Insoluble Complex Formation of Lipoprotein (a) with Low Density Lipoprotein in the Presence of Calcium Ions*

We investigated whether apolipoprotein B-contain- ing lipoproteins could bind to the insoluble complexes of lipoprotein (a) (Lp(a)) induced by Ca2+. Lp(a), but not low density lipoprotein (LDL), very low density lipoprotein (VLDL), or high density lipoprotein3 (HDL3) formed insoluble complexes at physiologic Ca2+ concentrations. Desialylation of Lp(a) dramatically decreased the ability of Lp(a) to aggregate, suggesting that sialic acids on Lp(a) were responsible for forming Ca2+ cross-bridges. Since a reduction of only 30% of the sialic acids on Lp(a) inhibited Ca2+-induced complex formation, it appears that only a small percentage of sialic acids on Lp(a) is involved in Ca2+-induced cross-bridging of Lp(a) particles. To determine whether other lipoproteins would complex to Lp(a) in the insoluble complexes, we mixed Lp(a) with LDL, VLDL, or HDL3 in the presence of Ca2+. Although both LDL and VLDL bound to the Lp(a) in the insoluble complexes, HDL3 not only did not bind, but it also prevented Lp(a) from forming insoluble complexes. LDL bound to Lp(a) in the insoluble complexes in a concentration-dependent manner, eventually reaching saturation at a molar ratio of 5:4 0.01% NaN3 7.2. purity of Lp(a) by SDS- PAGE under nonreducingconditions. Only those fractions completely free of HDL, were pooled. Purifed Lp(a) was stored at 4 "C in 10 mM Tris-HC1 containing 1 mg/ml Na,EDTA at pH 7.2, and other lipoproteins at 4 in 0.15 M NaCl containing mM Na2EDTA at pH


Insoluble Complex Formation of Lipoprotein (a) with Low Density Lipoprotein in the Presence of Calcium Ions*
Akira Yashiro, June O'Neil, and Henry F. HoffS We investigated whether apolipoprotein B-containing lipoproteins could bind to the insoluble complexes of lipoprotein (a) (Lp(a)) induced by Ca2+. Lp(a), but not low density lipoprotein (LDL), very low density lipoprotein (VLDL), or high density lipoprotein3 (HDL3) formed insoluble complexes at physiologic Ca2+ concentrations. Desialylation of Lp(a) dramatically decreased the ability of Lp(a) to aggregate, suggesting that sialic acids on Lp(a) were responsible for forming Ca2+ cross-bridges. Since a reduction of only 30% of the sialic acids on Lp(a) inhibited Ca2+-induced complex formation, it appears that only a small percentage of sialic acids on Lp(a) is involved in Ca2+-induced cross-bridging of Lp(a) particles. To determine whether other lipoproteins would complex to Lp(a) in the insoluble complexes, we mixed Lp(a) with LDL, VLDL, or HDL3 in the presence of Ca2+. Although both LDL and VLDL bound to the Lp(a) in the insoluble complexes, HDL3 not only did not bind, but it also prevented Lp(a) from forming insoluble complexes. LDL bound to Lp(a) in the insoluble complexes in a concentration-dependent manner, eventually reaching saturation at a molar ratio of 5:4 (LDL to Lp(a)). The interaction between LDL and Lp(a) appeared to be ionic, since increases in the positive charge on LDL by desialylation increased this interaction, whereas decreases in positive charge on LDL reduced this interaction. At higher Ca2+ concentrations, the binding of acetyl LDL to Lp(a) in the insoluble complexes was greater than that of LDL. Since more Ca2+ was required for concentration-dependent saturation of acetyl LDL binding, it i s likely that Ca2+ cross-bridging was responsible for this binding. Thus, LDL, especially its modified forms, could contribute to the formation of insoluble complex of Lp(a) with Ca2+ in atherosclerotic lesions and help explain its preferential accumulation there.
Although Lp(a)' is considered to be a risk factor for cardi-* This work was supported by National Institutes of Health Program Project Grants HL 29582 and HL 43339. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "adoertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.
$ T o whom correspondence and reprint requests should be addressed The Cleveland Clinic Foundation, Research Institute NCl, 9500 Euclid Ave., Cleveland, OH 44195.
'The abbreviations used are: Lp(a), lipoprotein (a); LDL, low density lipoprotein; VLDL, very low density lipoprotein; HDL,, high density lipoprotein 2; HI)L3, high density lipoprotein 3; apo(a), apolipoprotein (a); apoB, apolipoprotein B; PAGE, polyacrylamide gel electrophoresis. ovascular diseases (1-4), neither its potential atherogeneity nor its physiological function has been clearly elucidated. Lp(a) resembles LDL in composition except that it also possesses a unique protein, apolipoprotein (a) (apo(a)), which is covalently linked to apolipoprotein B (apoB) via a disulfide bond (5, 6). The presence of Nand 0-linked glycosylation sites in each kringle allows apo(a) to bear a high carbohydrate content. This carbohydrate has been shown to contain a high content of sialic acid, which probably contributes to the high negative charge of Lp(a) at physiologic pH (7)(8)(9)(10).
Apo(a) is characterized by possessing multiple kringle 4 repeats of plasminogen (7,8), and Lp(a) was shown to compete directly with plasminogen for sites on fibrin (11, 12) and vascular cells (13, 14) under in vitro conditions. These results suggest that Lp(a) may possess antifibrinolytic properties. In addition, Lp(a) could also be atherogenic by accumulating in atherosclerotic lesions and then contributing in some yet undetermined manner to accelerate the atherosclerotic process. We and others have shown previously that Lp(a) accumulates in human atherosclerotic lesions to an even greater extent than LDL, when normalized for equivalent plasma concentrations (15)(16)(17), Moreover, plasma and tissue Lp(a) contents correlated significantly (15). Some of this accumulation could have resulted from particle aggregation after complexing with Caz+. Although two reports show that Lp(a) undergoes aggregation in the presence of Ca2+ (18,19), no data were provided giving any indication of the underlying mechanism(s). Therefore, one of the aims of this current study was to determine the underlying mechanism of Ca2+induced aggregation of Lp(a). Since Lp(a) was shown in vitro to interact with other lipoproteins such as LDL and VLDL (20)(21)(22), we also asked whether the addition of LDL or VLDL to Lp(a) and Ca2+ would lead to the inclusion of these lipoproteins into insoluble complexes, thereby increasing the degree of total particle aggregation.
In this report we show that Lp(a) forms aggregates, even in the presence of physiologic Ca" concentrations, and that they result from Ca2+ cross-bridges formed between sialic acids on separate particles. Furthermore, we report that LDL or VLDL co-precipitate with Lp(a) in the presence of Ca2+, probably in the form of complexes, and that the putative LDL-Lp(a) interaction appears to be ionic in nature involving positive charges on apoB in LDL. Finally, we report that electronegative forms of modified LDL such as acetyl LDL interact more avidly than unmodified LDL with Lp(a) in the presence of Ca2+.

Methods
Isolation of Plasma Lipoproteins and Iodination of Lp(a) and LDL-Normal human plasma was screened for Lp(a) immunoreactivity by radioimmunoassay as described earlier (15). Plasma containing high levels of Lp(a) (42 mg of protein/dl) was used to isolate lipoproteins. Plasma was freshly obtained from a healthy fasting male donor by plasmapheresis in the presence of acid citrate-dextrose. Aprotinin (100 kallikrein-inactivating units/ml), leupeptin (0.5 pg/ml), pepstatin (0.7 pglml), NazEDTA (1 mg/ml), and NaN3 (0.01%) were added immediately after plasmapheresis. Plasma was subjected to sequential ultracentrifugation to isolate the desired lipoprotein fractions using the method of Hatch and Lees (23). VLDL, LDL, and HDLI were obtained as density fractions of d < 1.006 g/ml, 1.019 < d < 1.050 g/ ml, and 1.125 < d < 1.210 g/ml, respectively. Since contamination of LDL with Lp(a) was still observed on SDS-PAGE under nonreducing conditions, the LDL fraction was chromatographed on anti-apo(a)-Sepharose several times to remove any Lp(a) quantitatively. Polyclonal antibodies raised in goats against human plasma Lp(a) were obtained as described earlier (15). After removing all anti-apoB activity by applying an IgG fraction of anti-Lp(a) onto an LDL-Sepharose column as reported earlier (24), the anti-apo(a) fraction was bound to CNBr-Sepharose gel (Pharmacia) according to the manufacturer's recommendation. To separate Lp(a) from high density lipoprotein (HDL,), a density fraction of 1.050 < d < 1.120 g/ ml was subjected to gel filtration chromatography at room temperature on Sephacryl400 HR (2.5 X 95 cm) at a flow rate of 100 ml/h in 1 M NaCl and 5 mM Tris-HC1 containing 1 mg/ml Na2EDTA and 0.01% NaN3 at pH 7.2. The purity of Lp(a) was monitored by SDS-PAGE under nonreducingconditions. Only those fractions completely free of HDL, were pooled. Purifed Lp(a) was stored at 4 "C in 10 mM Tris-HC1 containing 1 mg/ml Na,EDTA at pH 7.2, and other lipoproteins were stored at 4 "C in 0.15 M NaCl containing 1 mM Na2EDTA at pH 7.2.
Lp(a) and LDL were labeled with NalZSI using the iodine monochloride procedure of McFarlane (25) as modified by Bilheimer et al. (26). The specific activities of labeled Lp(a) and LDL for the complexing interaction were 270 and 2.5 cpm/ng of protein, respectively, and that of labeled Lp(a) for radioimmunoassay was 4,000 cpm/ng of protein.
Chemical Modification of Lipoproteins-Desialylation of Lp(a) was performed as follows. 150 pg of Lp(a) protein in 5 mM Tris-HC1 (pH 7.2) was incubated with 0.05 unit of C. perfringens neuraminidase attached to beaded agarose at 37 "C for various incubation periods (3,6, and 9 h) with slow stirring. As a control, an equivalent amount of Lp(a) was treated similarly hut without enzyme. At the end of each incubation period, agarose-attached neuraminidase was first removed by low speed centrifugation, and the liberated sialic acid was then removed by dialysis against 5 mM Tris-HCI (pH 7.2) using Spectrapor 2 dialysis tubing. 270 pg of LDL was also treated with 0.05 unit of neuraminidase for 24 h in the same way. To estimate the extent of desialylation, the electrophoretic mobilities of desialylated Lp(a) and LDL were determined on 1% agarose gel as well as the actual amounts of residual sialic acid present by a modification of the method of Warren (27).
LDL was acetylated by the method of Fraenkel-Conrat (28). 18 mg of LDL protein was mixed at 4 "C with an equal volume of saturated sodium acetate, and 13.5 pl of acetic anhydride was added four times at 15-min intervals. After acetylation, LDL samples were dialyzed against 4 liters of buffer containing 0.15 M NaCI, 0.01% NazEDTA (pH 7.2) and then followed by filtration using a 0.2-gm filter. LDL was methylated by the method of Means and Feeney (29). 3 mg of LDL protein was diluted with 0.3 M sodium borate buffer (pH 9.0) to 1.5 times its original volume. Reductive methylation was carried out at 4 "C by the addition of 1 mg of sodium borohydride followed by six additions of 1 pl of 37% formaldehyde over a 30-min time interval. After methylation, LDL samples were dialyzed against 4 liters of buffer containing 0.15 M NaCI, 0.01% Na2EDTA (pH 7.2) followed by filtration using a 0.2-pm filter. The extent of lysine modification of LDL by acetylation or methylation was measured by changes in trinitrobenzene sulfonic acid reactivity (30) using sodium carbobenzoxy-L-lysine as a standard. They were 45 +: 13.3 and 47 k 8.8%, respectively.
Quuntitation of the Degree of Insoluble Complex Formation-Purified Lp(a) and other lipoproteins in stock solutions were dialyzed against 200-fold volumes of 5 mM Tris-HC1 (pH 7.2) using Spectrapor 2 dialysis tubing prior to experiments of insoluble complex formation. Lp(a) (final concentration, 100 pg of cholesterol/ml) was mixed with varying concentrations of CaClz in the presence or absence of appropriate amounts of VLDL, LDL, or HDL3 (final volume, 50 pl) and kept for 10 min at room temperature in microwell plates (96 wells). The amount of insoluble complex formed was then estimated by turbidity at 600 nm using a microplate reader (Bio-Rad model 3550). The turbidity of lipoprotein solutions containing no CaClz was measured as a blank. The amount of Lp(a) in the form of insoluble complexes was also determined by measuring the amount of Lp(a) cholesterol precipitated. Insoluble complex formation was carried out in microcentrifuge tubes (0.5 ml) instead of a microwell plate (final volume, 70 p l ) , and tubes were subjected to centrifugation (10,000 rpm for 10 min) to precipitate insoluble complexes. The precipitates were then washed twice with 5 mM of Tris-HC1 (pH 7.2) containing the same concentration of CaC12. Cholesterol contents in the precipitates were measured after resolubilizing with equivalent volumes of 0.5 M NaCl and 1 mM Na2EDTA. To determine the amount of Lp(a) and LDL (or modified LDL) in the insoluble complexes expressed as cholesterol content, '"I-Lp(a) or Y -L D L (or modified LDL) was used in separate experiments. After precipitating the insoluble complexes, the tips of tubes containing the precipitates were cut off with a blade after aspirating the supernatants. Subsequently, the radioactivities of precipitates and supernatants were individually counted and converted into cholesterol content.
Electrophoresis-The electrophoretic mobility of the lipoprotein was determined on premade 1% agarose gels following the manufacturer's instructions except that electrophoresis was performed at 90 volts for 90 min. Gels were stained for lipid using 0.025% Fat Red 7B in 60% methanol. The protein components of reduced and nonreduced lipoproteins were analyzed by SDS-PAGE. Lipoprotein samples were solubilized in the buffer containing 2% SDS, 60 mM Tris, 10% glycerin, and 0.01% bromphenol blue. Reduced samples were prepared by boiling at 100 "C for 7 min in the presence of 5% 8-mercaptoethanol. Appropriate amounts of samples were applied on premade 4-15% gradient acrylamide gels and electrophoresed at 12.5 mA for 30 min using the Phast Gel automated system (Pharmacia). After electrophoresis, proteins were visualized by silver staining.
Chemical Analysis-Protein was measured by the bicinchoninic acid assay as described by Smith et al. (31) except that a I-h 60 "C heating step was used. Bovine serum albumin was used as a standard. Cholesterol were determined by the procedure of Roeschlau et al. (32).

Insoluble Complex Formation of Lp(a) with Ca2+
--To determine the dependence of the insoluble complex formation of Lp(a) on Caz+ concentration, we mixed Lp(a) (100 pg of cholesterol/ml) with Ca2+ at concentrations ranging from 0 to 40 mM and estimated the degree of insoluble complex formation by solution turbidity. As low a concentration as 1.25 mM caused significant increases in solution turbidity, which reached a maximum at 20 mM and then decreased at higher Ca2+ concentrations (Fig. la). The same result was obtained with separate Lp(a) samples derived from different donors (not shown). To assess the specificity of such Ca2+-induced aggregation of lipoproteins, we also mixed VLDL, LDL, or HDL3 (100 Fg of cholesterol/ml) with Ca2+ under the same conditions as performed with Lp(a). However, none of these lipoproteins induced significant increases in solution turbidity ( Fig. l a ) , thus demonstrating the specificity of Lp(a) for such aggregation. To verify that the increase in turbidity reflected insoluble complex formation of Lp(a), we also determined the amount of Lp(a) cholesterol precipitated by low speed centrifugation. This index of particle aggregation closely mimicked solution turbidity, except that maximum precipitation of Lp(a) (86%) occurred at lower Ca2+ concentrations (2.5 mM) (Fig. Ib) than did turbidity. Given that physiologic concentrations of Ca2+ range from 2.5 to 5 mM (18,19), these results suggest that Lp(a) specifically forms insoluble complexes at physiologic Ca2+ concentrations under these experimental conditions. To determine whether the highly electronegative surface charge on Lp(a) was responsible for inducing particle aggregation in the presence of Ca2+ by forming ionic cross-bridges, we asked whether the reduction of the negative surface charge would inhibit particle aggregation. Since most of the electronegativity of Lp(a) is caused by the high sialic acid content of apo(a) (9), we studied whether neuraminidase digestion of sialic acids in Lp(a) would prevent insoluble complex formation. When Lp(a) was digested with neuraminidase, a timedependent decrease in the electronegative charge of Lp(a) was observed by agarose electrophoresis (Fig. 2). The sialic acid content of Lp(a) decreased from 85.4 to 59.4 pg/mg lipoprotein protein after a 6-h incubation, representing a 30% decrease. The sialic acid content of unmodified Lp(a) was similar to values reported previously (33). No degradation of apo(a) or apoB could be observed by SDS-PAGE (not shown), which would have indicated the presence of proteases in the neuraminidase samples being used or associated with Lp(a) particles. Neuraminidase treatment of Lp(a) for at least 6 h resulted in an inhibition of insoluble complex formation, especially at higher (5-10 mM) Ca2+ concentrations (Table I).
When excess sialic acid was added to the mixture of Lp(a) and Ca2+, the insoluble complex formation was also inhibited (not shown). These results strongly suggest that the mechanism of aggregation of Lp(a) is via Ca2+ cross-bridges involving sialic acids on different Lp(a) particles. However, since a reduction in sialic acid contents of 30% by neuraminidase was sufficient to inhibit Ca2+-induced aggregation, it would appear

TABLE I Effect of desiulylation of Lp(a) on the formation of insoluble complexes
Neuraminidase-treated Lp(a) (6 or 9 h) or sham-treated (9 h) Lp(a) (final concentration, 100 pg of cholesterol/ml) was mixed individually with 2.5, 5, or 10 mM CaC12, and the cholesterol contents in the insoluble complexes were determined as described under "Experimental Procedures." Untreated Lp(a) was used as a control. The sialic acid contents of sham-treated Lp(a) and Lp(a) treated for 6 h with neuraminidase were 85.4 and 59.4 pg of sialic acid/mg of lipoprotein, rewectivelv. remesentine a reduction of 30%. Involvement of Other Lipoproteins in The Insoluble Complex Formation of Lp(u) with Cu2+-Since Lp(a) has been recently shown to demonstrate specific binding with LDL and VLDL (20-22), we asked whether these lipoproteins might contribute to the bulk of aggregated Lp(a) by directly interacting with Lp(a) if present during Ca2+-induced aggregation. We therefore mixed Lp(a) with LDL, VLDL, or HDL3 (all at 100 pg of cholesterol/ml) in the presence of increasing concentrations of Ca2+ and estimated the degree of insoluble complex formation by turbidity. The addition of both VLDL and LDL increased the turbidity relative to Lp(a) alone at increasing Ca2+ concentrations (Fig. 3u). This increase was also found for individual LDL and VLDL samples derived from different donors (not shown). By contrast, the addition of HDL3 actually decreased the turbidity relative to Lp(a) alone at lower Ca2+ concentrations (Fig. 3a). We also measured the amount of cholesterol in the insoluble complexes formed in the presence of both LDL and Lp(a) with increasing Ca2+ concentrations (Fig. 3b) and compared these amounts with those obtained for Lp(a) alone (Fig. l b ) . The addition of LDL led to a substantial increase in cholesterol precipitated over the range of Ca2+ concentrations (Fig. 3b) relative to Lp(a) alone (Fig. l b ) . In particular, at physiologic Ca2+ concentrations, the precipitated cholesterol was twice as much as that of To determine the actual amounts of Lp(a) and LDL coprecipitated in such insoluble complexes and also to understand the complexing mechanism, we performed the complexing interactions with a fixed amount of Lp(a) and three different concentrations of LDL in the presence of increasing concentrations of Ca2+. The amounts of Lp(a) and LDL coprecipitated were determined using lZ5I-Lp(a) or lZ5I-LDL in separate experiments. The maximum precipitation of Lp(a) (>95%) was achieved between Ca2+ concentrations ranging from 2.5 to 15 mM (Fig. 4a). The amount of Lp(a) precipitated was independent of the amounts of LDL added. A gradual decrease in Lp(a) precipitated was observed a t concentrations of Ca2+ greater than 15 mM. When the amount of LDL coprecipitated with Lp(a) was assessed (Fig. 4b), the maximum was found at the same Ca2+ concentration (2.5 mM) and was independent of amounts of LDL added. The amount of LDL co-precipitated then sharply decreased at higher Caz+ concentrations as we showed previously for total cholesterol in the precipitate (Fig. 3b).
To determine the molar ratio of LDL to Lp(a) in the insoluble complexes at saturation, we carried out the complexing interactions with a fixed amount of Lp(a) and increasing concentrations of LDL in the presence of 2.5 mM Ca2+ as described previously. As we increased the amount of LDL added up to 600 pg of cholesterol/ml, the amount of Lp(a) precipitated remained the same (95 pg) (Fig. 5). By contrast, LDL showed a dose-dependent increase in incorporation into the precipitate, which reached saturation (126 pg A fixed amount of "'I-Lp(a) (or unlabeled Lp(a)) (final concentration, 100 pg/ml) and increasing concentrations of unlabeled LDL (or "' 1-LDL) cholesterol (final concentration, 0, 37.5, 75, 100, 150, 300, or 600 pg/ml) were mixed in the presence of 2.5 m M CaC12 in the separate experiments. Lp(a) and LDL cholesterol in the insoluble complexes were then individually determined from the amount of radioactivity precipitated as described under "Experimental Procedures." by the inhibition of LDL binding with increasing Ca2+ concentrations (Fig. 4b), increasing the positive charge on LDL should increase the interactions with Lp(a).
By contrast, decreasing the positive charge should decrease this interaction. T o verify this mechanism, we studied the effect of chemically modifying LDL on the formation of insoluble complexes. When LDL was digested with neuraminidase to remove sialic acids, thereby increasing the net positive charge as described previously for Lp(a), more than a %fold increase was found in the binding of desialylated LDL relative to unmodified LDL (Table 11). We then asked whether decreasing the positive charge of such basic amino groups as lysines on apoB in LDL would decrease the interaction with Lp(a). When lysine residues on apoB were modified by acetylation to neutralize the positive charge, the binding of acetyl LDL was reduced relative to unmodified LDL. However, since the  LDL modifications on the formation of Ca2+-induced   insoluble complexes Neuraminidase treatment, acetylation, and methylation of LDL were carried out as described under "Experimental Procedures." lZ5I-Lp(a) (or unlabeled Lp(a)) (final concentration, 100 pg of cholesterol/ ml) and unlabeled modified LDL (or modified lZ5I-LDL) (final concentration, 600 pg of cholesterol/ml) was mixed in the presence of 2.5 mM CaClz in the separate experiments. The cholesterol contents of Lp(a) and modified LDL in the insoluble complexes were then determined individually as described under "Experimental Procedures." The sialic acid contents of untreated LDL and LDL treated with neuraminidase for 24 h were 22.0 and 17.5 pg of sialic acid/mg of 1iDoDrotein. remectivelv. amount of Lp(a) precipitated was also reduced when acetyl LDL was added (Table 11), we cannot conclude whether the reduced acetyl LDL present in insoluble complexes was related to a reduced interaction with Lp(a) or to reduced precipitation of Lp(a). When the positive charge on LDL was further reduced by blocking arginine residues on apoB with cyclohexadione, no insoluble complex formation occurred with Lp(a) and 2.5 mM Ca2+ (not shown) as was found with acetylation of LDL. When lysine residues on apoB were modified by reductive methylation, which does not affect the surface charge on apoB, no change in the co-precipitation of LDL was found (Table 11). These results suggest that the interaction between Lp(a) and LDL is ionic and that the positive charges on LDL are responsible for the binding of LDL to Lp(a) during insoluble complex formation.
Since the interaction between Lp(a) and other lipoproteins containing apoB such as LDL and VLDL has been shown to be mediated by proline in apoB (22), we asked whether excess amounts of L-proline would also inhibit binding of LDL to Lp(a) in the insoluble complexes. We therefore formed insoluble complexes of Lp(a) and LDL with Ca2+ in the presence of increasing concentrations of L-proline. However, neither the amounts of LDL nor the amounts of Lp(a) in such insoluble complexes were affected by the presence of excess L-proline even up to 0.1 M (Fig. 6). This result suggests that the Lp(a)-LDL interaction mediated by proline residues on LDL was different from the one occurring in Ca2+-induced Lp(a) precipitates.

Enhanced Binding of Acetyl LDL to Lp(a) in Insoluble
Complexes-We initially employed acetylation of LDL to illustrate the importance of positive charges on LDL in the interaction between LDL and Lp(a). However, acetyl LDL inhibited the insoluble complex formation of Lp(a) at 2.5 mM Ca2+ (Table II), suggesting competitive binding with Lp(a) for Ca2+. This result led us to hypothesize that the binding of acetyl LDL may be enhanced if the Caz+ concentration was sufficiently high. To test this hypothesis, we determined the amounts of acetyl LDL and Lp(a) present in the insoluble complexes with increasing concentrations of Ca2+ (Fig. 7, a  and b ) in the same way as described in Fig. 4. These amounts were then compared with those obtained when unmodified LDL was employed (Fig. 4, a and b ) . Acetyl LDL inhibited Lp(a) insoluble complex formation in a concentration-dependent manner at low Ca2+ concentrations (1.25 and 2.5 mM) (Fig. 7a). However, a t higher Ca2+ concentrations (5-20 mM) the amount of Lp(a) precipitated was independent of the amount of LDL (Fig. 4a) or acetyl LDL (Fig. 7a) added.
When the amount of acetyl LDL in the precipitate was determined, major differences relative to LDL were found ( Figs. 4b and 76). LDL showed a maximum co-precipitation a t 2.5 mM Ca2+ which remained the same regardless of the LDL concentrations added (Fig. 4b). By contrast, the Ca2+ concentration at which co-precipitation of acetyl LDL reached saturation, increased with increasing acetyl LDL added (Fig. 7b). It is also of note that when 600 pg/ml acetyl LDL was added at the upper level of physiologic Ca2+ concentrations, e.g. 5 mM, 250 pg of acetyl LDL was present in the insoluble complex as compared with only 60 pg of LDL being present. At 15 mM Ca2+ the difference was even more dramatic, e.g. 380 uersus 50 pg, respectively. In the absence of Lp(a), acetyl LDL did not form insoluble complexes with Ca2+ (not shown).

DISCUSSION
In this study we have confirmed and extended previous reports (18,19) demonstrating that Lp(a) undergoes insoluble complex formation in the presence of Ca2+. Dahlen et al. (19) had previously reported aggregation of Lp(a) under similar conditions, but only 30% of the Lp(a) particles became insoluble, as contrasted to this study in which about 86% were aggregated. This discrepancy may be because of methodological differences in the two studies. We propose that insoluble complex formation of Lp(a) is induced by Ca2+ cross-bridges between different Lp(a) particles, mediated by sialic acids on apo(a). This is based on the fact that Lp(a) has a large number of sialic acids in apo(a) (9), and sialic acids have been shown to demonstrate a high binding affinity for Ca2+ (34). This interpretation is consistent with the observation that desialylation of Lp(a) with neuraminidase prevented insoluble complex formation. Even though other lipoproteins also possess sialic acids, they did not form insoluble complexes with Ca2+ ( Fig. 1). One of the reasons for this Lp(a) specificity could be the higher sialic acid content of Lp(a) compared with other lipoproteins (9, 33). A unique flexible extended open conformation of apo(a) (9) as contrasted to other apoproteins may be critical for this interaction. It is of note that only 30% of the sialic acids on Lp(a) appeared to be critical for Ca2+induced cross-bridging of Lp(a) particles, since 70% of the total sialic acid content still remained on neuraminidasetreated Lp(a) that did not undergo Ca2+-induced complex formation. The observations that highly electronegative molecules such as acetyl LDL or HDL3 inhibited Lp(a) insoluble complex formation at physiologic Ca2+ concentrations could be because of direct competition between sialic acids on apo(a) and the other highly electronegative lipoproteins for the limited amount of Ca2+. A decrease in Lp(a) insoluble complex formation at higher Ca2+ concentrations (Fig. Ib) could be because of the complete blockage of sialic acids on individual Lp(a) particles by Ca2+, thereby preventing Ca2+ from crossbridging.
This is the first study to report the binding of apoBcontaining lipoproteins such as LDL and VLDL to Lp(a) in Ca2+-induced insoluble complexes, as evidenced by their coprecipitation with Lp(a). The mechanism of binding appears to be ionic, and the positive charge on LDL could be responsible for the interaction. As such, it differs from the Lp(a)-LDL interaction reported by Trieu et al. (22), which requires the interaction of apo(a) with proline residues on LDL, rather than Ca2+. Our observation that excess L-proline failed to inhibit co-precipitation of Lp(a) and LDL in the presence of Caz+ is consistent with the conclusion that the mechanisms differ in the two situations. It is unlikely that Ca2+ forms cross-bridges between Lp(a) and LDL, since increasing the net positive surface charge on LDL by desialylation increased its co-precipitation with Lp(a) at physiologic Ca2+ concentrations. The observation that the binding of LDL to Lp(a) in the insoluble complexes was inhibited by higher Ca2+ concentrations also supports the importance of the positive charge on LDL for this interaction. Excess Ca2+ could compete with electropositive sites on apoB in LDL for electronegative sites on Lp(a), e.g. saturate them, thereby inhibiting the LDL binding.
The studies on the comparison of the interaction between Lp(a) with LDL or with acetyl LDL yielded several interesting observations. Although acetyl LDL, unlike LDL, appeared to bind strongly to Ca2+, probably because of its increased surface negative charge, it did not, by itself, form insoluble complexes with Ca2+, but rather bound to Lp(a) in the insoluble complexes. However, the mechanism for binding of acetyl LDL seems to differ from that of LDL and appears to be mediated by Ca2+ cross-bridges between negatively charged sites on acetyl LDL and on Lp(a). This interpretation is based on the observation that the Ca" concentration at which saturated binding of acetyl LDL occurred increased with increasing amounts of acetyl LDL added, as was reported previously between acetyl LDL and dextran sulfate (35). It is of note that at physiologic concentrations of Ca2+ (5 mM), more acetyl LDL than LDL bound to Lp(a) in the insoluble complexes.
The greater binding of acetyl LDL than LDL to Lp(a) during Ca2+-induced insoluble complex formation at physiologic Ca2+concentrations would suggest that chemically modified forms of LDL such as oxidized LDL found in the arterial wall (36-39) might demonstrate enhanced interaction with Lp(a). In fact, we found that oxidized LDL induced by incubating with 10 p~ Cu2+ for 24 h bound more avidly than LDL to Lp(a) in the insoluble complexes (not shown). We have demonstrated previously the accumulation of both Lp(a) and LDL in human atherosclerotic lesions (15). It is tempting to speculate that some of this accumulation might have been caused by Caz+-induced insoluble complex formation of Lp(a) and the co-precipitation of some of the accumulated LDL. Thus, such aggregation in the vessel wall could, in large part, be responsible for the accumulation in plaques of not only Lp(a), but a significant part of the LDL. It is of note that cardiovascular disease was shown to be particularly prevalent in individuals with both high plasma Lp(a) and high LDL as in heterozygous familial hypercholesterolemia (3, 40), in which more Lp(a) and LDL would be expected to accumulate in the arterial intima. If these in vitro observations reported in this study also occurred in the arterial wall, they could help explain the apparent atherogenicity of Lp(a).