Apolipoprotein A-I conformation in reconstituted discoidal lipoproteins varying in phospholipid and cholesterol content.

The effects of the size and cholesterol content on the conformation of apolipoprotein A-I (apoA-I) have been studied in reconstituted discoidal lipoproteins containing two apoA-I per particle (Lp2A-I). The immunoreactivity of a series of 13 epitopes distributed along the apoA-I sequence has been evaluated in Lp2A-I with a phospholipid/apoA-I molar ratio ranging from 31 to 156 and in Lp2A-I with constant phospholipids but varying in cholesterol content from 0 to 22 molecules. The results are compatible with a three domain structure in apoA-I in which the central domain is located between residues 99 and 143 and postulated to be a hinged domain that responds differentially to changes in phospholipid and cholesterol contents. Increasing the phospholipid content results in significant changes of epitope immunoreactivity throughout the N-terminal and central domains of apoA-I with fewer modifications in the C-terminal domain. In contrast, increasing Lp2A-I of two central epitopes, A11 (residues 99-132) and 5F6 (residues 118-148), and an extreme N-terminal epitope, 4H1 (residues 2-8). Interestingly, the effects of increasing cholesterol or phospholipids on these epitopes are opposite. This suggests a specific effect of cholesterol on the central domain tertiary structure between residues 99 and 143. Competition binding assays among pairs of antibodies binding to apoA-I on Lp2A-I are best explained by invoking inter- as well as intramolecular competitions. The specificity of the intermolecular competitions suggests an N to C termini arrangement of the two apoA-I molecules around the disc. Increasing the phospholipid content of Lp2A-I mainly increases the competitions between 3G10 and antibodies binding to most adjacent epitopes. Simultaneously as Lp2A-I enlarges, several of these antibodies also enhance the binding of 3G10. This has been interpreted as evidence of a structural rearrangement of apoA-I as a result of the size increase where the alpha-helix (residues 99-121) that contains the 3G10 epitope is increasingly interacting with lipids resulting in the enhanced expression of this epitope. The increasing interactions of apoA-I helices with lipids in the enlarging disc are compatible with previous reports of a greater apoA-I stability in the large discs. By contrast, cholesterol has limited but specific effects on antibody competitions and decreases the interaction of the N-terminal domain with the domain containing 3G10, either by direct cholesterol protein interaction or by modification of the lipid phase packing.

Apolipoprotein A-I (apoA-I) 1 is the primary protein component of plasma high density lipoproteins (HDL) (1,2). Like several other related exchangeable apolipoproteins, apoA-I contains multiple repeats of 22 amino acids, constituting amphipathic ␣-helices that are the major lipid-binding domains of the protein (3). These repeats are interrupted by proline or glycine residues that may act as helix-breakers by creating ␤-turns, providing more flexibility to the amphipathic helices, and allowing apoA-I to conform to the surface of HDL as reviewed by Segrest et al. (4). Differential interaction between monoclonal antibodies and HDL support the hypothesis that the conformation of apoA-I is not the same in all HDL subspecies, depending on size or lipid composition (5,6). This is in agreement with the demonstration that changes in reconstituted LpA-I composition modulate the net charge and the ␣-helicity of apoA-I and the stability of its helical segments (7). More recently, it has been demonstrated that a particular apoA-I conformation present in pre-␤ 1 -HDL could have a major role in promoting the efflux of cellular cholesterol (8). In general, data from the literature suggest that apoA-I conformation is both variable and critical for HDL metabolism, but study of its precise organization on the surface of native HDL remains limited by the great heterogeneity of this class of lipoproteins (9,10). Reconstituted complexes of isolated apoA-I with defined molar ratios of phospholipids and cholesterol have been used to prepare discoidal or spherical HDL in vitro (11)(12)(13). These complexes, similar to some HDL observed in vivo, exist as particles of very reproducible sizes and compositions and, of particular interest, as discretely sized particles containing a constant number of apoA-I/particle (reviewed in Refs. 14 and 15). As apoA-I conformation changes between particles, mobile regions (termed hinged domains) have been postulated to explain and regulate the variable size, apolipoprotein composition, and shape of HDL (10,16). This hypothesis is supported by physical and chemical modification and immunochemical studies of apoA-I within reconstituted complexes of different lipid composition (reviewed in Ref. 17) and more recently, by analogy to other apolipoproteins whose crystal structure has been solved (18,19). Nevertheless, little is known about the specific position of the domains and residues involved in the binding of phospholipids or cholesterol in these lipoproteins, and the existence of a hinge domain area is still a subject of speculation.
Epitope expression studies using reconstituted LpA-I have shown that binding to lipids produces drastic changes in apoA-I conformation and that particle size can also modify specific domains, such as a central hinged domain constituted by two adjacent antiparallel ␣-helices (16,20,21). The inhibitory or enhancing effects of anti-apoA-I mAbs on lecithin:cholesterol acyltransferase reaction with LpA-I also demonstrate the importance of the central domain of apoA-I (22). Although characterized by well defined molar ratios, the particles used in previous studies often vary in more than one component at a time, making the interpretation difficult. Recently, a new approach to prepare reconstituted HDL has been published (23). Simpler than previous methods, this approach is particularly accommodating for compositional manipulations of the particles and gives a larger scale for varying some components than that previously obtained (23). Using this method, we have undertaken a study of the immunoreactivity of apoA-I epitopes in order to define the changes of apoA-I conformation specifically associated with variations in either phospholipid or cholesterol contents of the particles. This is the first study to relate the contribution of individual lipids in reconstituted HDL to apoA-I epitope expression. The results confirm the mobility of the central area of apoA-I, which responds differentially and specifically to changes in phospholipid or cholesterol concentration. An unexpected implication of the N terminus region of the protein during size changes has also been observed, a region previously described by a complex tertiary structure (20).

EXPERIMENTAL PROCEDURES
Materials-Cholesterol and sodium cholate were purchased from Sigma. 1-Palmitoyl-2-oleoylphosphatidylcholine (POPC) and guanidine HCl (GndHCl) were obtained from Avanti Polar Lipids (Birmingham, AL) and from Life Technologies, Inc., respectively. All other reagents were analytical grade.
Purification of Apolipoprotein A-I-Human HDL was isolated from pooled plasma from normolipidemic volunteers by sequential ultracentrifugation as described previously (24). HDL was delipidated, and the apoA-I was prepared according to the method of Brewer et al. (25). The purity of the apoA-I preparations was confirmed by SDS-polyacrylamide gel electrophoresis. Aliquots of apoA-I were stored at Ϫ80°C in lyophylized form. Prior to use, purified apoA-I was resolubilized in 6 M GndHCl and dialyzed extensively against a 10 mM Tris buffer that contained 150 mM NaCl, 1 mM EDTA, and 1 mM NaN 3 , pH 8.0.
Preparation of Reconstituted Discoidal LpA-I-Discoidal reconstituted LpA-I were prepared by the cholate dispersion/Bio-Bead removal technique as described by Sparks et al. (23). Briefly, purified POPC (and unesterified cholesterol, if needed) in chloroform was dried under nitrogen in a 5-ml glass tube. Tris buffer was added, and the mixture vortexed thoroughly for 3 min, followed by the addition of sodium cholate in Tris buffer and vortexing for another 3 min. The dispersion was then incubated at 37°C and vortexed every 15 min for 1.5 h or until completely clear. ApoA-I and Tris buffer were added to obtain a final mixture diluted to 1 mg of protein/ml, which was incubated for 60 min at 37°C. Sodium cholate was removed from the mixture by incubation with hydrated Bio-Beads for 2.5 h at 4°C; the Bio-Beads were recovered from the sample by filtration through a 0.22-m filter. Final composition of discoidal LpA-I was evaluated after isolation by gel filtration on a Superose 6 column (2.5 ϫ 100 cm) and concentration by size exclusion with a porous membrane (Centriprep 30; molecular mass cutoff, 30 kDa; Amicon) under low speed centrifugation.
Determination of LpA-I Structural Characteristics-The homogeneity and hydrodynamic diameters of LpA-I particles were estimated by nondenaturing gradient gel electrophoresis using precast 8 -25% polyacrylamide gels (Pharmacia Biotech Inc.) and reference globular pro-teins (17.0 nm thyroglobulin, 12.2 nm ferritin, 10.4 nm catalase, 8.2 nm lactate dehydrogenase, and 7.1 nm albumin) provided by a high molecular weight calibration kit (Pharmacia, Uppsala) as described (26). The number of apoA-I molecules/LpA-I was determined by cross-linking the apoA-I with dimethyl suberimidate (27), and the products of the reaction were analyzed by SDS-polyacrylamide gel electrophoresis on 8 -25% acrylamide gel using the Phast System (Pharmacia Biotech Inc.). Cholesterol and phospholipid concentrations in pure isolated LpA-I particles were determined using enzymatic assays (Boehringer Mannheim). The concentration of apoA-I was measured by Lowry assay using bovine serum albumin as standard (28).
Monoclonal Antibodies-Antibodies 4H1, 2F1, 3G10, and 5F6 were prepared and previously characterized in our laboratory (29). Antibodies 2G11 and 4A12, first characterized by Petit et al. (30), were purchased from SANOFI Inc (Paris, France), whereas antibodies A05, A51, A16, A11, A17, A03, A07, and A44, also previously reported (20), were produced by the Institut Pasteur, Lille. The control mAb 2H2 is an antibody against rat synthetic atrial natriuretic factor (31). All mAbs were murine IgG, purified on protein G-Sepharose or protein A-Sepharose (Pharmacia) and proven free of murine apoA-I (not shown). The location of epitopes recognized by all these mAbs is summarized in Fig. 1.
Solid Phase Radioimmunoassay of apoA-I-Solid phase radioimmunoassays were done as described previously (21). Briefly, Immulon II Removawells were coated with 0.2 g of apo-HDL in 100 l of 15 mM Na 2 CO 3 /35 mM NaHCO 3 , pH 9.6, containing 0.02% NaN 3 , washed, and then saturated with 250 l of gelatin (0.5% in PBS, pH 7.2, and 0.02% NaN 3 ). Anti-apoA-I mAb (predetermined dilution) was mixed with serial dilutions of the competitive antigen (LpA-I) in reaction buffer (0.1% gelatin/0.02% NaN 3 in PBS, pH 7.2) and incubated in the coated and saturated wells for 1 h at room temperature. After three washes with 0.05% Tween and 0.02% NaN 3 in PBS, pH 7.2, 100 l of 125 I-labeled rabbit anti-mouse IgG in reaction buffer was added, and the wells were incubated for another 1 h at room temperature. Finaly, the wells were washed (3 times) and the cpm counted. Results were expressed as B/B 0 , where B and B 0 represent the cpm bound in the presence and absence of competitive antigen.
Competition between Anti-apoA-I mAbs-Competitive immunoassays between two antibodies were modified from those previously described (21). Immulon II Removawells were coated with 100 l of capturing mAb (also serving as one of the competing mAbs) diluted to a concentration of 5 g/ml with PBS containing 0.02% NaN 3 , pH 7.2, and incubated overnight at 4°C. After discarding the solution, the wells were washed once with PBS containing 1 mM EDTA and then saturated with 250 l of 0.5% gelatin in PBS, pH 7.2, for 1 h. In a 96-well microtiter plate (Linbro), previously saturated, serial dilutions (0.5% gelatin in PBS/EDTA) of competitor mAbs (starting concentration 60 g/ml) at 100 l/well were prepared; apoA-I in LpA-I was previously iodinated by Iodo-Bead technique (Pierce), and 50 l of 125 I-labeled LpA-I particles (3 ϫ 10 5 cpm/well) were mixed with competitor mAbs and incubated for 2 h. The reaction mixtures were then transferred to the precoated Removawells for an additional 2 h of incubation. The Removawells were washed five times with PBS/EDTA and counted. Results were expressed as B/B 0 (where B and B 0 represent the cpm in the presence and absence of competing antibody, respectively) at various molar ratios between competitor and capturing mAbs.

Characterization of Discoidal Lp2A-I Varying in Lipid
Composition-Five discoidal Lp2A-I particles were prepared using different starting POPC/apoA-I ratios and isolated as described by others (23). After isolation, they were analyzed in terms of homogeneity, size, and lipid composition (Table I). Chemical cross-linking of apoA-I shows that all particles contain two molecules of apoA-I (data not shown); all particles also exhibit only one band on nondenaturing gradient gel electrophoresis, which corresponds to homogeneous LpA-I preparations (not illustrated). As expected, the increase of POPC content from 31 to 156 molecules/apoA-I is positively correlated with the particle sizes, ranging from 8.8 to 10.2 nm (Table I). The sizes obtained are in general agreement with those reported for particles prepared with the same method (23).
Five other particles, varying in cholesterol content, have also been prepared (Table I). They show homogeneity on gradient gel electrophoresis (not illustrated). As shown in Table I, only the cholesterol composition is significantly different, ranging from 0 to 11 molecules/apoA-I; their sizes are not statistically different from what is obtained with similar POPC content alone.
Immunoreactivity of ApoA-I in Discoidal Lp2A-I Varying in Size-The immunoreactivities of 13 antibodies, directed against epitopes spanning most of the apoA-I sequence with isolated Lp2A-I varying in size are shown in Fig. 2A. The ED 50 calculated from displacement curves obtained with each Lp2A-I demonstrate significant difference in all domains of apoA-I.
In the N-terminal domain, the epitope for mAb 4H1 (residues 2-8) is progressively less reactive as the particle size of Lp2A-I increases. In contrast, four antibodies A16, A51, A05, and 2G11, which are specific for epitopes further in the N-terminal region and which overlap differently between residues 8 and 100 (Fig. 1), are more immunoreactive with large than small Lp2A-I.
In the middle of apoA-I, out of four epitopes tested (3G10, A11, 5F6, and A03), only the two central ones are modified when the particle size increases. These are the epitopes for mAbs A11 and 5F6, which overlap between residues 118 and 132 and show a decreased immunoreactivity ( Fig. 2A). Interestingly and logically, the two other epitopes (3G10 and A03) surrounding that region are unchanged. This is in agreement with what we observed previously for the same epitopes of apoA-I with another panel of reconstituted discoidal Lp2A-I, which, however, differed both in size and composition (21).
Therefore the present results prove that these changes in immunoreactivity are entirely related to a change in particle size and packing of phospholipids.
In the C-terminal half of apoA-I, we have tested three antibodies, two (A07 and A44) that react with overlapping epitopes situated between residues 149 -186 and 4A12, which recognizes the sequence 173-205. Despite the significant overlap of the two first epitopes, A07 is significantly less immunoreactive with the larger LpA-I as compared with A44, which is more reactive, a phenomenon previously reported (21). The epitope for mAb 4A12 is clearly unaffected by the size. As a consequence of this observation, we can now attribute the changes in reactivity of 4A12 with Lp2A-I particles observed in previous experiments (6,21) to changes in lipid composition rather than to changes in particle size.
The results obtained with the competitive RIA demonstrate that increasing the size of the discoidal particles induces a major change in the conformation of the entire apoA-I molecule as specific domains distributed along most of the sequence show modifications of their immunoreactivity. However, each domain appears differently affected at different stages of particle enlargement ( Fig. 2A). Starting with changes occurring between the small discs (ratios 31/1 to 94/1), the central part of apo A-I represented by the epitopes 5F6 and A11 seems modified with the initial size increments; with the larger discs (122/1 and 156/1), the changes of immunoreactivity appear predominantly in the N-terminal domain (A16, A51, A05, and 2G11), where we observe high increases in immunoreactivities, but excluding the extreme N-terminal epitope, 4H1, which exhibits a very significant decrease in immunoreactivity. The epitopes (A44 and A07) toward the C terminus show less drastic and more progressive modification of their immunoreactivity than epitopes in other domains. Furthermore, related to those changes of immunoreactivity, few RIAs have displayed significant differences in the calculated slopes, reflecting little differences in binding affinities of most mAbs for their specific epitopes with the various particles (Table II). Only in the largest particles (ratio 156/1) do we note significant changes in antibody affinities (mAbs 4H1, 2F1, and A16).
Comparison of the Immunoreactivity of apoA-I in Lp2A-I Varying in Cholesterol Content-Analysis of the immunoreactivities of this Lp2A-I panel shows very different results compared with what are obtained with varying POPC content. In the N-terminal region, the immunoreactivity of the epitope for 4H1 increases significantly with the addition of 2 mol or more of cholesterol, and in the central domain, we observed highly significant increases in immunoreactivity for only two overlapping epitopes, 5F6 and A11 (Fig. 2B). All of these changes of immunoreactivity are opposite to those that were obtained for FIG. 1. Epitope map of apolipoprotein A-I. The positions of epitopes recognized by these mAbs have been previously defined (20). The names of mAbs used in this study are placed above the solid bars, which represent the sequences recognized by these mAbs. The dashed lines on either side of the bars indicate that the antigenic recognition at this site may extend further.  b Discoidal complexes prepared by cholate dispersion/Bio-Beads removal method; composition after purification by Superose 6 chromatography. The number of apoA-I molecules/particle obtained by crosslinking with dimethyl suberimidate was two for each preparation. the same epitopes when Lp2A-I phospholipid content increases ( Fig. 2A). The immunoreactivities reached by 5F6, 4H1, and A11 with the discs containing the highest amount of cholesterol are of the same magnitude as those obtained with the small (4H1 and 5F6) and medium (A11) size discs, respectively, without any cholesterol. Except for 4A12, which does not show any change in immunoreactivity, the effect of cholesterol on those epitopes is quite similar to data that we obtained with native spherical HDL (6). Finally, none of the RIAs carried out with the available antibodies have displayed any statistically significant difference in the calculated slopes, suggesting that cho-lesterol induces no change in the relative binding affinities of these mAbs (Table III).
Competition between Antibodies for Binding to Lp2A-I-To obtain information on the relative position of epitopes as they appear on apoA-I associated with Lp2A-I, we have analyzed the competitions between pairs of mAbs for binding, first to an Lp2A-I with a ratio of POPC/apoA-I of 80/1, which is representative of a median-size Lp2A-I. The results are summarized in Table IV. Most competitions observed are compatible with the notion that mAbs that bind to epitopes that are close on the primary sequence should compete. This is the case for mAbs 2F1, A05, and 2G11 in the N-terminal domain and for their competition with 3G10, which binds to an epitope of the central domain. It is also the case for the competitions between mAbs binding to epitopes of the central domain (3G10, 5F6, and A03) and mAbs binding to epitopes of the adjacent C-terminal region (A07, A44, and 4A12). However to logically interpret the competitions between A05 and A16 binding to the N-terminal domain and A07 and 4A12 binding to the C-terminal domain, we need to take into account not only intramolecular competitions but also the presence of two apoA-I in Lp2A-I and the possibility of intermolecular competitions (see "Discussion").
Effect of Increasing the Lp2A-I Phospholipid/ApoA-I Ratio on the Competition between Antibodies-The competitions among pairs of mAbs for binding to Lp2A-I have been studied in particles where the POPC/apoA-I ratio increases progressively from 31/1 to 156/1, and the results are summarized in Fig. 3 (A-E). As expected the changes in the phospholipid to apoA-I ratio cause specific modifications in the competitions, reflecting the structural modifications induced on apoA-I bound to a progressively enlarging lipoprotein.
The increasing phospholipid content causes either decreases or increases in the competition between specific mAbs, and in the absence of any major change in the affinity of these mAbs (Table II), these changes are largely related to lipid-induced modifications of apoA-I tertiary structure that change the relative position of epitopes. However, in some cases, as we will discuss below, the binding of mAbs to Lp2A-I also appears to contribute to the conformational modification of apoA-I. There are few and rather limited structural modifications within the N-terminal domain. There is an increase in the competition of 2F1 with A05 (Fig. 3B). Further downstream, the central domain represented by the epitope for 3G10 appears to move closer to the N terminus as competitions increase between 3G10 and A05 or 2F1 (Fig. 3, A and B) and between 4H1 or A16 and 5F6 (Fig. 3D). At the same time there are increasing FIG. 2. Immunoreactivity of apoA-I epitopes in Lp2A-I as a function of increasing POPC content (A) or increasing cholesterol content (B). The composition of each Lp2A-I preparation is given in the insets, and the epitopes are positioned on the x axis as they appear in apoA-I sequence. They can be grouped as representative of the N-terminal (4H1 to 2G11), the central (3G10 to A03), and the C-terminal regions (A07 to 4A12). Each bar represents the ED 50 (inversely proportional to immunoreactivity) with its standard deviation (a, b, c, and d indicate the decreasing degree of significance relative to ED 50 for the first particle: a, p Ͻ 0.05; b, p Ͻ 0.01; c, p Ͻ 0.005; d, p Ͻ 0.001).
Ϫ1. competitions between the mAbs reacting to the central epitope 3G10 and the C-terminal epitopes A07 (not illustrated) and 4A12 (Fig. 3E), suggesting an increasing proximity for these domains as well. The above evidence, which indicates that Lp2A-I size increase appears to be accompanied by a conformational change of the molecule where the central domain appears to come closer to both the N-terminal and the C-terminal domains, is somewhat unexpected. However, as we will analyze these results under "Discussion," we attribute these changes to the increasing interaction of the ␣-helix 99 -121 that contains the 3G10 epitope with the lipids, thus causing a change in the orientation of 3G10 bound to its epitope.
As the phospholipid to apoA-I ratio increases, enhancement of binding of one mAb (i.e. the capture antibody) by another mAb has also been observed. In the case of 3G10 as the capture mAb, most mAbs reacting in the N-terminal region enhance its binding of Lp2A-I and in the largest particles this reaches significance for A16, A05, and 2G11 (Fig. 3C). These enhancements reflect conformational changes of apoA-I induced by the binding of the competing mAb, and these have been shown to be associated with modifications of LpA-I interactions with lecithin:cholesterol acyltransferase (22) or cholesterol transfers (32).
Effect of Increasing the Lp2A-I Cholesterol to ApoA-I Ratio on the Competition of Antibodies-Increasing the cholesterol content of Lp2A-I to 11 mol/apoA-I while maintaining POPC con-tent and size of Lp2A-I constant modifies only the immunoreactivity of 4H1, A11, and 5F6 (Fig. 2B). When we compared the competition between mAbs for binding to Lp2A-I with or without cholesterol, very unique changes were observed (Tables IV  and V). Most remarkably, the presence of cholesterol causes absolutely no change in the competitions between most pairs of mAbs but provokes a significant decrease in the enhancement normally exerted by four mAbs reacting to overlapping Nterminal epitopes on 3G10 binding (Table V). This appears to reflect a specific change in the relative position or interaction of the N-terminal domain in relation to the central domain where 3G10 is located, maybe as a result of a direct interaction of cholesterol with these domains or modified packing of the lipid phase.

DISCUSSION
Previous immunochemical studies from this laboratory on the effect of lipid composition on apoA-I structure in native HDL (6) and in discoidal reconstituted LpA-I (21) have demonstrated the significant effects of phospholipids and cholesterol in distinct domains. In discoidal LpA-I, the variation in phospholipid/apoA-I ratio causes an increase in epitope expression in two large domains of the N-terminal and C-terminal regions (spanning residues 14 -90 and 148 -209, respectively). Between these domains, the epitopes of a central region (residues 99 -143) decreased in immunoreactivity as the particle size increased (21). These observations first suggested to us (20) that these regions might constitute the two lipid-binding domains separating a hinge domain postulated by others to explain the existence of discrete size populations in HDL particles (16).
Here the progressive changes in the concentrations of either phospholipids or cholesterol in defined Lp2A-I provide a precise analysis of their respective effect on apoA-I conformation and confirm the notion of its three domain structure.
Effect of Phospholipids on the Immunoreactivity and Conformation of ApoA-I-As the POPC content and size of Lp2A-I increase, significant changes in immunoreactivity are observed for almost all epitopes. These results demonstrate that multiple regions are implicated in the structural adaptability of apoA-I in particles of increasing size. A summary of the immunoreactivity of the various epitopes as a function of the particle composition is presented in Fig. 4. Numerous epitopes increase in immunoreactivity with increasing POPC content; they are mostly located in the N-terminal end (A16, A51, A05, and 2G11), with another one in the C-terminal domain (A44). Two epitopes located in the middle area of apoA-I decrease in immunoreactivity (A11 and 5F6). Interestingly, the boundaries of this area are represented by two epitopes (3G10 and A03) that show no change in their immunoreactivity. Previous results (20,21) and the confirmation obtained here both support the The results are expressed as the percentages of competition (ϩ) or enhancement of the binding (Ϫ) and are presented as the means Ϯ S.D.
concept that the two central epitopes A11 and 5F6 probably identify a central mobile domain, constituted by a hinged pair of helices and located between the epitopes 3G10 and A03 (adjacent to ␤-turns), which seems involved in the process of disc expansion. For reasons still not understood, the expression of the extreme N-terminal epitope 4H1 parallels the variation in immunoreactivity of the epitopes of the central domain (A11 and 5F6), suggesting, as we discuss below, that the two share some structural and functional characteristics. In the N-terminal third of apoA-I, for which we earlier proposed a complex and condensed structure based on its high content of multiple overlapping discontinuous epitopes (20), all epitopes display very significant increases in immunoreactivity as the size of Lp2A-I increases. This effect could be mediated by the formation of short helices to accommodate the increasing disc size. Several studies have documented the increase in ␣-helix content with the size increase (33), which can be related to increase in the number of helical repeats (34) or to the formation of new short segments (23). The changes in lysine pK a reported by Sparks et al. (23) also implicated major changes in the N-terminal lysines that reflect the structural rearrangements of this domain with the increase in particle size.
Effect of Cholesterol on the Immunoreactivity of ApoA-I-In contrast to the widespread conformational changes introduced by increasing the phospholipid content, the changes in epitope immunoreactivity observed with increasing cholesterol content are very limited to a short area of the central domain, identified by epitopes A11 and 5F6, and to the extreme N-terminal epitope, 4H1 (Fig. 4). All increases in immunoreactivity of these epitopes are proportional to the cholesterol content. It must be noted that the change of immunoreactivity in response to cholesterol is opposite to that observed with POPC, demonstrating that the two lipid have distinct interactions with or effects on the protein resulting in different conformations (Fig.  4). Cholesterol can act directly by interaction with apoA-I and/or indirectly by modification of the physical state of the phospholipids. Our observations could be compatible with the suggestion of others (35, 36) that cholesterol exerts its effect directly by binding to apoA-I, and this should help us to localize the site of this interaction. Sparks and colleagues (37) have observed that although the increase of the phospholipid/apoA-I ratio in LpA-I increases the ␣-helicity of apoA-I, the addition of a small amount of cholesterol decreases ␣-helicity but increases the stability of the remaining helices. Also the addition of 1 mol of cholesterol/apoA-I is sufficient to induce a reduction in the surface potential of the LpA-I (37). This evidence of conformation changes exerted by cholesterol have been interpreted as the result of specific interaction of cholesterol with apoA-I polypeptide chain as first suggested by the energy transfer experiments of Massey et al. (35). Furthermore, titration of lysine pK a values using NMR spectroscopy is also in accord with our present results and demonstrates the differential effect of phospholipid and cholesterol (23,37). Indeed, increase in phospholipid content correlates with widespread lysine modifications, whereas the addition of cholesterol changes the pK a of only one lysine group, tentatively identified as Lys 106 , Lys 107 , Lys 238 , and Lys 239 ; in particles containing 2 mol of cholesterol/apoA-I, only some of the four lysines listed above are affected, whereas in particles with 10 mol of cholesterol/ apoA-I, the titrations of all four lysines are modified. Taken together with these results, the effect of cholesterol on the immunoreactivity of mAbs A11 and 5F6 with Lp2A-I suggests that their epitopes, which overlap between residues 99 and 141, have responded to the same modifications that involved Lys 106 and Lys 107 . In the absence of reporter epitopes for the C terminus, we cannot confirm the putative involvement of lysines 238 and 239. Two naturally occurring mutants involving apoA-I Lys 107 have been described, Lys 107 3 Met, which has a normal ability to interact with lipids, forms discoidal LpA-I and reacts with lecithin:cholesterol acyltransferase (38), and Lys 107 3 0, which forms abnormal LpA-I particles with decreased lecithin:cholesterol acyltransferase substrate activity (39) and decreased binding affinity to adipocytes (40). Therefore, the loss of the charge associated with Lys 107 is not essential to lipid binding properties of apoA-I, but deletion of Lys 107 changes the orientation of side chains in the ␣-helix and disrupts the lipid binding properties. The informative value of the observations cited above has been hindered by the susceptibility of apoA-I to oxidation and denaturation during purification (29,39), especially in the region of residues 99 -121 where multiple epitopes generated by oxidative processes have been demonstrated (29). Others have also shown the importance of apoA-I central domain in cellular cholesterol efflux (8) and in cholesterol esterification (41).
The linkage of the epitope at the N terminus, 4H1, with those in the central domain, A11 and 5F6, based on the similarity of their response to cholesterol is now well documented as a result of this and previous studies (21,42). The specificity of the FIG. 4. Location of apoA-I domains with differently immunoreactive epitopes depending on POPC or cholesterol content of discoidal Lp2A-I. On a planar representation of apoA-I supersecondary structure, the predicted elements of secondary structure are drawn on a line proportional to their position in the primary structure. Amphipathic ␣-helices, ␤-turns, and random coils are respectively indicated with boxes, curvilinear sections, and curves; the numbers along the line and the boxes refer to residues and are intended as approximate reference points for the elements of secondary structure. The location of epitopes is represented by a shaded area underlayed below the elements of secondary structure. The epitopes sharing the same change in immunoreactivity in response to phospholipids or cholesterol are given the same shading. The N-terminal sequence containing the 4H1 epitope has been positioned close to the central domain containing the epitopes for A11 and 5F6 to reflect the close structural (immunochemical) and functional properties of these domains as indicated in the text.  (3) Ϫ26 (4) b Ϫ5 (10) 4 (11) 13 (7) 3G10 49 (4) 38 (7) 82 (6) 23 (13) Ϫ3 (22) 83 (10) A11 6 (7) 2 (5) Ϫ9 (5) Ϫ5 (13) 7 (6) 7 (5) 5F6 Ϫ2 (6) Ϫ3 ( a The results are expressed as the percentages of competition (ϩ) or enhancement of the binding (Ϫ) and are presented as the means Ϯ S.D. b Indicates a significant difference (p Ͻ 0.01) compared with Lp2A-I without cholesterol in Table IV. cholesterol effect on the expression of the three epitopes is clear from the present study (Fig. 4). We have also shown previously that mAbs reacting with these three epitopes exert a similar enhancement on the reaction of lecithin:cholesterol acyltransferase with small Lp2A-I (22), a property that we tentatively attributed to conformational modifications of apoA-I that affect either its interaction with lecithin:cholesterol acyltransferase or the accessibility of cholesterol to the enzyme. In another study, we have shown that only 4H1, 3G10, and 5F6 (A11 was not included in the study) were found to have no effect on the passive transfer of cholesterol between LpA-I and low density lipoprotein, whereas all other mAbs tested (A05, A16, A51, 2G11, A03, A44 and 4A12) enhanced the transfer of cholesterol relative to that of phospholipids from Lp2A-I to low density lipoprotein (32). These results have been interpreted as evidence that the N terminus containing 4H1 shares a common structural association with the central domain containing 5F6. We postulate that the folding of the N-terminal region brings the N-terminal sequence in close proximity with the central domain located between residues 99 and 121, as shown on the apoA-I model in Figs. 5 and 6. The effects of cholesterol may be explained either by its heterogeneous distribution within the lipid phase and in relation to apoA-I chain or by changes in lipid phase fluidity or by a combination of both.

Competition among Pairs of mAbs as a Function of Lp2A-I Content in Phospholipids and Cholesterol: Information Derived on Changes in ApoA-I Conformation and Cooperativity between
Adjacent ␣-Helices-Our first report on the use of antibody competitions to derive information on the relative position of apoA-I epitopes and thus on the tertiary structure was based on the use of mAb 4H1 binding at the extreme N terminus as capture antibody for the Lp2A-I after reaction with two other competing antibodies (21). This study provided the first indirect evidence for the existence of a hinge domain in the central region of apoA-I that would explain the decrease in the competitions observed between mAbs in the small Lp2A-I (7.8 nm) compared with larger counterparts (9.6 nm). Here we have extended the initial observations with a more detailed study where the constituents of Lp2A-I are varied independently and where the competition assay has been modified to include only two competing mAbs without the complication of the possible competitions between the capture mAb and the other two antibodies.
For the interpretation of these competitions, we can consider that the observed competitions are either exclusively intramolecular or both intramolecular and intermolecular depending on the position of the epitopes and the relative position of the two apoA-I at the surface of the particle (Fig. 5). Considering first the competitions observed in a constant composition mode with the median size Lp2A-I, the results are consistent with the known position of epitopes in the primary sequence and suggest a rather compact structure with competitions within the N-terminal region, competitions between epitopes in the central domain and epitopes in the N-or C-terminal regions, and competitions within the C-terminal region. Most of these competitions can be logically interpreted as intramolecular competitions as presented in the model of Lp2A-I (Fig. 5). However, the competitions between mAbs A16 and A05, which react in the N-terminal domain, and mAb 4A12, which reacts with the farthest epitope in the C-terminal domain, are not compatible with an intramolecular competition in the apoA-I model of Fig. 5. However, in this Lp2A-I model where we only assume that the apoA-I follow one another in an N-to Cterminal arrangement around the discoidal particle, we can easily propose that the competitions between these mAbs are intermolecular. This model does not require a particular orien-  Fig. 4, and the N-terminal sequence including the 4H1 epitope has been positioned close to the central domain to reflect their common properties. Well defined epitopes are represented by different shaded areas underlayed below the elements of secondary structure where they are located. Several overlapping epitopes are grouped in the N-and Cterminal regions. The arrows linking epitopes identify a mAb competing significantly with another mAb, and the double-headed arrows indicate reciprocal competitions (data derived from Fig. 3, A-E). The intramolecular competitions are compatible with the position of the epitopes involved that are adjacent on the primary sequence. The existence of intermolecular competitions between mAbs reacting at N-and C-terminal epitopes supports the model proposed for the N-to C-terminal arrangement of the two apoA-I around the disc.
FIG. 6. Effects of increasing Lp2A-I POPC content on the observed competitions among pairs of mAbs. The planar model of apoA-I in an Lp2A-I particle is the same as that in Fig. 5. However, the arrows linking the epitopes recognized by competing mAbs indicate not only the existence of a competition or an enhancement of binding but also either a significant increase in the competitions (A) or a significant increase in enhancement of binding (B) as a function of the increase in POPC content and/or in the size of the particle. These changes, which are most evident in the central domain (residues 99 -121) containing the epitope for 3G10, are interpreted as evidence that as the particle enlarges, the orientation of this domain changes to become more closely associated with lipids. tation of the central domain and logically accommodates all competitions based on the positions of epitopes on the primary sequence. Therefore, these results are compatible with the model and provide the first indirect evidence documenting that the relative arrangement of apoA-I molecules in a discoidal Lp2A-I should be in an N-to C-terminal belt model around the disc.
As the ratio of phospholipids to apoA-I increases in Lp2A-I, the changes occurring in the competitions among pairs of mAbs are summarized in Fig. 6. The major changes involve antibody 3G10, whose competition with all other mAbs reacting at adjacent epitopes increases. There are also significant changes in the N-terminal domain where the competitions of 4H1 with 5F6 and 2F1 with A16 increase (Fig. 6A). The increase in phospholipid to apoA-I ratio is also accompanied by the enhancement of 3G10 binding by most mAbs reacting in the N-terminal domain (4H1, A16, A05, and 2G11) and by A03 reacting at an epitope downstream. In the absence of any significant modification of the affinity of the antibodies, these variations in immunoreactivity that appear centered on mAb 3G10 can be interpreted as follows. As the Lp2A-I particle enlarges, it is commonly assumed that more amphipathic ␣-helices become associated with the lipid phase (7, 14) by a process probably involving a hinged domain (16,21). Our results suggest that the epitope for 3G10, which spans the ␣-helix between residues 99 and 121, either represents or is contained in the hinge domain possibly constituted by the two central ␣-helices between residues 99 and 143. In small Lp2A-I the hinged domain is repulsed away from the lipid phase for lack of stabilizing interactions with adjacent helices, possibly also because of charge repulsions involving the proximity of glutamic and aspartic acid residues on the adjacent faces of the helices between residues 68 and 88 and between residues 100 and 119 (17). The particular position of these glutamic acid residues has also been proposed (43) to be important for lecithin:cholesterol acyltransferase activation. We also propose that the presence of a class Y amphipathic ␣-helix with an intermediate mean hydrophobic moment and an intermediate nonpolar face hydrophobicity (17) facilitates the repulsion of the hinged domain away from the lipid phase. The increase in POPC content of Lp2A-I facilitates the interaction of helix 99 -121 with the lipid phase, which in turn increases the competition of mAb 3G10 binding at this site with all mAbs reacting at adjacent epitopes. Simultaneously because of the decreased constraints in the center of the molecule nearly all mAbs reacting with adjacent epitopes upon binding to Lp2A-I can contribute to increase the interaction of helices with the lipids, including that of helix 99 -121, which increases both helix stability and binding of 3G10 to its epitope. We have earlier provided evidence for the increase in ␣-helical content of Lp2A-I upon binding of most mAbs (32), which supports this interpretation. The mobility of the domain containing the 3G10 epitope is also supported by its sensitivity to proteolytic cleavage (44) and by its high antigenicity (20). During this process of particle enlargement, other conformational modifications occur as indicated by the increase in the competitions between A03 and 2F1 and between A07 and 2F1.
This study implies a great sensitivity of apoA-I conformation to the lipid composition of reconstituted LpA-I. The modifications of these different domains, described here by an immunochemical approach, probably reflect the difference in hydrophobicity of the pairs of amphipathic ␣-helices distributed along the apoA-I sequence (17) and their ability to form stable structure with the adjacent helices. In small Lp2A-I with a limited amount of POPC, the central domain may be excluded from direct lipid interactions because of its low lipid affinity. As the amount of POPC increases, the interaction of that pair of helices with lipids is enhanced and a larger disc is formed also causing a reorganization of the N-terminal region. In contrast with increasing cholesterol, the ␣-helix content decreases (37) and the central domain is probably unstable, giving a higher expression of the central epitopes (more evident with 5F6 than A11, Fig. 2B).
Other evidence for the conformational sensitivity of the ␣-helix 99 -121 containing the epitope for 3G10 is the effect of increasing cholesterol content in Lp2A-I, which decreases the normal enhancement of 3G10 binding by mAbs binding to the N-terminal domain. Indeed, cholesterol would be expected to increase the acyl chain order and thus decrease the mobility of the lipid-associated protein domains (11,44).
In conclusion, this study confirms the notion that we previously introduced of a three domain structure for apoA-I, an N-terminal domain extending up to residue 99, a central domain centered between residues 99 and 121 but possibly extending to residue 143, and a C-terminal domain. We have demonstrated that variations in the lipid components of discoidal reconstituted Lp2A-I can have significant effects on the conformation of the entire apoA-I, particularly in its central domain. Cholesterol and phospholipid contents have specific and opposite effects on the conformation of this region of apoA-I, which has been suggested to be important in lecithin: cholesterol acyltransferase activation (22,41) and may also be implicated in apoA-I-mediated efflux of cellular cholesterol (8). Thus the lipid composition of LpA-I can modulate its metabolism through changes of apoA-I conformation. This appears to be mediated not only by the amount of cholesterol but also by the size of the disc. We also confirm here the existence of a link (structural, functional, or both) between the extreme N terminus, represented by the epitope for 4H1, and the central domain. Both appear significantly involved or affected by cholesterol. Finally, the domain for which we observed the least modification of conformation in this study is located in the C-terminal half of the protein, a region involved in lipid binding and possibly also in interactions with cellular binding sites. Thus, apoA-I conformation appears to be crucial in HDL metabolism, with different domains having specific functions.