The Central Helices of ApoA-I Can Promote ATP-binding Cassette Transporter A1 (ABCA1)-mediated Lipid Efflux

We have mapped the domains of lipid-free apoA-I that promote cAMP-dependent and cAMP-independent cholesterol and phospholipid efflux. The cAMP-dependent lipid efflux in J774 mouse macrophages was decreased by ∼80–92% by apoA-I[Δ(185–243)], only by 15% by apoA-I[Δ(1–41)] or apoA-I[Δ(1–59)], and was restored to 75–80% of the wild-type apoA-I control value by double deletion mutants apoA-I[Δ(1–41)Δ(185–243)] and apoA-I[Δ(1–59)Δ(185–243)]. Similar results were obtained in HEK293 cells transfected with an ATP-binding cassette transporter A1 (ABCA1) expression plasmid. The double deletion mutant of apoA-I had reduced thermal and chemical stability compared with wild-type apoA-I. Sequential carboxyl-terminal deletions showed that cAMP-dependent cholesterol efflux was diminished in all the mutants tested, except the apoA-I[Δ(232–243)] which had normal cholesterol efflux. In cAMP-untreated or in mock-transfected cells, cholesterol efflux was not affected by the amino-terminal deletions, but decreased by 30–40% and 50–65% by the carboxyl-terminal and double deletions, respectively. After adenovirus-mediated gene transfer in apoA-I-deficient mice, wild-type apoA-I and apoA-I[Δ(1–41)] formed spherical high density lipoprotein (HDL) particles, whereas apoA-I[Δ(1–41)Δ(185–243)] formed discoidal HDL. The findings suggest that although the central helices of apoA-I alone can promote ABCA1-mediated lipid efflux, residues 220–231 are necessary to allow functional interactions between the full-length apoA-I and ABCA1 that are required for lipid efflux and HDL biogenesis.

Apolipoprotein A-I (apoA-I) 1 is the major protein constituent of high density lipoproteins (HDL) and plays crucial roles in the synthesis, structure, and function of HDL (1). One important function of apoA-I and HDL is the induction of cholesterol efflux from cells, a process that is crucial for regulating cellular free cholesterol homeostasis (2,3), receptor endocytosis and recycling (4), and cell signaling (5). Four different kinds of interaction between extracellular acceptors have been proposed to cause cholesterol efflux. (a) Natural or reconstituted HDL particles or other lipophilic acceptor particles can cause a passive diffusion of cholesterol from cholesterol-poor membrane domains through the aqueous phase (2,6). (b) Interaction of cellular proteins (e.g. scavenger receptor B-I) with HDL facilitates the efflux of unesterified cholesterol to extracellular cholesterol acceptors (7)(8)(9). Activation of the cholesterol-esterifying enzyme lecithin:cholesterol acyltransferase by apoA-I also facilitates this process (10,11). (c) Lipid-free apolipoproteins with a high content of amphipathic ␣-helices, such as apoA-I, can associate with multiple molecules of cellular phospholipids and free cholesterol by a process that was termed microsolubilization (12,13). (d) Lipid-free apoA-I can also promote efflux of cellular phospholipids and cholesterol by an ATP-dependent process, which involves the Tangier disease gene product, i.e. the ATP-binding cassette transporter A1 (ABCA1) (14 -17). It has been proposed that initial interactions of apoA-I and ABCA1 promote efflux of membrane phospholipids, which results in the gradual extracellular lipidation of apoA-I and in the formation of pre-␤ 1 -HDL. This apoA-I-phospholipid complex promotes further efflux of cellular cholesterol (14, 18 -22) and the formation of discoidal pre-␤ 2 -HDL and, after cholesterol esterification by lecithin:cholesterol acyltransferase, spherical ␣-HDL particles (23)(24)(25)(26).
The mechanism of interaction between ABCA1 and apoA-I is not fully resolved. On one hand overexpression of ABCA1 was shown to enhance the cellular binding and uptake of apoA-I (27)(28)(29)(30). Because apoA-I can be cross-linked with ABCA1, a direct physical interaction between apoA-I and ABCA1 has been postulated as important for mediating the ABCA1-dependent lipid efflux. On the other hand it has also been suggested that ABCA1 does not interact physically with apoA-I but changes the membrane lipid microenvironment so that apoA-I can associate with the cell membrane and acquire phospholipids by binding in the proximity of ABCA1 (20). Analysis of five naturally occurring ABCA1 mutants, in the first and fourth extracellular domains, that are defective in cholesterol efflux showed that four of the five mutants failed to cross-link to apoA-I. The findings suggest that either direct binding of apoA-I to ABCA1 or insertion of apoA-I in the plasma membrane in close proximity to ABCA1 is necessary but not sufficient to promote cholesterol efflux (31). It has also been shown that ABCA1-mediated binding and lipid efflux are not specific for apoA-I but may also occur with other apolipoproteins that contain multiple amphipathic helical domains (27).
Previously, deletion mutants of apoA-I have been analyzed to unravel the structural requirements of apoA-I for inducing cholesterol efflux. Deletion of the carboxyl-terminal helices 8 -10 in apoA-I (apoA-I[⌬(187-243)] or apoA-I[⌬(210 -243)]) reduced the ability of the mutant protein to bind to macrophages and to promote efflux of cholesterol and phospholipid (32,33). Likewise, deletion of amino acid residues 211-243 and 222-243 impaired the ability of apoA-I to promote cholesterol efflux from cultures of hepatocytes and fibroblasts (13,34) (13,32,34).
In the present study we have analyzed the ability of various recombinant apoA-I mutants lacking amino-and carboxyl-terminal domains to promote efflux of cellular cholesterol and phospholipid from J774 mouse macrophages in the presence of a cAMP analog (i.e. ABCA1-catalyzed) or in the absence of a cAMP analog (i.e. at least partially independent of ABCA1). We have performed similar analysis in human embryonic kidney (HEK) 293 cells following transient transfection with an ABCA1 expression plasmid. Thereby we found evidence that the ABCA1-mediated cholesterol efflux by the full-length apoA-I requires a carboxyl-terminal domain in apoA-I, which encompasses amino acid residues 220 -231. The analysis of apoA-I mutants containing deletions in the amino-terminal region, carboxyl-terminal region, or both domains suggests that although there is ligand specificity of the full-length apoA-I, the central helices of apoA-I alone are capable of promoting ABCA1-mediated lipid efflux and the formation of discoidal HDL in vitro and in vivo.

Generation and Isolation of Recombinant Wild-type ApoA-I or ApoA-I Mutants in the Baculovirus System
To generate baculoviruses expressing wild-type (WT) and mutant apoA-I forms, a BamHI-SalI fragment containing human apoA-I cDNA was cloned into the polylinker region of pFASTBAC donor plasmid that contains the ampicillin and gentamycin resistance genes (Invitrogen). This recombinant plasmid also contains a histidine (His) tag and the TEV protease cleavage site. The apoA-I containing plasmid was used to transform DH10Bac Escherichia coli cells. These cells had been transformed previously with the baculovirus genome containing the lacZ and kanamycin resistance genes, along with a helper plasmid containing the tetracycline resistance gene and the transposase genes. Transposition of the apoA-I gene into the lacZ gene disrupts the expression of the lacZ and provides a recombinant plasmid named bacmid.
The generation of the donor plasmid expressing WT apoA-I and of plasmids expressing apoA-I[⌬ ], apoA-I[⌬(185-243)], and apoA-I[⌬(1-59)⌬(185-243)] mutants was described previously (35,36). The other mutant forms of apoA-I were produced in a similar manner. Briefly, the human apoA-I cDNA was mutagenized by PCR, using a set of flanking universal primers containing the restriction sites BamHI and SalI, using the pBluescript-AI cDNA plasmid as template. The set of specific primers used in mutagenesis, and the set of universal primers 5Ј(AIWC-5) and 3Ј(AIWC-3) are shown in Table I. Specifically, we used 5Ј(AIM1-5) and 3Ј(AIM1-3) primers for the generation of apoA-I[⌬(1-41)], and 5Ј(AIM1-5) and 3Ј(AIM3-3) primers for the generation of apoA-I[⌬(1-41)⌬(185-243)]. The DNA fragment containing the mutation of interest was digested with BamHI and SalI, and cloned into the corresponding sites of pFASTBAC donor plasmid. Cells containing recombinant bacmids were selected by kanamycin, tetracycline, and gentamycin resistance as white colonies due to the disruption of lacZ sequence in the recombinant bacmid. Recombinant bacmid DNA was isolated from minipreps and used to transfect a monolayer of Sf-9 insect cells (38 -40). Recombinant viruses were isolated, amplified, titrated, and used to infect larger Sf-9 cell cultures grown in suspension at 27°C. Sf-9 cells were pelleted and resuspended in a lysis buffer. The supernatant was used for the purification of apoA-I fusion proteins (His ϩ -apoA-I) using a nickel-nitrilotriacetic acid affinity column (41,42). The pure apoA-I without the His tag (His Ϫ -apoA-I), when needed, was obtained by cleavage with rTEV protease and purified by using again the nickel-nitrilotriacetic acid affinity chromatography.

Expression and Purification of Recombinant WT and Mutant Pro-apoA-I Isoforms in C127 Cells
Construction of plasmids expressing WT pro-apoA-I and pro-apoA-I[⌬(198 -243)] has been described previously (44,45). The pro-apoA-I[⌬(209 -243)] deletion mutant was constructed in a similar manner. Briefly, the fourth exon of the human apoA-I gene was amplified and mutagenized by PCR, using a set of specific mutagenic primers (209S and 209A) containing the mutation of interest (a stop codon at nucleotide 209) and a set of flanking universal primers (AINotI and AIXhoI) containing the restriction sites NotI and XhoI. The sequences of the primers are shown in Table I. The pUCA-I N * vector, which contains a NotI site in intron 3 and a XhoI site in the 3Ј-end of the apoA-I gene, was used as a template in the amplification reactions (43). The DNA fragment containing the mutation of interest was digested with NotI and SalI and cloned into the NotI and XhoI sites of the pBMT3X-AI vector.
Permanent cell lines in mouse mammary tumor C127 cells expressing the WT pro-apoA-I and the carboxyl-terminal truncation pro-apoA-I[⌬(198 -243)] and pro-apoA-I[⌬(209 -243)] were generated as described previously (43). C127 cell clones overproducing the WT and the variant apoA-I forms were grown in roller bottles on collagen-coated lead microspheres (Verax Corp.), and the proteins were purified from the serum-free medium as described previously (43) with some modifications. Briefly, the medium collected from the roller bottles was dialyzed against 0.01 M Tris, pH 8, and filtered. Up to 350 ml of the medium were passed through a 5-ml HiTrap Q HP column or up to 1200 ml were passed through a 25-ml Q-Sepharose high performance column at a rate of 1 ml/min. Both columns had been equilibrated with the Tris buffer. The proteins were eluted with a linear gradient of 1 M NH 4 HCO 3 in the Tris buffer (15 volumes of elution buffer for HiTrap Q column and 5 volumes for Q-Sepharose high performance column) at the same flow rate. The fractions were analyzed by SDS-PAGE, and those containing the apoA-I protein were pooled and further concentrated down to 2 ml, using a Centriplus concentrator (Amicon) with a molecular weight cut-off of 10,000. The concentrated sample was applied to a gel filtration column HiPrep Sephacryl S-200 and eluted with 1 column volume (120 ml) of 0.15 M NH 4 HCO 3 buffer at a rate of 0.4 ml/min. The purity of the apoA-I preparation was assessed by SDS-PAGE, and fractions greater than 95% pure were recovered.

Expression of WT and Mutant Pro-apoA-I Isoforms in an Adenovirus System
The construction of recombinant adenoviruses carrying the genomic sequence for the WT pro-apoA-I and pro-apoA-I[⌬(220 -243)] has been described before (45). The adenovirus expressing pro-apoA-I[⌬(232-243)] was generated in a similar way. The DNA fragments containing the mutations of interest were produced as described in the previous paragraph, using the 220S and 220A primers for the mutagenesis of pro-apoA-I[⌬(220 -243)] and the 232S and 232A primers for pro-apoA-I[⌬(232-243)] ( Table I). The recombinant adenoviruses expressing pro-apoA-I[⌬(1-41)] and pro-apoA-I[⌬(1-41)⌬(185-243)] were generated using cDNAs for apoA-I deletion mutants. The human apoA-I cDNA was mutagenized by PCR, using a set of primers containing the restriction sites BamHI and SalI and using the pBluescript-AI cDNA plasmid as template. The set of specific primers used in mutagenesis and the set of primers containing the restriction sites BamHI and SalI, designated AI(1-59)F and AI(1-59)R, are shown in Table I as template. Following amplification and mutagenesis, the DNA fragments were digested with BamHI and SalI and subcloned into the pCA13 vector under the control of the cytomegalovirus promoter (Microbix Biosystems Inc.). Recombinant adenovirus constructs carrying this cDNA were prepared following co-transfection of E1 transformed 911 cells with the plasmid pJM17 (Microbix Biosystems Inc.) and the appropriate pCA13 plasmid (45).
Human HTB13 cells (SW 1783, human astrocytoma grown to 80% confluence in Leibovitz's L-15 medium containing 10% (v/v) fetal bovine serum (FBS) in roller bottles) were infected with adenoviruses expressing WT pro-apoA-I, pro-apoA-I[⌬(220 -243)], and pro-apoA-I[⌬(232-243)] at a multiplicity of infection of 20. After 24 h of infection, cells were washed twice with serum-free medium and preincubated in serum-free medium for 30 min, and fresh serum-free medium was added. After 24 h the medium was harvested, and fresh serum-free medium was added to the cells. The harvests were repeated 8 -10 times.
The proteins were purified from the serum-free medium by ion exchange chromatography followed by gel filtration, as described above for C127 cells.

ABCA1 Expression Plasmid
The expression plasmid of human ABCA1 was a generous gift of Dr. M. W. Freeman and was constructed as described (46).

Cell Cultures and Lipids Efflux Assays
Studies in J774 Mouse Macrophages-J774 mouse macrophages were grown in RPMI 1640 with 10% (v/v) FBS and antibiotics. J774 macrophages were plated in 12-well plates at density of 5 ϫ 10 5 cells/ well. The following day, cells were labeled with 1 ml of labeling medium (6 Ci/ml 1,2-[ 3 H]cholesterol or 1 Ci of methyl[ 3 H]choline chloride dissolved in 0.1% (v/v) ethanol as the final concentration and dispersed into RPMI 1640 supplemented with 0.2% (w/v) BSA). Following 24 h of labeling, cells were washed twice with serum-free medium and equilibrated for 24 h with or without 0.3 mM cpt-cAMP in 1 ml of RPMI 1640 supplemented with 0.2% (w/v) BSA. At the end of the treatment period with cpt-cAMP, cells were washed twice with serum-free medium and incubated with 1 ml of RPMI 1640, supplemented with 0.2% (w/v) BSA, with or without 1 M WT apoA-I or mutant forms at 37°C. For cholesterol efflux experiments, at different time points up to 6 h, 55 l of medium were collected and clarified by centrifugation in a microcentrifuge for 2 min. The radioactivity in 40 l of the supernatant was determined by liquid scintillation counting. For phospholipid efflux experiments, after 4 h of incubation, the media were collected and clarified by centrifugation in a microcentrifuge for 2 min. A 100-l aliquot of the supernatant was first extracted with chloroform and methanol (47), and then the radioactivity of the chloroform phase was determined by liquid scintillation counting. At the end of the incubation, cells were lysed by 800 l of lysis buffer (PBS containing 1% (v/v) Triton X-100) for 30 min at room temperature. For cholesterol efflux experiments, the radioactivity in 40 l of each cell lysate was determined by scintillation counting. For phospholipid efflux experiments, the phospholipids in 100 l of each lysate were extracted with chloroform and methanol (47), and the radioactivity was determined by scintillation counting. The percentage of secreted 3 H-lipid was calculated by dividing the medium-derived counts by the sum of the total. To calculate the net cpt-cAMP-dependent efflux, the lipid efflux of the untreated  c Amino acid position (ϩ) relative to the mature apoA-1 (i.e. ϩ1 from mature apoA-I protein). d The stop codon is underlined.
cells was subtracted from the lipid efflux of the cells treated with cpt-cAMP. Studies in HEK293 Cells following Transfection with an ABCA1 Expression Plasmid-HEK293-EBNA cells were cultured in DMEM (high glucose) with 10% (v/v) FBS and antibiotics (31,46). One day before transfection, the cells were plated on 24-well plates coated with poly-D-lysine at density of 2 ϫ 10 5 cells/well in the same medium without antibiotics. The next day, cells at about 95% confluence were transfected with LipofectAMINE 2000 and ABCA1 or pcDNA1 (mock) (Invitrogen) plasmids for 16 h. The DNA-LipofectAMINE 2000 complexes were formed in serum-free Opti-MEM I. After this period of time, cells were labeled with 0.5 Ci/ml 1,2-[ 3 H]cholesterol in DMEM (high glucose) supplemented with 10% (v/v) FBS. Following 24 h of labeling, cells were washed twice with PBS and incubated for 1 h with DMEM (high glucose) supplemented with 0.2% (w/v) fatty acid-free BSA. At the end of this period, cells were incubated with 0.5 ml of DMEM (high glucose) and supplemented with 0.2% (w/v) fatty acid-free BSA with or without 1 M WT apoA-I or mutant forms at 37°C for 4 h. The media were collected and clarified by centrifugation in a microcentrifuge for 2 min, and the radioactivity in 300 l of the supernatant was determined by liquid scintillation counting. The cells were lysed with 0.1 M NaOH, and the radioactivity in a 300-l aliquot of the cell lysate was determined by scintillation counting. [ 3 H]Cholesterol efflux was expressed as the percentage of the radioactivity released in the medium relative to the total radioactivity in cells and medium. To calculate the net ABCA1mediated efflux, the cholesterol efflux of the cells transfected with the control plasmid (mock) was subtracted from the cholesterol efflux of the cells transfected with the ABCA1 plasmid.

Animal Studies
Male apoA-I-deficient mice (apoA-I Ϫ/Ϫ ) (ApoA1 tm1Unc ) C57BL/6J mice (48) were purchased from The Jackson Laboratories (Bar Harbor, ME). The mice were maintained on a 12-h light/dark cycle and standard rodent chow. The mice were 6 -8 weeks old at the time of injection of the adenoviruses. All procedures performed on the mice were in accordance with National Institutes of Health and institutional guidelines. The A-I Ϫ/Ϫ mice were injected via the tail vein with 1 ϫ 10 9 plaque-forming units of recombinant adenovirus per animal, and the animals were sacrificed 4 days post-injection following a 4-h fast. The plasma of the mice was adjusted to 1 mM phenylmethylsulfonyl fluoride, 0.2% (w/v) aprotinin, 0.1% (w/v) EDTA, and 0.002% (w/v) sodium azide prior to analysis of lipids, lipoproteins, and apoA-I.

Cholesterol, Triglycerides, and Plasma ApoA-I Levels and
Steady-state Hepatic Human ApoA mRNA Levels The concentration of cholesterol and triglycerides in plasma was determined using the INFINITY cholesterol reagent (Sigma) and INFINITY triglyceride reagent (Sigma) according to the manufacturer's instructions. Cholesterol and triglyceride concentrations were determined spectrophotometrically at 490 nm. The concentration of human apoA-I in plasma was determined by sandwich ELISA (36). Mouse monoclonal anti-human apoA-I antibodies 5F6 (University of Ottawa Heart Institute) (the epitope is amino acids 118 -141 of the human apoA-I) were used for coating microtiter plates, and sheep polyclonal anti-human apoA-I coupled to horseradish peroxidase (Biodesign International) was used as a secondary antibody. The immunoperoxidase procedure was employed for the colorimetric detection of apoA-I at 490 nm, using o-phenylenediamine dihydrochloride as substrate. The concentration was estimated based on standard curves constructed either with a human serum standard (Cardiolipid Level 1, Sigma) or with the purified lipid free WT and mutant forms of apoA-I. The concentration of human apoA-I in plasma was also estimated by Western blot analysis of the apoA-I samples. The membrane was probed with a goat polyclonal anti-human apoA-I antibody (Chemicon International). The proteins were visualized by chemiluminescence following incubation with rabbit anti-goat IgG (Biodesign International) coupled to horseradish peroxidase. ApoA-I was quantitated by scanning densitometry using different concentrations of purified WT apoA-I so that the intensities of the signal of plasma WT and mutant apoA-I forms were similar and in the linear range.
Total cellular RNA was isolated from the mouse liver using the Trizol reagent (Invitrogen) as recommended by the manufacturer. 15 g of RNA were separated on 1.0% agarose-formaldehyde gels, transferred to Hybond-N ϩ nylon membrane (Amersham Biosciences), and cross-linked to the filter by UV irradiation (Stratalinker, Stratagene). The apoA-I probe for hybridization contained 290 bp of exon 4 of human apoA-I and 148 bp of the intergenic sequence between the apoA-I and apoCIII genes. The mouse glyceraldehyde-3-phosphate dehydrogenase probe (905 bp) was obtained from Ambion. The probe was labeled with 32 P using the Multiprime DNA labeling system (Amersham Biosciences). Quantitation of the x-ray film was performed by a PhosphorImager (Amersham Biosciences) using the ImageQuant program. The apoA-I mRNA signal was normalized for the glyceraldehyde-3-phosphate dehydrogenase mRNA signal.

Fractionation of Plasma by Equilibrium Density Gradient Centrifugation and Electron Microscopy Analysis of the ApoA-I-containing Fractions
An aliquot of 250 l of plasma was adjusted to 2 ml with PBS, and the lipoproteins were separated on 10 -20% sodium bromide gradients (49). The tubes were spun for 66 h at 32,000 rpm on an SW-41 rotor (Beckman Instruments), and fractions (0.4 ml) were collected from top to bottom. The densities of the fractions were determined by comparing the readings measured with a refractometer (American Optical Corp.) to values obtained for solutions with known densities. The fractions were dialyzed against ammonium acetate and carbonate buffer (126 mM ammonium acetate, 2.6 mM ammonium carbonate, 0.26 mM EDTA, pH 7.4). Aliquots of the fractions were subjected to 13% SDS-PAGE, and the gels were transferred to nitrocellulose membrane for Western blot analysis. The membrane was probed with the mouse monoclonal antihuman apoA-I antibody 5F6 (University of Ottawa Heart Institute). The proteins were visualized by chemiluminescence following incubation with goat anti-mouse IgG 1 (Roche Molecular Biochemicals) coupled to horseradish peroxidase. Fractions containing peak concentration of apoA-I were analyzed by electron microscopy.

Electron Microscopy
Aliquots of the fractions from equilibrium density gradient centrifugation after dialysis against ammonium acetate and carbonate buffer were stained with sodium phosphotungstate, visualized in the Phillips CM-120 electron microscope (Phillips Electron Optics, Eindhoven, Netherlands), and photographed as described previously (50). The photomicrographs were taken at ϫ75,000 magnification and enlarged 3 times.

Preparation and Electron Microscopy Analysis of rHDL
Containing POPC:Cholesterol:ApoA-I rHDL particles were prepared by the sodium cholate dialysis method (51), using POPC:cholesterol:apoA-I:sodium cholate in a molar ratio of 100:10:1:100, as described previously (52). The lipid-free apoA-I that was not incorporated in the lipid-protein complexes was removed by dialysis overnight against PBS buffer at 4°C with at least 3 changes of 4 liters each, using dialysis tubing of 50,000 daltons molecular weight cut-off. The rHDL samples were adjusted either by dilution or by concentration to a final protein concentration of ϳ1 mg/ml. A 5-l aliquot of each sample was used for electron microscopy as described above. The photomicrographs were taken at ϫ45,000 magnification and enlarged 2.5 times.

CD Spectroscopy
CD measurements were performed on an AVIV 62 DS spectropolarimeter (AVIV Associates, Inc.), equipped with a thermoelectric temperature control and calibrated with d-10-camphorsulfonic acid in 5-, 2-, or 1-mm path length quartz cuvettes, at protein concentration 20 -100 g/ml. Far-UV CD spectra were recorded as described previously (44,53). The spectra were normalized to molar residue ellipticity using a mean residue weight of 115.0 for WT apoA-I, 115.

Thermal and Chemical Unfolding Monitored by CD
Monitoring of thermal and chemical unfolding of apoA-I by ellipticity at 222 nm was performed at protein concentrations 25-80 g/ml as described previously (44,53). The midpoint (melting temperature), T m , and van't Hoff enthalpy, ⌬H v , of the temperature-induced transitions were determined from van't Hoff plots (55), ln K eq versus 1/T, where K eq is the apparent equilibrium constant, and T is temperature in degrees Kelvin. The equilibrium constant was calculated as K eq (T) ϭ (⌰ F Ϫ ⌰ obs )/(⌰ obs Ϫ ⌰ U ), where ⌰ obs is the observed ellipticity, and ⌰ F and ⌰ U are base lines for the protein folding and unfolding states determined by linear extrapolations of the pre-and post-transitional regions. In the transition region, the plots were fitted by the linear function, ln K eq ϭ ⌬S/R Ϫ (⌬H v /R) ϫ 1/T, where R ϭ 1.98 cal/mol ϫ K is the universal gas constant, and ⌬S is the transition entropy.
For chemical unfolding, the free energy of denaturation (conformational stability), ⌬G D 0 , the midpoint of denaturation, D1 ⁄2 , and m value that reflects the steepness of the denaturation curve in the transition region were determined using a linear extrapolation method (56). A plot of Gibbs free energy ⌬G D ϭ Ϫ RT ln K eq , where T ϭ 25°C ϭ 298.15 K, versus denaturant concentration, [D], was fitted by the linear function, . The apparent equilibrium constant, K eq , was calculated as shown above, except ⌰ obs ([D]) is the observed ellipticity at a given concentration of GdnHCl, [D], and ⌰ F and ⌰ U are the ellipticity in the folded and unfolded states. ⌰ F and ⌰ U were taken as constants.

DMPC Clearance Kinetic Analysis
The initial rate of binding of WT and mutant proteins to DMPC multilamellar liposomes was studied using a kinetic-turbidimetric method (57). Briefly, 250 g of DMPC in chloroform:methanol (2:1) solution were dried under nitrogen and resuspended in 900 l of PBS by vortexing to form liposomes. 100 g of apoA-I in 100 l of PBS were added to the lipid emulsion to give DMPC:apoA-I ratio of 2.5:1 (w/w) and apoA-I concentration of 0.1 mg/ml. The DMPC clearance was followed by monitoring the decrease in absorbance at 325 nm at 5-min intervals at 24°C. A plot of absorbance at 325 nm versus time in min, A(t), was fitted by an exponential decay function: A(t) ϭ A ϱ ϩ ae Ϫt/k , where A ϱ , a, and k are fitting parameters. Fig. 1 shows the apoA-I forms used in this study and the expression system used for their production. The WT and the variant apoA-I forms were produced using three different expression systems (baculovirus, adenovirus, and permanent cell lines). SDS-PAGE analysis showed that all apoA-I forms were apparently free of any contaminations (Fig. 1).

Production and Purification of WT and Mutant ApoA-I Forms Used in Lipid Efflux Studies-
Effect of Amino-, Carboxyl-terminal and Double Deletion Mutants of ApoA-I on cAMP-dependent and cAMP-independent Efflux of Cellular Cholesterol from J774 Mouse Macrophages and the Influence of the His Tag-To determine the potential domains of apoA-I that contribute to cholesterol efflux, J774 mouse macrophages were labeled with [ 3 H]cholesterol for 24 h and then incubated in serum-free culture medium for 24 h in the presence or absence of 0.3 mM of the cAMP analog cpt-cAMP. cAMP analogs have been reported to enhance the expression of ABCA1 (18,58,59). Cholesterol efflux was monitored as a function of time in the presence of 1 M lipid-free WT or mutant forms of apoA-I, as described under "Experimental Procedures." The apoA-I concentration used is higher than that required for saturable cholesterol or phospholipid efflux in untreated or cpt-cAMP-treated J774 macrophages (60) and in transiently or stably transfected cells expressing the ABCA1 transporter (27,28).
The time course of cholesterol efflux from untreated (cAMPindependent cholesterol efflux) and cpt-cAMP-treated (total cholesterol efflux) cells, respectively, using apoA-I forms lacking the amino-terminal 1-41 or 1-59 segments or the carboxylterminal 185-243 segment or both the amino-and carboxylterminal segments, was monitored over a 6-h period. The time course of the net cpt-cAMP-dependent cholesterol efflux was obtained by subtracting the efflux curves of the untreated cells from the treated cells. This analysis indicated that the efflux curves (total, cpt-cAMP-independent (data not shown) and cpt-cAMP-dependent ( Fig. 2A)) are almost linear over a 4-h period.
To facilitate the analysis of the cholesterol efflux by the different apoA-I forms, we tabulated total, cpt-cAMP-independent, and cpt-cAMP-dependent efflux after 4 h of incubation with 1 M WT apoA-I and mutants at 37°C, setting the efflux values for WT apoA-I to 100% (Fig. 2B) decreased the cpt-cAMP-independent cholesterol efflux by 50 -65% (Fig. 2B, white bars). The cpt-cAMP-dependent cholesterol efflux was reduced by 80% compared with WT apoA-I when the carboxyl-terminal apoA-I[⌬(185-243)] deletion mutant was used as the cholesterol acceptor. Much less pronounced impairments of cholesterol efflux were seen in the presence of apoA-I variants with amino-terminal deletions since removal of the amino terminus decreased cholesterol efflux by 15%. Unexpectedly, cpt-cAMP-dependent cholesterol efflux was restored to 78% of WT apoA-I when the cholesterol efflux assay was performed in the presence of double deletion mutants lacking both the aminoand carboxyl-terminal domains (Fig. 2B, shaded bars). These findings indicate that the various apoA-I mutants affect differently cpt-cAMP-dependent and -independent cholesterol efflux.
The baculovirus system produces a fusion of apoA-I-(1-243) with a 28-residue amino-terminal segment with the sequence MSYYHHHHHHDYDIPTTFNLYFQGAMGS. This segment contains 6 His residues, a spacer region, and the TEV protease cleavage site. Removal of the His tag by rTEV protease generates a His Ϫ -apoA-I, which still contains an additional 5 aminoterminal sequence (GAMGS). To assess the potential interfer- respectively, as compared with the corresponding His Ϫ -apoA-I forms (data not shown). The presence of His tag also reduced 27-53% the cpt-cAMP-independent efflux capacities of aminoterminal deletion mutants but had no significant effect on the double mutants (data not shown). These observations led us to utilize the His Ϫ -apoA-I forms in the preceding and the following experiments.

Effect of Amino-, Carboxyl-terminal and Double Deletion Mutants of ApoA-I on ABCA1-mediated Efflux of Cellular Cholesterol in HEK293 Cells Transfected with an ABCA1 Expression
Plasmid-To confirm that cAMP-dependent cholesterol efflux of amino-, carboxyl-terminal and double deletion mutants of apoA-I was mediated by ABCA1, we compared the cAMP-dependent cholesterol efflux to the ABCA1-dependent efflux. HEK293-EBNA cells were transfected with empty vector or an ABCA1 expression plasmid for 16 h, as described under "Experimental Procedures." Following transfection, the cells were labeled with [ 3 H]cholesterol for 24 h and then incubated with or without 1 M lipid-free WT or mutant forms of apoA-I for 4 h. The efflux values obtained for WT apoA-I were set as 100%.
The cholesterol efflux pattern in HEK293-EBNA cells transfected with an ABCA1 plasmid was similar to that of cAMPdependent cholesterol efflux in J774 macrophages (compare Fig. 2, B and C). In mock-transfected cells, efflux of cellular cholesterol followed the same pattern as in cAMP-untreated cells (compare Fig. 2, B and C). Similar to the findings of The cpt-cAMP-independent cholesterol efflux values for the WT pro-apoA-I produced by either the permanent cell lines or the adenovirus system were much lower than the efflux values observed for the WT apoA-I produced with the baculovirus system (compare Fig. 2, B and D) but were at the same levels as the efflux values obtained with human apoA-I isolated from plasma (data not shown). By contrast to cpt-cAMP-independent, the cpt-cAMP-dependent cholesterol efflux values for all four WT apoA-I were similar (Fig. 2, B and D).
The cpt-cAMP-independent cholesterol efflux was very low for the WT pro-apoA-I and barely detectable in the pro-apoA-I mutants with carboxyl-terminal truncations (Fig. 2D, white   bars). The cpt-cAMP-dependent cholesterol efflux was 20, 33, 9, and 99% of the WT control value for the mutants lacking the carboxyl-terminal domains 198 -243, 209 -243, 220 -243, and 232-243, respectively (Fig. 2D, shaded bars). These data suggest that amino acid residues 220 -231 of apoA-I form a domain that is important for the ability of the intact apoA-I to induce cpt-cAMP-dependent cholesterol efflux. The analysis showed that the cpt-cAMP-dependent phospholipid efflux followed similar patterns to those described for cholesterol efflux. Compared with WT apoA-I, cpt-cAMP-dependent phospholipid efflux was 8% in the presence of apoA-I[⌬(185-243)] and 72 and 102% in the presence of apoA-I[⌬ ] and apoA-I[⌬(1-59)], respectively. As in the cholesterol efflux experiments, the deletion of both the amino-and the carboxyl-terminal segments restored the phospholipid efflux capacity of apoA-I to 75% (compare Fig. 2, B and E).

Effect of Amino-, Carboxyl-, and Double Deletion Mutants of
The cpt-cAMP-independent phospholipid efflux was low for WT apoA-I, was decreased approximately by about 50% for the two amino-terminal deletion mutants, and was barely detectable in the other mutants.
The thermal unfolding curves monitored by the ellipticity at 222 nm, [⌰ 222 ] (T), are shown in Fig. 3A; the corresponding van't Hoff plots, ln K eq (1/T), are shown in the inset; and the values of the melting temperature, T m , and the effective enthalpy, ⌬H v , determined by the van't Hoff analysis are given in Table II. This analysis did not show a significant effect of the ⌬(1-41) truncation on the melting temperature (T m ) but did show an ϳ9°C reduction in T m as a result of the double truncation ⌬(1-41)⌬(185-243), indicating a significant destabilizaof 0.3 mM cpt-cAMP followed by incubation with 1 M of WT and mutants apoA-I forms for 4 h at 37°C. Media and cells were collected separately; the 3 H-phospholipids were first extracted, and then the radioactivity was measured as described under "Experimental Procedures." The percent of 3 H-phospholipid efflux represents the amount of the radioactivity released in the medium divided by the total radioactivity present in the culture medium and the cell lysate. Black, white, and shaded bars represent total, cpt-cAMP-independent, and cpt-cAMP-dependent 3 H-phospholipid efflux, respectively, obtained as described in B. The numbers on top of the bars represent the phospholipid efflux relative to the WT control set to 100%. Values are the means Ϯ S.D. from two independent experiments performed in duplicate. tion of the double truncated mutant protein. Both mutations lead to a reduction in ⌬H v , suggesting a lower cooperativity of the thermal transitions for the mutants compared with WT apoA-I.
The chemical unfolding curves for WT apoA-I and the mutant proteins tested, monitored by ellipticity at 222 nm, are shown in Fig. 3B, and the conformational stability, ⌬G D 0 , and midpoint of chemical denaturation, D1 ⁄2 , and m values determined from the curves are listed in Table II Kinetics of Binding of ApoA-I to DMPC Liposomes-The kinetics of association of lipid-free WT apoA-I, amino-terminal deletion and double deletion mutants of apoA-I with DMPC multilamellar vesicles was followed by the decrease in the turbidity at 325 nm, which reflects the formation of DMPC-apoA-I complexes. These data for WT and each mutant apoA-I are shown in Fig. 4A. It was found that both the apoA-I[⌬(1-41)] mutant and the double deletion mutant apoA-I[⌬(1-41)⌬(185-243)] have a much lower rate of association with DMPC liposomes as compared with the WT apoA-I (Fig. 4A).

Formation of Discoidal POPC-ApoA-I Particles by the WT ApoA-I and the Truncated ApoA-I Form ApoA-I[⌬(1-41)⌬(185-243)]-
The sodium cholate dialysis method allowed the formation of discoidal particles with the WT and the double-truncated apoA-I form (Fig. 4, B and C). Under the negative staining conditions used, the electron micrographs showed that these particles form the typical "rouleaux," indicating that they are discoidal and have the thickness of a phospholipid bilayer. The number of rouleaux observed depends on the concentration of the sample on the carbon grid. In less concentrated samples, a large number of round particles that lie flat on the grid were also observed. The findings indicate that the double truncation  of apoA-I did not affect the ability of the central helices to form discoidal particles in vitro.
Analysis of the Ability of the Amino-terminal Deletion and Double Deletion Mutant of ApoA-I to Form HDL in Vivo-The ability of the WT and truncated mutants of apoA-I to form HDL in vivo was tested by adenovirus-mediated gene transfer in apoA-I Ϫ/Ϫ mice. Four days post-infection, plasma was collected and analyzed for plasma lipid and apoA-I levels and the presence of HDL particles. Livers were also collected and analyzed for human apoA-I mRNA levels to assess the expression of the apoA-I. Table III shows the plasma cholesterol, triglycerides, and apoA-I levels and the hepatic apoA-I mRNA levels of the mice 4 days post-infection. The apoA-I concentration was estimated by three different approaches as described under "Experimental Procedures," and the results are shown in Table III. These analyses showed that after normalization for the mRNA levels, there is still an ϳ15-fold difference between the observed and expected plasma concentration of apoA-I between the WT and the double deletion mutant. A similar difference is also observed between the WT and the amino-terminal deletion mutant. It appears that the low levels of the double deletion mutant may be the result of both faster catabolism and lower hepatic expression of the recombinant adenovirus. On the other hand, the low plasma levels of apoA-I[⌬(1-41)] can only be attributed to the faster catabolism because the hepatic apoA-I mRNA levels of this mutant are high.
Analysis of the apoA-I distribution following equilibrium density gradient centrifugation of plasma showed that the WT apoA-I and the apoA-I[⌬(1-41)] mutant were distributed predominantly in the HDL 2 and HDL 3 region (Fig. 5A), whereas the double deletion mutant apoA-I[⌬(1-41)⌬(185-243)] was shifted more toward the HDL 2 region (Fig. 5A). Similar results were obtained by analysis of the apoA-I distribution in the gel filtration fractions of plasma (data not shown).

DISCUSSION
Lipid-free apoA-I has been shown to induce efflux of phospholipid and cholesterol from cellular plasma membranes by two processes, one independent of cellular proteins (i.e. "microsolubilization") and one dependent on ABCA1 (12)(13)(14)(15)(16)(17). In this study we performed systematic deletions of helices of apoA-I to localize domains that are required for the fulfillment of these two lipid efflux functions. To differentiate ABCA1mediated and ABCA1-independent cholesterol efflux mechanisms, we performed the experiments in the presence and absence of a cAMP analog that was shown previously to be a potent inducer of ABCA1 gene expression and activity (18,58,59,61). To prove that cAMP-dependent lipid efflux represents ABCA1-mediated lipid efflux, we also investigated the ability of apoA-I variants to induce cholesterol efflux from cells that are transfected to overexpress ABCA1. For the interpretation of the data on ABCA1-independent lipid efflux it is, however, important to bear in mind that in the absence of cAMP, J774 cells have some ABCA1 activity. Thus cAMP-independent lipid efflux probably includes some ABCA1-mediated processes.
The present study shows that in cAMP-stimulated cells and  (33). The objective of the present study was to map the domain(s) of apoA-I that are required for cholesterol efflux. For this purpose we analyzed several additional apoA-I mutants containing deletions in the carboxyl-terminal region. This analysis showed that the cAMP-dependent cholesterol efflux capacity of several apoA-I variants truncated in the carboxyl-terminal region remained decreased for all the mutants that lacked the 220 -231 amino acid sequence of apoA-I. However, efflux was restored to normal levels when the 232-243 region of apoA-I was deleted. This analysis indicated that the region 220 -231 of the intact apoA-I molecule is required for the ABCA1-mediated cholesterol efflux. Interestingly, this is a domain within apoA-I that shows a much higher degree of evolutionary conservation than other domains of apoA-I (62). It has also been shown that substitution of the hydrophobic amino acids L222K, F225K, and F229K of apoA-I in this region diminishes the ability of apoA-I to associate with HDL and phospholipids in vitro (43). Furthermore, expression of this mutant in apoA-I-deficient mice by adenovirus gene transfer prevents the maturation of HDL in vivo and results in the accumulation of discoidal HDL particles in plasma (45). Similarly, it has been shown that expression of a carboxyl-terminal apoA-I[⌬(220 -243)] mutant, which lacks the 220 -231 region in apoA-I deficient mice, produced only small numbers of discoidal HDL particles in vivo (45). These data are also in agreement with the virtual HDL deficiency of a patient in whom a frameshift mutation resulted in the synthesis and secretion of a truncated apoA-I mutant with a changed sequence of amino acids 203-229 (63). Surprisingly however, the lipid efflux de-fect of apoA-I was restored when, in addition to the carboxyl terminus, the amino terminus of apoA-I was also deleted. A possible explanation of these data could be that conformational constraints brought about by the deletion of the carboxyl-terminal region diminish the functional interactions between apoA-I and ABCA1. Further deletion of both the amino-and carboxyl-terminal regions may relieve these conformational constraints to allow the double deletion mutant of apoA-I to interact functionally with the ABCA1 and thus promote cholesterol efflux from cell cultures.
To gain further insight into the relationship between the structure and function of apoA-I, we studied the physicochemical properties of the double deletion mutant of apoA-I and compared them with those of WT apoA-I and the amino-or the carboxyl-terminal deletion mutants. Structural analysis showed that neither deletions of amino-terminal amino acid residues 1-41 performed in this study nor of the carboxylterminal amino acid residues 185-243 studied previously (44) changed significantly the ␣-helical content. The carboxyl-terminal truncation did not interfere with the stability of the mutant protein (44). On the other hand, based on chemical denaturation data the amino-terminal truncation destabilizes the protein.
The functional studies showed that the carboxyl-terminal deletion diminished greatly the ABCA1-mediated lipid efflux, as well as the ability to solubilize multilamellar DMPC phospholipid vesicles (43). In contrast, the amino-terminal deletion has only small effect on lipid efflux and solubilizes with a slower rate multilamellar DMPC phospholipid vesicles. These data indicate that the carboxyl-terminal domain of apoA-I is required to enable the WT apoA-I to induce cAMP-and hence ABCA1-dependent lipid efflux from cells and to solubilize multilamellar DMPC phospholipid vesicles, whereas the aminoterminal domain has a smaller effect on both processes.
The current study also showed that the apoA-I[⌬(1-41)⌬(185-243)] mutant which lacks both the amino-and the carboxyl-terminal domain shows a ϳ7% reduction in its ␣helical content and has greatly reduced chemical and thermal stability and cooperativity of unfolding as compared with WT apoA-I. When tested in vitro and in vivo, the double deletion mutant has the ability to form discoidal HDL particles. Physicochemical analysis of rHDL particles containing the apoA-I[⌬(1-41)⌬(185-243)] mutant showed that the lipid-bound form of this mutant has 0.4 kcal/mol increased chemical stability compared with the lipid-free form (data not shown). Thus the lipid-bound form of this mutant appears to be thermodynamically more favorable than the lipid-free form. This may explain the partial ability of this mutant to solubilize multilamellar DMPC vesicles and to promote ABCA1-mediated lipid ] expressing adenoviruses was adjusted to 2 ml with PBS, and the lipoproteins were separated on 10 -20% sodium bromide gradients. The apoA-I in the fractions was analyzed by Western blot, as described under "Experimental Procedures." B-E, electron photomicrographs of apoA-I-containing fractions. The indicated fractions obtained by equilibrium density gradient centrifugation shown in A, which contain the highest concentration of apoA-I, were examined by electron microscopy as described under "Experimental Procedures." The photomicrographs were taken at ϫ75,000 magnification and enlarged 3 times. The fractions from the plasma of mice expressing WT apoA-I were diluted 3 times.

TABLE III
Comparison of plasma cholesterol, triglycerides, and apoA-I levels and hepatic mRNA levels of apoA-I Ϫ/Ϫ mice prior to or 4 days post-infection with recombinant adenoviruses expressing the WT apoA-I or truncated forms of apoA-I Plasma obtained from apoA-I Ϫ/Ϫ mice prior to or after infection with 1 ϫ 10 9 plaque-forming units of adenoviruses expressing WT apoA-I or the indicated truncated forms of apoA-I, 4 days post-infection, were analyzed for cholesterol and triglyceride levels, as well as apoA-I levels (determined by ELISA using as standard either human serum or purified lipid-free WT and mutant apoA-I forms or by immunoblotting) as described under "Experimental Procedures." Liver human apoA-I mRNA levels were determined by Northern blot analysis as described under "Experimental Procedures." The mRNA levels are expressed relative to that in mice infected with WT apoA-I virus. Values are the means Ϯ S.D. (n ϭ 3). Previous studies (19,22) have indicated that ABCA1 is not directly involved in the efflux of free cholesterol. Other groups have proposed that the interaction of apoA-I with ABCA1 or a partner protein stimulates translocation of intracellular cholesterol and phospholipids from the Golgi apparatus to the plasma membrane ABCA1, where it associates with apoA-I to form the precursors HDL particles, which subsequently mature to spherical HDL (61,64). One study has suggested that apoA-I-ABCA1 complexes may recycle between plasma membrane and intracellular lipid pools where apoA-I may be lipidated by the action of ABCA1 and subsequently exocytosed (61,64,65). Our data show that the ABCA1-dependent cholesterol and phospholipid effluxes follow similar patterns, suggesting that the two processes are kinetically indistinguishable.
It was observed previously (13,27,66) that water-soluble apolipoproteins as well as artificial peptides with amphipathic ␣-helices induce cholesterol efflux from cells. Because the number of amphipathic ␣-helices was correlated roughly with the efficacy of cholesterol efflux, it was believed that apoA-I does not contain stringent structural domains that are important for this process. The impaired ABCA1-mediated cholesterol efflux of apoA-I variants with carboxyl-terminal deletions does not seem to be a consequence of alterations in the structure of apoA-I. Thus, it is reasonable to assume that stringent structure-function relationships between apoA-I and ABCA1 are required to enable apoA-I to induce cAMP-and hence ABCA1dependent lipid efflux from cells. The present study indicates that these specific functional interactions between the fulllength apoA-I and ABCA1 require the carboxyl-terminal residues 220 -231 of apoA-I.
Carboxyl-terminal truncation interfered much less with the ability of apoA-I to induce cAMP-independent efflux (Fig. 2B). Moreover, and in contrast to the cAMP-dependent lipid efflux, further removal of the amino terminus did not restore the defect in the double deletion mutant but rather reduced further the cAMP-independent cholesterol efflux. This ABCA1-independent cholesterol efflux pathway, i.e. microsolubilization, appears to be largely a function of the apolipoprotein content of amphipathic ␣-helices rather than of a specific domain. Because the number of amphipathic ␣-helices is a major determinant of the phospholipid binding capacity of apoA-I, these data also indirectly support the previous notion that physicochemical interactions between apoA-I and plasma membrane phospholipids rather than biological ligand/receptor interactions may underlie "microsolubilization" (12,13).
Our in vivo data suggest that the specific, i.e. cAMP-and ABCA1-dependent lipid efflux pathway, and the nonspecific, i.e. cAMP-and ABCA1-independent lipid "microsolubilization" pathway, have different impacts on HDL biogenesis. WT apoA-I or apoA-I mutants with a preserved carboxyl terminus such as apoA-I[⌬(1-41)] maintain their capacity to induce ABCA1-mediated cholesterol efflux and to form spherical HDL in vivo. Previous studies also showed that the apoA-I[⌬(7-43)] mutant forms spherical particles and few discoidal particles, whereas the apoA-I[⌬(7-65)] mutant and various internal deletion mutants form discoidal HDL particles (33,67). By contrast, apoA-I variants with carboxyl-terminal truncations such as apoA-I[⌬(220 -243)], which lacks the 220 -231 domain, interfere with ABCA1-mediated cholesterol efflux and, although not strongly interfering with ABCA1-independent cholesterol efflux, do not form HDL in vivo (45). The inability of the carboxyl-terminal truncated apoA-I variants to form HDL most likely reflects their limited capacity to promote cholesterol efflux and form pre-␤-HDL precursors, as well as their reduced affinity for phospholipids and pre-formed HDL particles. Previous studies have shown that carboxyl-terminal truncated apoA-I forms fail to associate with preformed HDL particles following density gradient ultracentrifugation (43). Similarly, carboxyl-terminal truncated apoA-I forms, produced by proteolysis of HDL by metalloproteinases secreted by macrophage cultures, are found in lipid-free fractions after gel filtration (68). Thus, the lipid-poor particles or lipid-free mutant apoA-I forms generated by adenovirus-mediated gene transfer most likely are catabolized rapidly by the cubilin receptor or other mechanisms (69,70). This may explain our findings that indicate that the steady-state apoA-I mRNA levels of the WT and mutant apoA-I forms do not correlate with the plasma apoA-I levels. The combined deletion of the carboxyl and amino termini partially restores the inability of the carboxyl-terminal deletion variant to induce ABCA1-dependent cholesterol efflux and to form lipidated discoidal HDL precursors, despite the fact that this variant has an impaired ability to release cellular lipids by cAMP-independent mechanisms. We therefore conclude that not apolipoprotein-mediated lipid efflux per se, but rather the mechanism of lipid efflux which involves functional interactions between ABCA1 and apoA-I, is essential for the formation of HDL precursors which can mature to HDL. This process has at least two prerequisites, namely the cellular abundance of ABCA1 and the presence of a full-length apoA-I with an intact carboxyl-terminal domain.
Overall, our findings suggest that the central region of apoA-I alone, which contains helices 3-7, has the capacity to promote ABCA1-mediated lipid efflux. Following acceptance of cholesterol and phospholipids, the lipidated particles containing the central apoA-I helices 3-7 may proceed to form discoidal particles in vivo.
When carboxyl-terminal segments that contain residues 220 -231 are deleted, the apoA-I cannot function in ABCA1mediated cholesterol efflux. This may lead to the formation of unstable HDL precursors that may be catabolized rapidly by the kidney.
It is possible that deletion of the carboxyl-terminal domain hinders the functional interactions of the mutant apoA-I with ABCA1, whereas this hindrance may be limited when both the amino-and carboxyl-terminal domains are deleted.
The 220 -231 domain of apoA-I has a smaller effect on the ability of apoA-I to induce ABCA1-independent efflux, i.e. microsolubilization, of cellular lipids.
The restoration of ABCA1 lipid efflux in the apoA-I[⌬(232-243)] mutant, which contains the 220 -231 region, as well as in the double deletion mutant, which lacks both the amino-and the carboxyl-terminal domains, indicates that specific interactions of the 220 -231 region with amino-terminal domain of apoA-I may be necessary to allow functional interactions be-tween the full-length apoA-I and ABCA1 that are required for cholesterol efflux and HDL biogenesis.