Decreased Binding of Apolipoprotein (a) to Familial Defective Apolipoprotein B-100 ( A q f 5 O o + Gln) A STUDY OF THE ASSEMBLY OF RECOMBINANT APOLIPOPROTEIN (a) WITH MUTANT LOW DENSITY LIPOPROTEINS*

In familial defective apolipoprotein B-100 (FDB), glu- tamine is substituted for arginine at position 3500 of the amino acid sequence. “his mutation alters the structure of low density lipoproteins (LDL) and reduces their binding to LDL receptors. We studied the assembly in vitro of FDB-LDL with two recombinant apo(a) (r- apo(a)) isoforms containing 17 or 18 kringle %type repeats, respectively. R-apo(a) complexed to LDL in a con- centration- and time-dependent manner. When we mixed normal LDL at protein concentrations from 1 to 10 mglliter with 200 &liter r-apo(a) and incubated for 20 h, 1544% of r-apo(a) were bound to LDL, forming an artificial Lp(a)-like particle. With LDL from a homozygous FDB patient, only 2-16%

particle and a specific antigen, apo(a) (5). Apo(a) exists in more than 30 genetic isoforms which differ by their size and number of kringle IV-like repeats (6); the apo(a) size is negatively correlated with plasma Lp(a) levels (7)(8)(9)(10). The mechanism underlying this association is still unknown. We showed in humans that Lp(a) concentrations strongly correlate with the rate of synthesis, but not with the fractional catabolic rate (11). In an extension of this work, Rader et al. (7) reported that the variation of Lp(a) levels among individuals with the same phenotype was also caused by differences in the Lp(a) production rate. Subsequent studies in homozygotic twins and sib pair analyses claimed that the heritability of apo(a) levels was more than 90% (9, 10). There exists, however, a great number of nongenetic factors affecting plasma Lp(a) concentrations: Lp(a) increases in acute phase reactions (12), during pregnancy (131, in many forms of kidney disease (14, X ) , and in endocrine disorders (16,17). Hormonal contraceptives, testosterone, and smoking appear to lower Lp(a) (18)(19)(20). The mechanisms of all these changes are unknown.
The role of the LDL receptor in the clearance of Lp(a) is controversial. Lp(a) was found increased in familial hypercholesterolemia (21,22), suggesting that LDL receptors were involved in the catabolism of Lp(a). However, up-regulation of LDL receptors by hydroxymethylglutaryl-CoA reductase inhibitors (23- 25) or cholestyramine (26) did not reduce Lp(a) and other investigators reached the conclusion that defective LDL receptors had no effect on Lp(a) (27)(28)(29)(30). There seems to be agreement now that the LDL receptor binds Lp(a) with a significantly lower affinity as compared to LDL (31) and that the LDL receptor most probably plays a minor role for the catabolism of Lp(a) (32). Because the fractional catabolic rate of Lp(a) is approximately equal to that of LDL, alternative pathways must contribute to the clearance of Lp(a), one of which possibly involving the LDL receptor-related proteinla,-macroglobulin receptor (33).
Familial defective apolipoprotein B-100 (FDB) denotes a genetic disorder of lipoprotein metabolism leading to high plasma cholesterol and LDL concentrations (34)(35)(36)(37). FDB is caused by an arginine to glutamine mutation at codon 3500 of the apoB gene (35). This substitution leads to a defective binding of apoB-100 to the LDL receptor. We show here that LDL purified from a homozygous FDB patient exhibits an approximately 50% reduced capacity to assemble with recombinant apo(a) (r-apo(a)) in vitro.

EXPERIMENTAL PROCEDURES
Materials-COS-7 cells were obtained from the American Type Culture Collection (ATCC CRL 1651). Antibodies specific for apo(a) and apoB were produced in our laboratory (11,31). Horseradish peroxidaselinked protein A and a n enhanced chemiluminiscent detection kit (ECL) were obtained from Amersham. Lp(a) isoform standard and reference sera were purchased from Immuno (Vienna, Austria). Replicationdefective biotinylated adenovirus (dl 3121, streptavidin-polylysine conjugate, and human transferrin-polylysine conjugate were produced at the Institute of Molecular Pathology Vienna, Austria. The apoB-100 specific antibody MB47 was provided by Dr. Linda K. Curtiss (The Research Institute of Scripps Clinic, La Jolla, CAI. Recombinant apo(a) containing 17 KIV-type units was obtained from Dr. Dan Eaton (Genentech Inc., San Francisco, CA). Nitrocellulose was from Hoefer Scientific. All other materials were from Sigma.
Patients-We studied LDL from four patients with FDB, one homozygote and three heterozygotes. The homozygous patient (F.B.) was the 55-year-old male who was described previously (37,38). The three heterozygous patients (R.W., R.O., and I.K.) belonged to one family; they were on normal diet.
Electron Microscopy-LDL were dialyzed against 0.125 moVliter ammonium acetate, 2.6 mmobliter ammonium carbonate, and 0.26 mmoV liter EDTA.Na,. They were mixed with sodium phosphotungstate (final concentration 10 gfliter), applied to Formvar-carbon coated grids, and examined in a Philips EM 300 electron microscope at a n accelerating voltage of 100 kV.
Construction of an Apo(u) Expression Vector4DNA clones reported by McLean et al. (44) were used for the production of the apo(a) expression plasmid lpSG5-18) containing DNA sequences coding for 18 KIVlike domains, for the kringle V (KV)-like and the protease domain, following standard recombinant DNA techniques (45, 46). As described elsewhere, a construct was generated encompassing the apo(a) cDNA 5'-untranslated region, the signal sequence, KIV-1 through KIV-5, 294 bp of KIV-6, the last 48 bp of KIV-3, the complete KIV-4,294 bp of KIV-5, the last 48 bp of KIV-27, KIV-28 through KIV-37, KV, the protease domain, and 67 bp of the 3'-untranslated region. This fragment was ligated into the EcoRI site of the expression vector pSG5 (46). High level expression of r-apo(a) was obtained in COS-7 cells, cultured in Dulbecco's modified Eagle's medium supplemented with 10% (v/v) lipoproteindeficient serum, using the "transferinfection" method (47) as described (48). The r-apo(a) containing medium was harvested after 48 h and the amount of r-apo(a) was determined with a sandwich DELFIA. The recombinant protein migrated slightly faster than the F-isoform of apo(a) in the phenotyping standard (Immuno, Austria) upon SDS-agarose gel electrophoresis. In some experiments, r-apo(a) obtained from Dr. Dan Eaton (Genentech Inc., San Francisco, CA) which contained 17 copies of KIV-type repeats (49) was investigated for comparison.
Isolation ofLDL-LDL were prepared by ultracentrifugation (1.023-1.0630 kghiter) from fasting plasma (50). LDL fractions were subjected to immunoabsorption for removal of any contaminating Lp(a). To avoid LDL oxidation, all steps of the LDL preparation were performed under nitrogen and in the presence of 0.5 ghiter of EDTA.Na,. The LDL were >97% pure as judged by electrophoresis, chemical, and immunochemical analyses.
Binding of r-Apo(a) to LDL in Vitro-Medium of transfected COS-7 cells containing 200 pghiter r-apo(a) was incubated for 20 h at 37 "C with LDL at different concentrations which were calculated on the basis of the apoB content. 0.2 pmobliter aprotinin, 50 pmoVliter leupeptin, and 1 pmoVliter phenylmethylsulfonyl fluoride were added to inhibit proteolysis. After incubation, aliquots of the mixtures were immediately analyzed by DELFIAand SDS-agarose gel electrophoresis and immunoblotting.
Electrophoresis a n d Western Blotting-Samples were mixed with SDS (final concentration 10 glliter) in the absence of any sulfhydryl reducing agent and heated for 5 min at 100 "C. SDS-agarose gel electrophoresis was performed as described (51). After electrophoresis, proteins were transferred to a nitrocellulose membrane by electroblotting overnight at 4 "C in 10 mmoVliter Tris-HC1 and 40 mmoVliter glycine buffer, pH 7.4. The membrane was blocked for 60 min in 5% (w/v) powdered skim milk, then incubated for 3 h with rabbit anti-apo(a) antibody (diluted 1:1000), washed extensively in Tris-buffered saline, pH 7.4, containing 0.05% (w/v) Tween 20, and incubated for 2 h with horseradish peroxidase-labeled protein A. After washing, as described above, the membrane was incubated with the ECL Western blotting detection reagent for 2 min and subjected to autoradiography according to the manufacturer's instructions. The sensitivity of this method was 5 ng.
Immunoquantitation of Free r-Apo(a) and r-Apo(ai.LDL Complexes by DELFIA-Lp(a) was determined with a sandwich DELFIA. 96-well plates (Costar, Cambridge, MA) were coated with polyclonal affinity purified anti-apo(a) from sheep. Nonspecific binding sites were blocked with 250 pl of 0.5% (v/v) bovine serum albumin for 30 min. 200-pl aliquots of the samples were added to the wells and incubated for 2 h a t 20 "C. After three washing steps with 50 mM Tris-HC1, pH 7.7, a polyclonal detection antibody (either anti-apo(a) or anti-apoB from rabbits), labeled with Eu, was added to the wells and further incubated for 2 h at 20 "C. Excess antibody was removed by two washing steps with 50 mM Tris-HC1, pH 7.7. 200 pl of enhancement solution (Pharmacia) were added, and after 15 min, fluorescence was determined in a DELFIA reader. For the determination of total apo(a), Eu labeled anti-apo(a) was used; for the determination of r-apo(a).LDL complexes, Eu labeled anti-apoB was used. Standard curves were produced with the Lp(a) Reference Standard from Immuno Diagnostika and included in each experiment. The assay was linear between 1 and 100 ng of Lp(a) per well; the coefficient of variation was 2.6%.
Binding ofApoB-100-specific Monoclonal Antibody MB47 to LDL and r-Apo(a).LDL Complexes-MB47 reactivity with LDL and r-apo(a).LDL complexes was determined by means of a solid-phase competitive enzyme immunoassay. Microplates (Nunc Immuno Plates I) were coated with 150 p1 of control LDL (10 mghiter protein in 0.2 mobliter carbonate buffer, pH 10.0), blocked with 10 ghiter bovine serum albumin (in carbonate buffer), and washed with 200 pl of phosphate-buffered saline containing 0.05% Tween 20.
LDL were desalted on Pharmacia PD-10 columns equilibrated with phosphate-buffered saline containing EDTA.Na, (0.3 mmobliter), assayed for protein, and adjusted to 10 g/liter bovine serum albumin and 0.05% Tween 20. LDL (final concentration 70 mghiter apoB) were preincubated with KIV-17 r-apo(a) (5 mg/liter, 20 h). Control incubations were performed with exactly the same dilutions of LDL, but without r-apo(a). The mixtures were incubated overnight with MB47 (ammonium sulfate precipitate of mouse ascites fluid) diluted 1:4000 in phosphate-buffered saline containing 0.3 mmoVliter EDTA, 10 glliter bovine serum albumin, and 0.5% Tween 20, and 100 pl were loaded into LDLcoated microplate wells. After 2 h, the plates were washed three times and the amount of antibody bound to immobilized LDL was determined with peroxidase-conjugated anti-mouse immunoglobulin G (Boehringer Mannheim, diluted 1:750 in phosphate-buffered saline-Tween-bovine serum albumin). Color was developed with o-phenylenediamine.

RESULTS
We obtained LDL from one homozygous and three heterozygous patients with FDB (Table I). The morphology of FDB-LDL as studied by negative staining electron microscopy revealed no significant differences from normal (not shown). However, slight differences were observed when the chemical compositions of the patients' LDL were compared to normal LDL. The calculated particle radius of LDL from F.B. (homozygous FDB) was smaller than for normal LDL. Consistently, LDL from F.B. contained less phospholipid molecules per particle than normal. This reflects the preponderance of small dense LDL in the patient, which has been described previously (37). On the other hand, LDL from the three FDB heterozygotes turned out slightly larger, due to an increased cholesteryl ester content.
In previous experiments, we studied the assembly of Lp(a) in vitro by incubation of r-apo(a) with LDL from normal healthy donors. This way, we generated Lp(a)-like complexes which were indistinguishable from native Lp(a) with respect t o hydrated density, composition, and morphology (48, 53). Com-

Lp(a) Assembly in Familial
Defective ApoB-100 plexes were formed whether or not living cells were present; the degree of complex formation was time dependent. Fig. 1 displays a time course experiment in which 200 pgfliter r-apo(a) from COS-7 cells transfected with pSG5-18 were incubated for 0.5-24 h with 5 mgfliter LDL-protein and the degree of Lp(a) assembly was measured. A plateau was reached after approximately 16 h. Thus, all subsequent experiments were performed a t 20 h incubation. The efficacy of Lp(a) assembly also depended on the ratio of apoB to r-apo(a) in the medium, increasing with higher ratios. For reasons of practicability we carried out most studies a t a fixed concentration of 200 pgfliter r-apo(a) and variable apoB concentrations ranging from 0.5 to 10 mg/ liter. We were interested to see whether FDB-LDL behaved similar to normal LDL with respect to complex formation with r-apo(a). To study this, LDL from a homozygous FDB patient and from a normal individual were incubated for 20 h with 200 pgfliter r-apo(a). The complex formation was monitored by submarine SDS-agarose gel electrophoresis and Western blotting using anti-apo(a) as detecting antibody (Fig. 2). With normal LDL, two major bands were observed, one migrating to the position of S4 and the other slightly faster than the F isoform of the phenotyping standard. In addition to anti-apo(a), the slow migrating band also reacted with anti-apoB (not shown). The fast migrating band reacted with anti-apo(a) only. As shown elsewhere (48,531, addition of reducing agents, e.g. mercaptoethanol, caused the slow migrating band to disappear, while only one fast migrating apo(a) band was seen. Thus, the slow migrating band consisted of the apo(a).LDL heterodimer, which is dissociated by mercaptoethanol, but not by SDS alone. When the same experiment was performed with FDB-LDL, only small amounts of heterodimers were found (Fig. 2).
In another approach to examine the assembly of r-apo(a) with LDL, we added r-apo(a) at a final concentration of 5 mg/ liter to apo(a)-negative plasma, incubated at room temperature, and separated the mixture by fast flow size exclusion chromatography. We then analyzed the column fractions for r-apo(a).LDL complexes. As expected, r-apo(a)-LDL complexes were produced, eluting at slightly higher apparent molecular masses than LDL (Fig. 3). In the FDB homozygous plasma, the concentration of LDL (LDL-C 2.52 gfliter) was approximately 2-fold that in the normal plasma (LDL-C 1.16 gfliter). In spite of this, only half the amount of r-apo(a).LDL complexes was formed in the patient's plasma, compared to the normal plasma.
Both SDS-agarose gel electrophoresis and size exclusion chromatography are semi-quantitative means to study Lp(a) assembly. To approach the problem in quantitative terms, similar incubations as described in Fig. 2 were performed with different concentrations of normal and mutant LDL, followed by the measurement of free and LDL-complexed apo(a) using DELFIA (Fig. 4). At all LDL concentrations, the amounts of r-apo(a) found in the artificial Lp(a) complex were significantly lower with mutant LDL than with normal LDL. For FDB-LDL the amount of r-apo(a) complexed to LDL ranged between 6.0 f 0.8 and 21.2 2 3.2% of total r-apo(a), as opposed to normal LDL where values between 18.5 2 2.1 and 42.4 2 5.1% were found.
To confirm these findings, a different batch of r-apo(a) containing 17 kringgle IV-like domains (described in Ref. 49) was studied (Fig. 5). In the incubation mixtures containing 10 mg/ liter normal LDL-protein and 200 pgfliter r-apo(a), 44% of total r-apo(a) was complexed with LDL and formed an Lp(a)-like particle; heterozygous FDB-LDL bound 20-36% r-apo(a), and homozygous FDB-LDL only 16%. At lower LDL concentrations, the amount of r-apo(a) complexed to FDB-LDL was always significantly smaller than the amount of r-apo(a) complexed to normal LDL.
The monoclonal antibody MB47 recognizes an epitope near the receptor binding domain of apoB-100. We reasoned, if rapo(a) indeed bound to this region, MB47 reactivity might be affected by the attachment of apo(a). Therefore, we studied the interaction of normal LDL and FDB-LDL with MB47 before and after adding r-apo(a) (17 KIV). As expected, FDB-LDL exhibited a higher affinity for MB47 than normal LDL. When r-apo(a) was complexed to normal LDL, the resulting complexes displayed a higher affinity for MB47 than the LDL alone. In contrast, no change in MB47 reactivity was observed when FDB-LDL were preincubated with r-apo(a) (Fig. 6).

DISCUSSION
In studies which are published elsewhere (48, 531, we investigated the assembly of Lp(a) in eukaryotic cells transfected with expression vectors containing variable lengths of apo(a) cDNA. All data we obtained were compatible with a two-step model of extracellular Lp(aj assembly. This assumption is in line with reports from other groups using systems similar to ours (54,55), transgenic mice (561, or cultured primary baboon hepatocytes (57). Considering all these earlier findings to-  gether, it appears that apo(a) is synthesized in the liver and, in a first step, loosely complexes to LDL outside the hepatocyte. In that complex, the sulfhydryl group of cysteine 4057 comes close t o a free sulfhydryl group of apoB-100 and forms a stabilizing disulfide bond in the second step, The latter assumption is based on the observation that the substitution of cysteine residue 4057 for serine (55) or for arginine (48) by site-directed mutagenesis led to a significant decrease of stable Lp(a) assembly. We reasoned, if our two-step model for Lp(a) assembly was correct, then not only the substitution of cysteine residue 4057 in apo(a), but also a conformational change in apoB-100 which interferes with the noncovalent attachment of apo(aj to the LDL surface should reduce complex formation and Lp(a) synthesis. TO evaluate this hypothesis, we examined the binding of r-apo(a) to LDL from patients with familial defective apoB-100 ( A r g l t S o 0 + Gln). Using different strategies and r-apo(a) from two independent sources containing either 17 or 18 KIV-like repeats, we demonstrated that the association of r-apo(a) with FDB-LDL is greatly reduced. Circumstantially, these findings would imply that the apoB-100 receptor binding domain is involved in the binding of apo(a) to LDL. When tested in vitro, Lp(a) exhibits reduced binding to the LDL receptor compared to LDL (31,32,58). This concurs with the idea that apo(a) partially masks the apoB-100 receptor binding domain. The latter contains two lysine-rich segments extending from residues 3147 to 3157 and from residues 3359 to 3367 (59,60). Its conformation is profoundly altered in FDB (61)(62)(63). We, therefore, speculated that the conformational change in the apoB-100 receptor binding domain was responsible for the reduced ability of FDB-LDL to interact with apo(a).
To throw additional light on this hypothesis we examined whether the binding of r-apo(a) to LDL affected the affinity for the apoB-100 specific monoclonal antibody MB47. The MB47 epitope consists of two non-linear domains including amino acid residues 3429-3453 and 3507-3523; it is located in the vicinity of the apoB-100 receptor binding domain (64,65). We fully confirmed previous work showing that MB47 binds to FDB-LDL with higher affinity than to normal LDL (37,66). Interestingly, complexation of r-apo(a) to normal LDL increased their reactivity with MB47, suggesting that the MB47 epitope was involved in the binding of apo(a) andor underwent a conformational change on apo(a) attachment. In contrast, incubation of r-apo(a) with FDB-LDL did not influence MB47 immunoreactivity. This may be due to the reduced ability of FDB-LDL to interact with apo(a). Alternatively, FDB-LDL may already possess a conformation, which is similar to that induced by the binding of r-apo(a) to normal LDL.
We wish to reconcile our findings with a report by Perombelon et al. (67) who studied Lp(a) in 31 members of two FDB families. They found higher Lp(a) concentrations in FDB heterozygotes, compared to unaffected relatives. They also analyzed Lp(a) levels of family members with identical apo(a) phenotypes, with or without FDB. In five cases, where a direct comparison was possible, FDB patients again exhibited higher Lp(a) levels than the healthy family members. Obviously, this would be in line with the idea that Lp(a) is metabolized by LDL receptors. However, Perombelon et al. (67) also performed a detailed analysis of FDB-Lp(a) in vitro which was in sharp contrast to this assumption. When they removed the apo(a) moiety of Lp(a) from heterozygous FDB patients, the resulting LDL-like particles (i.e. Lp(a)-) contained a smaller proportion of defective particles (50%) than LDL from the same patient (60-75%) (67). Finally, when the same families were re-examined, no differences in Lp(a) concentrations were detected between affected and unaffected siblings with the same apo(a) allele (29, 681, suggesting that LDL receptors do not significantly contribute to Lp(a) catabolism in humans. Further support for this view comes from work showing that defective LDL receptors have almost no effect on Lp(a) concentration (27)(28)(29)(30) and that hydroxymethylglutaryl-CoA reductase inhibitors do not lower Lp(a) (23-25).
In heterozygous FDB, the ratio of mutant to normal apoB-100 ranges from 60:40 to 7525 (34,67,69). If Lp(a) assembly was in fact occurring in the plasma compartment, the ratio of Lp(a) particles containing mutant apoB-100 to those with "wild type" apoB-100 should be close to that for LDL, i.e. between 60:40 and 75:25. However, according to Perombelon et al. (67) this ratio is 50:50. Our results may provide a plausible solution of this paradox: although circulating a t a higher concentration than normal LDL, the defective LDL complexes less apo(a) in FDB heterozygotes. As a net effect, the concentration of Lp(a) molecules endowed with the defective apoB-100 is shifted toward approximately equal concentration of Lp(a) with normal apoB-100.
Plasma Lp(a) concentrations are highly heritable and the apo(a) locus accounts for more than 90% of the interindividual variation in plasma Lp(a) concentrations (9). It is, therefore, believed that sequence differences at the apo(a) locus determine the rate of Lp(a) biosynthesis. Rader et al. (7), however, reported differences in the Lp(a) production rate among individuals with the same phenotype, and a large number of nongenetic factors is gradually becoming recognized which modify Lp(a) concentrations (12)(13)(14)(16)(17)(18)(19)(20). Here we demonstrate experimentally that genetic variation altering apoB-100 structure and metabolism may influence Lp(a) biosynthesis.
In summary, we show that FDB-LDL exhibits reduced capacity to assemble with two different recombinant apo(a) molecules containing 17 or 18 kringle IV repeats. These findings not only provide a possible explanation for the observations of Perombelon et al. (67), but in extension of this suggest that the LDL surface, and in particular the epitopes of apoB-100 which are important for LDL receptor binding, are involved in Lp(a) assembly as well. This hypothesis which has a potential bearing on Lp(a) homoeostasis in vivo is now pursued further in our laboratories.