Antibodies to the Carboxyl Terminus of Human Apolipoprotein A-I THE PUTATIVE CELLULAR BINDING DOMAIN OF HIGH DENSITY LIPOPROTEIN 3 AND CARBOXYL-TERMINAL STRUCTURAL HOMOLOGY BETWEEN APOLIPOPROTEINS A-I AND A-11*

We have studied the binding of ‘2SI-labeled high density lipoproteins (HDLJ) to liver plasma membranes, which are thought to contain specific HDL receptor sites, using anti-peptide antibodies directed against two sites in the carboxyl-terminal region of human apoA-I. Two distinct antibody populations raised to peptides corresponding to amino acid residues 205-220 and 230-243, respectively, recognized regions of apoA-I that are exposed in the lipid environment of HDL3. However, anti-AI[230-2431 IgG, but not anti- AI[205-2201 IgG, recognized HDL2, suggesting that residues 205-220 of apoA-I are expressed differently in the two HDL populations. In addition, anti-AIr230-2431 IgG showed strong cross-reactivity toward apoA-11. Epitope mapping studies showed that anti-AIr230- to an epitope located in the carboxyl-ter-minus of apoA-11, significant structural homology between the carboxyl-terminal of apoA-11, significant structural homology be- tween the carboxyl-terminal regions of apoA-I and A-11, two candidate proteins for mediating the specific cellular interaction of HDL3. Fab fragments from anti-AI[205-2201 and the binding of ’“I-HDL3 to plasma by approx- imately and istry on an Applied Biosystems Model 430A peptide synthesizer. To facilitate conjugation of the peptides to carrier proteins, Cys-Gly spacers were added to the amino-terminal ends. Cleavage and depro- tection were performed using hydrogen fluoride. The crude peptides were then purified by reversed-phase (RP)-HPLC, and the amino acid sequences were validated using an Applied Biosystems Model 470A protein Sequencer, equipped with an on-line Model 120A PTH analyzer. For the immunization procedures described below, peptides were coupled to ovalbumin (OVA) with maleimidobenzoyl-n-hydrox- ysuccinimide ester (Pierce). The specificity of each antisera for AI[205-2201 and AI[230-243] were determined by ELISA (described below) using the peptides coupled to keyhole limpet hemocyanin (KLH) as antigens. To avoid cross-reaction with maleimidobenzoyl- n-hydroxysuccinimide ester, the coupling agent succinimidyl 3-(2-pyridy1dithio)propionate (Pierce) was used to prepare the KLH con- jugates. Production and Purification of Anti-peptide Antibodies-Antisera against the peptides AI[205-2201 and AI[230-2431 (conjugated to OVA) were produced in 5-6-month-old rabbits which received 0.5 mg of conjugate emulsified in 1 ml of phosphate-buffered saline (PBS), pH 7.4, and Freund's complete adjuvant, by subcutaneous injection at six different sites. The animals were boosted 3 weeks after the first immunization and bled 2 weeks thereafter. Each antiserum was subjected to affinity chromatography on Protein A-Sepharose (Phar-macia) (33). Antibody eluted from Protein A was dialyzed against PBS, then passed down an OVA-Sepharose 4B column to remove anti-OVA reactivity (determined by ELISA). Specific antibody was then eluted from a human apoA-I-Sepharose Results were expressed as a percentage of the total lZ5I-labeled HDL3 bound in the absence of antibody. Other Procedures-Isoelectric focusing was performed in 7.5% polyacrylamide gels containing 6.8 M urea and 2% ampholine pH 4-6.5 (LKB-Pharmacia, Sweden), using a Bio-Rad Mini Protean I1 Dual Slab Gel apparatus. Protein bands were stained with 0.1% Coomassie Brilliant Blue R-250 in 45% ethanol, 10% acetic acid. Lipoprotein and protein concentrations were determined according to the method described by Lowry et al. (42) using bovine serum albumin as standard; peptide concentrations were determined by amino acid analysis.

Synthetic peptides representing selected regions of a protein sequence can elicit antibodies capable of reacting with the whole protein (21,22). We have generated two populations of anti-peptide antibodies recognizing residues 205-220 and 230-243 of human apoA-I, respectively. This report describes the ability of these site-specific antibodies to inhibit the binding of HDLs to liver plasma membranes, which are thought to contain specific receptors for HDL3 (2,23). These studies extend our recent findings using purified cyanogen bromide digest fragments of apoA-I (24,25) and suggest that residues 205-243 contain, or form part of, the binding domain of HDL3. We propose that the interaction of HDL3 with liver plasma membranes is mediated by HDL receptor sites which are specific for a region in the carboxyl-terminal portion of apoA-I.
Peptide Synthesis and Conjugation-Two peptides, denoted AI[205-2201 and AI[230-2431, were synthesized using the human apoA-I sequence reported by Brewer et al. (31) and correspond to the amino acid residues 205-220 and 230-243, respectively. Both peptides represent regions of apoA-I that are predicted to be exposed in a lipid environment (32). The peptides were synthesized using t-Boc chem-istry on an Applied Biosystems Model 430A peptide synthesizer. To facilitate conjugation of the peptides to carrier proteins, Cys-Gly spacers were added to the amino-terminal ends. Cleavage and deprotection were performed using hydrogen fluoride. The crude peptides were then purified by reversed-phase (RP)-HPLC, and the amino acid sequences were validated using an Applied Biosystems Model 470A protein Sequencer, equipped with an on-line Model 120A PTH analyzer. For the immunization procedures described below, peptides were coupled to ovalbumin (OVA) with maleimidobenzoyl-n-hydroxysuccinimide ester (Pierce). The specificity of each antisera for AI [205-2201 and AI[230-243] were determined by ELISA (described below) using the peptides coupled to keyhole limpet hemocyanin (KLH) as antigens. To avoid cross-reaction with maleimidobenzoyln-hydroxysuccinimide ester, the coupling agent succinimidyl 3-(2-pyridy1dithio)propionate (Pierce) was used to prepare the KLH conjugates.
Production and Purification of Anti-peptide Antibodies-Antisera against the peptides AI[205-2201 and AI[230-2431 (conjugated to OVA) were produced in 5-6-month-old rabbits which received 0.5 mg of conjugate emulsified in 1 ml of phosphate-buffered saline (PBS), pH 7.4, and Freund's complete adjuvant, by subcutaneous injection at six different sites. The animals were boosted 3 weeks after the first immunization and bled 2 weeks thereafter. Each antiserum was subjected to affinity chromatography on Protein A-Sepharose (Pharmacia) (33). Antibody eluted from Protein A was dialyzed against PBS, then passed down an OVA-Sepharose 4B column to remove anti-OVA reactivity (determined by ELISA). Specific antibody was then eluted from a human apoA-I-Sepharose 4B column using 0.1 M glycine HC1, pH 2.8, and neutralized immediately with 1 M Tris buffer, pH 8.0. Purified human apoA-I and OVA (Sigma) were coupled to CNBr-activated Sepharose 4B as recommended by Pharmacia. Antibody specific for apoA-I (designated anti-apoA-I) was also obtained from rabbits immunized with purified human apoA-I, as previously described (3). For the production of Fab fragments, affinitypurified antibody was subjected to papain digestion according to the procedures of Gorini et al. (34). Fab fragments were removed from undigested IgG and Fc fragments by passage through Protein A-Sepharose. The activity and purity of isolated Fabs were assessed by ELISA and SDS, 10-15% polyacrylamide gradient gel electrophoresis (Phastgel system, Pharmacia), respectively. Fab fragments were also prepared from two apoA-I-specific monoclonal antibodies, denoted AI-1 and AI-3, which recognize epitopes exposed on the surface of HDL:, particles (35) that have recently been localized to residues 28-47 and 140-147 of apoA-I, respectively (36).
Enzyme-linhd Zmmunosorbent Assays-Titration curves of the anti-peptide antibodies toward different antigens were determined by ELISA. Briefly, 96-well plates (Immulon 11) were coated with 10 pg/ ml antigen in 0.05 M sodium carbonate buffer, pH 9.6, for 1 h, a t room temperature (100 pl/well). The wells were then washed three times in PBS containing 0.05% Tween 20, followed by the addition of 100 pl of serially diluted (1:2, from 50 pg/ml IgG), affinity-purified antibody. After 1 h at room temperature, the wells were washed as before, then incubated for another 1 h with 100 pl of goat anti-rabbit IgG (H + L) horseradish peroxidase conjugate (Bio-Rad) diluted 1 in 2000 in PBS/O.O5% Tween 20. After three washes with PBS-Tween, 100 pl/well of 0.1% ABTS (2',2-azinobis(3-ethylbenzothiazoline-6sulfonic acid)), 0.02% H202 in 0.1 M citrate buffer, pH 4.0, was added for 30 min, and the color which developed was quantitated using a Titertek Multiscan (Flow Laboratories) with a filter setting of 414 nm. To compare antibody reactivities toward isolated apoA-I, HDLZ, or HDL:!, a competitive ELISA system was employed that followed the same coating and washing procedures as described above; however, the incubation buffer included l % skim milk rather than Tween 20, which has been shown to alter the immunoreactivity of apoA-I in HDL (18,19,35). These duplicate wells of plates coated with apoA-I received 50 pl of serially diluted (1:2) antigen, followed by 50 pl of anti-peptide antibody (diluted 1/1000), for 1 h at room temperature. Specifically bound antibody was then detected as described above.
Binding Studies-Binding assays were performed in triplicate. Assay tubes contained 200 pg of rat liver plasma membrane protein, 0.2 pg of 1251-labeled HDL3, and varying amounts of purified Fab fragments in a final volume of 200 pl. Incubations were performed in buffer containing 100 mM NaCI, 50 mM Tris-HCI, 0.01% EDTA, and 0.1% casein, pH 7.4. Rat liver plasma membranes were prepared with modifications (23) of the method described by Fleischer and Kervina (39). HDL3 was radiolabeled with ' ' ' I by the McFarlane procedure (40), as described previously (41), to a specific activity of 300-400 cpm/ng. Fab fragments were preincubated with lZ5I-HDL3 for 2 h at 37 "C in 1.5-ml Microfuge tubes (Eppendorf); following the addition of liver plasma membrane, the tubes were incubated for a further 4 h at 37 "C. 170 pl of each incubation mixture was transferred onto a vacuum filter manifold fitted with GF/C glass fiber filters (Whatman) presoaked for 3 h in 0.1% casein. The membranes were washed under vacuum with 6 X 1 ml incubation buffer and transferred to tubes for counting. Nonspecific binding to GF/C filters (measured in the absence of plasma membranes) represented 5-8% of the total counts bound to the filters. Results were expressed as a percentage of the total lZ5I-labeled HDL3 bound in the absence of antibody.
Other Procedures-Isoelectric focusing was performed in 7.5% polyacrylamide gels containing 6.8 M urea and 2% ampholine pH 4-6.5 (LKB-Pharmacia, Sweden), using a Bio-Rad Mini Protean I1 Dual Slab Gel apparatus. Protein bands were stained with 0.1% Coomassie Brilliant Blue R-250 in 45% ethanol, 10% acetic acid. Lipoprotein and protein concentrations were determined according to the method described by Lowry et al. (42) using bovine serum albumin as standard; peptide concentrations were determined by amino acid analysis.

RESULTS
Specificity of the Anti-peptide Antibodies-To select antibodies immunoreactive toward the parent protein, human apoA-I, anti-peptide antibodies were purified by affinity chromatography on apoA-I-Sepharose 4B as described under "Experimental Procedures." The recovery of apoA-I-specific immunoglobulin from antisera raised to peptide AI[230-2431 was 5-fold greater than the amounts recovered from anti-AI[205-2201 sera (data not shown). The antigenic reactivities of antisera and affinity-purified antibody were determined by ELISA.
The purified anti-peptide antibodies bound only to their respective peptides, with no cross-reactivity observed toward the other peptide (Fig. 1). Although both antibody populations recognized the parent protein (apoA-I), of particular interest was the ability of anti-AI[230-2431 IgG to recognize human apoA-I1 with an unusually high cross-reactivity (Fig. 1, panel  B ) . In contrast, anti-AI[205-2201 IgG was unable to bind apoA-11. To determine whether the anti-AI[205-2201 and anti-AI[230-2431 antibodies recognized the prominent apoA-I isoforms found in HDL, immunoblotting was performed following isoelectric focusing of apoA-I (and apoA-11) derived from human HDL (d 1.063-1.210 g/ml). The immunoblot patterns for each antibody were identical, both clearly detecting the major apoA-I isoforms (Fig. 2). Furthermore, the reactivity of anti-AI[230-2431 toward the apoA-I1 isoforms confirmed the cross-reactivity identified by the ELISA method.
particles. However, different reactivities of the anti-peptide antibodies toward isolated apoA-I, HDLz, and HDLB suggested structural differences in the expression of these two regions of apoA-I. Anti-AI[230-2431 showed reactivity toward all three lipid-bound forms of apoA-I, whereas anti-A1[205-2201 had little or no reactivity toward apoA-I in HDL2 (Fig.  3). Both anti-peptide populations were also compared for their ability to immunoprecipitate "'I-labeled HDLB, relative to anti-apoA-I IgG. Assigning an arbitrary 100% for the maximum immunoprecipitation produced by anti-apoA-I IgG, the  Epitope Mapping of ApoA-II Using Anti-AI/230-243]-As described above, anti-AI[230-2431, although raised to a peptide representing a portion of human apoA-I, displayed strong reactivity toward purified apoA-I1 using both ELISA and immunoblotting techniques. Close inspection of the primary structures of apoA-I (31) and apoA-I1 (38) reveals a 41% sequence homology between the last 17 carboxyl-terminal residues of both apolipoproteins (Fig. 4). Most of the region showing this homology is included in the peptide AI12. 30-2431. To confirm that the carboxyl-terminus of apoA-I1 contains the epitope(s) recognized by anti-AI[230-243] IgC, tryptic fragments of apoA-I1 were prepared and used in a competitive ELISA. The RP-HPLC chromatograph of the generated peptides is shown in Fig. 5. The identities of the peaks, determined by amino acid analysis and amino-terminal sequencing, agreed with the expected peptides previously reported by Lux et al. (38). Selected peptides were then compared for their abilities to inhibit the binding of anti-AIl230-2431 to apoA-11, as described under "Experimental Procedures." Only peptide AII[56-771 could reduce the binding of anti-AI[230-2431 to immobilized apoA-I1 (Fig. 6). The inhibition produced by peptide AII[56-77] was identical with that produced by whole apoA-11, confirming that the AI/AII cross- reactivity is due to similarities which reside in the carboxylterminal regions.
Inhibition of "'I-HDL3 Binding to Rat Liver Plasma Membranes-To determine the effects of the antibodies on the binding of HDL3 to rat liver plasma membranes, lZ5I-labeled-HDL, particles were preincubated with each antibody prior to the addition of plasma membranes. Initial studies using whole antibody molecules resulted in enhanced levels of lZ5I-HDLs binding to membranes, presumably due to the formation of antibody-HDL3 aggregates (3). T o avoid this effect, and to minimize the possibility of steric inhibition resulting from the use of whole IgG molecules, Fab fragments were prepared. Fabs of anti-apoA-I, anti-AI[205-2201, and anti-AI[230-2431, with similar reactivities toward apoA-I, all produced inhibition of binding, whereas Fabs from an unrelated anti-peptide antibody with no specificity toward apoA-I, had little or no effect (Fig. 7). Anti-AI[205-2201 and anti-apoA-I at concentrations of 300 pg/ml Fabs reduced the binding of '"I-HDL3 to less than 80% of control values. At similar concentrations, anti-AI[230-2431 inhibited binding by approximately 60%. Under the same conditions, Fab fragments from the two monoclonal antibodies, AI-1 and AI-3, had no significant effect on the binding of lZ5I-HDL3 to the hepatic membranes (Fig. 7).

DISCUSSION
Anti-peptide antibodies generated to two peptides, synthesized from two distinct regions of the carboxyl-terminal region of human apoA-I, were found to inhibit the binding of HDL3 to liver plasma membranes. In addition to supporting the proposal that apoA-I can act as a specific ligand for HDL3 cellular binding sites (2)(3)(4)(5)(6), the present studies further suggest that a specific region in the carboxyl-terminus of apoA-I is responsible for the cellular binding of HDL3. Furthermore, characterization of the specificities displayed by the antipeptide antibodies has identified a region of structural homology between human apoA-I and apoA-11, two proteins previously implicated in mediating the binding of HDL to human (6) and rat (2) liver plasma membranes.
Peptides AI[205-220] and AI[230-2431, corresponding to residues 205-220 and 230-243 of apoA-I, respectively, generated two distinct populations of anti-peptide antibodies which could recognize the major isoforms of apoA-I from HDL (Fig.  2). Anti-AI[205-2201 and anti-AI[230-2431 also recognized apoA-I associated with apoA-I-DMPC and HDL3 particles (Fig. 3), which is consistent with the proposed orientation of apoA-I in the lipid environment (32), in which both regions are thought to contain sites exposed on the lipoprotein surface. Such sites are potentially available for interactions with enzymes, receptors, or other blood components involved in lipid metabolism; therefore, these antibodies may provide useful tools for further probing the structural-functional properties of apoA-I. However, apparent structural differences between the expression of apoA-I in HDL, and HDL3 were identified by the inability of anti-AI[205-2201 to recognize HDLz particles. It is possible that residues 205-220 become hidden in the lipid environment of the larger HDLz particles due to conformational changes of apoA-I, or, alternatively, these residues may be masked by other protein moieties on the particle surfaces.
Anti-AI[205-2201 and anti-AI[230-2431 Fab fragments significantly reduced the interaction between HDL3 and rat liver plasma membranes (Fig. 7), whereas two monoclonal antibodies, AI-1 and AI-3, recognizing epitopes positioned toward the amino-terminal (residues 28-47) and middle (residues 140-147) portions of apoA-I, respectively (37), had little or no effect. Thus, both anti-peptide antibodies may recognize epitopes which contain, or lie close to, a cellular binding domain located in the carboxyl-terminal region of apoA-I. The higher levels of inhibition observed with anti-AI[205-2201 may indicate that residues within 205-220 are more specifically involved in, or lie closer to, the actual binding region. The region of apoA-I recognized by anti-AI[205-2201 IgG is thought to include a p-conformation between two amphipathic a-helical regions (32). Recent studies involving chemical modification of lysine or arginine residues of HDLB (43, Human apoA-I 13261 45), or tetranitromethane treatment of HDL3 (6,(44)(45)(46), suggested that the amphipathic a-helical regions of apoA-I mediate the cellular binding of HDL. The present findings cannot rule out this possibility as anti-AI[205-2201 may hinder the adjacent amphipathic regions from interacting with the cellular binding sites, although the inhibition observed suggests that the cellular binding of HDL3 does not require complete accessibility of all the apoA-I a-helical repeats. Following our previous identification of HDL-binding proteins (putative receptors), present in human and rat liver plasma membranes that recognize HDL, and purified apoA-I (23), and a more recent demonstration that these HDLbinding proteins are specific for cyanogen bromide fragment 4 (residues 149-243) of apoA-I (24,25), we postulated that a binding domain within this region mediates the interaction between apoA-I and specific HDL receptor sites (24). The ability of the anti-peptide antibodies to inhibit HDL, binding to liver plasma membranes is consistent with this proposal.
One unexpected finding from this study was the high level of cross-reactivity shown by anti-AI[230-2431 to apoA-11, detected by means of ELISA and immunoblotting (Figs. 1 and 2). Immunological cross-reactivity between apoA-I and apoA-I1 has been previously reported by Silberman et al. (18), where two monoclonal antibodies showed high reactivity with apoA-I1 and slight cross-reactivity with apoA-I. Comparison between the amino acid sequences of both apolipoproteins shows sequence homology over several amino acid residues at the carboxyl-terminal regions (Fig. 4). By epitope mapping, we then confirmed that the antigenic site of apoA-I1 recognized by anti-AI[230-2431 IgG is contained in the carboxyl-terminal portion (residues 56-77) (Figs. 4 and 5). Previous studies have shown that this region of apoA-I1 corresponds with the most antigenic part of the molecule (47), which is not masked when associated with lipids in HDL. The significance of such structural homology between apoA-I and apoA-I1 is unclear, although it is interesting to speculate that similar structural conformations residing in the carboxyl-terminal portions may account for the ability of apoA-I and apoA-I1 to recognize the HDL-binding proteins identified in both human and rat liver plasma membranes (23).
In summary, the finding that two anti-peptide antibodies, both directed toward the carboxyl-terminus of apoA-I, can inhibit the interaction between HDL, and liver plasma membranes provides strong evidence that the carboxyl-terminus of apoA-I mediates the cellular binding of HDL, possibly through specific plasma membrane receptors. Further, we anticipate that these new antibodies will provide useful tools for immunochemical characterization of previously unknown structural and functional domains of apoA-I involved in lipid metabolism.