Effects of deletion of the carboxyl-terminal domain of ApoA-I or of its substitution with helices of ApoA-II on in vitro and in vivo lipoprotein association.

In the present study, the lipoprotein association of apoA-I, an apoA-I (ΔAla190-Gln243) deletion mutant and an apoA-I (Asp1-Leu189)/apoA-II (Ser12-Gln77) chimera were compared. At equilibrium, 80% of the 125I-labeled apolipoproteins associated with lipoproteins in rabbit or human plasma but with very different distribution profiles. High density lipoprotein (HDL)2,3-associated fractions were 0.60 for apoA-I, 0.30 for the chimera, and 0.15 for the deletion mutant, and corresponding very high density lipoprotein-associated fractions were 0.20, 0.50, and 0.65. Clearance curves after intravenous bolus injection of 125I-labeled apolipoproteins (3 μg/kg) in normolipemic rabbits could be adequately fitted with a sum of three exponential terms, yielding overall plasma clearance rates of 0.028 ± 0.0012 ml·min−1 for apoA-I (mean ± S.E.; n = 6), 0.10 ± 0.008 ml·min−1 for the chimera (p < 0.001 versus apoA-I) and 0.38 ± 0.022 ml·min−1 for the deletion mutant (p < 0.001 versus apoA-I and versus the chimera). Fractions that were initially cleared with a t1/2 of 3 min, most probably representing free apolipoproteins, were 0.30 ± 0.04, 0.50 ± 0.06 (p = 0.02 versus apoA-I), and 0.64 ± 0.07 (p = 0.002 versus apoA-I), respectively. At 20 min after the bolus, the fractions of injected material associated with HDL2,3 were 0.55 ± 0.06, 0.25 ± 0.03 (p = 0.001 versus apoA-I), and 0.09 ± 0.01 (p < 0.001 versus apoA-I and versus the chimera), respectively, whereas the fractions associated with very high density lipoprotein were 0.15 ± 0.006, 0.25 ± 0.03 (p = 0.008 versus apoA-I), and 0.27 ± 0.03 (p = 0.003 versus apoA-I), respectively. The ability of the different apolipoproteins to bind to HDL3 particles and displace apoA-I in vitro were compared. The molar ratios at which 50% of 125I-labeled apoA-I was displaced from the surface of HDL3 particles were 1:1 for apoA-I, 3:1 for the chimera and 12:1 for the deletion mutant, indicating 3- and 12-fold reductions of the affinities for HDL3 of the chimera and the deletion mutant, respectively. These data suggest that the carboxyl-terminal pair of helices of apoA-I are involved in the initial rapid binding of apoA-I to the lipid surface of HDL. Although the lipid affinity of apoA-II is higher than that of apoA-I, substitution of the carboxyl-terminal helices of apoA-I with those of apoA-II only partially restores its lipoprotein association. Thus, this substitution may affect cooperative interactions with the middle amphipathic helices of apoA-I that are critical for its specific distribution over the different HDL species.

Low plasma levels of high density lipoproteins (HDL) 1 and of their major protein component, apolipoprotein A-I (apoA-I), correlate with an increased risk for coronary heart disease (1), and family and twin studies have suggested that decreased HDL levels are partially hereditary (2)(3)(4)(5). In addition, HDL and their apolipoproteins increase the net efflux of cellular unesterified cholesterol (6 -8) and remove excess cholesterol from peripheral (nonhepatic) cells (9), which may explain the inverse correlation between risk of coronary heart disease and HDL levels (10).
ApoA-I, the major protein component of HDL, is an important determinant of the concentration of HDL in plasma (11). ApoA-I binds and transports plasma lipids, serves as a cofactor for the enzyme lecithin:cholesterol acyltransferase, and increases cholesterol efflux from peripheral tissues (12)(13)(14). In addition, apoA-I is an important ligand in the binding of HDL to cell membranes (15,16). These characteristics contribute to the ability of HDL to induce reverse cholesterol transport and thus to the protective effect of HDL on cardiovascular disease.
ApoA-I is synthesized as a prepropeptide, which is cotranslationally cleaved to pro-apoA-I, and then cleaved during secretion to form the mature 243-amino acid apoA-I protein (17). The secondary structure of apoA-I contains amphipathic helices composed of hydrophilic and hydrophobic surfaces (18 -20). It has been demonstrated that the carboxyl-terminal domain of apoA-I plays an important role in lipid binding and in the interaction with cell membranes (16,(21)(22)(23)(24)(25). Analysis of patients with reduced plasma concentration of HDL revealed that accelerated catabolism, possibly due to enzymatic degradation, of apoA-I is the most common cause of low HDL levels (26 -29). Schmidt et al. (30) demonstrated that deletion of the carboxylterminal domain of apoA-I results in decreased in vivo lipoprotein association, in an altered distribution pattern in HDL, and in an increased clearance rate. Recently, we have demonstrated (25) that deletion of the carboxyl-terminal domain of apoA-I decreases the rate but not the extent of in vitro phospholipid association and that this interaction results in the formation of larger discoidal particles with increased apoA-I/ phospholipid ratios.
The present study compares the in vitro and in vivo lipopro-tein binding properties of wild-type human apoA-I, an apoA-I (⌬Ala 190 -Gln 243 ) carboxyl-terminal deletion mutant, and an apoA-I (Asp 1 -Leu 189 )/apoA-II (Ser 12 -Gln 77 ) chimera, in which the carboxyl-terminal pair of helices of apoA-I have been substituted with the pair of helices of apoA-II.

EXPERIMENTAL PROCEDURES
Reagents-Oligonucleotides were obtained by custom synthesis (Pharmacia, Brussels, Belgium). DNA sequencing was performed on a Pharmacia ALF DNA sequencer. Chromatography materials were obtained from Pharmacia.
All cDNA constructs were confirmed by DNA sequencing, as described previously (25).
Expression and Purification of Apolipoproteins-Apolipoproteins were expressed in the periplasmic fractions of E. coli WK6 host cells as described (25).
Standard apoA-I was isolated from normolipemic human plasma as described previously (17). The purity of proteins was established by SDS-gel electrophoresis (31) and immunoblotting (32).
Pharmacokinetics-Proteins were iodinated by the Bolton and Hunter method (33). The pharmacokinetic properties of wild-type apoA-I, apoA-I (⌬Ala 190 -Gln 243 ), and apoA-I (Asp 1 -Leu 189 )/ApoA-II (Ser 12 -Gln 77 ) in New Zealand White rabbits were determined by measurement of the residual radioactivity after bolus injection of 125 I-labeled proteins (3 g/kg) in blood samples that were taken at times 1, 2, 5, 10, 20, 30 min and at 1, 2, 3, 4, 5, 6, 7, 8, 24, 28, and 31 h. The results were plotted semilogarithmically, and the curves were fitted with a sum of three exponential terms C(t) ϭ Ae Ϫ-␣t ϩ Be Ϫ␤t ϩ Ce Ϫ␥t , by graphical curve peeling (34). The coefficients A, B, and C were calculated from the intercepts on the ordinate, whereas the exponents ␣, ␤, and ␥ were calculated from the slopes. The following clearance parameters were calculated using standard formulas derived by Gibaldi and Perrier (34): total volume of distribution V D ϭ dose/C; extrapolated area under the curve (AUC)ϭ A/␣ ϩ B/␤ ϩ C/␥, and plasma clearance rate Clp ϭ dose/AUC. Statistical differences between these parameters were calculated using the Student t test.
Density Gradient Ultracentrifugation and Gel Filtration Chromatography-Continuous density gradient ultracentrifugation (35) was performed using a table top T-100 ultracentrifuge (Analis, Namur, Belgium) in 5-ml tubes. Four hundred l of rabbit plasma were analyzed by gel filtration on a Superose 6HR column equilibrated with 20 mM Tris-HCl buffer, pH 8.1, containing 0.15 M NaCl, 1 mM EDTA, and 0.02 mg/ml sodium azide in a fast protein liquid chromatography system (Waters Associates, Milford, MA). The levels of phospholipids and cholesterol were determined using standard enzymatic assays (Biomérieux, Marcy, France, and Boehringer Mannheim, Meylon, France, respectively), and the protein levels were determined according to Bradford (36).
Displacement of Radiolabeled ApoA-I from the Surface of HDL-Human HDL 3 , containing 240 g of apoA-I in 200 l, was incubated with 30 g of 125 I-labeled apoA-I for 1 h at 37°C. Nonbound radiolabeled apoA-I was separated from the HDL 3 fraction by filtration on a Centricon 100 filter (Amicon, Beverly, MA). Radiolabeled HDL 3 was diluted 8-fold in 20 mM Tris-HCl buffer, and 50-l aliquots, containing 0.80 g of radiolabeled apoA-I, were mixed with 50 l of solutions that contained 0.24, 0.48, 0.96, 1.92, 3.84, 7.68, or 15.36 g of apoA-I, deletion mutant, or chimera. After 1 h of incubation at 37°C, free apolipoproteins were separated from the HDL 3 by filtration. The HDL 3 -associated radioactivity and the radioactivity in the filtrate were measured. The percentage of HDL 3 -associated radiolabeled apoA-I was determined as a function of the amount of competing apolipoprotein in the incubation mixture.

RESULTS
The predicted amphipathic helical regions in apoA-I, the apoA-I (⌬Ala 190 -Gln 243 ) carboxyl-terminal deletion mutant, apoA-II, and the apoA-I (Asp 1 -Leu 189 )/apoA-II (Ser 12 -Gln 77 ) chimera, according to Brasseur et al. (37,38), are illustrated in Fig. 1. Edmundson wheel diagrams confirmed that the aligned carboxyl-terminal segments in apoA-I and the chimera indeed may form amphipathic helices and that the orientation of their respective hydrophilic and hydrophobic surfaces is very similar (Fig. 2). The hydrophobicity of the carboxyl-terminal helix of the chimera was higher than that of the original ninth helix of apoA-I (mean residue hydrophobicities were 0.038 for apoA-I and 0.060 for the chimera).
Apolipoproteins were expressed in the periplasmic space of E. coli cells and purified to homogeneity as revealed by SDSpolyacrylamide gel electrophoresis (Fig. 3). Each of the proteins migrated as a single band with the expected molecular masses on polyacrylamide gels: 28.3 kDa for wild-type recombinant apoA-I, 29.8 kDa for the chimera, and 22.0 kDa for the deletion mutant. The identity of each band was confirmed by immunoblot analysis, using polyclonal sheep anti-human apoA-I and sheep anti-human apoA-II antibodies (data not shown). As previously shown, the molecular masses and the in vitro phospholipid binding properties of wild-type recombinant apoA-I and plasma apoA-I are identical (25).
To assess the impact of the deletion or the substitution of the carboxyl-terminal domain on protein self-association, the lipidfree apolipoproteins were solubilized in phosphate-buffered saline at three concentrations (0.125, 0.25, and 0.50 mg/ml) and treated with the cross-linking agent bis(sulfosuccinimidyl)suberate. Fig. 4 illustrates the oligomerization patterns of the different apolipoproteins. At the lowest concentration all apolipoproteins existed predominantly as monomers. At the highest concentration, apoA-I exhibited monomeric (38% of total protein), dimeric (22%), trimeric (16%), and tetrameric (14%) forms ( Fig. 4). At the same concentration, the percentage of apoA-I (⌬Ala 190 -Gln 243 ) that existed as monomer was higher (monomer, 75%; dimer, 16%; trimer, 6%; and tetramer, 3%). At all concentrations, Ͼ90% of the chimera existed as monomer (Fig. 4). Radiolabeled apolipoproteins with specific radioactivities of 3 ϫ 10 6 cpm/g of protein were used within 24 h after radiolabeling and analyzed by SDS-polyacrylamide gel electrophoresis to exclude degradation and self-association (not shown). The association of radiolabeled apolipoproteins with lipoprotein particles was studied following in vitro incubation in rabbit plasma for 60 min. Fractions of added apolipoproteins that were associated with HDL 2,3 particles were 0.60 Ϯ 0.05 for apoA-I (mean Ϯ S.E.; n ϭ 6), 0.32 Ϯ 0.02 for the chimera (p ϭ 0.007 versus apoA-I), and 0.14 Ϯ 0.005 for the deletion mutant (p Ͻ 0.001 versus apoA-I and versus the chimera). Fractions of added apolipoproteins associated with VHDL particles were 0.20 Ϯ 0.01 for apoA-I, 0.48 Ϯ 0.05 for the chimera (p Ͼ 0.001 versus apoA-I), and 0.65 Ϯ 0.06 for the deletion mutant (p Ͻ 0.001 versus apoA-I and p ϭ NS versus the chimera). Fractions present as free apolipoproteins in the plasma were 0.20 for all three compounds (data not shown). The density distributions of radiolabeled apolipoproteins associated with lipoprotein particles in human plasma are illustrated in Fig. 5. Fractions that were associated with HDL 2,3 particles were 0.55 Ϯ 0.06 for apoA-I, 0.30 Ϯ 0.04 for the chimera (p ϭ 0.006 versus apoA-I), and 0.15 Ϯ 0.02 for the deletion mutant (p Ͻ 0.001 versus apoA-I, p ϭ 0.007 versus the chimera). Fractions that were associated with VHDL particles were 0.25 Ϯ 0.02 for apoA-I, 0.50 Ϯ 0.06 for the chimera (p Ͻ 0.003 versus apoA-I), and 0.65 Ϯ 0.05 for the deletion mutant (p Ͻ 0.001 versus apoA-I and p ϭ NS versus the chimera). Fractions that were present as free apolipoproteins in the plasma again were 0.20 for all three compounds.
The plasma clearance of radiolabeled wild-type apoA-I, apoA-I (⌬Ala 190 -Gln 243 ), and apoA-I (Asp 1 -Leu 189 )/apoA-II (Ser 12 -Gln 77 ) was analyzed in rabbits (Fig. 6). The disappearance rates of all proteins could be described by a sum of three exponential terms by graphical curve peeling. The calculated pharmacokinetic parameters are summarized in Table I. The plasma clearance rate was 0.028 Ϯ 0.0012 ml⅐min Ϫ1 for 125 Ilabeled apoA-I as compared with 0.025 Ϯ 0.0011 ml⅐min Ϫ1 for apoA-I antigen, as determined in an enzyme-linked immunosorbent assay based on monoclonal antibodies specific for apoA-I following injection of human apoA-I. These data suggest that clearance of apoA-I was not affected by the labeling procedures. Plasma clearance rates of the chimera and of the deletion mutant, respectively, were 3.6-fold and 13.6-fold higher than that of apoA-I (Table I). Values of t1 ⁄2 ␣, t1 ⁄2 ␤, and t1 ⁄2 ␥ were, however, very similar for all apolipoproteins: 3, 220, and 2,300 min, respectively, suggesting that the differences in clearance resulted from differences in the lipoprotein profiles. This is illustrated in Fig. 7, which shows the distributions of cholesterol and phospholipids (upper panel) and of radiolabeled apolipoproteins (lower panel) in the different lipoprotein fractions at 20 min postinjection. Ninety percent of 125 I-labeled apoA-I was associated with HDL 2 and HDL 3 particles (fractions 15-25), whereas only 10% was associated with smaller, phospholipid-rich VHDL particles (fractions 26 -35). Corresponding values were 60 and 40% for the chimera and 30 and 70% for the deletion mutant.
The lipoprotein binding properties of the three apolipopro-teins were further investigated by analyzing the density distribution of radiolabeled apolipoproteins. The distribution patterns at 20 min postinjection are illustrated in Fig. 8. At 20 min postinjection, HDL 2,3 -associated fractions were 0.55 Ϯ 0.06 for apoA-I, 0.25 Ϯ 0.03 for the chimera (p ϭ 0.001 versus apoA-I) and 0.09 Ϯ 0.01 for the deletion mutant (p Ͻ 0.001 versus apoA-I and versus the chimera) ( Fig. 9). At 4 h postinjection HDL 2,3 -associated fractions were 0.43 Ϯ 0.04, 0.19 Ϯ 0.02 (p Ͻ 0.001 versus apoA-I) and 0.08 Ϯ 0.006 (p Ͻ 0.001 versus apoA-I and versus the chimera), respectively (Fig. 9). At 24 h postinjection, HDL 2,3 -associated fractions were 0.34 Ϯ 0.05, 0.14 Ϯ 0.02 (p ϭ 0.004 versus apoA-I), and 0.06 Ϯ 0.008 (p Ͻ 0.001 versus apoA-I and p ϭ 0.004 versus the chimera), respectively (Fig. 9). Estimated half-lives of HDL 2,3 -associated protein were 2,200 Ϯ 130 min for apoA-I, 2,500 Ϯ 300 min (p not significant) for the chimera and 2,400 Ϯ 240 min Incubation of 50-l aliquots of radiolabeled HDL 3 , containing 0.8 g of radiolabeled apoA-I, with 50 l aliquots containing increasing amounts (0.24, 0.48, 0.96, 1.92, 3.84, 7.68, or 15.36 g) of apoA-I, the chimera, or the deletion mutant resulted in a concentration-dependent displacement of radiolabeled apoA-I from the surface of HDL. Fifty percent displacement was obtained with 0.9 g of apoA-I, 3.0 g of the chimera, and 8.7 g of the deletion mutant, thus at 1:1, 3:1, and 12:1 molar ratios of added protein to HDL 3 -associated 125 I-labeled apoA-I, respectively (Fig. 10). These data indicate that the affinity for HDL 3 of 125 I-labeled apoA-I was identical to that of nonlabeled apoA-I, whereas the affinities of, respectively, the chimera and the deletion mutant were 3-and 12-fold lower. DISCUSSION The variability in HDL cholesterol levels is largely determined by differences in the fractional catabolic rate of apoA-I, which is inversely correlated with HDL particle size (39). In order to analyze the role of the carboxyl-terminal domain of apoA-I in phospholipid binding, lipoprotein association, HDL particle size distribution, and clearance, mutants of apoA-I have been generated. In the present study, the in vitro and in vivo lipoprotein binding properties of apoA-I, an apoA-I (⌬Ala 190 -Gln 243 ) deletion mutant, and an apoA-I (Asp 1 -Leu 189 )/apoA-II (Ser 12 -Gln 77 ) chimera were compared. In the apoA-I (Asp 1 -Leu 189 )/apoA-II (Ser 12 -Gln 77 ) chimera, the Ala 190 -Gln 243 carboxyl-terminal domain of apoA-I was substituted with the Ser 12 -Lys 28 and Pro 51 -Val 68 ␣-helical segments of apoA-II. Previously, it has been shown that synthetic peptides overlapping with these apoA-II sequences associated with phospholipids, suggesting that these helical segments consti- tute phospholipid binding domains (40,41). Furthermore, epitope-mapping studies showed that antibodies to the carboxyl-terminal domain of apoA-I bound to an epitope in the Gln 36 -Gln 77 segment of apoA-II, demonstrating significant structural homology between these domains (42). Finally, apoA-II can displace apoA-I from the surface of recombinant HDL particles without loss of phospholipids or major change in particle size (43). 125 I-Labeled apoA-I associated preferentially (60%) with HDL 2 and HDL 3 particles both in rabbit and human plasma. Deletion of the Ala 190 -Gln 243 carboxyl-terminal domain of apoA-I did not reduce the extent (80%) of in vitro lipoprotein association in rabbit and human plasma under equilibrium conditions but altered its distribution profile. Only 15% of the apoA-I (Ala 190 -Gln 243 ) deletion mutant associated with HDL 2,3 particles. Although the predicted secondary structure and the amphipathicity of the chimera were very similar to those of apoA-I, only 30% of the chimera associated with HDL 2,3 particles. Because lipoprotein distribution profiles in human and rabbit plasma were very similar, rabbits were used as model animals to investigate the effects of changed lipoprotein distribution of both the chimera and the deletion mutant on their pharmacokinetic properties.
Following bolus injection of 125 I-labeled apoA-I in rabbits, it was cleared with a rate of 0.028 ml⅐min Ϫ1 . This value is very similar to that determined by Ikewaki et al. (39) following bolus injection of either exogenously or endogenously labeled apoA-I in humans. Deletion of the carboxyl-terminal domain of apoA-I resulted in a 13.6-fold increased clearance rate. This was most probably due to an enhanced clearance in the ␣-phase (64% as compared with 30% for apoA-I) of free apolipoprotein in solu-  tion, suggesting a slower rate of in vivo lipoprotein association of the deletion mutant. These data are in agreement with earlier findings that deletion of the carboxyl-terminal domain of apoA-I reduced the rate of in vitro phospholipid binding (25). Accelerated clearance may further result from a decreased association with HDL 2,3 particles (9% as compared with 55% for apoA-I) and from an increased association with smaller, phospholipid-rich VHDL particles (30% as compared with 15% for apoA-I) that were cleared more rapidly from the circulation. Indeed, estimated half-lives of HDL 2,3 particles were approximately 2,200 min as compared with approximately 260 min for VHDL particles. These data are in agreement with earlier findings of Schmidt et al. (30) that progressive carboxyl-terminal domain truncation of apoA-I resulted in a progressive increase of its clearance rate, resulting from an increased association with VHDL. In aggregate, these data suggest that deletion of the carboxyl-terminal domain of apoA-I correlated with a decreased rate of lipoprotein association and/or increased rate of dissociation and a preferential association with smaller, phospholipid-rich VHDL particles that are cleared more rapidly.
The plasma clearance rate of the chimera was 3.6-fold faster than that of apoA-I but was 3.8-fold slower than that of the deletion mutant. The differences between the chimera and apoA-I could be explained by a decreased rate of lipoprotein association, decreased association with HDL 2,3 particles (25% as compared with 55% for apoA-I), and increased association with VHDL (25% as compared with 15%). The differences between the chimera and the deletion mutant could be explained essentially by an increased association of the chimera with HDL 2,3 particles (25% as compared with 9%). The finding that ␣-, ␤and ␥-half-lives of all apolipoproteins were very similar and corresponded with half-lives of free apolipoprotein, VHDL, and HDL 2,3 particles, respectively, indicated that the differences in clearance rate indeed could be explained by differences in distribution profile. These differences in distribution profile could be explained by a 3-fold (for the chimera) and a 12-fold (for the deletion mutant) lower affinity for HDL 3 particles.
Cross-linking experiments demonstrated that the self-association of both the deletion mutant and the chimera were lower than that of apoA-I. Human apoA-II self-associates in lipid-free solution (44,45), resulting in dimerization of the disulfidelinked monomer, but it does not seem to aggregate in a monomer-dimer-tetramer fashion as does apoA-I. It has been suggested that this greater ability of apoA-I to self-associate may shield hydrophobic surfaces within the molecule better and allow apoA-I to be thermodynamically stable when not bound to lipid (46). On the other hand, apoA-II must bind to lipid in order to form a stable entity and will therefore bind to a more diverse group of lipids (46). Despite the lack of self-association of the chimera (Fig. 3), the lipoprotein affinity of this molecule was lower than that of apoA-I (Fig. 10). It seems unlikely that the decreased lipoprotein affinity of the chimera results from a stronger intramolecular association of the apoA-II-(51-68) helix with the apoA-II- (12)(13)(14)(15)(16)(17)(18)(19)(20)(21)(22)(23)(24)(25)(26)(27)(28) helix relative to that between the apoA-I-(223-239) and apoA-I-(190 -206) helices because this putative association does not affect the lipoprotein association of apoA-II. An alternative explanation is that although the carboxyl-terminal domain of apoA-I is important for its initial rapid binding to the lipid surface of HDL, cooperative interactions with the middle six amphipathic helices of apoA-I may be important for its HDL subspecies distribution. Recently such a model has been proposed (47) on the basis of the finding that although only the end 22-residue helices of apoA-I have significant lipid affinity, apoA-I is, on a molar basis, about 10 times more effective than the most effective 22-residue peptide in reducing the enthalpy of the gel-to-liquid crystalline phase transition of L-␣-dimyristoylphosphatidylcholine multilamellar vesicles.
In summary, deletion of the carboxyl-terminal domain of apoA-I results in a significant decrease of its lipoprotein affinity, a reduction of its HDL 3 association, and thus a more rapid clearance. Although substitution of this domain with helices of apoA-II restores the number of amphipathic helices and increases the hydrophobicity of the carboxyl-terminal domain, it does not restore completely its lipoprotein affinity and its HDL 3 association, possibly because this substitution affects cooperative interactions with the middle amphipathic helices of apoA-I that are critical for its specific distribution over the HDL subspecies.