INTRODUCTION

Lipoprotein(a) (Lp(a))¹ is an atherogenic lipoprotein that is formed by the disulfide linkage of apolipoprotein (apo) B100 in a low density lipoprotein particle to apo(a) (1-3). Lp(a) is present only in the plasma of mammals that synthesize apo(a), which include humans, old world monkeys, great apes, and hedgehogs (4-7). In many epidemiological studies, high plasma levels of Lp(a) have been associated with an increased risk of atherosclerotic coronary heart disease (8-10). Why Lp(a) promotes atherosclerosis and coronary heart disease is not fully known, although several possible mechanisms have been suggested in recent years (for reviews, see Refs. 1-3).

The structural features of apoB that allow it to form a disulfide linkage with apo(a) are only partially understood (11). Site-directed mutagenesis studies have shown that the carboxyl-terminal cysteine of apoB100, cysteine 4326, is required for Lp(a) formation (12-14), almost certainly because it is the site of attachment for apo(a). However, it is obvious that other features of the apoB molecule, aside from a single free cysteine residue, must play a role in the binding of apoB and apo(a). It seems likely that an initial noncovalent interaction brings the two molecules into close proximity so that the disulfide bond can form (11). In this study, we sought to define the sequences, aside from the critical free cysteine residue, that are important for apoB's interaction with apo(a).

The simplest hypothesis regarding the noncovalent interaction event would be that apo(a) binds to apoB sequences near the critical free cysteine residue, cysteine 4326. In considering this hypothesis, we examined the amino acid sequences surrounding residue 4326, both for human and mouse apoB. The mouse apoB sequence was examined because there is some evidence that mouse apoB can bind to apo(a) noncovalently (19), although it lacks a cysteine at residue 4326 (13) and therefore cannot form bona fide Lp(a). Interestingly, although the 50 amino acids carboxyl-terminal from cysteine 4326 are reasonably well conserved (66% identical) between the human and mouse sequences, the 50 amino acids amino-terminal to residue 4326 are poorly conserved (34% identical). These observations led us to hypothesize that the sequences carboxyl-terminal from cysteine 4326 might play a role in Lp(a) assembly.

To test the hypothesis that sequences carboxyl-terminal to cysteine 4326 might facilitate the interaction between apo(a) and apoB, we generated transgenic mice expressing two truncated forms of human apoB, apoB95 (4330 amino acids), and apoB97 (4397 amino acids). We then compared the abilities of apoB95, apoB97, and apoB100 to bind covalently to apo(a) and form Lp(a). These studies provided new insights into the apoB sequences that are required for the interaction of apoB and apo(a).

MATERIALS AND METHODS

To generate constructs coding for apoB95 (4330 amino acids) and apoB97 (4397 amino acids), we used "pop-in, pop-out" gene targeting (13, 15, 16) to introduce nonsense mutations into a 108-kilobase (kb) YAC that contained the entire human apoB gene (13). The gene-targeting vectors were generated by introducing nonsense mutations into a 2.8-kb XbaI fragment from the apoB gene (from intron 28 to just past the end of the exon 29 coding sequences) that had been cloned into the URA3-containing yeast integrating vector (pRS406, Stratagene, La Jolla, CA) (13). To generate the targeting vectors, we introduced subtle mutations into exon 29 of the apoB gene using the Excite polymerase chain reaction (PCR)-based site-directed mutagenesis kit from Stratagene. For the apoB95 gene-targeting vector, we used oligonucleotide B95 (5-CCTTAATCTTCAcTAGTTCAATGAAT-3); for the apoB97 gene-targeting vector, we used oligonucleotide (5-GGACTTCCATTCTGAATAAATTGTCAagGCCTCTAACTTTACTTCC-3). In each oligonucleotide, the boldface and italic nucleotide introduced a nonsense mutation. Additional nucleotide substitutions (shown in lowercase) were introduced to create new restriction sites (SpeI for the apoB95 mutation and StuI for the apoB97 mutation). Mutagenesis reactions were confirmed by DNA sequencing. Both gene-targeting vectors were linearized at a unique EcoRI site in exon 29 and transformed into yeast spheroplasts harboring the 108-kb human apoB YAC. Transformants were selected initially on plates lacking uracil and later on plates lacking uracil, tryptophan, and lysine. Transformants in which the gene-targeting vector had integrated correctly into exon 29 of the apoB YAC (the "pop-in" step) were identified by pulsed field gel electrophoresis and PCR analysis (see below). For the "pop-out" step, targeted transformants were grown in medium lacking lysine and tryptophan and subsequently plated onto 5-fluoroorotic acid-containing plates (17). Colonies were then plated onto plates lacking lysine and tryptophan, and correctly targeted pop-out clones were identified by pulsed field gel electrophoresis and PCR analysis.

To analyze gene-targeting events at both the pop-in and pop-out steps, high molecular weight DNA (18) was prepared from yeast clones and electrophoresed on a 1% pulsed-field agarose gel in 0.5 × TBE (45 mM Tris, 45 mM boric acid, 1 mM EDTA, pH 8.3) at 6 V/cm for 15.16 h at 14 °C, with initial and final switching times of 0.22 and 21.79 s, respectively. The YAC bands were visualized after staining with ethidium bromide.

To confirm the introduction of the apoB95 nonsense mutation into the YAC, targeted YAC clones were identified by enzymatically amplifying a 354-base pair (bp) fragment of exon 29 (apoB cDNA nucleotides 13,067-13,421) with PCR primers 5-CATTAAACAGCTGAAAGAGATGA-3 and 5-GTTAGAGGCACTGACAATATATTC-3. The PCR product was cleaved with SpeI and analyzed on an ethidium bromide-stained 2% agarose gel. Cleavage of the mutant apoB 354-bp DNA fragment with SpeI yields 222- and 132-bp fragments. To confirm the introduction of the apoB97 mutation, a 480-bp fragment of exon 29 (apoB cDNA nucleotides 13,295-13,775) was amplified with PCR primers 5-GATCCAAGTATAGTTGGCTGG-3 and 5-GTTCATGACTGTGGTTGATTGC-3. The PCR product was cleaved with StuI and analyzed on an ethidium bromide-stained 2% agarose gel. Cleavage of the mutant apoB 480-bp DNA fragment with StuI yields 362- and 118-bp fragments.

The human apo(a) transgenic mice used in this study were originally generated and characterized by Chiesa et al. (19). Transgenic mice expressing wild-type human apoB100, human apoB90, and a mutant human apoB (C4326G) have been described previously (13, 20, 21). To generate mice expressing apoB95 and apoB97, mutant YAC DNA was purified from pulsed field agarose gels (22), adjusted to a concentration of 2.0 ng/µl, and microinjected into F2 C57BL/6 × SJL fertilized mouse eggs. Transgenic founders were identified by the presence of human apoB in the plasma, either by Western blot analysis or by a human apoB-specific radioimmunoassay (RIA) (21).

To assess the size of the human apoB proteins in the plasma of transgenic founder animals, fresh plasma was obtained, and the plasma proteins were size-fractionated on SDS-4% polyacrylamide gels under reducing conditions. The separated proteins were transferred to a sheet of nitrocellulose membrane for Western blot analysis, using the human apoB-specific monoclonal antibody 1D1, which binds between amino acids 474 and 539 of apoB100 (23). For these experiments, plasma samples from human apoB100 (20) or human apoB90 (21) transgenic mice were used as controls. The distribution of apoB and lipids in the plasma of transgenic mice was analyzed by fast phase liquid chromatography (FPLC), using a Superose 6 10/30 column (20).

The ability of apoB95 and apoB97 to bind covalently to apo(a) and form Lp(a) was assessed using the Western blot assay described by Chiesa et al. (19). The plasma from a human apo(a) transgenic mouse (19) was incubated with plasma samples from a transgenic mouse expressing wild-type human apoB100 (19), a transgenic mouse expressing a mutant human apoB100 (C4326G) (13), a human apoB95 transgenic mouse, and a human apoB97 transgenic mouse, according to methods described previously (13, 21). Recombinant human apo(a) from transgenic mouse plasma (3.0 µl) was incubated with samples of wild-type and mutant forms of human apoB contained in transgenic mouse plasma. Incubations were performed in 0.15 M NaCl in a total volume of 40 µl for 4 h at 37 °C. Each incubation contained an identical amount of apoB, as judged by a monoclonal antibody-based human apoB RIA. Samples (5 µl) were removed from the incubations after 10, 30, 60, 120, and 240 min and subjected to electrophoresis on SDS-4% polyacrylamide gels under nonreducing conditions. The separated proteins were transferred to a nitrocellulose membrane for Western blot analysis with the human apo(a)-specific antibody IgG-a5 (24, 25), which was conjugated to horseradish peroxidase. Lp(a) formation was also assessed with a specific "sandwich" RIA using 96-well polyvinylchloride plates, as described previously.

The human apoB95 and human apoB97 transgenic mice were bred with mice that were hemizygous for a human apo(a) transgene (19). The offspring were screened for apoB expression by Western blot analysis with monoclonal antibody 1D1. To identify mice expressing human apo(a) and to document the amount of apo(a) that was covalently linked to apoB, Western blots of nonreduced SDS-4% polyacrylamide gels were performed with apo(a)-specific monoclonal antibody IgG-a5.

RESULTS

To assess the importance of the carboxyl terminus of apoB in Lp(a) formation, we compared the ability of apoB95 and apoB97 to bind to apo(a) and form Lp(a). The apoB95 and apoB97 nonsense mutations were introduced into exon 29 of the apoB gene by pop-in, pop-out gene targeting in a 108-kb YAC containing the entire human apoB gene (Fig. 1). After the pop-in step, the size of the YAC increased to 115 kb, reflecting integration of the 7-kb gene-targeting vector (Fig. 2A, lanes 2 and 3). To document that the apoB95 mutation had been retained within the pop-in YAC, we enzymatically amplified a segment of the apoB gene flanking the apoB95 nonsense mutation and then cleaved the amplified DNA with SpeI (Fig. 2B, lane 2); for the apoB97 experiments, the PCR product was cleaved with StuI (Fig. 2B, lane 4). The gene-targeting process was completed by growing yeast harboring the apoB95 and apoB97 pop-in YACs in 5-fluoroorotic acid, which selects for growth of clones that had lost the URA3 gene through an intrachromosomal recombination event. After this pop-out step, the apoB95 and apoB97 YACs returned to the original size of 108 kb (Fig. 2A, lanes 4 and 5). Retention of the targeted mutations after the pop-out step was documented by digesting enzymatically amplified fragments of exon 29 DNA with either SpeI or StuI (Fig. 2B, lanes 5 and 6). The overall efficiencies for both the pop-in and pop-out steps were high. Approximately 30% of the pop-in colonies were appropriately targeted and contained a single integrated vector. Most of the other colonies had two or three copies of the gene-targeting vector. Approximately 25% of the pop-out colonies had undergone the predicted intrachromosomal gene-targeting event and retained the targeted mutation.

To generate transgenic mice, the apoB95 and apoB97 YACs were purified from pulsed field agarose gels and microinjected into fertilized mouse eggs. Eleven of 58 newborn pups expressed apoB95, and 4 of 46 pups expressed apoB97. The founders expressing the highest levels of apoB were bred with C57BL/6 mice to establish transgenic lines. Fig. 3 shows apoB95 and apoB97 in the plasma of founder mice. The plasma of the apoB95 and apoB97 transgenic mice contained elevated levels of cholesterol and triglycerides in the low density lipoprotein fraction, as judged by FPLC fractionation studies (20) (data not shown).

To test the ability of apoB95 and apoB97 to bind covalently to apo(a) and form Lp(a), plasma samples from these mice were incubated with apo(a), and the incubation mixtures were analyzed by Western blots of SDS-polyacrylamide gels with an apo(a)-specific monoclonal antibody (Fig. 4). Lp(a) is easily distinguished from apo(a) by Western blot analysis because the covalently bound apoB-apo(a) complex is much larger and migrates more slowly on SDS-4% polyacrylamide gels. These experiments revealed that Lp(a) formation was detectable with both apoB95 and apoB97; however, a relatively small amount of Lp(a) was formed with apoB95, suggesting that Lp(a) assembly with apoB95 was less efficient than with apoB100.

To explore Lp(a) formation with apoB95, apoB97, and apoB100 further, we used the same Western blot assay to examine the kinetics of Lp(a) formation. In these experiments, apoB95, apoB97, wild-type apoB100, and a mutant apoB100 (C4326G) lacking the site of attachment for apo(a) (Fig. 5A) were incubated with apo(a), and the formation of Lp(a) was assessed at 10, 30, 60, 120, and 240 min. Each of the incubation mixtures contained an identical amount of apoB, as judged by a monoclonal antibody-based RIA (Fig. 5B). With wild-type human apoB100, Lp(a) formation was evident by 10 min, and by 30 min, all of the apo(a) was covalently bound to apoB100 (Fig. 5A, lane 2). As expected, no Lp(a) was formed with the mutant apoB100 (C4326G) (Fig. 5A, lane 3). With apoB95, there was no evidence of Lp(a) formation at 10 or 30 min, and only trace amounts were observed at 60 min. Even after 120 and 240 min, large amounts of free apo(a) were still present (Fig. 5A, lane 4). ApoB97 formed Lp(a) more rapidly than apoB95, with clear cut Lp(a) formation after 10 min. However, the Western blots suggested that apoB97 might form Lp(a) less efficiently than apoB100, in that small amounts of free apo(a) could be detected at the later time points (Fig. 5A, lane 5).

To further analyze whether apoB97 and apoB100 truly differed in their efficiency of interacting with apo(a) and forming Lp(a), we used a solid-phase RIA to assess the kinetics of Lp(a) formation. In these experiments, equal amounts of apo(a) and apoB were incubated for 10-240 min, as described above, and then placed in wells of 96-well polyvinylchloride plates coated with an apo(a)-specific monoclonal antibody, LPA-6. The plates were washed, and the amount of Lp(a) formation was judged by the binding of a ¹²⁵I-labeled monoclonal antibody, C1.4, that is specific for the amino-terminal portion of the apoB molecule (26). As shown in Fig. 6, this RIA provided further evidence that Lp(a) formation was slower with apoB97 than with the wild-type apoB100. Once again, the amount of apoB97 and apoB100 added to the RIA wells was identical (Fig. 5B).

To test the ability of the apoB95 and apoB97 variants to form Lp(a) in vivo, we examined Lp(a) formation in mice expressing human apo(a) and either apoB95 or apoB97. In previously published experiments (20, 21), we had crossed mice expressing high levels (~50 mg/dl) of human apoB100 or human apoB90 (4084 amino acids) with the apo(a) transgenic mice. In mice expressing both apo(a) and apoB100, all of the apo(a) in the plasma was covalently complexed to apoB in the form of Lp(a) (20), whereas all of the apo(a) circulated freely in mice expressing both apo(a) and apoB90 (21). In the current experiments, we mated the human apo(a) transgenic mice with mice expressing high levels (~50 mg/dl) of either apoB95 or apoB97. Offspring from these matings were screened for human apoB and apo(a) by Western blot analysis. Transgenic mice expressing both apo(a) and apoB95 (Fig. 7A, lanes 4, 8, and 9) had only a very small amount of Lp(a) in their plasma, and most of the apo(a) in the plasma was free. In contrast, mice expressing both apo(a) and apoB-97 (Fig. 7B, lanes 3 and 8) had large amounts of Lp(a) in their plasma and virtually no free apo(a).

DISCUSSION

We have been interested in defining the structure of Lp(a), with an emphasis on defining the structural features of apoB that are required for its interaction with apo(a). Our earlier site-directed mutagenesis experiments (13) and those of Callow and Rubin (14) demonstrated that the carboxyl-terminal cysteine of apoB, cysteine 4326, is the site of attachment for apo(a). Although the latter studies represented a significant step forward in understanding Lp(a) structure, it is plain that other features of the apoB molecule, aside from a single free cysteine residue, must be important for apoB's interaction with apo(a). It is generally assumed that Lp(a) assembly involves an initial noncovalent interaction between apo(a) and apoB, which brings the two molecules together so that a disulfide bond can form (11). However, there has been a paucity of data regarding the apoB sequences that are important for facilitating the formation of the disulfide bond. In this study, we sought to approach that issue by examining the interaction between apo(a) and two truncated apoB's, apoB95 and apoB97. ApoB95 contains cysteine 4326, followed by four additional residues (for a total of 4330 amino acids). ApoB97 contained the amino-terminal 4397 amino acids of apoB100. The ability of the two different truncated apoB molecules to interact with apo(a) was assessed by monitoring the rate of the disulfide linkage between the two molecules. Interestingly, apoB95 formed bona fide, disulfide-linked Lp(a), although the efficiency of Lp(a) formation was quite poor, as judged both by in vitro assays of Lp(a) assembly and by the analysis of transgenic mice expressing both apo(a) and apoB95. These studies demonstrated that the residues carboxyl-terminal to the last amino acid of apoB95 (amino acids 4331-4536) are not absolutely required for the Lp(a) assembly but contribute significantly to the efficiency of Lp(a) formation. Lp(a) assembly with apoB97 (4397 amino acids) was very robust when compared with apoB95 and only slightly less efficient than with apoB100.

The observation that lysine analogues (e.g., -amino-n-caproic acid) can block the formation of Lp(a) in vitro (19) has led many to assume that lysine residues must be involved in the initial noncovalent interaction between apoB100 and apo(a). -Amino-n-caproic acid binds to the lysine-binding site of plasminogen kringle 4 (27), and it presumably abrogates Lp(a) formation by binding to the analogous site on apo(a). There is little doubt that apo(a) can bind to lysine, since Lp(a) can actually be purified by taking advantage of apo(a)'s ability to bind to lysine-Sepharose affinity columns (28, 29). Apo(a)'s lysine-binding properties led us to hypothesize that apo(a) initially interacts with apoB lysine residues near the site of attachment for apo(a), apoB cysteine 4326. To assess the plausibility of that hypothesis, we used the computer program LOCATE (30) to predict the structure of the amino acids surrounding apoB cysteine 4326. Interestingly, apoB amino acid residues 4315-4338 are predicted to form a class A amphipathic -helix with high lipid affinity (Fig. 8, A and B). Within this -helix, the side chains of cysteine 4326 and three lysine residues (residues 4319, 4322, and 4331) project into the aqueous phase, while the side chains of many hydrophobic residues are buried within the lipid phase. When one looks "down the barrel" of this -helix (Fig. 8A), lysine 4322 and lysine 4319 flank cysteine 4326. As illustrated in the helical net diagram of the -helix (Fig. 8B), lysine 4322 and lysine 4319 are spatially near cysteine 4326 (6.0 and 10.5 Å away, respectively, along the long axis of the -helix). Lysine 4331 is also near cysteine 4326 (7.5 Å along the long axis of the -helix but on the opposite water-lipid interface) (Fig. 8, A and B). As noted earlier, the -helix formed by residues 4315-4338 is predicted to bind lipids with high affinity. However, in the case of apoB95, the truncation after residue 4330 is predicted to substantially reduce the ability of the -helix to associate stably with a lipid interface (see legend to Fig. 8). Thus, our analysis suggested two factors that could potentially underlie the inefficient Lp(a) formation with apoB95: the amphipathic -helix lacks lysine 4331, and its ability to stably associate with a lipid interface is diminished as a result of the truncation.

An amphipathic -helix (class A) spanning residues 4315-4338.

The aforementioned analysis is obviously predicated on the assumption that the sequences adjacent to cysteine 4326 are important for the initial interaction of apo(a) and apoB. However, our data cannot exclude the possibility that the apoB truncations result in conformational changes that indirectly affect the efficiency of Lp(a) assembly. For example, the slightly lower efficiency of Lp(a) formation with apoB97 than with apoB100 could be due to conformational changes that limit the access of apo(a) to its binding site on apoB. In addition, the markedly impaired efficiency of Lp(a) formation with apoB95 could result (at least in part) from conformational changes that prevent the binding of apo(a) to a binding site that is distant from cysteine 4326.

The principal finding of this study, that the efficiency of Lp(a) formation is dependent on the carboxyl terminus of the molecule (particularly residues 4331-4397), provides a new foundation for studying the structural features of apoB required for Lp(a) assembly. The techniques used in this study, mutagenesis of a human apoB-YAC followed by expression studies in transgenic mice, provide a practical paradigm for future investigations to further define the nature of the interaction between apo(a) and apoB. The YAC gene-targeting system is effective and versatile and can be used to eliminate the lysines that surround cysteine 4326, to scramble stretches of amino acid residues, to replace human sequences with mouse sequences, or even to move residues 4315-4338 (the amphipathic -helix containing cysteine 4326) to another site on the apoB molecule. The efficiency of generating human apoB transgenic mice with YAC DNA is high (equivalent in our hands to that obtained with plasmid or P1 DNA) (12, 13), and the ability to mate the apoB transgenic mice with apo(a) transgenic mice provides an important in vivo tool for assessing the importance of specific apoB sequences in Lp(a) assembly. In addition, we believe that this series of experiments and analyses opens the door for other experimental strategies, such as testing whether certain apoB polypeptides can interfere with Lp(a) formation. Delineating the apoB residues that cause it to bind to apo(a) could lead to the development of pharmaceutical agents that would interfere with the assembly of Lp(a).