Binding and cross-linking studies show that scavenger receptor BI interacts with multiple sites in apolipoprotein A-I and identify the class A amphipathic alpha-helix as a recognition motif.

Scavenger receptor, class B, type I (SR-BI) mediates the selective uptake of high density lipoprotein (HDL) cholesteryl ester without the uptake and degradation of the particle. In transfected cells SR-BI recognizes HDL, low density lipoprotein (LDL) and modified LDL, protein-free lipid vesicles containing anionic phospholipids, and recombinant lipoproteins containing apolipoprotein (apo) A-I, apoA-II, apoE, or apoCIII. The molecular basis for the recognition of such diverse ligands by SR-BI is unknown. We have used direct binding analysis and chemical cross-linking to examine the interaction of murine (m) SR-BI with apoA-I, the major protein of HDL. The results show that apoA-I in apoA-I/palmitoyl-oleoylphosphatidylcholine discs, HDL(3), or in a lipid-free state binds to mSR-BI with high affinity (K(d) congruent with 5-8 microgram/ml). ApoA-I in each of these forms was efficiently cross-linked to cell surface mSR-BI, indicating that direct protein-protein contacts are the predominant feature that drives the interaction between HDL and mSR-BI. When complexed with dimyristoylphosphatidylcholine, the N-terminal and C-terminal CNBr fragments of apoA-I each bound to SR-BI in a saturable, high affinity manner, and each cross-linked efficiently to mSR-BI. Thus, mSR-BI recognizes multiple sites in apoA-I. A model class A amphipathic alpha-helix, 37pA, also showed high affinity binding and cross-linking to mSR-BI. These studies identify the amphipathic alpha-helix as a recognition motif for SR-BI and lead to the hypothesis that mSR-BI interacts with HDL via the amphipathic alpha-helical repeat units of apoA-I. This hypothesis explains the interaction of SR-BI with a wide variety of apolipoproteins via a specific secondary structure, the class A amphipathic alpha-helix, that is a common structural motif in the apolipoproteins of HDL, as well as LDL.

High density lipoprotein (HDL) 1 donates cholesteryl esters (CE) to cells via the selective CE uptake pathway, a process in which CE are transferred from the HDL particle to the plasma membrane without the concomitant uptake and degradation of HDL apolipoproteins (1). In rodents, the HDL-CE selective uptake pathway plays a major role in plasma cholesterol metabolism by delivering HDL-CE to the liver in the final steps of reverse cholesterol transport (2)(3)(4)(5). HDL-CE selective uptake also occurs prominently in steroidogenic cells of the adrenal gland and ovary where it provides cholesterol for steroid production and for the accumulation of cytoplasmic CE storage droplets (6 -10). Despite extensive studies of this widespread process, little is known about the biochemical mechanism by which CE molecules are transferred from the HDL particle to the cell.
Recent studies identified a cell surface receptor, scavenger receptor BI (SR-BI), that binds HDL particles and facilitates the selective uptake of HDL-CE in transfected cells (11). Murine (m) and rat SR-BI are strongly expressed in those tissues that exhibit the selective uptake pathway in vivo, namely the liver, adrenal gland, ovary, and testis (11)(12)(13). A similar distribution of SR-BI expression occurs in human tissues (14,15). Immunohistochemical studies show that SR-BI is present on the surface of steroidogenic cells in these tissues of rats and mice, and its expression is regulated by tropic hormones in concert with the regulation of steroid production (12,13,16,17). Immunolocalization at the electron microscopic level in rat ovarian luteal and testicular Leydig cells show that SR-BI is present on microvillar membrane domains that form channels in which HDL particles are sequestered (16,18). These microvillar channels are believed to be the site at which the selective uptake of HDL-CE occurs (19). These data provide strong circumstantial evidence that SR-BI is the cell surface receptor responsible for HDL-CE selective uptake. Direct evidence for SR-BI function is provided by studies in which antibody to the extracellular domain of mSR-BI blocked HDL-CE selective uptake and the delivery of HDL cholesterol to the steroidogenic pathway in cultured murine adrenocortical and ovarian cells (20,21). In addition, inactivation of the SR-BI gene in mice increased plasma HDL cholesterol levels and reduced neutral lipid stores in the adrenal glands (22). Similarly, mice carrying an induced SR-BI mutation that reduced hepatic SR-BI expression levels by 50% showed a similar reduction in hepatic HDL-CE selective uptake (23). Taken together, these observations indicate that SR-BI plays a key role in mediating HDL-CE selective uptake in the liver and in steroidogenic cells.
The molecular basis for the recognition of HDL particles by SR-BI is unknown. In transfected cells mSR-BI binds HDL (11), LDL and modified LDL (24), and protein-free lipid vesicles containing anionic phospholipids (25). In earlier studies with the murine Y1-BS1 adrenocortical cell line, it was shown that recombinant HDL particles containing apoA-I, apoC proteins, or apoE were all active in HDL-CE selective uptake (26). More recently, recombinant bilayer phospholipid discs containing apoA-I, apoA-II, or apoCIII were shown to bind to Chinese hamster ovary cells expressing transfected mSR-BI (27). These results do not provide a clear picture of the HDL-SR-BI interaction and are compatible with models in which HDL recognition by SR-BI is due to interaction with HDL lipids, or with common motifs shared among HDL (and LDL) apolipoproteins, or to a combination of such interactions. In the present study we have explored this issue by using chemical cross-linking to study the interaction between mSR-BI and apoA-I as it occurs in recombinant phospholipid discs, in spherical HDL 3 particles, and in lipid-free apoA-I. The results show that each of these species binds and cross-links efficiently to mSR-BI. The N-and C-terminal cyanogen bromide fragments of apoA-I each bind with high affinity and cross-link to SR-BI, indicating that apoA-I:SR-BI interaction is not due to a unique binding site in apoA-I. In addition, a model class A amphiphathic helix binds and cross-links to mSR-BI. These results indicate that mSR-BI makes direct protein-protein contacts with apoA-I and point to the amphipathic helix as the primary recognition motif for the HDL-mSR-BI interaction.
Preparation of Discoidal Recombinant HDL Complexes-ApoA-I⅐POPC⅐free cholesterol (FC) discoidal complexes (referred to as AI⅐POPC discs) were prepared and characterized as described (30) using an initial molar ratio of 1/100/5, respectively. Measured molar composition after preparation was 1/99/2.5, apoA-I/POPC/FC. Analysis by non-denaturing 8 -25% polyacrylamide gel electrophoresis showed a single band comigrating with the catalase standard with an apparent Stokes' diameter of 10.4 nm (data not shown).
ApoA-I fragments were obtained by the method of Morrison et al. (31). CNBr digestion of apoA-I produces four fragments with molecular weights of 9880 (N-terminal), 3190, 4250, and 10,700 (C-terminal). In brief, apoA-I (30 mg) was digested with 3 ml of (300 mg/ml) CNBr in 70% trifluoroacetic acid for 24 h at room temperature in the dark. The reaction mixture was diluted with water, lyophylized, and the residue subjected to reversed-phase HPLC (Vydac, 22 mm (inner diameter) ϫ 25 cm, 10-m resin) using acetonitrile, water, and 0.1% trifluoroacetic acid as solvent with a gradient of 0 -60% acetonitrile for 90 min at a flow rate of 5 ml/min. Fractions at retention times 55.3 min (fragment 1), 57 min (fragment 2), 62 min (fragment 3), 65 min (fragment 4), and 68 min (fragment 5) were subjected to analytical HPLC and mass spectral analysis using a PE-Sciox APT-III triple-quadrupole ion spray mass spectrometer (at the University of Alabama at Birmingham Mass Spectroscopy Core Facility analysis core facility). Fragments 1, 2, and 3 had respective masses of 3190, 4250, and 7427, corresponding, respectively, to the two middle CNBr fragments and the middle fragments covalently bound. Fragments 4 and 5 corresponded to the N-terminal and C-terminal fragments, with masses of 9880 and 10,700, respectively. The analytical HPLC profiles of these fragments correlated well with the published fragmentation pattern (31). N-terminal and C-terminal fragments were dialyzed extensively in 0.1 M guanidine HCl. DMPC⅐CNBr fragment complexes were made at a DMPC:protein ratio of 2:1 (w/w). In most cases 4 mg of DMPC dissolved in CHCl 3 was dried onto a borosilicate tube and subsequently vacuum-dried for 30 min. It was then suspended in 150 mM NaCl, 0.25 mM EDTA, pH 7.4, by vortexing. After equilibrating the DMPC multilamellar vesicles at 24°C, 2 mg of the peptide, also equilibrated at 24°C, was added to the DMPC preparation, and incubated for 30 min at 24°C. A subsequent water bath sonication of 10 min followed this incubation. The complexes were then dialyzed extensively against 150 mM NaCl, 0.25 mM EDTA, pH 7.4. Formation of complexes was assessed by non-denaturing polyacrylamide gel electrophoresis and negative stain electron microscopy.
Discoidal complexes containing DMPC and the model class A amphipathic helix, 37pA (32) were prepared by incubation of the peptide with DMPC multilamellar vesicles at the transition temperature as described (33). Peptide (0.5 mg) was dissolved in 0.5 ml of 0.15 M NaCl, 1 mM EDTA, pH 8.3, 0.01% NaN 3 overnight at 4°C followed by bath sonication for 3 h. DMPC (20 mg) was dissolved in CHCl 3 /methanol (2/1), dried to a film on glass tube, and held under vacuum overnight. The DMPC was then suspended in 2 ml of the above saline solution by vortexing, bath sonicated for 30 min, and equilibrated at 24°C. DMPC (1.25 mg) was added to the 37pA peptide also equilibrated at 24°C, and incubated for 48 h at 24°C to allow formation of the discoidal complexes.
Expression of mSR-BI-COS-7 cells were maintained in DMEM supplemented with 10% FBS, 1 mM sodium pyruvate, and antibiotics in T-75 flasks and were subcultured once a week using a 1:20 split ratio. Cells were transiently transfected with an expression vector for mSR-BI (35) or with vector alone. Cells (1.5 ϫ 10 6 ) were seeded in 100-mm plates in DMEM supplemented with 10% FBS and incubated for 18 h at 37°C in a humidified 95% air, 5% CO 2 incubator. A mixture of 10 g of the desired plasmid, diluted in serum-free DMEM, and 30 l of Fugene-6 (Roche Molecular Biochemicals) prepared in a sterile polystyrene tube (Falcon 2058) was added dropwise to the plated cells. After incubation (18 -24 h, 37°C), transfected cells were trypsinized, pooled, suspended in growth medium, and replated in multi-well plates as needed for experiments. Experiments were carried out 24 h after replating.
Cross-linking-Except as noted in figure legends, the following protocol was used for cross-linking. mSR-BI-or vector-transfected COS-7 cells in six-well plates were placed on ice and washed three times with 3 ml of MEM/HEPES, pH 7.4, 1% BSA. Ligand was added in 1 ml of MEM, 1% BSA, and the cells were incubated for 60 min at 37°C under an atmosphere of 95% air, 5% CO 2 . Cells were cooled on ice, washed two times with 2 ml of MEM/HEPES, pH 7.4, 1% BSA, and three times with 3 ml of PBS. Cross-linker was dissolved immediately before use at 10 mg/ml in dimethyl sulfoxide, diluted to 2.5 ϫ 10 -4 M with PBS, and 1 ml was added per well. Cells were incubated for 45 min at room temperature after which cells were cooled on ice, and the incubation medium was removed. Cells were lysed in 400 l of 0.02 M sodium phosphate, pH 7.4, 1 mM MgCl 2 , 0.5% Nonidet P-40, 10 g/ml aprotinin, 1 g/ml leupeptin, 1 g/ml pepstatin, 0.2 mM phenymethylsulfonyl fluoride, and 10 mM glycine, and centrifuged for 10 min at 10,000 ϫ g at 4°C. An aliquot of the supernatant was removed for ␥ counting, and the remain-der was used for immunoprecipitation or frozen on dry ice and stored at Ϫ80°C for later analysis.
Immunoprecipitation-Antiserum (7.5 l) raised against the C-terminal cytoplasmic tail of mSR-BI (17) or control antiserum and lysis buffer to bring the volume to 150 l were added to cell supernatant (100 -150 l). To this was added 150 l of 2ϫ immunoprecipitation buffer (200 mM NaCl, 100 mM LiCl, 10 mM EDTA, 100 mM Tris, pH 7.4, 1% Triton X-100, 1% sodium deoxycholate, 0.1% sodium dodecyl sulfate, containing protease inhibitors as noted above). The sample was incubated at 4°C for 3-18 h, after which protein A-agarose was added, and the incubation was continued for 1 h. The tube was rotated during the final hour to keep the protein A beads in suspension. The tube was centrifuged for 2 min at 10,000 ϫ g, and a measured aliquot of the supernatant was removed for ␥ counting. The pellet was washed twice with 1 ml of 1ϫ immunoprecipitation buffer, and the beads were transferred to a tube for ␥ counting. In some experiments, the beads were eluted with 150 l of 60 mM Tris, pH 6.8, 6% glycerol, 2.4% SDS by boiling for 5 min. After centrifugation, a measured aliquot was removed for ␥ counting and another for analysis by non-reducing 7.5% SDSpolyacrylamide gel electrophoresis with the buffer system of Laemmli (36) using Kaleidoscope (Bio-Rad) prestained molecular weight standards. Gels were dried, and radioactivity was imaged with a Molecular Dynamics PhosphorImager.
Ligand Binding-Transfected cells, plated in 12-well plates, were placed on ice and washed two times with 1 ml of MEM/HEPES, pH 7.4, 1% BSA. Ligand in 0.25 ml of chilled MEM/HEPES, pH 7.4, 1% BSA was incubated with the cells for 2 h at 4°C. An aliquot of medium was removed, centrifuged at 10,000 ϫ g for 5 min to pellet cell debris, and subjected to ␥ counting to determine the free ligand concentration. Cells were washed three times with 2 ml of MEM/HEPES, pH 7.4, 1% BSA and three times with 2 ml of PBS. Cells were solubilized in 1 ml of 0.1 N NaOH, and samples were removed for ␥ counting and for protein determination (37). In some experiments, washed cells were solubilized in 0.5 ml of 0.5% Nonidet P-40 in PBS, and transferred to a tube for ␥ counting and protein measurement. SR-BI-specific binding of apoA-I was determined by subtracting values from vector-transfected cells from the SR-BI-expressing cells to generate an SR-BI-vector curve. Binding parameters for B max and K d were obtained by nonlinear regression (GraphPad Prism) using a one-or two-site binding isotherm as indicated (20,38).

Interaction of ApoA-I⅐POPC Discoidal Complexes with mSR-
BI-In order to examine the interaction between apoA-I⅐POPC discoidal complexes and mSR-BI, COS-7 cells were transiently transfected with vector alone or with an mSR-BI expression vector. Fig. 1 shows that mSR-BI-expressing cells bound apoA-I⅐POPC complexes with a concentration dependence indicative of both high and low affinity components, whereas vectortransfected cells showed only a low affinity binding component. Resolution of the mSR-BI binding isotherm by nonlinear regression analysis showed a saturable high affinity component with an apparent K d ϭ 6 g/ml (apoA-I protein), a value similar to that reported for the binding of HDL 3 to mSR-BI on Y1-BS1 adrenocortical cells (20,38) or on transiently transfected COS-7 cells (35). Interestingly, the slope of the low affinity component for the mSR-BI-expressing cells (1.18 Ϯ 0.14 ng bound/mg of cell protein per g/ml apoA-I) at high ligand concentrations (50 -200 g/ml) was much greater than the slope of the low affinity component in vector-transfected cells (0.23 Ϯ 0.07 ng bound/mg of cell protein per g/ml apoA-I). This result suggests that the low affinity binding component seen at high ligand concentrations (Ͼ100 g/ml) in the mSR-BI-expressing cells is due to mSR-BI expression and is different from the low affinity binding seen with the vector-expressing cells. The basis for this low affinity binding component is unknown but might reflect SR-BI-dependent changes in plasma membrane lipid domains (39,40).
To test for direct interactions between mSR-BI and apoA-I, mSR-BI-expressing or vector-transfected cells were incubated with 125 I-apoA-I⅐POPC discs for 90 min at 37°C, washed to remove unbound ligand, and subsequently treated with one of two homobifunctional N-hydroxysuccinimide esters, both of which have a 12-Å spacer arm. DSP or Lomant's reagent (41) is very lipid-soluble, whereas DTSSP is its water-soluble analog. To test for cross-linking between apoA-I and mSR-BI, cell extract was immunoprecipitated with antibody raised against the C-terminal cytoplasmic tail of mSR-BI, and the immunoprecipitate was assayed for the presence of 125 I-apoA-I. The data in Table I (part A) show results from a typical experiment. The columns labeled "cpm/well" show the 125 I-apoA-I remaining on the cells after incubation with apoA-I⅐POPC discs, washing, and reaction with or without a cross-linker. As expected from the previous binding analysis (Fig. 1), mSR-BI-expressing cells bound 10 -20-fold more apoA-I as compared with vector-expressing cells. When cell extracts from mSR-BI-expressing cells that had not been treated with cross-linker were immunoprecipitated with either anti-mSR-BI or control antibody, little or no 125 I-apoA-I was recovered in the pellet. In contrast, when mSR-BI-expressing cells were treated with either DSP or DTSSP, 125 I-apoA-I was recovered in the immunoprecipitate with anti-mSR-BI but not with control antibody. The percentage of the bound 125 I-apoA-I recovered in the immunoprecipitate (Table I, part A, column designated "Percentage crosslinked") was 27% for DTSSP and 37% for DSP, indicating that cross-linking of bound apoA-I to mSR-BI is quite efficient. The specificity of the cross-linking reaction for monitoring apoA-I-mSR-BI interaction is evident from the result that recovery of FIG. 1. Binding of apoA-I⅐POPC discs to mSR-BI-expressing COS-7 cells. SR-BI-expressing and vector-tranfected COS-7 cells in 12-well plates were incubated with 125 I-apoA-I⅐POPC discs for 2 h at 4°C as described under "Experimental Procedures." Bound apoA-I is plotted versus the concentration of free apoA-I at the conclusion of the assay. The curve for SR-BI-expressing cells is a nonlinear regression fit to a two-site model. The curve for vector-transfected cells is a linear regression fit. 125 I-apoA-I in the immunoprecipitate required both reaction with cross-linker and antibody to mSR-BI.
To ensure that cross-linking of apoA-I⅐POPC discs did not in some way generate species that could be recognized by the antipeptide mSR-BI antibody, apoA-I⅐POPC discs were crosslinked in solution in the absence of cells, and then reacted with anti-mSR-BI. As shown in Table I, part B, only trace amounts of 125 I-apoA-I were recovered in the immunoprecipitate, and the presence or absence of cross-linker had no effect. Thus, 125 I-apoA-I recovered in the anti-mSR-BI immunoprecipitate with extracts from mSR-BI-expressing cells reflects cross-linking of the ligand to cell surface mSR-BI. In other experiments we compared the efficiency of cross-linking by DSP when mSR-BI-expressing cells were incubated for 60 min with 125 I-apoA-I⅐POPC discs at either 0°C or 37°C; in both cases, when cross-linking was carried out for 45 min at either 0°C or 23°C, only minor differences in cross-linking were seen with the efficiencies ranging from 37% to 45% (data not shown).
The cross-linked complex formed between apoA-I and mSR-BI was examined by 7.5% SDS-PAGE. As shown in Fig. 2 , lanes 1, 3, and 4). Thus, the band migrating at approximately 225 kDa represents apoA-I crosslinked to cell surface mSR-BI.
Interaction of Lipid-free ApoA-I and HDL 3 with mSR-BI-In order to test whether the lipidation state of apoA-I was crucial for interaction with mSR-BI, the binding and cross-linking of apoA-I⅐POPC discs was compared with that of lipid-free apoA-I. Fig. 3 shows the binding of apoA-I⅐POPC discs (panel A) to SR-BI-expressing and vector-transfected cells in comparison to the binding of lipid-free apoA-I (panel B) to the same batch of transiently transfected COS-7 cells. As is apparent, lipid-free apoA-I binds in a saturable, high affinity manner to SR-BIexpressing cells. In comparison to the binding of apoA-I⅐POPC discs, one difference with lipid-free apoA-I is much greater binding to vector-transfected cells, a result most likely reflecting nonspecific lipid-free apoA-I interactions with cell surface phospholipid. SR-BI-specific binding of apoA-I⅐POPC discs and lipid-free apoA-I was obtained by subtracting the apoA-I bound to vector-transfected cells from that bound to SR-BI-expressing cells. The resulting binding isotherms (Fig. 3, dashed lines) showed that lipid-free apoA-I (panel B) bound with a K d (ϳ4 g/ml) similar to that of apoA-I⅐POPC discs (ϳ6 g/ml) (panel A).
Cross-linking of lipid-free apoA-I to mSR-BI was tested in comparison to the cross-linking of apoA-I⅐POPC discs and HDL 3 . COS-7 cells expressing mSR-BI or transfected with vector were incubated with ligands at 20 g/ml followed by thorough washing, cross-linking with DSP, and immunoprecipitation of the cell lysate with anti-mSR-BI. As anticipated, cellassociated ligand was dramatically greater in the mSR-BIexpressing cells as compared with vector-transfected cells with apoA-I⅐POPC discs as the ligand (Table II). This was also the case with HDL 3 , which showed 25 times more association to mSR-BI-expressing cells as compared with vector-transfected cells. In contrast, and consistent with the binding analysis of Fig. 3B, cell association of lipid-free apoA-I to mSR-BI-expressing cells was only about 2-fold greater than to vector-transfected cells. Because of differences in the background binding of lipid-free apoA-I to vector-transfected cells, the cross-linking efficiency (percentage of cell-associated cpm precipitated by anti-SR-BI) was corrected for the binding to vector-transfected cells (see Table II, legend). After this correction, the percentage of the SR-BI-specific apoA-I that was cross-linked to SR-BI by DSP was 36% for apoA-I⅐POPC discs, 24% for HDL 3 , and 25% for lipid-free apoA-I. Thus, 25% of the lipid free-apoA-I bound to mSR-BI was cross-linked to the receptor. Note that less than TABLE I Cross-linking of apoA-I to mSR-BI In part A, quadruplicate wells of a 12-well plate containing cells expressing mSR-BI or transfected with vector were incubated at 37°C for 90 min in 0.25 ml of MEM/1% BSA containing 125 I-apoAI ⅐ POPC discs (660 cpm/ng) at 10 g/ml. Cells were washed and incubated in 1 ml of PBS without or with cross-linker (2.5 ϫ 10 Ϫ4 M) for 45 min at room temperature as described under "Experimental Procedures." Cells were placed on ice, incubation medium was removed, duplicate wells were lysed and pooled in a total volume of 400 l of lysis buffer, and the lysate was cleared by centrifugation at 12,000 ϫ g for 10 min at 4°C. An aliquot of each lysate was counted, and an aliquot (150 l) of the mSR-BI lysate was reacted overnight at 4°C with either anti-mSR-BI or with control antibody. After another 1-h incubation with protein A-agarose and centrifugation, the supernatant and the protein A-agarose pellet were counted. The percent cross-linked represents the percentage of the input counts that were recovered in the protein A-agarose pellet using anti-mSR-BI. In part B, 125 I-apoAI ⅐ POPC discs (660 cpm/ng) at 10 g/ml were incubated with or without cross-linker (2.5 ϫ 10 Ϫ4 M) in PBS for 45 min at room temperature. Triplicate samples approximating the amount of 125 I-apoAI ⅐ POPC discs used for immunoprecipitation in part A were reacted with anti-mSR-BI as in part A. Values are the mean Ϯ S.D. (n ϭ 3). 0.5% of the lipid-free apoA-I associated with vector-transfected cells was recovered in the anti-mSR-BI immunoprecipitate (data not shown). These data indicate that apoA-I binds directly to mSR-BI irrespective of its lipidation state or whether it is present on discoidal or spherical HDL particles. However, the differences in cross-linking efficiencies with apoA-I⅐POPC discs, lipid-free apoA-I, and apoA-I in HDL 3 suggest that the conformation of apoA-I may play a role in defining the exact nature of the interaction between the receptor and apoA-I.

Interaction of N-and C-terminal Domains of ApoA-I with mSR-BI-
To determine whether apoA-I has a unique site for interaction with SR-BI, the N-and C-terminal CNBr fragments were prepared, complexed with DMPC, and tested for binding and cross-linking to mSR-BI. As shown in Fig. 4, both the N-terminal (residues 1-86) (panel A) and the C-terminal (residues 148 -243) (panel B) fragment bound in a high affinity saturable manner to SR-BI-expressing cells. Both fragments also bound to a lesser extent to vector-transfected cells, and this binding also appeared to be saturable. SR-BI-specific binding was obtained by subtracting the values obtained with vector-transfected cells from the values obtained with SR-BI-expressing cells. These data were analyzed by non-linear regression using a one-site binding model to obtain SR-BIspecific binding isotherms (SR-BI-vector, dashed lines, panels  A and B). This analysis showed that the K d for binding of the C-terminal fragment (11 g/ml) was severalfold lower than for the N-terminal fragment (39 g/ml) and the B max values for

TABLE II
Cross-linking of ligands to SR-BI Duplicate wells of a six-well plate containing cells expressing mSR-BI or transfected with vector were incubated at 37°C for 60 min in 1.0 ml of MEM, 1% BSA containing 125 I-ligands at 20 g/ml. Cells were washed and incubated in 1 ml of PBS with DSP (2.5 ϫ 10 Ϫ4 M) for 45 min at room temperature as described under "Experimental Procedures." Lysates were immunoprecipitated as described in Table I. Immunoprecipitates with control antibody were all at background radioactivity levels (data not shown). The percentage of apoA-I cross-linked to mSR-BI represents the percentage of the mSR-BI-specific counts that were recovered in the protein A-agarose pellet using anti-mSR-BI. The mSR-BI-specific counts were determined by subtracting the counts bound to vector-transfected cells from counts bound to mSR-BI-expressing cells. both fragments were similar (N-terminal, 440 ng/mg of cell protein; C-terminal, 410 ng/mg of cell protein). When this analysis was repeated with a separate batch of transiently transfected cells, the K d for the C-terminal fragment (5 g/ml) was approximately 3-fold lower than for the N-terminal fragment (17 g/ml) and the B max values were similar (N-terminal, 220 ng/mg of cell protein; C-terminal, 166 ng/mg of cell protein). Thus, these data indicate that multiple domains of apoA-I can interact with mSR-BI, with the C-terminal domain having a somewhat higher affinity. The internal CNBr fragments of apoA-I are too small to form stable complexes with DMPC, and, therefore, were not analyzed.
To test for direct interactions between SR-BI and the Nterminal and C-terminal apoA-I fragments, 20 g/ml apoA-I fragment⅐DMPC complexes or apoA-I⅐POPC complexes were bound to vector-transfected and SR-BI-expressing cells and cross-linked with DSP. Measurement of SR-BI-specific cellassociated ligand (Table III, ng/well) showed that the C-terminal apoA-I fragment was bound to the same extent as intact apoAI, whereas considerably less N-terminal fragment was bound. This result is consistent with the binding analysis in Fig. 4 showing a greater affinity of the C-terminal fragment for mSR-BI. After correcting for the amount of each ligand bound to vector-transfected cells, the percentage of the SR-BI-specific ligand that was cross-linked to SR-BI by DSP was 32% for the C-terminal fragment, 25% for the N-terminal fragment, and 38% for intact apoA-I. Thus, both the N-and C-terminal fragments cross-link efficiently to SR-BI.
Interaction of a Model Amphipathic ␣-Helix with mSR-BI-ApoA-I contains 10 tandem repeats of 11 or 22 amino acid units, which have the properties of amphipathic ␣-helices (42). To determine whether mSR-BI recognizes an amphipathic ␣-helical motif, the class A ␣-helix, 37pA, was complexed with DMPC and tested for binding to mSR-BI-expressing COS-7 cells. This model ␣-helix is a dimer of 18A linked by a single proline residue and shows no amino acid sequence relatedness to apoA-I (32). As shown in Fig. 5, 37pA⅐DMPC discs bind with high affinity to mSR-BI, but also show marked low affinity association with vector-transfected cells. Subtraction of the 37pA values bound to vector-transfected cells from the values for SR-BI-expressing cells yielded a saturable binding curve (Fig. 5, SR-BI-vector) with K d ϭ 0.4 g/ml. When tested for cross-linking to mSR-BI, 37pA⅐DMPC cross-linked to mSR-BI with an efficiency similar to that of lipid-free apoA-I (Table II). After correction for background binding to vector-transfected cells, the efficiency of cross-linking of 37pA⅐DMPC to mSR-BI (28.4%) was similar to that of either HDL 3 (24.1%) or lipid-free apoA-I (24.8%) but somewhat less than that of apoA-I⅐POPC discs (36.7%). These results indicate that mSR-BI can recognize and interact directly with a class A amphipathic ␣-helix.

DISCUSSION
The results of the present study show that apoA-I in apoA-I⅐POPC discs, HDL 3 , or in a lipid-free state binds to mSR-BI with high affinity. The similarity of the binding affinities (K d Х 5-8 g/ml) of apoA-I in HDL 3 (20,35,38) or apoA-I⅐POPC discs or in the lipid-free state (Fig. 3) suggests that direct proteinprotein contacts are the predominant feature that drives the interaction between HDL and mSR-BI. Additionally, the similar binding affinities of these different forms of apoA-I suggest that the conformational changes that accompany the lipidation of apoA-I (43) are not essential in defining the interaction with mSR-BI. The corollary conclusion is that the mSR-BI recognition motifs within apoA-I must be surface exposed to an adequate degree for receptor recognition in lipid-free apoA-I as well as in lipidated apoA-I on bilayer discs and spherical HDL 3 particles.
The conclusion that direct protein-protein interaction is the predominant feature that defines the HDL-mSR-BI interaction is supported independently by the results of cross-linking experiments. These results showed that apoA-I in apoA-I⅐POPC discs cross-linked efficiently (35-45% of bound apoA-I) to mSR-BI, indicating that the interaction involves contacts that bring reactive lysine residues on apoA-I and mSR-BI within the 12-Å   Table I. The percentage of apoA-I cross-linked to mSR-BI represents the percentage of the mSR-BI-specific counts that were recovered in the protein A-agarose pellet using anti-mSR-BI. The mSR-BI-specific counts were determined by subtracting the counts bound to vector-transfected cells from counts bound to mSR-BI-expressing cells. Anti-mSR-BI immunoprecipitates of lysates from vector-transfected cells were all at background levels of radioactivity (data not shown). spacer length of the cross-linker. The cross-linking efficiencies of apoA-I on HDL 3 and lipid-free apoA-I were the same and also high (25%), although somewhat less than that of apoA-I in the bilayer disc. These differences in cross-linking efficiencies may reflect subtle differences in mSR-BI-apoA-I interactions depending on the lipidation state of apoA-I. Nevertheless, the major conclusion is that apoA-I cross-links efficiently to mSR-BI irrespective of its lipidation state.
The finding of similar binding affinities and cross-linking efficiencies among the three forms of apoA-I examined suggests that the recognition motif on apoA-I is not dramatically altered by the lipidation state. The amphipathic ␣-helical repeats of apoA-I are a potential recognition motif for interaction with mSR-BI. The ␣-helix contents of apoA-I in the apoA-I⅐POPC discs studied here and in the lipid-free state are 75% and 46%, respectively (43). To test whether mSR-BI could recognize a generic form of the type of amphipathic ␣-helix (class A) most common in apoA-I (44), 37pA was complexed with DMPC to form bilayer discs, and the interaction of 37pA⅐DMPC with mSR-BI was studied. The results showed that 37pA⅐DMPC bound to mSR-BI with high affinity and cross-linked to mSR-BI with an efficiency (28%) similar to that of lipid-free apoA-I or HDL 3 . This result identifies the class A amphipathic ␣-helix as one motif recognized by mSR-BI and leads to the hypothesis that mSR-BI recognizes apoA-I via the multiple class A amphipathic ␣-helical repeats in this apolipoprotein. This hypothesis is supported by the finding that the N-terminal and Cterminal CNBr fragments of apoA-I each bound to mSR-BI, indicating that apoA-I has multiple sites for receptor interaction. Interestingly, the C-terminal CNBr fragment showed a higher affinity for mSR-BI (Fig. 4) and a higher cross-linking efficiency (Table III). This result is consistent with the above hypothesis in that the C-terminal CNBr fragment has more predicted class A amphipathic ␣-helical repeats than the Nterminal fragment (44). Measurements of ␣-helix by attenuated total reflection infrared spectroscopy confirms that the C-terminal CNBr⅐DMPC complex contains more ␣-helical content than the N-terminal CNBr⅐DMPC complex (45). The Cterminal CNBr fragment also contains two segments of predicted class Y amphipathic helix (44), which might contribute to interaction with mSR-BI. Additional studies will be required to test whether the multiple class A amphipathic ␣-helices in apoA-I differ in their interaction with mSR-BI and whether there are significant differences among the types of amphipathic ␣-helix found in apoA-I.
There are two interesting features of the hypothesis that mSR-BI recognizes apoA-I via the amphipathic ␣-helical repeats. First, mSR-BI recognizes recombinant HDL particles containing a variety of different apolipoproteins including apoA-I, apoA-II, apoC-III, or apoE (26,27). These apolipoproteins are related evolutionarily and contain multiple amphipathic ␣-helical repeats. mSR-BI also recognizes native LDL as well as modified LDL (24), the apolipoprotein of which, apoB100, contains abundant amphipathic ␣-helix (46,47). Recent studies indicate that mSR-BI mediates selective cholesteryl ester uptake from LDL particles (48). Thus, recognition of these diverse apolipoproteins by mSR-BI may occur through a common secondary structure motif, the class A amphipathic ␣-helix, and be relatively independent of amino acid sequence variations in the amphipathic ␣-helical segments. Additionally, the class A amphipathic ␣-helix has negatively charged amino acids clustered in the center of the polar face which would be predicted to be the helical face that interacts with SR-BI. It is interesting to speculate that anionic phospholipid vesicles interact with SR-BI (25) through a similar clustering of negative charge.
The second feature of this hypothesis is that apoA-I contains multiple class A amphipathic ␣-helices. Because HDL 3 has multiple copies of apoA-I and each apoA-I has 6 class A ␣-helical repeats, this mode of interaction may serve to rapidly and efficiently dock the HDL particle without the need for a unique receptor-HDL orientation. As previously noted (38), the affinity of mSR-BI for HDL 3 (K d Х 200 -300 nM apoA-I) is relatively weak compared with the affinity of the LDL receptor for LDL (K d Х 2 nM apoB100), consistent with the idea that HDL comes on and off mSR-BI quite rapidly. Such rapid on and off interactions may be ideally suited for rapid cholesteryl ester selective uptake as well as rapid release of the processed HDL particle to permit another cycle of receptor-HDL interaction and lipid transfer.
Although SR-BI binds to many HDL and LDL apolipoproteins, it is not clear whether binding to SR-BI as mediated by different apolipoproteins translates to equivalent cholesteryl ester selective uptake. Comparison of LDL and HDL, for example, showed that mSR-BI-mediated selective cholesteryl ester uptake from LDL was 6 -7-fold less efficient than uptake from HDL (48). In addition, studies with apoA-I-deficient mice suggest a stringent requirement for apoA-I for CE accumulation in steroidogenic cells, whereas apoA-II deficiency has little effect on CE accumulation despite having similarly reduced levels of plasma HDL CE (10). This in vivo requirement for apoA-I might reflect a role of apoA-I in the selective uptake process in FIG. 5. Binding of the class A amphipathic ␣-helix, 37pA⅐DMPC complexes to SR-BI-expressing and vector-transfected cells. SR-BI-expressing and vector-transfected cells in 12-well plates were incubated for 2 h at 4°C with 125 I-37pA⅐DMPC discs as described under "Experimental Procedures." Bound 37pA is plotted versus the concentration of free 37pA at the conclusion of the experiment. The SR-BI-vector curve (dashed line) was obtained by subtracting the mean vector values from the mean SR-BI values and fitting the resultant data via nonlinear regression using a one-site binding isotherm. addition to its binding to mSR-BI. Additional studies with model peptides as well as apoA-I mutants will likely prove informative in extending our understanding of the relationship between apolipoprotein binding to SR-BI and the actual transfer of lipid from the HDL core to the plasma membrane.
The mechanism of mSR-BI-mediated selective CE uptake is not well understood. Recent studies indicate that HDL CE moves down its concentration gradient through a nonaqueous pathway or "channel" from the HDL particle to the plasma membrane (38). Additionally, the extracellular domain of SR-BI does more than simply bind HDL with high affinity and tether the particles close to the plasma membrane (35). The extracellular domain appears to be required for efficient CE transfer, possibly by forming the nonaqueous pathway for CE movement. We speculated that the extracellular domain of mSR-BI may form this pathway via homomeric or heteromeric interactions with other membrane proteins (35). The results of the cross-linking analysis in Fig. 2 (lane 4) support this idea. These data show a cross-linked mSR-BI species with an apparent molecular weight of approximately 225,000. Assuming that 60,000 daltons of this complex is due to apoA-I cross-linked to itself (Fig. 2, lane 8), approximately 165,000 daltons is due to the mSR-BI multimer. Given an mSR-BI apparent molecular weight of 82,000 (11,13), this is sufficient mass to reflect an mSR-BI dimer or a single mSR-BI protein complexed with one or more additional membrane proteins. Additional studies to elucidate the nature of this complex are clearly warranted.