The Carboxyl-terminal Hydrophobic Residues of Apolipoprotein A-I Affect Its Rate of Phospholipid Binding and Its Association with High Density Lipoprotein*

We performed a series of mutations in the human apolipoprotein A-I (apoA-I) gene designed to alter specific amino acid residues and domains implicated in lecithin:cholesterol acyltransferase (LCAT) activation or lipid binding. We used the mutant apoA-I forms to establish nine stable cell lines, and developed strategies for the large scale production and purification of the mutated apoA-I proteins from conditioned media. HDL and dimyristoyl phosphatidylcholine binding assays using the variant apoA-I forms have shown that replacement of specific carboxyl-terminal hydrophobic residues Leu222, Phe225, and Phe229 with lysines, as well as replacement of Leu211, Leu214, Leu218, and Leu219 with valines, diminished the ability of apoA-I to bind to HDL and to lyse dimyristoyl phosphatidylcholine liposomes. The findings indicate that Leu222, and Phe225, Phe229 located in the putative random coil region, and Leu211, Leu214, Leu218, and Leu219 located in the putative helix 8, are important for lipid binding. In contrast, substitutions of alanines for specific charged residues in putative helices 7, 8, or 9 as well as various point mutations in other regions of apoA-I, did not affect the ability of the variant apoA-I forms to bind to HDL or to lyse dimyristoyl phosphatidylcholine liposomes. Cross-linking experiments confirmed that the carboxyl-terminal domain of apoA-I participates in the self-association of the protein, as demonstrated by the inability of the carboxyl-terminal deletion mutants Δ185–243 and Δ209–243 to form higher order aggregates in solution. Lecithin:cholesterol acyltransferase analysis, using reconstituted HDL particles prepared by the sodium cholate dialysis method, has shown that mutants (Pro165 → Ala,Gln173 → Glu) (Leu311 → Val,Leu214 → Val,Leu318 → Val,Leu319 → Val), Leu222 → Lys,Phe255 → Lys,Phe290 → Lys) and Δ209–243 reduced LCAT activation (38–68%). Mutant (Glu191 → Ala,His195 → Ala,Lys196 → Ala) enhanced LCAT activation (131%), and mutant (Ala162 → Leu,Leu189 → Trp) exhibited normal LCAT activation as compared with the wild type proapoA-I and plasma apoA-I forms. The apparent catalytic efficiency (V max(app)/K m (app)) of the apoA-I mutants ranged from 17.8 to 107.2% of the control and was the result of variations in both the K m and theV max in the different mutants. These findings indicate that putative helices 6 and 7, and the carboxyl-terminal helices 8 and 9 contribute to the optimum activation of lecithin:cholesterol acyltransferase. In addition to their use in the present study, the variant apoA-I forms generated will serve as valuable reagents for the identification of the domains and residues of apoA-I involved in binding the scavenger receptor BI, and facilitating cholesterol efflux from cells as well as aid in the structural analysis of apoA-I.

that putative helices 6 and 7, and the carboxyl-terminal helices 8 and 9 contribute to the optimum activation of lecithin:cholesterol acyltransferase. In addition to their use in the present study, the variant apoA-I forms generated will serve as valuable reagents for the identification of the domains and residues of apoA-I involved in binding the scavenger receptor BI, and facilitating cholesterol efflux from cells as well as aid in the structural analysis of apoA-I.
Apolipoprotein A-I (apoA-I) 1 is the major protein constituent of HDL and plays an important role in HDL stability, lipid transport, and metabolism (1). As a component of HDL, apoA-I is the principal physiological activator of lecithin:cholesterol acyltransferase (LCAT), particularly with physiological lecithins (2,3). Reconstituted HDL (rHDL) particles, formed in vitro by mixing apoA-I with phospholipid-cholesterol vesicles, serve as substrates of LCAT and are converted into cholesteryl ester containing spheres upon incubation with LCAT (4). It was shown that rHDL particles have different sizes and that their LCAT activation ability correlates with the conformation of apoA-I on these particles (5)(6)(7). ApoA-I also promotes the efflux of cholesterol from peripheral cells, thus providing a substrate for the LCAT reaction (8). Finally, apoA-I plays a role in receptordependent or receptor-independent binding of HDL to cell surfaces (9 -12). Such binding may contribute to either cholesterol efflux (8) or selective lipid uptake (13). As a result of these activities, apoA-I may play an important role in regulating the cholesterol content of peripheral tissues through the reverse cholesterol transport pathway (4,8,14).
Amino acid (15) and nucleotide (16,17) sequence analyses of apoA-I have shown that both the protein and the gene contain repeated units. The protein contains 22-or 11-residue repeated units, which are organized into amphipathic ␣-helices (15)(16)(17). Models of secondary structure have been proposed that predict the presence of 8 or 9 helical regions in the apoA-I molecule (18,19). The model proposed by Atkinson predicts the existence of 9 antiparallel helices and is consistent with the formation of an antiparallel ␣-helical bundle structure in solution (18).
Several studies have examined the ability of synthetic peptides and fragments of apoA-I to associate with phospholipid vesicles and to activate LCAT (20 -22). Moderate reduction in LCAT activation in the range of 40 -70% of normal was ob-served for a few naturally occurring mutants (23)(24)(25)(26)(27)(28), whereas more substantial reduction was observed with deletion mutants in one or more of the apoA-I helices (29 -31). These results, as well as studies using monoclonal antibodies (32,33), suggest that several domains within the central region of apoA-I, between residues 95 and 185 are important for the activation of LCAT. These domains contain the putative helices 6 and 7 and the hinged domain, which includes the putative helices 4 and 5. The carboxyl-terminal region of apoA-I (residues 190 -243) has been shown to be important for lipid and HDL binding as well as for binding to cell membranes (12, 29 -31, 34, 35).
Although the data obtained with the carboxyl-terminal apoA-I mutants are informative, mapping of the apoA-I domains that are functionally important requires more precise point mutagenesis that disturbs minimally the three-dimensional structure of apoA-I. In the present study we describe the generation and characterization of novel apoA-I variants. Functional analysis established that specific hydrophobic residues in the putative loop region (223-231) and within putative helix 8 (187-223) are important for binding to HDL and for the initial association of apoA-I with multilamellar phospholipid vesicles. A relatively small reduction in LCAT activation was observed for several point mutants as a result of variations in both the apparent K m and V max .

Materials
The Klenow fragment of DNA polymerase I, T4 ligase, polynucleotide kinase, and restriction enzymes were purchased from New England Biolabs. Calf intestinal alkaline phosphatase was from Stratagene (La Jolla, CA). [ 35 S]Methionine (Ͼ1000 Ci/mmol) and [ 14 C]cholesterol (45-60 mCi/mmol) were from NEN Life Science Products. The Sequenase sequencing kit was from U. S. Biochemical Corp. Materials for the polymerase chain reaction were from Perkin-Elmer. Bactotryptone and Bacto-yeast extract were from VWR (Pittsburgh, PA). Dulbecco's modified Eagle's medium (DMEM) and methionine-free DMEM were from Life Technologies, Inc. Materials for the two-dimensional polyacrylamide gel electrophoresis were described previously (36). IgSorb was from the Enzyme Center (Boston, MA). BSA, POPC, DMPC, sodium cholate, aprotinin, and benzamidine were from Sigma. Materials for oligonucleotide synthesis were from Applied Biosystems.

Methods
Mutagenesis and Plasmid Construction-The oligonucleotides utilized as primers for the in vitro mutagenesis of the apoA-I gene were synthesized by the solid-phase phosphate triester method using an Automated Oligonucleotide Synthesizer (Applied Biosystems, model 380-B), according to the instructions provided by the manufacturer. The oligonucleotides were purified by electrophoresis on 20% polyacrylamide, 7 M urea gels.
The PstI-PstI fragment of the apoA-I gene was ligated to SalI linkers and inserted into the SalI polylinker site of the pUC19 vector. This new plasmid designated pUCA-I N (37) was mutagenized by polymerase chain reaction to introduce two new restriction sites, in intron 3 and at the 3Ј end of the apoA-I gene: a NotI site was introduced at nucleotides 1166 -1173, and an XhoI site was introduced at nucleotides 2191-2196 of the pUCA-I N plasmid, creating plasmid pUCA-I N *.
The SalI-SalI fragment encompassing the apoA-I gene was excised from the pUCA-I N * plasmid and inserted into the unique XhoI site of the pBMT3X vector, thus placing the apoA-I gene under the control of the mouse metallothionine I promoter. This new vector, designated pBMT3X-AI, was digested with NotI and XhoI and was used to replace the normal with the corresponding mutated apoA-I segment, as described below.
The fourth exon of the human apoA-I gene was amplified and mutagenized by polymerase chain reaction, using a set of specific mutagenesis primers, containing the mutation of interest, and a set of flanking universal primers, containing the restriction sites NotI and XhoI, using the pUCA-I N * vector as a template (38). 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. The variant apoA-I sequences were verified by DNA sequencing.
Generation of Stable Cell Lines-The C127 cell line (ATCC CRC 1616) was maintained in DMEM supplemented with 10% fetal calf serum (Sigma), and grown at 37°C, in 5% CO 2 . This cell line is a suitable host for transfection with the bovine papilloma virus-containing plasmids, described above.
To generate stable cell lines expressing the apoA-I variants, cells were transfected by the calcium chloride co-precipitation method (39). After selection for 10 -15 days in media containing 10 M CdCl 2 , surviving colonies were isolated with cloning cylinders. For protein labeling, 60-mm diameter cell cultures were rinsed twice with methioninefree DMEM containing 2 mM glutamine and 10 M CdCl 2 , and preincubated in the same media for 2 h. After two additional rinses, the cells were labeled overnight with 0.25 mCi of [ 35 S]methionine. Media were collected, immunoprecipitated with rabbit anti-human apoA-I antibodies, and analyzed by two-dimensional SDS-PAGE and autoradiography as described (40).
Large Scale Growth of Cell Cultures-Cell clones overproducing the variant apoA-I forms were grown for 5-7 days in two T-75 flasks, containing DMEM plus 10% fetal calf serum and 10 M CdCl 2 . Confluent flasks were trypsinized, and the cells were placed into 850-cm 2 roller bottles. The bottles were rotated at 1 rpm, to allow the cells to attach to the surface of the bottle. When the cells had grown to 80 -85% confluence, 10 ml of packed Verax microcarriers were added and the speed of rotation was increased to 7 rpm. Prior to use the microspheres were autoclaved in phosphate-buffered saline solution, and subsequently incubated in media containing 1% fetal calf serum.
In the roller bottle system, the cells grow on 100-m diameter collagen-coated porous lead microspheres. The cells attach to the collagen matrix and grow both on the surface and inside the beads. The porous nature of the microspheres greatly increases the surface area available to the cells allowing very high cell density. The constant rotation of the bottles, combined with the increase in the surface area, allows gas exchange to take place at an increased rate.
The cells were fed twice a week with 300 ml of DMEM medium containing 5% fetal calf serum and 10 M CdCl 2 . For protein purification, the cells were rinsed twice with serum-free medium, and preincubated in the same medium for 2 h. After one additional rinse, the cells were incubated overnight in serum-free media. The conditioned medium was collected and stored in 1 mM EDTA, 0.01% NaN 3 , 10 M aprotinin, and 10 M benzamidine. Collection of media was repeated every 3-4 days.
Purification of Variant ApoA-I Forms-Medium (1 liter) collected from the roller bottles was concentrated down to 50 ml using an Amicon Ultrafiltration Cell and membranes with a 10,000 molecular weight cut-off. The concentrated medium was then dialyzed against 0.01 M Tris, pH 8, filtered, and passed through a DEAE HiTrapQ column (Pharmacia Biotech Inc.), which had been equilibrated with the same buffer. The protein was then eluted with a step gradient (5% 3 20% 3 50% 3 100%) of 1 M NH 4 CO 3 in the Tris buffer. 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 Centricon 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 (Pharmacia), at a rate of 0.1 ml/min, and eluted with one column volume (320 ml) of 0.15 M NH 4 CO 3 , 0.02% NaN 3 buffer at a rate of 0.1-0.5 ml/min. The purity of the apoA-I preparation was assessed by SDS-PAGE. Fractions greater than 95% pure were recovered. The concentration of the apolipoprotein was determined by the absorbance at 280 nm and an extinction coefficient of 1.15 mg Ϫ1 cm 2 .
Flotation Properties of the ApoA-I Variants-Cell cultures expressing the variant apoA-I proteins were labeled with [ 35 S]methionine in methionine-free medium as described (41). A 2-ml aliquot of the culture medium was adjusted to a density of 1.21 g/ml with 0.65 g of potassium bromide, placed in a cellulose nitrate tube, mixed with 100 g of human HDL, and overlaid sequentially with 1.75 ml of a potassium bromide solution of d ϭ 1.15 g/ml, 3 ml of each of potassium bromide solutions of d ϭ 1.063 g/ml and d ϭ 1.019 g/ml, followed by normal saline.
The tubes were then centrifuged in a Beckman SW41 rotor at 35,000 rpm for 22 h. After centrifugation, 12 1-ml fractions were collected from the top of the tube using a Haake/Buchler fraction collector. The samples were then extensively dialyzed against a 1 mM solution of cold methionine, dried up in a Speed-Vac, resuspended in one-dimensional SDS-PAGE sample buffer, and analyzed by one-dimensional SDS-PAGE and autoradiography. To quantify the 35 S-labeled apolipoproteins, the protein bands of the one-dimensional gels were excised and solubilized in 2.5 ml of 30% (w/v) H 2 O 2 , at 60°C in scintillation vials. The solubilized acrylamide was then mixed with 15 ml of scintillation fluid and counted in an LKB scintillation counter.
Binding of ApoA-I to DMPC Liposomes-The binding of the wild type and the various mutant forms of apoA-I to DMPC multilamellar liposomes was studied by kinetic-turbidimetric methods (42). DMPC, dissolved in a glass-distilled chloroform:methanol (2:1) solution, was placed in a glass tube. The sample was dried under nitrogen, and the appropriate amount of a 5 mg/ml solution of apoA-I in buffer (10 mM Tris-HCl, pH 8, 150 mM NaCl, 1 mM NaN 3 , 0.01% EDTA) was added to it, to give a final DMPC:apoA-I ratio of 2.5:1 (w/w) at a final protein concentration of 0.2 mg/ml. The experiment was performed at 24°C, and the absorbance at 325 nm was monitored at 5-min intervals using a Perkin-Elmer Lambda 3A spectrophotometer.
ApoA-I Cross-linking-The optimal concentration for cross-linking lipid-free apoA-I was 1.2-2.0 mg/ml. Prior to cross-linking the sample was dialyzed overnight in phosphate buffer (0.02 M sodium phosphate, 0.9% NaCl, 0.01% EDTA, 0.02% NaN 3 , pH 7.4) using dialysis tubing with a molecular weight cut-off of 3,500. An aliquot of 100 l of each sample was placed into Eppendorf tubes and 50 l of 10 mM bis(sulfosuccinimidyl)suberate (BS 3 ) solution was added. The solution was vortexed, covered loosely, and incubated at 4°C, for 3.5 h. At the end of the incubation time, 10 l of 250 mM ethanolamine quenching solution was added to the reaction mixture. The solution was then concentrated to 1/4 of its original volume, using a Speed-Vac, and run on an 8 -25% gradient Phast gel apparatus (Pharmacia) using SDS buffer strips, and electrophoresed for 76 V-h.
Preparation of rHDL-The rHDL complexes were prepared by the original sodium cholate dialysis method (43) using a molar ratio of 100:10:1:100 of POPC:cholesterol:apoA-I:sodium cholate. In a typical experiment, 0.14 mg of cholesterol (5,000 -7,000 cpm of 14 C-labeled cholesterol/nmol of cold cholesterol) and 2.71 mg of POPC were placed in glass tubes, vortexed gently, and dried under nitrogen. The dried lipid was dissolved in a 10 mM Tris-HCl, pH 8, 150 mM NaCl, 1 mM NaN 3 , and 0.01% EDTA buffer by repeated vortexing. The suspension was stored on ice for 1 h, sodium cholate was added, and the solution was kept on ice for 1 h more. Finally the apoA-I was added, and the incubation was continued for another 1 h. Sodium cholate was removed by dialysis at 4°C against 5-6 liters of the 10 mM Tris-HCl, pH 8, 150 mM NaCl, 1 mM NaN 3 , and 0.01% EDTA buffer, using membranes with a molecular weight cut-off of 12,000 -14,000. The rHDL particles were analyzed on a native 8 -25% acrylamide gradient gel at 15°C on a Pharmacia Phast-gel system. The rHDL particles were stored at 4°C in nitrogen to prevent the oxidation of lipids.
Electron Microscopy-rHDL particles prepared with different variant apoA-I forms were negatively stained with potassium phosphotungstate using carbon-coated grids, and photographed with a Philips CM12 electron microscope (Philips Electron Optics, Eindhoven, The Netherlands).
LCAT Assay-The human LCAT enzyme was purified from normal plasma by ultracentrifugal flotation, followed by chromatography on Affi-Gel Blue, DEAE-Sepharose, and Phenyl-Sepharose columns as described (44). For LCAT analysis, the substrate rHDL particle was diluted in buffer (10 mM Tris-HCl, pH 8, 150 mM NaCl, 1 mM NaN 3 , and 0.01% EDTA), to give final apoA-I concentrations ranging from 10 Ϫ7 to 10 Ϫ6 M. Each reaction mixture contained 50 l of a 40 mg/ml BSA, 20 l of 100 mM ␤-mercaptoethanol, and 50 l of a 1-2 g/ml LCAT solution.
The reaction was carried out for 30 min at 37°C.

Characterization of Secreted Wild Type and Variant ApoA-I Forms-
To characterize the mutant apoA-I forms, cell clones expressing the mutant apoA-I genes (Table I) were labeled with [ 35 S]methionine, immunoprecipitated, and analyzed by twodimensional PAGE and autoradiography, using the plasma wild type apoA-I as an internal marker. The Coomassie-stained gel obtained from this analysis shows the position of the plasma apoA-I isoforms that were included in the sample, and the autoradiogram shows the position of the newly synthesized apoA-I. Superimposition of the gel on the autoradiogram establishes the charge and size differences between the plasma apoA-I and the newly synthesized variant apoA-I forms (37). Thus, using this analysis, mutants with charge or size differences from the wild type can be distinguished unequivocally. In panel A, the secreted radiolabeled variant form is smaller in size, and is two charge units more negative than the wild type proapoA-I (isoform 2). This is consistent with the net loss of two positive charges; the deleted sequence contains 8 negatively charged and 10 positively charged residues. The change in size is consistent with the deletion of 58 amino acids. In panel B, the secreted radiolabeled variant form is smaller in size, and is one charge unit more positive than the wild type proapoA-I. This is consistent with the net loss of one negative charge; the deleted sequence contains five negatively charged and four positively charged residues. The change in size is consistent with the deletion of 35 amino acids. In panel C, the secreted radiolabeled variant form has three additional positive charges as compared with the corresponding wild type proapoA-I. This is consistent with the acquisition of three positive charges due to the replacement of three neutral residues with three positively charged lysines. The additional acidic forms with the same M r present on the gel are the result of deamidation or carbamylation of the major form, as described previously (45). The more basic, higher M r isoform observed in this panel and in panels E, F, G, H, and I has been observed previously and represents the pre proapoA-I form (37). The acidic higher M r isoforms of the newly secreted apoA-I present in this panel have been observed previously and may represent post-translationally modified forms of unknown  nature (37,40,46). In panels D, G, and I, the secreted radiolabeled variant forms overlap completely with the wild type plasma proapoA-I. This is consistent with the lack of change in the total charge of these variant proteins due to the mutations.
In panel H, the secreted radiolabeled variant form has one additional positive charge as compared with the wild type proapoA-I. This is consistent with the net loss of one negative charge due to the mutation. In panels E and F, the secreted radiolabeled variant forms have one additional negative charge as compared with the corresponding wild type proapoA-I form. This is consistent with the acquisition of one extra negative charge, due to the mutations.

Isolation of the Wild Type and Variant ApoA-I Forms from
Conditioned Medium-To obtain large quantities of apoA-I protein, the permanent cell lines expressing the different apoA-I forms were grown on a large scale using collagen-coated microspheres in roller bottles.
For protein purification, the cells were preincubated in 100 ml of serum-free media overnight. Approximately 1-1.5 liters of media thus collected was concentrated and passed through an anion exchange column (HiTrap-Q) on an FPLC, to remove the majority of the residual BSA. Fractions containing the majority of the apoA-I protein with the least amount of contaminants were pooled and concentrated to a small volume (1-2 ml) and chromatographed on a HiPrep Sephacryl S-200 column. Protein purity was 95-99% as determined by SDS-PAGE. Typical purification profiles of the different apoA-I mutants are shown in Fig. 2 (A-D).
Binding of the Variant ApoA-I Forms to HDL-Previous studies have demonstrated that when soluble, lipid-free apoA-I is incubated with excess unlabeled plasma HDL, the exogenous apolipoprotein is incorporated into HDL by displacing another apoA-I molecule from the HDL surface (37). It has also been shown that radiolabeled apoA-I secreted by C127 cell lines can be incorporated into HDL particles, but in the absence of added HDL recombinant apoA-I expressed in C127 cells is distributed mostly in the lipoprotein-free fraction (d Ͼ 1.21 g/ml) (47).
The ability of the variant apoA-I forms to bind to HDL was analyzed by KBr density gradient ultracentrifugation (41). The distribution of the radiolabeled recombinant variant and wild type apoA-I in different lipoprotein fractions, relative to the plasma apoA-I, is shown in Fig. 3 (A-I). The location of the mutations, in the model proposed by Nolte and Atkinson (18), described in this and subsequent figures is shown in panel J. The density distribution of the recombinant wild type apoA-I was similar to that of plasma apoA-I, suggesting an equilibrium between the recombinant and the plasma wild type forms. The majority of this recombinant protein floats in the HDL region, between densities 1.06 and 1.21 g/ml.
Deletion of the putative helices 8 and 9 (⌬185-243), or part of helix 8 and helix 9 (⌬209 -243), affected dramatically the flotation properties of the mutant proteins. The majority of the protein was recovered in the non-lipoprotein fraction (d Ͼ 1.21 g/ml) (Fig. 3, A and B). The findings confirm previous findings (29) that the carboxyl-terminal region of apoA-I is necessary for its ability to bind to the lipoprotein surface.
Point mutations were used to identify the specific residues within the carboxyl terminus of apoA-I that are involved in HDL binding. Substitution of alanines for specific charged residues (Glu 191  Note that in panel A, the newly synthesized apoA-I is smaller in size, and is two charge units more negative than the wild type proapoA-I, consistent with the net loss of two positive charges. In panel B, the newly synthesized apoA-I is smaller in size, and is one charge unit more positive than the wild type proapoA-I, consistent with the net loss of one negative charge. In panel C, the newly synthesized apoA-I has three additional positive charges as compared with the corresponding wild type proapoA-I, confirming the replacement of three non-charged residues with three positively charged lysines. In panels D, G, and I, the newly synthesized apoA-I overlaps completely with the plasma proapoA-I, confirming the lack of change in the total charge, due to the mutations. In panel E, the newly synthesized apoA-I has one additional negative charge as compared with the wild type proapoA-I, confirming the net loss of one negative charge due to the mutation. In panel F, the newly synthesized apoA-I has one additional negative charge as compared with the corresponding wild type proapoA-I form, confirming the acquisition of one extra negative charge, due to the mutations. cial for the ability of the protein to associate normally with HDL ( Fig. 3, E, H, and I).
However, substitution of lysines for three specific hydrophobic residues (Leu 222 , Phe 225 , and Phe 229 ) located in the predicted random coil region between putative helices 8 and 9, altered dramatically the ability of the mutant protein to bind to HDL. As shown in Fig. 3C, the majority of the mutant protein is recovered in the lipoprotein-free fraction (d Ͼ 1.21 g/ml), indicating that these specific hydrophobic residues are critical for the ability of the protein to bind HDL.
Similarly, replacement of a series of leucine residues (Leu 211 , Leu 214 , Leu 218 , and Leu 219 ) in putative helix 8 by valines, which have similar hydrophobicity but are less bulky, resulted in the same dramatic alteration in the flotation properties of the variant protein (Fig. 3D), indicating the importance of these residues for the binding of apoA-I to HDL.
Substitution of alanine for proline between putative helices 6 and 7 and the alteration of a neighboring glutamine to glutamate did not affect the binding of the variant apoA-I to HDL (Fig. 3F). These mutations were predicted to alter the orientation of these helices due to the elimination of the helix-breaking proline residue, and the change of the predicted A type halfrepeat to a B type half-repeat (18).
Finally, a mutation in helix 6, which introduces amino acids at positions that are expected to disrupt the formation of the helical bundle in solution (Ala 152 3 Leu, Leu 159 3 Trp) (48,49) does not affect the flotation properties of the mutant protein (Fig. 3G), suggesting that these substitutions do not affect the binding of apoA-I to HDL.
Lipid Binding Properties of the Variant ApoA-I Forms-DMPC binding experiments were performed to assess the effect of the mutations on the kinetics of interaction of apoA-I with DMPC multilamellar vesicles. The rate of the interaction was monitored by the change in absorbance at 325 nm. The experiments were performed at 24°C, the transition temperature of the lipid, where the gel and liquid-crystalline phases co-exist, and where defects in the lipid matrix make it easier for apoA-I to interact.
As illustrated in Fig. 4 (A-G), both plasma and recombinant wild type proapoA-I bind and solubilize DMPC rapidly, as indicated by the dramatic decrease in turbidity of the DMPC dispersions. In contrast, the apoA-I mutants in which the putative helices 8 and 9 (⌬185-243), or part of helix 8 and helix 9 (⌬209 -243) were deleted, interacted extremely slowly with the phospholipid (Fig. 4, A and B). After incubation at 24°C for 8 h, the reaction mixture was adjusted to a density of 1.21 g/ml and was fractionated by ultracentrifugation. Approximately 85% of the wild type apoA-I was recovered in the top fraction, bound to the DMPC, whereas only approximately 20% of the deletion mutant (⌬209 -243) protein was found in the top fraction (data not shown).
The same slow kinetics of interaction were observed for the mutant in putative helix 8, where four valines were substituted for leucines (Leu 211 3 Val,Leu 214 3 Val,Leu 218 3 Val,Leu 219 3 Val) (Fig. 4D). It is worth noting that even when the ratio of protein to lipid was doubled (2.5:2, DMPC:protein) the reaction was still extremely slow (data not shown).
Similarly slow kinetics of interaction with DMPC were obtained for the mutant in which hydrophobic residues in the predicted random coil region between helices 8 and 9 were changed to charged residues (Leu 222 3 Lys,Phe 225 3 Lys, Phe 229 3 Lys) (Fig. 4C). This variant associates initially with the phospholipid at a rate that is slightly higher than the rate of association of the mutants of Fig. 4 (A, B, and D). However, the rate of association is still slow compared with that observed for the plasma and the recombinant wild type apoA-I.
In contrast to the slow rate of solubilization of multilamellar vesicles of DMPC observed in apoA-I mutants with substitutions of hydrophobic carboxyl-terminal amino acids, other point mutants in the central and carboxyl-terminal region of the molecule do not affect the ability of the protein to solubilize efficiently the multilamellar vesicles of DMPC. Fig. 4 (E-G) shows the kinetics of interaction of the mutant apoA-I forms (Glu 191 3 Ala,His 193 3 Ala,Lys 195 3 Ala), (Pro 165 3 Ala,Gln 172 3 Glu), and (Ala 152 3 Leu,Leu 159 3 Trp) with DMPC. All of these mutants interact spontaneously and solubilize DMPC with a rate comparable to those of the recombinant wild type proapoA-I.
The results obtained with the DMPC binding assay are in agreement with the results from the HDL binding assay and indicate that those mutants (both the deletion and the point mutations) that lost their ability to associate normally with the HDL, also displayed slow kinetics of association with the multilamellar vesicles of DMPC. The findings show that the carboxyl terminus of apoA-I is important for lipid binding, and indicate for the first time that specific hydrophobic residues in the predicted random coil region between putative helices 8 and 9, as well as the leucine residues present in putative helix 8, are critical for the ability of the protein to bind to both the HDL surface as well as to bind and solubilize multilamellar vesicles of DMPC.
Self-association Properties of the Variant ApoA-I Forms-To study the self-association properties of the mutants in comparison with the plasma apoA-I protein, we performed cross-link- ing experiments using the cross-linking reagent BS 3 .
Plasma apoA-I formed dimers, trimers, and tetramers in addition to monomers, when it was present in solution at a concentration of 1.5 mg/ml (Fig. 5). In contrast, both of the carboxyl-terminal deletion mutants (⌬185-243) and (⌬209 -243) formed predominantly monomers, and a few dimers. The formation of the dimer was concentration-dependent, but even at concentrations as high as 2 mg/ml, the dimer represented only a minor component.
The mutants ( The variant (Leu 222 3 Lys,Phe 225 3 Lys,Phe 229 3 Lys) in the random coil region between putative helices 8 and 9, which did not bind to HDL and which interacted slowly with multilamellar vesicles of DMPC, self-associated to form dimers, trimers, and tetramers. This suggests that either these residues are not involved in the self-association of apoA-I or that selfassociation requires more than one region of the apoA-I molecule.
The observation that deletion of a small part of the carboxylterminal region of apoA-I prevents the protein from self-associating at high concentrations in solution is of great importance. In the future it may be possible, using these mutants or derivatives, to determine the structure of apoA-I by x-ray crystallography or NMR spectroscopy.
Generation of rHDL Substrates for the LCAT Reaction-The mutant proteins were reconstituted in particles containing POPC, and cholesterol (cold and labeled) at a ratio of 100:10:1 of POPC:cholesterol:apoA-I using the sodium cholate dialysis method (43), and were used as substrates for the LCAT reaction. The sodium cholate dialysis method allowed the formation of HDL particles, even with the mutants which interacted very slowly with phospholipid. These particles were sized by native gradient gel electrophoresis (Fig. 6). This analysis identified two populations of particles at a ratio of approximately 3:1, with diameters of 96 Å and 109 Å. LCAT assays were performed with the mixed particle population.
The rHDL particles were also negatively stained with potassium phosphotungstate, overlaid on carbon-coated grids, and photographed with a Philips CM12 electron microscope. Fig. 7 (A-H) shows formation of the rHDL particles with all of the variant apoA-I protein samples tested for LCAT activation. Under the negative staining conditions used, these particles form the typical "rouleaux" indicating that they are discoidal in shape and that they have the thickness of a phospholipid bilayer. The number of rouleaux observed depends on the concentration of the sample on the carbon grid. In the samples that are less concentrated, a large number of round particles that lie flat on the grid are also observed. These particles do not seem to pack hexagonally on the grid during aggregation (a characteristic of spherical structures), providing further evidence that these particles are, in fact, discoidal in shape.
Activation of LCAT by the Variant ApoA-I Forms-Even though other apolipoproteins like apoC-I, apoA-IV, and apoE can activate the LCAT reaction in vitro using rHDL particles as substrates, none is as effective as apoA-I, when physiological lecithins are used as substrates (3). To identify the domains and residues of apoA-I that are responsible for LCAT activation, the mutant proteins were reconstituted in rHDL particles. The LCAT activity was assayed as the rate of production of labeled cholesterol esters from the rHDL particles. The labeled cholesterol esters were separated from the free cholesterol by thin layer chromatography. All LCAT assays were standardized by adding fixed amounts of apoA-I reconstituted in the HDL particles, and LCAT enzyme. The ability of the wild type proapoA-I secreted from C127 cells to activate LCAT is comparable with that of plasma apoA-I (Fig. 8). The mutant in helix 6, which contains amino acid substitutions designed to destabilize the bundle structure of apoA-I in solution (Ala 152 3 Leu,Leu 159 3 Trp), was also able to activate LCAT to levels comparable with those of plasma apoA-I. However, the mutant apoA-I form (Pro 165 3 Ala,Gln 172 3 Glu), activates LCAT to approximately 55% as compared with the wild type apoA-I. The substitutions of alanine for proline and glutamate for glutamine are predicted to change the orientation of these helices. These results indicate that the putative helices 6 and 7 contribute to the efficient activation of LCAT.
The mutant form (Glu 191 3 Ala,His 193 3 Ala,Lys 195 3 Ala) activates LCAT to levels slightly higher than the ones observed for the wild type apoA-I.
The apparent kinetic parameters for LCAT activation were also calculated by measuring the initial velocity of the LCAT reaction as a function of the apoA-I concentration in the rHDL substrate particles. Four different concentrations of apoA-I were used ranging from 4.57 ϫ 10 Ϫ6 to 5.71 ϫ 10 Ϫ7 M apoA-I. The kinetic parameters are summarized in Table II. The apparent K m values reflect the binding affinity of the enzyme for the rHDL substrate, and the apparent V max values reflect the activation of the enzyme and the catalytic rate constants (50). The apparent catalytic efficiency (apparent V max /K m ) of the apoA-I mutants ranged from 17.8 to 107.2% of the control and was the result of variations in both the K m and the V max (Table II). DISCUSSION Background-Epidemiological and genetic data have shown convincingly that low levels of HDL or apoA-I are associated with an increased risk of developing coronary heart disease (51). In a systematic effort to map the domains of apoA-I important for its functions, we have mutagenized the human apoA-I gene and we have created permanent cell lines expressing the variant apoA-I proteins. Large scale cultures of the permanent cell lines in roller bottles containing collagen-coated microspheres allowed us to obtain sufficient quantities of the apoA-I substrate and analyze its functions.
Lipid and Lipoprotein Binding Properties of the Variant ApoA-I Forms-Consistent with previous observations, deletion of residues 185-243 (⌬185-243) or 209 -243 (⌬209 -243) of apoA-I severely altered the ability of the mutant proteins to bind to HDL (29). To identify specific residues within the car- The LCAT activity was assayed as the rate of production of labeled cholesterol esters from the rHDL vesicles, as described under "Methods." The labeled cholesterol esters were separated from the free cholesterol by thin layer chromatography. All LCAT assays were standardized by adding fixed amounts of apoA-I (reconstituted in the HDL particles), and LCAT enzyme. Error bars represent standard deviation for n ϭ 3 or n ϭ 4 experiments.
boxyl terminus of apoA-I involved in lipid and/or lipoprotein binding, we introduced a series of point mutations in this region. Analysis of the ability of these mutants to bind to HDL has shown that substitution of a series of charged amino acids between residues 195 and 238 did not affect the ability of the mutant proteins to bind to HDL, indicating that inter-or intrahelical ionic interactions may not be essential for the binding of apoA-I to HDL. This is consistent with the presence, in the general population, of several substitutions of charged for neutral amino acids that do not affect HDL levels (52). This finding indicates that apoA-I may have the ability to tolerate substitutions of charged amino acids without adverse physiological consequences.
In contrast, substitution of the positively charged lysine for specific hydrophobic residues (Leu 222 , Phe 225 , and Phe 229 ) located in the predicted random coil region (18), altered dramatically the binding of the mutant protein to HDL. It is reasonable to assume that this highly hydrophobic domain of apoA-I has the ability to insert into the phospholipid component of HDL and initially anchor the protein on to the lipoprotein surface. Following the initial attachment, the rest of the molecule might subsequently "wrap" around the HDL particle with the non-polar phases of the ␣ helices oriented inside toward the phospholipid side chains, and the polar phases oriented outside facing the aqueous environment. The involvement of this highly hydrophobic region of apoA-I in lipid binding was also suggested by Fourier power spectra analysis of the hydropathy profile along the apoA-I sequence (18). This analysis showed that the sequence between residues 222-230 has a hydropathy profile between what is expected for an ␣ helix and a ␤ sheet structure (18). This region that does not participate in the formation of a defined secondary structure, may serve to attach apoA-I to HDL. Similar conclusions were reached by CD and denaturation studies of deletion mutants of apoA-I, which showed that the COOH-terminal region of apoA-I is largely unstructured and assumes an amphipathic ␣-helical structure upon binding to lipid (53).
X-ray crystallography has demonstrated that leucine residues are located in topologically distinct positions and contribute to the stabilization of the helical bundle structure of apoE in solution (54). It has also been shown that helices which participate in a four-helix bundle structure display a pattern of hydrophobic and hydrophilic residues, which can be described as a seven-residue repeat of the type (a, b, c, d, e, f, g, h) n (48,49). Leucine residues are found mostly at positions a and d and may serve to stabilize the tertiary structure of the bundle by hydrophobic interactions. Computer analysis of the apoA-I sequence indicated that such heptad repeats are encountered in the apoA-I sequence between residues 115 and 180 (helices 4 -7) and residues 188 and 243 (helices 8 and 9). 2 Replacement of Leu 211 , Leu 214 , Leu 218 , and Leu 219 by valines in helix 8 altered dramatically the ability of the mutant protein to bind to HDL. The presence of valines in these positions, which have similar hydrophobicity to but are less bulky than leucines, might prevent leucine zipper type hydrophobic interactions between juxtapositioned leucine residues (55). This may cause conformational changes in the random coil region that diminish the ability of this region to attach to the surface of HDL. The structural alterations associated with these mutants are the subject of ongoing research. Preliminary analysis of several other point mutations, along the predicted helices 1-6 of apoA-I, has shown that these mutations did not have any effect on the ability of the protein to bind to HDL, thus reinforcing the notion that the carboxylterminal region of apoA-I plays a unique role in the binding of apoA-I to HDL. 3 The results obtained from the kinetic analysis of DMPC binding are consistent with the results obtained from the HDL binding assay. Mutants that failed to bind to HDL also lysed multilamellar vesicles of DMPC very slowly, pointing to the possibility that the mechanism of the interaction of apoA-I with the HDL surface and the multilamellar phospholipid vesicles may be similar.
The carboxyl-terminal domain (putative helices 8 and 9 and the random coil region), however, is most likely only involved in the initial penetration of the protein into the phospholipid bilayer, since the mutant proteins that did not bind to HDL and interacted very slowly with the DMPC bilayers were able to form rHDL particles when the sodium cholate reconstitution method was used. These results are in agreement with previously published studies, where a proteolytic fragment of apoA-I (1-192) interacted slowly with DMPC, but did form rHDL particles when the sodium cholate method was used (34).
Activation of LCAT by Variant ApoA-I Forms-rHDL particles formed using the sodium cholate dialysis method have been shown before to be excellent substrates for the LCAT reaction (56). In the present study, these particles were visualized by electron microscopy and sized by native gradient gel electrophoresis. This latter analysis identified two populations of particles with diameters of 96 Å and 109 Å. LCAT assays were performed with the mixed particle population. The ratio of the 96-Å to the 109-Å particle present in the mixture is 3:1 and is the same for the plasma, the recombinant wild type, and the mutant pro-apoA-I proteins. Therefore these two different size particles contribute equally to the total LCAT activity in the various samples. Since the ability of the 96-Å particles to activate LCAT is 10-fold higher than that of the 109-Å particle, the contribution of the 109-Å particle is only 1/30th of the overall activation. Thus our results, expressed as percent activation relative to the wild type apoA-I, reflect the relative change in LCAT activation of the 96-Å substrate.  Previous studies have pointed out the importance of putative helices 6 and 7 (residues 145-183) of apoA-I in the activation of LCAT (21,29,30). Specifically, deletion of residues 148 -186 or residues 143-164 and 165-185 reduced the capacity of the mutant proteins to activate LCAT to background levels (29,30). The possibility exists, however, that these results were due to dramatic alterations in the protein structure caused by the large deletions.
The point mutant (Pro 165 3 Ala,Gln 172 3 Glu) eliminated a helix-breaking proline residue, and also changed a glutamine to glutamate. According to the model proposed by Nolte and Atkinson (18), type A half-repeats have neutral residues at the eighth position, whereas type B have negatively charged residues. Thus, the glutamine to glutamate substitution converts the predicted type A repeat to a type B repeat. It has been suggested that ␤ turns occur mostly between A and B, rather than AA or BB repeats (18). Thus, the 45% reduction in the LCAT activation ability of this mutant suggested that the two amino acid substitutions may have distorted the orientation of the putative helices 6 and 7, which were shown previously to be important for LCAT activation (29,30,33). Consistent with this finding, a naturally occurring apoA-I variant (Pro 165 3 Arg) has a 45-55% LCAT activation ability as compared with wild type apoA-I (57). It is possible that one or both of the helices interact electrostatically with the polar face of the predicted amphipathic ␣-helical segment found close to the active site of the LCAT enzyme (residues 151 and 174 of LCAT), as it was proposed by Fielding (14). Ionic interactions between LCAT and apoA-I have also been suggested by the quantitative dissociation of LCAT from HDL in concentrated salt solutions (58). Alterations in the ionic interactions between apoA-I and the LCAT enzyme may lead to a less efficient activation.
The 42% decrease in the LCAT activation ability of the deletion (⌬209 -243) mutant could be attributed mainly to the change in the apparent V max . This is consistent with previous studies (29,30,31), which showed that deletion of residues 209 -219, 220 -241, 212-233, and 213-243 resulted in LCAT activation of 11.2, 16, 28, and 13%, respectively, as compared with the wild type apoA-I. The variation in the extent of inhibition is probably due to either the difference in the phospholipid used for the formation of the reconstituted particles (egg phosphatidylcholine in the previous studies, versus POPC in this study), or the difference in the size of the particles formed. In addition, the deletion (⌬213-243) mutant used in the previous study contained 12 residues of unrelated carboxyl-terminal sequence (29). Substitution of valines for leucines 211, 214, 218, and 219 or substitution of lysines for leucine 222 and phenylalanines 225 and 229, resulted in 32 and 68% reduction, respectively, in the capacity of the mutant proteins to activate LCAT. This impairment was the result of a decrease in the apparent V max for the former mutant and mainly due to an increase in the apparent K m for the latter mutant. The combined data from this and the previous studies (29 -31) suggest that domains and residues within the carboxyl-terminal region of apoA-I contribute to the optimal activation of LCAT. It is possible that the presence of the carboxyl-terminal domain allows apoA-I to acquire a proper conformation, which facilitates the postulated specific interaction of the middle region (residues 145-183) of apoA-I with the LCAT enzyme.
A mutant in the putative helix 6 containing amino acids in positions that are expected to destabilize the bundle structure in solution (49) binds normally to HDL and DMPC and also activates LCAT normally. These findings suggest that potential structural alterations of the bundle structure in solution have no consequences for lipid binding and LCAT activation, where the lipid bound protein assumes a new conformation.
Substitution of alanines for charged residues in putative helix 8 enhanced slightly the capacity of the variant protein to activate LCAT as compared with the wild type proapoA-I. The significance of this increased activation is not clear. It is possible that the removal of charged residues in putative helix 8 may allow a stronger electrostatic interaction to occur between the putative ␣-helical segment of LCAT (residues 151 and 174) and putative helices 6 and 7 of apoA-I. Alternatively, the absence of these charged residues may allow the cholesterol substrate to position itself more favorably relative to the active site serine residue of the LCAT enzyme. The catalytic efficiency of the last two mutants was comparable with that of the wild type pro apoA-I (Table II) due to concomitant increases in both the V max and the K m .
The lack of dramatic changes in the apparent V max for any of the mutations tested strongly suggests that none of the apoA-I residues altered participate in the catalytic mechanism of LCAT.
Domains of ApoA-I Involved in Self-association-In addition to its role in the initial anchoring of the protein into the phospholipid bilayer, the carboxyl-terminal domain of apoA-I also appears to participate in the self-association of the protein. The deletion mutants (⌬185-243) and (⌬209 -243) existed only as monomers and dimers in solution rather than as a mixture of oligomers. Similar conclusions were reached with the aminoterminal proteolytic fragment (1-192) of apoA-I (34). It has been suggested that self-association of apoA-I may promote stabilization of the potential amphipathic ␣-helical segments of the carboxyl-terminal region which is less organized in the monomeric form (53). Overcoming the oligomerization problem and achieving high concentrations of the monomeric apoA-I fragment in solution can facilitate efforts to derive the threedimensional structure of apoA-I by x-ray crystallography or NMR spectroscopy.
Overall, the present study shows that specific hydrophobic residues in the predicted random coil region between helices 8 and 9, and in the putative helix 8 (Fig. 3), are critical for the ability of apoA-I to bind to HDL and to lyse multilamellar vesicles of DMPC. It is possible that the formation of a "leucine zipper-like" structure between the putative helices 8 and 9 may stabilize this random coil region during its interaction with lipids and lipoproteins and allow its association with the phospholipid surface. Although the carboxyl-terminal region of apoA-I is required for both the lipid binding and self-association, the residues which participate in these two functions appear to be different.
Several amino acid substitutions in the carboxyl-terminal domain of apoA-I cause a moderate reduction in the catalytic efficiency of LCAT, suggesting that residues in this region contribute to optimum activation of LCAT, without a direct participation in the catalytic mechanism.
In addition to their use in the present study, the variant apoA-I forms generated will serve as valuable reagents for future studies to identify the domains and residues of apoA-I involved in binding to scavenger class B1 receptor, in promoting cholesterol efflux from cells, as well as for the structural analysis of apoA-I.