Probing the C-terminal domain of lipid-free apoA-I demonstrates the vital role of the H10B sequence repeat in HDL formation

apoA-I plays important structural and functional roles in reverse cholesterol transport. We have described the molecular structure of the N-terminal domain, Δ(185-243) by X-ray crystallography. To understand the role of the C-terminal domain, constructs with sequential elongation of Δ(185-243), by increments of 11-residue sequence repeats were studied and compared with Δ(185-243) and WT apoA-I. Constructs up to residue 230 showed progressively decreased percent α-helix with similar numbers of helical residues, similar detergent and lipid binding affinity, and exposed hydrophobic surface. These observations suggest that the C-terminal domain is unstructured with the exception of the last 11-residue repeat (H10B). Similar monomer-dimer equilibrium suggests that the H10B region is responsible for nonspecific aggregation. Cholesterol efflux progressively increased with elongation up to ∼60% of full-length apoA-I in the absence of the H10B. In summary, the sequential repeats in the C-terminal domain are probably unstructured with the exception of H10B. This segment appears to be responsible for initiation of lipid binding and aggregation, as well as cholesterol efflux, and thus plays a vital role during HDL formation. Based on these observations and the Δ(185-243) crystal structure, we propose a lipid-free apoA-I structural model in solution and update the mechanism of HDL biogenesis.

structure assignments for apoA-I in 10 nm discoidal and spherical HDL (36), which showed complete helical structure. The HX-derived secondary structure for lipid-free apoA-I suggested that the whole C-terminal domain lacks defined structure (12). In addition, lipid-free apoA-I models based on cross-linking experiments suggested that part of the H10 region formed helical structure (37,38). Most recently, MD simulation of full-length apoA-I based on Borhani et al.'s (35) and our crystal structure suggested that parts of H8 and H10 form helical structure after equilibrium in solution (30). These divergent results imply that the C-terminal domain of apoA-I is a very dynamic region and may adopt different conformations and play important roles during the lipid-binding process.
In order to examine the C-terminal domain of apoA-I, successive elongation constructs of (185-243), namely (198-243), (209-243), (220-243), and (231-243) with 11 residue increments of the sequence repeats were generated to probe the structure and function of the C-terminal domain of apoA-I in the lipid-free state and HDL formation as shown in Fig. 1.
apoA-I is believed to possess a two-domain structural organization with the N-terminal domain (residues 1-189) forming a helix bundle, while the C-terminal domain (190-243) forms a less organized structure (31). Initial binding to a lipid surface is thought to occur through the amphipathic C-terminal domain, followed by opening of the N-terminal helix bundle. This mechanism may represent a general feature for lipid interaction by exchangeable apolipoproteins because apoE behaves similarly. The two-domain structure of apoE has been verified by the fulllength apoE3 NMR structure (32). Based on our previous mutation analysis (23)(24)(25)33) and considering that residues 185 and 186 are both glycine, we proposed a slightly different two domain structure: N-terminal domain (1-184) and C-terminal domain (185-243) (34). This analysis led to the successful crystallization and crystallographic determination of the molecular structure of the N-terminal domain (185-243) at 2.2 Å resolution (34).
Different laboratories have produced divergent results concerning the structure of the C-terminal domain. Borhani et al.'s (35) crystal structure that was thought to represent the lipid-bound state showed complete helical formation for the C-terminal domain except the first half of the H10 region. This helical conformation of the C-terminal domain is closely consistent with HX-derived secondary Fig. 1. Design of apoA-I elongation constructs extended by 11 residue repeating sequence based on consensus sequence analysis and secondary structure of crystal structure (185-243). Green, exon-3 encoded region; ice-blue:,exon-4 encoded region; purple, CSP A homology sequence; cyan, CSP B homology sequence; yellow, proline.
concentrations of BOG, from 0 to 50 mM, after allowing the protein to reach equilibrium (27). For each protein, three samples were prepared independently and each sample was scanned three times with 5 s accumulation time for each data point. Three scans for each sample were averaged, the baseline subtracted, and -helical content estimated.

Fluorescence spectroscopy
Fluorescence measurements were performed on a fluroMax-2 fluorescence spectrometer (Instruments S.A. Inc.) at 25°C using 5 nm excitation and 2.5 nm emission slit widths (33). The ANS fluorescence emission was recorded in the 5 mM phosphate buffer at pH 7.4 in the presence of WT apoA-I or apoA-I mutants with a protein concentration of 0.05 mg/ml and ANS concentration of 0.25 mM. ANS fluorescence was excited at a wavelength of 395 nm and the emission spectra were scanned at wavelengths from 400 to 600 nm. As a control, ANS fluorescence was also scanned in the same buffer in the absence of any protein or with 0.05 mg/ml BSA that is known to have hydrophobic binding pockets, as references.

DMPC turbidity clearance kinetic analysis
The kinetics of binding and remodeling of DMPC multilamellar liposomes by the WT and apoA-I elongation constructs was studied with a DMPC turbidity clearance method, as previously established (24). DMPC, 10 mg powder, was dissolved in a chloroform/methanol (2:1) solution and dried under nitrogen followed by resuspension in 2 ml phosphate buffer by vortexing to give 5 mg/ml DMPC stock. DMPC stock suspension (50 l, 5 mg/ml, 250 g total) was added into 1 ml, 0.1 mg/ml, WT or mutant apoA-I protein solution to achieve a lipid protein ratio of 2.5:1 (w/w). The DMPC clearance then was monitored by the decrease in absorbance at 325 nm at 24°C for 1 h.

Cholesterol efflux assay
Human embryonic kidney (HEK)293 cells were maintained in DMEM containing 10% FBS, 100 U/ml penicillin, and 100 g/ml streptomycin. Mouse macrophage-derived J774 cells were maintained in RPMI containing 10% FBS, 100 U/ml penicillin, and 100 g/ml streptomycin. ABCA1-mediated efflux from HEK293 or J774 cells was determined as described with slight modifications (43,44). HEK293 cells were plated in 24-well collagen-coated plates. Twenty-four hours later, cells were transiently transfected with cDNA encoding ABCA1 or the corresponding control empty vector using Genjet according to the manufacturer's instructions. Cellular cholesterol pools were labeled 24 h posttransfection with 1 M BODIPY-cholesterol in DMSO for 24 h. Cells were then washed and incubated with 10 g/ml apoA-I or apoA-I mutant forms for 24 h. J774 cells were plated in 24-well plates. Twentyfour hours later, the cellular cholesterol pools were labeled with BODIPY-cholesterol complexed to methyl--cyclodextrin at a 1:30 ratio for 90 min in medium A (RPMI containing 0.2% BSA supplemented with 2 g/ml ACAT inhibitor). ABCA1 was then unregulated by incubating the cells in medium A containing 0.3 mM cpt-cAMP for 20-24 h. Cells were washed and incubated with 10 g/ml apoA-I forms in medium B (MEM-HEPES containing 0.01% BSA supplemented with 2 g/ml ACAT inhibitor and 0.3 mM cpt-cAMP) for 24 h. Control cells were incubated in medium lacking cpt-cAMP. Efflux media from HEK293 and J774 cells were filtered to remove cell debris and cells were solubilized in 1% sodium cholate. Fluorescence in cells and media was measured using a plate reader (Tecan Infinite M1000). Percent efflux promoted by the apoA-I forms was calculated from the net fluorescence in medium after subtraction of fluorescence effluxed from cells incubated in medium not containing apoA-I as percent of total fluorescence in medium and cells in each well. Percent efflux RPMI 1640, MEM-HEPES, FBS, penicillin, and streptomycin were purchased from Invitrogen. BSA, methyl--cyclodextrin, 8-(4-chlorophenylthio)adenosine 3′,5′-cyclic monophosphate (cpt-cAMP), ACAT inhibitor (Sandoz 58-035), and HRP-conjugated antibodies to goat were purchased from Sigma. BODIPY-cholesterol (Topfluor) was purchased from Avanti. Genjet was purchased from SinaGen Laboratories. All chemicals used were of analytical grade purity. All solutions were made from doubly distilled water and were filtered by 0.2 m Millipore filters.

Generation of expression plasmids, expression and purification of apoA-I WT and mutant proteins
Elongation mutations were done following the QuickChange site-directed mutagenesis protocol. Truncated mutations were confirmed by DNA sequencing. Recombinant human WT apoA-I and elongation mutants were expressed in Escherichia coli and purified as described previously (34) with an additional glycine residue at the N terminus. Protein, at concentrations of <0.3 mg/ml, was denatured in 6 M GdnHCl and then dialyzed extensively against 5 mM phosphate buffer at pH 7.4 in the cold room to make sure it was properly folded.

Circular dichroism spectroscopy
Circular dichroism (CD) spectra were measured with an Aviv 62DS or Aviv 215 spectropolarimeter equipped with thermoelectric temperature control (Aviv Associates, Lakewood, NJ). Protein solutions at concentrations of 0.05-0.3 mg/ml in 5 mM phosphate buffer at pH 7.4 were placed in 1 mm quartz cuvettes for far-UV spectra measurements. Our previous studies (23) have demonstrated that at this protein concentration (e.g., <0.3 mg/ ml), plasma apoA-I shows no sign of self-association. Far-UV (250-185 nm) spectra were recorded at 1 nm bandwidth and 1 nm step size with 5 s accumulation time for each data point at 25°C (25). Near-UV (325-250 nm) spectra were recorded at protein concentrations of 0.1-0.3 mg/ml in 10 mm quartz cuvettes at 1 nm bandwidth and 1 nm step size with 90 s accumulation time for each data point at 5°C (7). The spectra were normalized to protein concentration and expressed as mean residue ellipticity [] using a mean residue weight. The -helix content was estimated from the mean residue ellipticity at 222 nm [ 222 ] according to a previous protocol (39). ORIGIN 7.5 (Microcal, Amherst, MA) was used for the CD data display and analysis.
Thermal and chemical unfolding. Thermal unfolding was monitored by the protein ellipticity at 222 nm while heating the samples from 5 to 95°C with a temperature step size of 1°C with 90 s accumulation time for each data point. Van't Hoff enthalpies (H v s) of the transitions were determined by a conventional van't Hoff analysis (40) with the assumption of a two-state unfolding model, with baselines that were obtained by linear extrapolation of the pre-and posttransitional regions, as previously described (25). The value of Tm was determined from the peak position of the first derivative func- Chemical unfolding was monitored by the protein ellipticity at 222 nm at 25°C, following incubation of protein samples with various concentrations of GdnHCl from 0 to 2 M to allow proteins to reach equilibrium. A linear extrapolation method (42) under the assumption of the two-state unfolding model was used to determine the apparent free energy difference between the native state and unfolded state of the protein in the presence of denaturant, G D 0 , as previously described (25).

-Helical induction by BOG
-Helical content was monitored by protein ellipticity at 222 nm at 25°C, following incubation of protein samples with various from the -helical content for each recombinant protein.
As shown in Fig. 2A and Table 1, with the sequential elongation of the C-terminal domain up to residue 230, the helical content progressively decreased, while the number of residues in helical conformation remained approximately unchanged at 125 residues. These observations suggest that most of the residues in the C-terminal domain between residues 185 and 230 lack defined structure. The (231-243) construct had similar helical content to WT full-length apoA-I, but the same number of helical residues (125) as the shorter constructs. Addition of the last 13 residues to give the full-length WT protein results in an increase in the number of helical residues to 132, suggesting that the last 11 residue repeat (H10B) contains helical structure.
As shown in Fig. 2B and Table 1, the successive C-terminal elongation constructs up to residue 230 exhibited similar unfolding cooperativity to (185-243) that was greater than WT full-length apoA-I. These observations suggest that the final H10B sequence repeat in the C-terminal domain has an effect on the N-terminal helical bundle. Thus, the helical conformation in this region may result in interaction of H10B with the N-terminal helix bundle. Previous research has also suggested that the C-terminal helix is located near the N-terminal helix bundle and there is interaction between them (25,26,45). Due to the increase in the unfolding cooperativity, the stability of the elongation apoA-I variants is increased or remains similar compared with WT apoA-I (Fig. 2C, Table  1). Previous research also showed that (191-220) exhibited lower -helix content and thermal stability compared with from J774 or HEK cells upregulated to ABCA1 (where ABCA1 was not upregulated or transfected with control empty vector, respectively) was lower than the relative efflux calculated for cells not incubated with apoA-I. The HDL particles formed through the efflux in the media were separated by 4-15% native gel and immunoblotted with apoA-I antibodies.

Gel filtration assay
To determine the dimer and monomer distribution, purified protein constructs (0.5-1 mg/ml) were applied on the Superdex 75 10/300 gel filtration column equilibrated with 25 mM Tris and 150 mM NaCl at pH 8.0. The elution was monitored at UV 280 nm.

Cross-linking assay
Concentrations of purified protein were 1 mg/ml in 25 mM phosphate and 150 mM NaCl (pH 8.0) buffer for the cross-linking assay. Freshly solubilized BS3 was added to the protein solutions at a final concentration 0.25 mM. The mixed solution was cross-linked at room temperature for 30 min then 1 M Tris-HCl (pH 7.5) was added to quench the reaction for 15 min at room temperature and then loaded on a 4-15% SDS-PAGE.

-Helical content and stability in the lipid-free state
Normalized far-UV spectra of the WT and the apoA-I elongation variants were used to estimate the -helical content in the secondary structure of the proteins. The number of residues in helical conformation was estimated presence of WT and the apoA-I elongation variants, BSA, and in the buffer alone are shown in Fig. 3A. The wavelength of maximum fluorescence and relative intensity values determined from the spectra are listed in Table 2. In phosphate buffer, ANS fluorescence has a very low intensity and an emission maximum at 517 nm. In the presence of BSA, a fatty acid transporter that has multiple hydrophobic binding pockets, there is a significant (54 nm) blue shift and almost 7-fold enhancement in ANS fluorescence compared with the ANS in phosphate buffer alone. In the presence WT apoA-I, ANS emission shows a 40 nm blue shift and 4-fold increase in the intensity compared with the ANS in the buffer alone. In the presence of all the C-terminal elongation constructs up to residue 230, ANS emission shows only a 27-29 nm blue shift and similar fold (2-fold) increase in the intensity compared with the ANS in the buffer alone. These observations suggest that the C terminus of apoA-I has exposed hydrophobic surface due to the 231-241 residue (H10B) region. The C-terminal domain of apoA-I is generally considered to lack structure and be responsible for the aggregation and lipid binding (20). The combination of the CD data and the ANS fluorescence data suggests that formation of the helical structure in the 231-241 residue, H10B, region is required to obtain a large exposed hydrophobic surface or binding region because the preceding residues of the C-terminal domain have less effect. Previous research has also demonstrated that (190-243) (31) and single (L230P) or triple (L230P/L233P/Y236P) proline insertion into the C-terminal helix decreases the ANS fluorescence (47), which is consistent with our data, suggesting that helical structure is required in the H10B region to obtain a large hydrophobic surface or binding region.

Interaction with BOG
BOG has been widely used as a mild lipid-mimicking detergent to induce and stabilize amphipathic -helical structure in solution and during protein crystallization (48). The detergent has been used in numerous studies to investigate the effects of lipid environment in the helical WT apoA-I (45). Considering that the 191-220 segment may form a loop, this deletion might shift the H10B helix interaction with the N-terminal helix bundle causing the destabilization, while the elongation apoA-I variants remain similar or have increased stability.

Near-UV spectra
Near-UV spectra of the variant forms of apoA-I are shown in Fig. 2D. The major contribution to the near-UV spectra comes from the aromatic side chains (7). WT apoA-I and the C-terminal elongation constructs contain four Trp residues (W8, 50, 72, and 108), which are all located at the N-terminal domain helical bundle. All the elongation constructs and WT apoA-I have similar near-UV spectra after normalization to protein concentration and show a large negative peak at 292 nm (7) that corresponds to the Trp residues, suggesting that the WT and all the elongation constructs have similar tertiary structure. The C-terminal elongation constructs still possessed all four Trp, and showed similar near-UV spectra after normalization to protein concentration to WT apoA-I, but with a deeper peak at 292 nm. Thus, elongation of the C-terminal region decreased the Trp signal in the near-UV spectra, suggesting a more relaxed environment of the Trp (7). This observation suggests that elongation of the C terminus did not result in a major change in the N-terminal helical bundle organization, but made it somewhat less compact, consistent with the thermal unfolding data, with less unfolding cooperativity compared with (185-243).

ANS fluorescence analysis
Fluorescence of ANS in the presence of the lipid-free WT and variant forms of apoA-I was measured to determine whether the C-terminal elongations affected the exposure of hydrophobic surfaces or cavities of apoA-I. The intrinsic fluorescence of ANS has been shown to be enhanced and blue shifted upon binding to hydrophobic surfaces or cavities, while the water-phase dye does not contribute to the emission (46). ANS fluorescence spectra recorded in the structure in this region, again reaching maximum effect after BOG micelle formation. Most interestingly, (185-243) lost 5% helical content at the CMC, suggesting that BOG may partially disrupt the more compact structure in the N-terminal domain by binding with an exposed nonstructured region and affecting the structured region. All of the C-terminal elongation constructs form additional helical structure after binding to BOG. With the successive C-terminal elongation, more helical residues can be induced, suggesting that most of the C-terminal domain lacks helical structure, as shown in Fig. 3C. In addition, the BOG concentration to induce 50% of the total helical conformation of the apolipoproteins (27). The critical micelle concentration (CMC) of BOG at 25°C is 20-25 mM. The -helical content of the apoA-I variants as a function of BOG concentration is shown in Fig. 3B. The -helical content of WT apoA-I increased from 54% to 64% as the BOG concentration increased from 0 mM to 50 mM and reached a maximum at the BOG CMC concentration, 25 mM. These results suggest that BOG can bind to WT apoA-I and induce -helical structure by 10%. Induction of -helical structure by BOG was also observed for the C-terminal elongation apoA-I variants, suggesting that BOG may bind to the C-terminal region and induce helical   data suggest that the H10B region not only forms helical structure but is also responsible for nonspecific aggregation of apoA-I.

Cholesterol efflux
In order to test the role of the C-terminal domain of apoA-I in ABCA1-mediated cholesterol efflux, efflux assays were conducted using macrophages and transfected HEK cells. As shown in Fig. 5A, B, with the elongation of the C-terminal domain, the cholesterol efflux progressively increased. However, the cholesterol efflux ability of (231-243) only achieved 60% compared with WT apoA-I. Again, these observations suggest that the last 11 residues, the H10B region, play a vital role during the interaction with ABCA1 to facilitate the cholesterol efflux. This result is consistent with previous research that showed that deletion [(223-243)] or disruption of the C-terminal lipid binding domain (L230P or L230P233PY236P) of apoA-I drastically reduced cholesterol efflux, indicating that the C-terminal helix is essential during the cholesterol efflux process (50). Western blot analysis using anti-apoA-I antibodies shows that (231-243) forms HDL-sized particles similar to WT apoA-I, whereas the shorter C-terminal constructs form less well-defined HDL particles due to decreased efflux, as shown in Fig. 5C.

C-terminal domain structure in the lipid-free state
Based on the CD analysis, we concluded that most of the C-terminal domain of apoA-I, residues 185-243, adopts unstructured conformation with the exception of the final H10B repeat, residues 231-241, that forms helical structure. MMP-8-treated apoA-I digested at positions Glu191-Tyr192 and Val221-Leu222 also suggests that this region lacks structure in lipid-free apoA-I (51). Additionally, this analysis is consistent with the low-resolution crystal structure of (1-43) from Borhani et al. (35). As shown in Fig. 6A, the H10B region is in helical conformation with a large exposed hydrophobic surface in this crystal structure. The helical-wheel representation of the H10 region shown in structure (D 1/2 ) of the (198-243), (209-243), (220-243), and (231-243) forms are similar, 20 mM as shown in Fig. 3B. With the addition of the last H10b region, the D 1/2 of the WT apoA-I significantly decreases to 15 mM BOG. This observation suggests that H10B with an exposed hydrophobic surface due to helical structure facilitates the binding with BOG and thus induces helical formation in the rest of the C-terminal domain of apoA-I.

DMPC turbidity clearance
Association of lipid-free apoA-I with DMPC multilamellar vesicles causes a decrease in the turbidity at 325 nm. This process reflects the kinetics of formation of DMPC-apoA-I complexes (49). Figure 3D shows the clearance curves. Compared with WT apoA-I, the C-terminal elongation constructs all showed decreased initial rate of clearance of DMPC liposome turbidity that almost removed the DMPC binding ability completely. This observation is consistent with data from other laboratories suggesting that the C-terminal domain of apoA-I is vital to initiate lipid binding (31). The elongation constructs demonstrate, for the first time, that the H10B region in the C-terminal domain is the vital region to initiate the DMPC binding and perhaps facilitate opening of the N-terminal folding domain for lipid solubilization. This result is consistent with the BOG binding and ANS fluorescence experiments.

Gel filtration and cross-linking
Previously, we have shown that after deletion of the C-terminal domain, (185-243) undergoes a monomerdimer equilibrium that facilitates the crystallization of (185-243). As shown in Fig. 4A, the C-terminal elongation constructs show a similar monomer-dimer equilibrium to (185-243) and progressively migrate toward the full-length apoA-I elution position due to the increment in the molecular weight. Full-length apoA-I exhibits a complex elution profile as a result of nonspecific self-association. In addition, cross-linking of the C-terminal elongation constructs with BS3 shows similar dimer/monomer forms on SDS-PAGE to (185-243), while WT apoA-I can be crosslinked to higher oligomers, as shown in Fig. 4B. These of an apoA-I model by Segrest et al. (30) also suggested that the H10 region (residues 226-239) forms helical structure. Electron paramagnetic resonance studies of the C-terminal domain of apoA-I suggested that residues 221-236 form helical structure, while the rest lack definite structure in Fig. 6B demonstrates that the H10B region has a larger hydrophobic surface compared with other helical regions (52). Thus, not only hydrophobic residues, but also helical structure, are required to obtain a large exposed hydrophobic surface. A full-length molecular dynamics simulation  lipid binding. This result is consistent with previous studies that show that a peptide representing the apoA-I -helix composed of amino acids 220-241 has the highest affinity for lipids (19) compared with other sequence repeats, and an apoA-I mutation lacking the C-terminal residues (residues 187-243) has reduced ability to bind with lipids to form particles (22).

The H10B region is vital for the formation of HDL through cholesterol efflux
The large hydrophobic surface area of the H10B region plays a vital role during cholesterol efflux. The elongation constructs up to residue 230 restore some cholesterol efflux ability compared with (185-243), but only achieve approximately 60% of full-length apoA-I efflux ability. This observation suggests that the H10B region with exposed hydrophobic surface facilitates the interaction with lipid membrane. WT apoA-I has significantly lower BOG concentration required to induce the secondary structure of apoA-I. This observation suggests that the H10B region forming hydrophobic helix facilitates the integration between apoA-I and the cell membrane, thus facilitating cholesterol efflux. Recently, a novel apoA-I truncation (apoA-I Mytilene), due to a heterozygous nonsense mutation (Gln216), has been reported to be associated with the lipid-free state (14). The entire C-terminal domain converts into helical structure upon lipid binding (14). Residues 191-220 play a role in enhancing the ability of apoA-I to bind to and solubilize lipids by forming helix upon lipid interaction (45). All of these studies suggest that the C-terminal domain has a dynamic structure and plays an important role during the lipid-binding process by shifting from disordered into helical structure.

The H10B region is responsible for lipid binding and aggregation
All of the elongation constructs lacking the H10B region show decreased hydrophobic exposure to ANS and slow DMPC clearance similar to (185-243). These results are consistent with previous research demonstrating that the C-terminal domain is responsible for lipid binding (20). Here we pinpoint this function to the H10B region. The helical structure in this region in solution forms a large exposed hydrophobic surface that is responsible for lipid binding.
In addition, gel filtration and cross-linking demonstrate that all the constructs lacking the H10B region exhibit monomer and dimer equilibrium in contrast to the nonspecific aggregation of WT apoA-I. Thus, for the first time, we identify that the region of apoA-I responsible for aggregation corresponds to that responsible for initiation of Fig. 7. Proposed full-length apoA-I in the lipid-free state and a possible three-state mechanism of HDL formation. Full-length apoA-I in lipid-free state: residues 1-184 from (185-243) crystal structure (Protein Data Bank accession number: 3R2P), residues 185-243 from (1-43) crystal structure (Protein Data Bank accession number: 1AV1) with relaxation from helix to random coil except the H10B region. Initial state: Monomeric apoA-I adopts the helix bundle conformation with the C-terminal helical H10B region forming exposed hydrophobic surface. Intermediate state: Two monomeric apoA-I close enough to facilitate the domain swap interaction and result in dimerization due to binding to lipids spontaneously or facilitated through ABCA1. Final state: Lipids are solubilized by the double belt dimer with additional helical conformation forming in the unstructured region of the C-terminal domain further stabilizing the interaction and resulting in the release of discoidal HDL from the cell membrane.