apoA-II, or apoE-3 were incubated at 30 g of protein/ml with either SR-BI-transfected or vector-transfected cells for 2 h at 37°C. The cell monolayer was washed and solubilized as described above. Selective uptake of the CE was calculated as described previously (16). Cellular uptake of various types of lipid from discoidal rHDL was measured in COS-7 cells with and without SR-BI expression. rHDL particles with 14 C/ 3 H-labeled pairs of lipids, prepared as described above, were incubated with the cells at 30 g of protein/ml at 37°C for 1 h. Cells were washed as described above and lysed in 0.1 N NaOH. An aliquot of the lysate was taken to determine the cell associated lipid uptake, which was expressed as percent of label present in the incubation medium taken up by the cell per h/mg of cellular protein.
Selective CE uptake from [ 3 H]COE/ 125 I-apoA-I rHDL is ϳ95% in COS cells expressing SR-BI so that at least 95% of the cell associated [ 3 H]COE is due to the selective uptake pathway and not whole particle uptake. TLC analysis of the extracellular medium after the incubation period did not show any detectable decomposition of the lipid labels used.
Dual-labeled phospholipid vesicles and lipid emulsion particles prepared as described above were incubated with SR-BI-expressing and control COS-7 cells. For most experiments, the vesicles were incubated at 100 g of PL/ml for 1 h at 37°C. The cells were then washed and lysed as described above. Control experiments were also conducted by varying the temperature, time of incubation, and concentration of the vesicles in the incubation medium.

RESULTS
Our goal in this study was to examine the association between ligand binding and cellular accumulation of lipids mediated by SR-BI. There are numerous studies on SR-BI/ligand interaction indicating that the receptor can bind apolipoproteins as well as certain PL (for reviews, see Refs. 21 and 35). SR-BI was initially described as a receptor for anionic PL but not neutral PL (20). In agreement with this, the binding isotherm in Fig. 1 show that anionic PC/PS-SUV bind better (K d ϳ 14 g of PL/ml) and in greater amounts (B max ϳ 2800 ng of PL/mg of cell protein) than neutral POPC-SUV (K d ϳ 50 g of PL/ml and B max ϳ 900 ng of PL/mg of cell protein). Even at concentrations of 1 mg of PL/ml, POPC-SUV bound at Ͻ1000 ng of PL/mg of cell protein (data not shown). Since PC and PS form SUV of similar sizes (36), it is apparent that on a per particle basis, the PS-containing vesicles bind with higher affinity and to a greater degree. Since the data presented in Fig.  1 represent the difference in ligand binding between control and SR-BI-expressing COS cells, the measured SUV binding is attributable directly to SR-BI. For comparison, the binding of apoA-I rHDL occurs with similar affinity (K d ϳ 5 g of protein/ml and 11 g of PL/ml) as PC/PS-SUV but with similar capacity (B max ϳ 720 ng of PL/mg of cell protein) to that of POPC-SUV when expressed on the basis of PL. Since SUV contain ϳ10-fold more PL molecules than discoidal rHDL particles, the K d for the PC/PS-SUV is lower when expressed on a per particle basis. Comparison of the B max values shows that fewer PC/PS-SUV or PC-SUV can be accommodated at saturation on the SR-BI at the cell surface.
The Influence of Apolipoprotein Content on SR-BI-mediated CE Uptake-Since SR-BI displays a broad ligand specificity, it is important to examine in detail the effects of donor particlereceptor interaction on the cellular uptake of lipids. In particular, the roles of different apolipoproteins in this process need to be compared. Fig. 2 shows that the major exchangeable apolipoproteins of HDL are all able to bind to SR-BI and mediate CE selective uptake. In these experiments, dual-labeled rHDL ( 125 I-apolipoprotein, [ 3 H]cholesteryl oleoyl ether (COE)) were incubated with COS-7 cells with and without SR-BI expression at 30 g of protein/ml for 2 h at 37°C. The inability of apoC to form stable PL complexes of comparable sizes precluded their use in this study. The data summarized in Fig. 2 show that apoA-I is the most effective donor of CE to cells via SR-BI. Compared with vector-transfected cells there was a 7-fold increase in binding of apoA-I-containing rHDL to SR-BItransfected cells (from 36 to 277 ng of protein/mg of cell protein) and a concomitant 19-fold increase in CE selective uptake. ApoA-II rHDL had a 3-fold increase in binding to SR-BI-expressing cells compared with vector cells (from 53 to 172 ng of protein/mg of cell protein) and a corresponding 17-fold increase of CE selective uptake. ApoE-3 rHDL exhibited the highest level of binding but the incremental increase due to SR-BI was only 3-fold, suggesting that the particles also bind to sites other than SR-BI. The control cells showed a significantly higher binding of apoE-3 rHDL compared with apoA-I or apoA-II rHDL; this difference is probably a reflection of apoE binding to proteoglycans or LDL receptor family members (37).
When comparing the relative efficiencies of selective CE uptake promoted by the various apolipoproteins, the data (Fig.  2) need to be evaluated in several ways; for example, factors such as the amount of CE per rHDL particle and the number of particles bound to the receptor have to be taken into account. Taking these factors into account indicated that for each nanogram of rHDL CE bound to SR-BI at steady state, there were 24, 41, and 11 ng of CE taken up selectively in 2 h from rHDL containing apoA-I, apoA-II, and apoE-3, respectively. Thus, when the data are normalized for the amount of CE uptake per equivalent of CE bound, apoA-II rHDL shows the most efficient CE selective uptake. However, when analyzed in terms of CE uptake per ng of protein bound, apoA-I rHDL and apoA-II rHDL show similar efficiencies (0.1 ng of CE/ng of protein bound) while apoE-3 is the least efficient (0.05 ng of CE/ng of protein bound). Since the rHDL particles had similar diameters and similar protein to PL ratios, normalizing to the amount of protein bound is equivalent to normalizing to the number of particles bound. Interestingly, de Beer et al. (25) reported that apoA-II is more efficient at mediating selective CE uptake via SR-BI than apoA-I using rHDL containing both apoA-I and apoA-II. It is difficult to compare our results with their findings because we used rHDL containing only apoA-I, apoA-II, or apoE-3. Nonetheless, it is clear that the nature of the apolipoprotein molecule can have a significant influence on SR-BImediated selective CE uptake.
Specificity of SR-BI-mediated Lipid Selective Uptake-Having demonstrated that all major apolipoproteins of HDL can mediate selective CE uptake, we next focused our attention on the various classes of lipid molecules in HDL (e.g. free cholesterol (FC), triglycerides, and phospholipids) that can potentially be metabolized by cells. Ji et al. (38) have published results indicating that SR-BI facilitates the uptake of CE and FC from HDL into the liver but a quantitative comparison of the intrinsic rates of uptake of different classes of lipid via SR-BI is not available. To address the question of preferential HDL lipid uptake by SR-BI, we examined the relative uptake of different HDL lipids from rHDL in COS-7 cells transiently transfected with either SR-BI or the empty vector. The rHDL used contained apoA-I as the sole apolipoprotein and this was complexed with POPC plus different radioactive lipid tracers of interest. Thus, each preparation of rHDL had essentially the same structure and bound to SR-BI to the same extent permitting direct comparison of the rate constants for uptake of each lipid examined. Fig. 3 compares the uptake of CE and FC in control and SR-BI-expressing cells; it is apparent that FC is a better substrate for SR-BI than CE. Examination of the relative selective uptake of various lipid classes (Table I) demonstrates that SR-BI can facilitate selective uptake of all the major HDL lipid classes. In general, neutral lipids have higher rates of selective uptake compared with PL. Our findings here are in general agreement with Greene et al. (39) who showed that SR-BI can mediate the selective uptake of TG. It is interesting to note that SR-BI mediates uptake of two plant sterols, sitostanol and sitosterol, although the rate constants are about half the value for FC. For all classes of lipids examined, FC has the highest selective uptake rate constant.
As seen from Table I, the rate constant for SR-BI-mediated uptake of CE is at least 5-fold greater than the rate constants for the PL examined. Our data show that within the individual classes of PL, the uptake of SM was reduced compared with the uptake of PC, PI, or PS. It is important to note that all rHDL particles used in the experiments were essentially identical in composition and size so the rates of PL uptake observed can be compared directly. Interestingly, SR-BI does not show preference for anionic PS and PI over zwitterionic PC in light of the fact that anionic PL are a better ligand for the receptor (Fig. 1). We observed previously that PS uptake in Y1-BS1 adrenal cells was significantly higher than that of other classes of PL (16). The reason for this discrepancy between COS cells and Y1-BS1 cells is not entirely clear, but it is possible that Y1-BS1 cells express a PS translocase that serves to draw more PS from the rHDL by removing PS from the outer leaflet of the plasma membrane.
Lipid Uptake from Apolipoprotein-free Donor Particles- Fig.  1 shows that SUV containing PS bind very well to SR-BI whereas SUV composed exclusively of POPC do not bind to the same extent. Little is known about the consequences of this binding for SR-BI-mediated lipid uptake from apolipoproteinfree PL vesicles. To understand this process better, we generated SUV composed of 98:1:1 mol ratios of POPC:FC:CO (POPC-SUV) and containing [ 14 C]SM and [ 3 H]COE. The uptake of labeled lipid molecules from these dual-labeled SUV was compared with the uptake of the same lipids from apoA-I rHDL. Fig. 4 panel A, shows that expressing SR-BI in COS-7 cells enhanced uptake of both SM and COE from rHDL compared with control cells, with the uptake of COE being significantly higher than that of SM. After subtracting the values for control cells, the SR-BI contribution to COE and SM uptake was roughly 7-fold higher for the former lipid. In contrast, the uptake of COE and SM from SUV (Fig. 4, panel B) showed no preferential uptake of COE relative to SM although the uptake of both lipids was enhanced by SR-BI. The temperature dependence of SUV lipid uptake showed that there was a nearly 10-fold increase in SUV lipid uptake when the experiments were conducted at 37°C versus 4°C (an increase from 0.5 to 4.9% of lipid label uptake/mg of cell protein/h). The amount of SR-BI at the cell surface was not temperature-dependent since the binding of 125 I-HDL was similar at 4 and 37°C (data not shown). Therefore, the cell-associated lipids measured at 37°C in the SUV uptake experiments were due to intracellular accumulation of lipids and not simply due to SUV bound to receptor on the cell surface. It should be noted that the masses of CE taken up from the rHDL (Fig. 4, panel A) and the SUV (Fig. 4, panel B) are similar under the experimental conditions used. However, comparison of lipid mass uptake between an SUV and a discoidal rHDL does not readily provide information about the mechanism of lipid uptake because the structure, size, and composition of the two lipid donors are not identical.
Panel C in Fig. 4 compares the isotopic ratio of lipid labels   present in the extracellular incubation medium (i.e. associated with the lipid donor particle) and the corresponding isotopic ratio associated with the cell lysate after washing. The data show that for a SUV donor particle the COE/SM isotopic ratio for the medium and the cells was the same, indicative of a stoichiometric uptake of lipid labels. In contrast, when the donor particle was rHDL the COE/SM isotopic ratio was dramatically increased in the cell lysate, consistent with selective transfer of CE relative to SM. Addition of apoA-I to POPC-SUV (500:1 POPC to apoA-I final mole ratio) caused selective uptake of [ 3 H]COE over [ 14 C]SM (1.8-fold more CE than SM) (Table II). To confirm the observation that the presence of apoA-I on SUV allowed selective lipid uptake, [ 3 H]COE-labeled POPC-SUV were complexed with 125 I-apoA-I and incubated with COS-7 cells expressing SR-BI. Selective CE uptake was observed from SUV that contained apoA-I although the amount of CE uptake due to selective uptake was 60 -65% (data not shown) compared with 95% for apoA-I-containing rHDL particles (see Fig.  2).
When SUV composed of equimolar POPC and PS (PC/PS-SUV) were used and uptake of [ 3 H]COE and [ 14 C]SM was examined, there was a greater uptake of lipid labels as compared with uptake from SUV composed exclusively of POPC (Table II). This effect is probably due to the fact that PC/PS-SUV bind better than POPC-SUV to SR-BI (Fig. 1). Interestingly, the fractional uptake of both labels in the SUV ([ 3 H]COE and [ 14 C]SM was the same (Table II). Note also that stoichiometric uptake was apparent when the lipid label pairs were FC and SM (Table II). This result is identical to that seen with POPC-SUV indicating that the mechanism of lipid uptake is similar regardless of the overall SUV charge. Substituting PS with another anionic phospholipid, PI, gave similar results (data not shown). As observed with POPC-SUV, preincubation of apoA-I with PC/PS-SUV prior to presentation to the SR-BI-expressing cells enhanced the preferential uptake of [ 3 H]COE over [ 14 C]SM (1.4-fold).
Since SUV do not contain a hydrophobic lipid core, we wanted to test whether a lipid emulsion containing a core of triolein could give rise to a similar stoichiometric uptake of lipids. Lipid emulsion particles (POPC:triolein, 1:2, w/w) were labeled with either [ 3 H]COE or [ 3 H]FC and [ 14 C]SM and presented to SR-BI-expressing cells. The fractional uptake of both lipid labels via SR-BI was similar (Table III), as was observed when SUV were the donor particles. It is apparent from the data in Tables II and III, that for SUV and emulsion particles that do not contain any apolipoprotein, the lipid uptake via SR-BI is stoichiometric. Addition of unlabeled rHDL (which binds with higher affinity than SUV to SR-BI (Fig. 1)) completely abolished the uptake of lipids from labeled SUV (Fig. 5), which indicates that uptake of lipids from SUV involves direct interaction of the donor SUV with SR-BI. The inhibition by rHDL in SR-BI-containing cells reduced the level of SUV lipid uptake to that seen with control cells in the absence of competitor (data not shown).
In light of the above findings, we examined if the SR-BImediated uptake of lipids from PL vesicles was due to uptake of only the lipid components or to uptake of the intact particle including the inner aqueous compartment. To achieve this, [ 3 H]inulin was encapsulated in the inner aqueous compartment of [ 14 C]SM-labeled LUV. Inulin is a carbohydrate that has been shown to be a good marker for whole particle uptake of vesicles since it remains trapped in a nondegraded state inside cells if the vesicle enters the cell via whole particle endocytosis (40). Since uptake of LUV lipids by fusion of the LUV PL bilayer with the cell plasma membrane involves disruption of the vesicle structure, release of inulin from the aqueous compartment of the vesicle into the extracellular medium is possible. Control experiments showed that SR-BI-  (Table  III). Thus, the lipid constituents of the LUV are taken into the cells but the contents of the trapped aqueous phase are excluded. This indicates that intact vesicles are not endocytosed. Taken together, these results support a model in which SR-BI promotes fusion of apolipoprotein-free PL vesicles (and emulsion particles) with the cell plasma membrane.

DISCUSSION
A major finding in this report is that the nature of the lipid donor/SR-BI interaction is critical in determining the mechanism of lipid uptake via SR-BI. When the lipid donor does not contain an apolipoprotein and the binding to SR-BI has to be mediated by the lipid itself, then donor particle fusion occurs leading to stoichiometric uptake of the various classes of lipids in the donor. One explanation for this effect could be that the extracellular domain of SR-BI contains a fusogenic structural motif that destabilizes the lipid donor particle and promotes fusion with the cell plasma membrane. When an amphipathic ␣-helix is present on the surface of the lipid donor and mediates the binding to SR-BI, the fusion event does not occur. Elabo-ration of the exact reasons for this difference will require additional studies. However, it is possible that either apolipoproteins bind to a different, nonfusogenic site on SR-BI or apolipoprotein binding to the fusogenic site on SR-BI prevents its direct interaction with lipids in the donor particle and the subsequent fusion event. Alternatively, the presence of apolipoprotein alters the lipid packing in the lipid donor particle and inhibits fusion while allowing selective uptake.
The interaction between apolipoprotein and SR-BI is crucial in binding the donor lipid particle to the receptor and initiating selective lipid uptake. Three apolipoprotein-dependent variables, in part, influence the structure and function of the lipid donor particle with regard to selective uptake of a lipid such as CE. These variables are: 1) the binding of the lipid donor particle to SR-BI; 2) the lipid packing in the donor particle; and 3) the CE content of the donor particle. It is difficult to examine independently the influence of these variables on the apolipoprotein/SR-BI interaction. We observed that at equal apolipoprotein concentration, rHDL particles containing apoA-II bind significantly less than rHDL containing either apoA-I or apoE-3 (Fig. 2). Thus, it is not surprising that the absolute amount of CE uptake from the apoA-II particles is less than from either the apoA-I or apoE-3 rHDL because the binding of the lipid donor is a necessary step for CE diffusion into the cell. On the other hand, although apoA-I rHDL and apoE-3 rHDL bind to the same extent, the amount of CE selective uptake from apoA-I rHDL is ϳ2 times higher than that from apoE-3 rHDL despite the fact that the CE/PL ratio is ϳ1.3 times lower in the apoA-I rHDL. Since the same concentration of lipid donor particles is bound to SR-BI, it follows that apoA-I mediates CE selective uptake the most effectively. The nature of the apolipoprotein can significantly affect the rate of CE transfer from the bound HDL particle to the plasma membrane. This result shows that the detailed nature of the apolipoprotein/ SR-BI interaction directly influences CE selective uptake, with factors such as the position of the apolipoprotein on SR-BI probably being important.
The rate constant for lipid uptake from the bound lipid donor particle into the plasma membrane is also important for explaining the different rates of selective uptake of different classes of lipid. The rate constants in Table I reflect the relative ease of transfer of each lipid from the bound rHDL particle into the plasma membrane via SR-BI. Since the order of rate constants is FC Ͼ CE Ͼ TG Ͼ PL, it is clear that hydrophobic neutral lipids (e.g. FC, CE, and TG) can transfer more readily than more polar PL molecules which have bulky charged head groups. Detailed knowledge of the structure of the extracellular domain of SR-BI is required to understand this effect at the molecular level. Previously, we have postulated that facilitation of lipid uptake occurs because SR-BI creates a nonpolar channel between a reversibly bound HDL particle and the plasma membrane (16). According to this model, neutral FC, CE, and TG molecules can desorb from the surface of the bound HDL particle and be accommodated in the channel better than more polar PL molecules. The selectivity arises from the relative abilities of the lipid molecules to partition out of the HDL particle and properties of the channel such as its size, and the distribution of polar and nonpolar sites within it. If an HDLplasma membrane hemifused state is involved in lipid selective uptake (17,41), then the discrimination between lipid classes arises from differences in lateral mixing rates. In both models, the direction of net transfer would be dictated by the relevant concentration gradient between the bound HDL particle and the plasma membrane. However, it is not obvious in the hemifusion model why the lateral diffusion coefficients of FC, PC, and SM would be different enough to account for the relative rate constants listed in Table I. The relative rate constants for selective uptake of the different classes of HDL lipids can be used to predict the relative clearance rates of these lipids via SR-BI. The rate constants in Table I relate to selective uptake from discoidal rHDL particles rather than from the spherical HDL particles found in plasma. However, the values probably apply to transfer from the surfaces of both types of HDL particle. The reason for this is that the molecular mechanism is essentially the same, as reflected in the identical activation energies measured for CE selective uptake from discoidal rHDL and spherical HDL 3 (16). Making this assumption, it is possible to estimate relative selective uptake rates for the lipids in an HDL 3 particle. The lipid composition of a typical human HDL 3 particle is 55, 10, 29, and 7 weight % for PC, TG, CE, and FC, respectively (42). The relative fluxes of lipid mass can be calculated using the relative rate constants in Table I and the relationship, flux ϭ (pool size) ϫ (rate constant). On this basis, the relative mass fluxes of PC, TG, CE, and FC from HDL via SR-BI-mediated selective uptake are 1, 0.6, 2.6, and 1, respectively. Therefore, SR-BImediated selective uptake should be the greatest for cholesterol (FC and CE), accounting for about 70% of the total lipid uptake from human HDL 3 . However, it is apparent that significant catabolism of HDL PL and TG can occur via SR-BI-mediated uptake. Of course, this process requires eventual movement of the lipids to the cell interior (43), and in the case of polarized cells such as hepatocytes HDL cholesterol can be transported across the cell (44).