Expression of Human Apolipoprotein B90 in Transgenic Mice DEMONSTRATION THAT APOLIPOPROTEIN B90 LACKS THE STRUCTURAL REQUIREMENTS TO FORM LIPOPROTEIN(a)*

Lipoproteida) (Lp(a)) is a lipoprotein formed by the disulfide linkage of apolipoprotein(a) (apota)) to the apoBlOO of a low density lipoprotein particle. Earlier site-directed mutagenesis studies of apo(a) demonstrated that apo(a) cysteine 4057 is required for the di- sulfide linkage; however, the cysteine residue within apoBlOO that is involved in the disulfide bond has not been identified. We previously demonstrated that the apoBlOO produced by human apoB transgenic mice binds to apo(a) and forms Lp(a) (Linton, M. F., Farese, R. V., Jr., Chiesa, G., Grass, D. S., Chin, P., Hammer, R. E., Hobbs, H. H., and Young, S. G. (1993) J. Clin. Invest. 92, 3029303’7). To further explore the structural features of human apoB that are required for the formation of Lp(a), we used a transposon-interrupted human apoB

consists of a complex between the apolipoprotein (apo) BlOO of a low density lipoprotein (LDL) particle and apolipoprotein(a) (apo(a)), a glycoprotein that is homologous to plasminogen, a serine protease. High concentrations of Lp(a) in the plasma have been shown to be a strong risk factor for premature coronary heart disease in many human populations (4, 5), although it is noteworthy that several recent studies have failed to identify Lp(a) as a significant risk factor for coronary disease (6)(7)(8). The reason(s) for the atherogenicity of Lp(a) are not known with certainty, but several possible mechanisms have been suggested (3).
The structural features of apo(a) and apoB that lead to their association and formation of Lp(a) are incompletely understood. Strong evidence exists, however, that the association of apo(a) and apoB involves a disulfide bond between cysteine residues of the two molecules. Although Lp(a) cannot be dissociated into apo(a) and LDL-apoB100 by either chaotropic agents or by heating to 100 "C (9), Lp(a) can be completely dissociated by disulfide reducing agents such as dithiothreitol (10,11). Studies with reducing agents cannot, however, prove the existence of a sulfhydryl linkage because both apoBlOO and apo(a) also have multiple internal disulfide bonds (12,13), and it could be argued that reducing agents could dissociate apo(a) and apoBlOO by breaking internal disulfide bonds, thereby altering the conformation of one or both molecules and interfering with a noncovalent interaction. Stronger evidence in favor of a disulfide linkage has come from recent functional studies of mutant apo(a) molecules in which the best candidate for a free cysteine residue, Cys4057 (13), was changed to other amino acid residues (14, 15). Kochinsky et al. (15) and Brunner et al. (14) mutated Cys4057 t o several other amino acid residues and then expressed the mutant apo(a) proteins in cultured cells. Unlike native apo(a), the mutant apo(a) did not bind to human LDL-apoBlOO and form Lp(a). Although one still could argue that the substitution of other amino acid residues for Cys405' might alter the conformation of the apo(a) molecule, these mutagenesis experiments, taken together with the earlier studies, constitute persuasive evidence that apo(a) and apoBlOO are disulfide-linked.
The structural features of apoBlOO that are essential for its interaction with apo(a) have not yet been identified. Most investigators presume that a specific domain of apoBlOO binds to apo(a), thereby bringing Cys4057 of apo(a) into close proximity with a free cysteine of apoBlOO to facilitate the formation of a disulfide bond. Trieu and McConathy (16) have reported the existence of hydrophobic interactions between LDL and apo(a) that may be relevant to the formation of Lp(a), but to date neither the specific region of the apoBlOO molecule that binds to apo(a) nor the apoB cysteine residue involved in the disulfide linkage has been identified with certainty. Molecular modeling Human ApoB90 Cannot Form Lp(a) studies (17), immunochemical studies using apoB peptide-specific antibodies (18), and studies involving the fluorescent labeling of the free cysteines of Lp(a) (19) have implicated apoB Cys3734 as a residue that might be involved in the disulfide linkage. However, no one has definitively identified the apoBlOO cysteine by protein sequencing or by examining the functional characteristics of mutant apoB molecules.
In 1993, we demonstrated that the human apoB produced by human apoB transgenic mice is capable of interacting with apo(a) to form bona fide Lp(a) particles (20). Expression of human apoB in the mouse is useful for studying apo(a1-human apoB interactions because the endogenous mouse apoB lacks the structural features required for complexing with apo(a) (9).
To examine the structural features of apoB required for the formation of Lp(a), we generated transgenic mice expressing a truncated form of human apoB, apoB90, and then tested the capacity of apoB9O to form Lp(a) particles.

MATERIALS AND METHODS
Dansgenic Mice-Human apo(a) transgenic mice (9) were obtained from Robert Hammer (University of Texas Southwestern Medical Center, Dallas, TX). We previously generated transgenic mice expressing human apoBlOO by microinjecting murine zygotes with a P 1 bacteriophage clone, p158, that spanned the human apoB gene (20). For the in vitro assays of Lp(a) formation, we used the plasma of the F, offspring of human apoB transgenic founder 1102; the mean concentration of human apoBlOO in the plasma of chow-fed female and male mice was 63 and 39 mg/dl, respectively. For the breeding experiments with apo(a) transgenic mice, we used the F, offspring of the human apoB transgenic founder 1095. The human apoBlOO level in these mice was -8 mg/dl.
To generate transgenic mice expressing human apoB90, we interrupted p158 (the P1 clone containing the apoB gene) (20) with a minimid pZTneo5, obtained from Nat Sternberg of the DuPont Merck TnlO transposon containing a neomycin resistance gene (from the plas-Pharmaceutical Company, Glenolden, PA) according to techniques outlined in a recent review (21).
Briefly, Escherichia coli cells (strain NS3529) harboring the P1 clone, p158, were grown in Luna-Bertani broth (LB) containing 50 pg/ml of kanamycin (the P1 vector contains a kanamycin resistance gene). The cells were made competent and transformed with the transposon-containing plasmid, pZTneo5, which contains an ampicillin resistance gene. The transposable element within pZTneo5 contains a chloramphenicol resistance gene and a n RSVneo gene. The transposable element also contains a Not1 restriction site, which is useful for localizing the site of transposition. Following the transformation of the P1-containing E. coli with pZTneo5, multiple drug-resistant colonies were obtained on LB agar plates containing chloramphenicol, ampicillin, and kanamycin. A single colony was grown in 10 ml of LB containing chloramphenicol (25 pg/ml) and kanamycin (50 pg/ml) until an A, , of 0.05 was reached. Then, isopropyl-1-thio-& D-galactopyranoside was added to the culture (final concentration, 1 m), and the cells were grown for 3 h. The cells were then pelleted by centrifugation and resuspended in 1 ml of LB containing 5 mM CaCl, (LB-Ca). The isopropyl-1-thio-p-D-galactopyranoside-induced cells were infected with a virulent strain of the P1 bacteriophage (provided by N. Sternberg) at a multiplicity of infection of 2 and were incubated at 37 "C. After 10 min, 2 ml of LB-Ca were added, and the incubation was continued for 2 h. To generate a P1 phage lysate, the cells were lysed by the addition of several drops of chloroform. To prepare the phage lysate for infection of E. coli cells, 200 pl of the phage lysate was added to a sterile tube, and the chloroform was allowed to evaporate for 10 min at 37 "C. To infect NS3529 E. coli with the phage lysate, the cells were grown in 10 ml of LB to a n A,,, of 0.4, pelleted by centrifugation, and resuspended in 1 ml of LB-Ca. A total of 200 p1 of the NS3529 cells were added to 200 pl of the lysate and incubated for 2 h at 37 "C. The E. coli cells were harvested by centrifugation and resuspended in 400 pl of LB. To isolate E. coli colonies containing a transposon-interrupted p158 plasmid, various amounts of the phage-infected cells were spread onto LB agar plates containing both chloramphenicol and kanamycin. Single colonies were grown in LB containing kanamycin, and P1 plasmid DNA was isolated by a standard alkaline lysate technique using lysozyme (22). NotYSalI restriction digestion of the miniprep DNAfrom 40 transposon-interrupted P1 clones resulted in the identification of one clone, p158Ineo8, that appeared to have a n exon 29 interruption. Automated DNA sequencing of p158/neo8 using exon 29 oligonucleotide primers revealed that the transposon was inserted into exon 29 of the apoB gene (at apoB cDNA nucleotide 12461); on the basis of the DNA sequencing results, p158/neo8 was predicted to encode a truncated apoB protein, apoB90, containing 4084 amino acids.
We isolated the p158/mo8 DNAfromE. coli strain NS3529 by alkaline lysis (22) using lysozyme (Boehringer Mannheim) and purified it from the lysate using QIAGEN-tip 500 columns (Qiagen, Inc., Chatsworth, CA) according to the manufacturer's procedures. To purify the insert of p158/neo8 for microinjection of murine zygotes, the DNA was digested with MluI, a restriction enzyme that cleaves a t four sites within the P1 vector sequences but does not cleave within the insert of p158/neo8. A 91-kilobase MluI fragment containing the transposon-interrupted apoB gene (and approximately 8 kilobases of P1 vector sequences) was purified from a 1% pulsed-field agarose gel by digesting the agarose with P-agarase according to the procedures described by Schedl et al. (23) for the purification of yeast artificial chromosome DNA. The P-agarase and all restriction enzymes were purchased from New England Biolabs (Beverly, MA). After digestion with agarase, the DNA-containing solution was spun in a microcentrifuge a t 14,000 rpm for 10 min to sediment undigested agarose gel fragments; the DNAwas dialyzed against a TE (10 mM Tris, 1 mM EDTA) buffer containing 100 mM NaCl, 30 spermine, 70 p~ spermidine and then adjusted to a concentration of 3 ng/pl. The purified p158/neo8 MluI fragment was microinjected into ICR zygotes in the transgenic core facility of the Gladstone Institute of Cardiovascular Disease and by David Grass of DNX Biotherapeutics, Inc., a t Princeton, NJ, according to standard techniques (24).
To identify transgenic founder animals, we analyzed mouse plasma for the presence of human apoB production by Western blot analysis or by radioimmunoassay (RIA), as described below. All mice were fed a mouse chow diet.
Monoclonal Antibodies-To identify human apoB in mouse plasma samples, we used two different apoB100-specific monoclonal antibodies, MB47 and MB43, which we previously had generated and characterized (5,26). Antibody MB47 binds near amino acid 3500; antibody MB43 binds between amino acids 4027 and 4081 (27). The apoB-specific monoclonal antibody C1.4 (281, which has an epitope near amino acid 500, was provided by Gustav Schonfeld of the Washington University School of Medicine in St. Louis, MO. The apoB-specific monoclonal antibodies 1D1, Bsol 16, and 2D8 (27) were obtained from Ross Milne and Yves Marcel of the University of Ottawa Heart Institute (Ontario, Canada). Antibody 1D1 binds between amino acids 474 and 539; antibody 2D8 binds between amino acids 1438 and 1480; Bsoll6 binds between amino acids 4157 and 4189. A horseradish peroxidase-labeled monoclonal antibody specific for human apo(a), I&-1A2 (9), was obtained from Francesco Mancini (University of Texas, Southwestern).
We generated another apo(a)-specific antibody, LPA6, for use in this study. To develop antibody LPA6, we isolated human LDL (d = 1.019-1.063 g/ml) from a human subject with high plasma levels of Lp(a). The LDL was then adjusted to d = 1.060 g/ml and subjected to ultracentrifugation for 24 h; the "dense" LDL that floated to the top of the ultracentrifuge tube was used to immunize mice. Fusions of splenic lymphocytes with P3Ag8.653.1 myeloma cells were performed as described previously (29). Hybridoma colonies were screened initially for binding to the immunogen by solid phase RIA. An apo(a)-specific antibody, LPA6, was subsequently identified by its ability to bind to recombinant apo(a) (a gift of Richard Smith of J & J Biotechnology, La Jolla, CA). Antibody LPA6 is a n IgG2a K antibody and does not bind to human plasminogen i n a solid phase RIA.
Radioimmunoassays of Human ApoB-To measure the concentration of human apoB in the plasma of transgenic mice, human plasma lipoproteins, or the plasma fractions obtained from Superose 6 chromatography, we used the solid phase "sandwich RIA" described in detail by Linton et al. (20). Briefly, this RIA utilized two human apoB-specific monoclonal antibodies, MB47 and C1.4. Ninety-six-well polyvinyl chloride plates were coated with antibody MB47. After blocking the plates with bovine serum albumin (BSA), we added the samples to be tested to the plate and incubated them for 16 h a t 4 "C. After the plates were washed, the human apoB that had been captured by the immobilized antibody MB47 was detected with 1261-labeled antibody C1.4. In experiments to assess the human apoB concentration in the in vitro incubations with mouse plasma containing apo(a) and apoB90, the 96-well polyvinyl chloride plates were coated with phosphate-buffered saline containing 2 pg/ml of monoclonal antibody 2D8 rather than antibody MB47. In this RIA, lZ5I-C1.4 was also used to detect the apoB captured by the immobilized antibody 2D8.
Western Blot Analysis-To identify human apoB in transgenic mouse plasma, 1 pl of plasma was size-fractionated on a 4% SDS-polyacrylamide gel under reducing conditions (3% 2-mercaptoethanol in the Human ApoB90 Cannot Form Lp(a) sample buffer). The separated proteins were electrophoretically transferred onto a nitrocellulose membrane for Western blot analysis (30). Chromatographic Separation of the Plasma Lipoproteins-To analyze the distribution of cholesterol, triglycerides, and human apoB in the plasma of transgenic mice, plasma samples were fractionated on a Superose 6 10/50 column (Pharmacia Biotech Inc.) (20,31). The cholesterol and triglyceride concentration in each fraction was determined with kits from Abbott Diagnostics (North Chicago, IL) and Boehringer Mannheim, respectively. The relative content of human apoB in each fraction was assessed by the solid phase sandwich RIA (20).
Incubation of Dansgenic Mouse Plasma with Apola&The formation of Lp(a) in vitro was assessed according to the techniques described by Chiesa et al. (9). In this assay, recombinant human apo(a) from human apo(a) transgenic mouse plasma (5 p1) was incubated with a sample of human apoB-containing lipoproteins (1 pg) in a 0.15 M NaCl solution (total incubation volume, 40 111); after a 4-h incubation a t 37 "C, the mixture was size-fractionated on a 4% SDS-polyacrylamide gel under nonreducing conditions as well as reducing conditions (incubation with 3% 2-mercaptoethanol for 10 min at 90 "C). The separated proteins were then electrophoretically transferred to a sheet of nitrocellulose membrane for Western blotting with a horseradish peroxidase-labeled apo(a)-specific monoclonal antibody, IgG-lA2. Because Lp(a) migrates at a much higher molecular weight than the free apo(a), the formation of Lp(a) can be detected on an IgG-1A2 Western blot by the appearance of the high molecular weight band (and the disappearance of the lower molecular weight, free apo(a) band). We tested four samples for their capacity to form Lp(a) in this assay system. First, we tested the plasma of a human subject with homozygous hypobetalipoproteinemia (32) who exclusively synthesized apoB45.2 (and absolutely no apoB48 or apoB100). The plasma of this subject did not contain detectable levels of Lp(a) or apo(a). We also tested the plasma of human apoBlOO transgenic mice and the plasma of human apoB9O transgenic mice. As a positive control for Lp(a) formation, the apo(a) transgenic mouse plasma was incubated with human d e 1.063 g/ml lipoproteins, which were prepared from the plasma of a normolipidemic human subject by ultracentrifugation (33). Finally, as a negative control, the apo(a) transgenic mouse plasma was either incubated alone or incubated with nontransgenic mouse plasma. Before setting up the incubations, the concentration of apoB in each sample was determined by both the sandwich RIA and a Western blot analysis using monoclonal antibody 1D1 or MB43. Each of the incubations of apo(a) with the human apoB-containing samples contained equal amounts of apo(a) (5 p1 of plasma) and apoB (1 pg). The monoclonal antibodies we used to assess apoB concentration bound apoB9O and apoBlOO with equal affinity because no differences in the slopes of the apoB9O and apoBlOO binding or displacement curves were noted in three different RIA formats. The first was a competitive RIA that tested the capacity of the apoB samples to compete with '"I-LDL for binding to immobilized antibody MB47 (20). The second was a sandwich RIA using antibody MB47 as the capture antibody and '251-C1.4 as the detecting antibody (20), and the third was an otherwise identical sandwich RIA using monoclonal antibody 2D8 as the capture antibody and 12s11-C1.4 as the detecting antibody.
All in vitro apo(a) incubations also were performed in the presence of 100 mM E-aminocaproic acid (Sigma), a lysine analog that interferes with the binding of human LDL to apo(a) (34).
Lipoprotein (a) Sandwich RIA-To assess the formation of Lp(a) during the incubation of apo(a) with human apoB-containing lipoproteins, we used a sandwich RIA, using antibody LPAG as the "capture" antibody and 12sI-C1.4 to detect the Lp(a) that was captured by LPA6. Each well of the flexible polyvinyl chloride 96-well plates was incubated with 50 pl of phosphate-buffered saline containing LPAG (5 pg/ml) for 16 h a t 4 "C.
The plates were washed four times with phosphate-buffered saline containing 0.1% RIA-grade bovine serum albumin, 0.05% Tween 20, and 0.04% sodium azide (SPRIA) and then incubated with 200 pl of SPRIA containing 2.0% BSA (SPRIA-BSA). A total of 5 p1 from each of the in vitro Lp(a) incubation mixtures (described in the preceding two paragraphs) were adjusted to a volume of 50 pl with SPRIA-BSA and added to the LPA6-coated plates in triplicate. After a 16-h incubation at 4 "C, the plates were washed six times with SPRIA. To detect the Lp(a) that had been captured by the immobilized antibody LPA6, the plates were incubated with 50 pl of SPRIA-BSA containing 12s11-C1.4 (8,000 c p d p l ) for 4 h a t 4 "C. The plates were washed six times with SPRIA, and the wells were counted in a y counter.

RESULTS AND DISCUSSION
In 1993 we generated human apoBlOO transgenic mice, using a P1 bacteriophage clone, p158, that spanned the entire

Apo-090
Apo-048---1 2 1 2 1 2 1 2 FIG. 1. Western blots of the plasma of human apoBl00 and human apoB9O transgenic mice, using different human apoBspecific monoclonal antibodies. The plasma (1 pl) from an apoBlOO (lane I ) and an apoB9O (lane 2 ) transgenic mouse was subjected to electrophoresis on a 4 6 SDS-polyacrylamide gel under reducing conditions; the separated proteins were then electrophoretically transferred to a nitrocellulose membrane. The membrane was probed with four different apoB-specific monoclonal antibodies: 1D1, MB47, MB43, and Bsol 16. human apoB gene; these mice were bred with apo(a) transgenic mice to develop mice expressing high levels of Lp(a) (20). In this study, we interrupted p158 with a transposon to generate a mutant apoB gene clone, p158/neo8, encoding a carboxyl-terminally truncated apoB protein, apoB9O. Microinjection of the 91-kilobase MluI fragment of p158/neo8 into fertilized ICR eggs yielded 92 offspring, 12 of which expressed human apoB9O in their plasma. An additional six offspring were obtained from microinjected C57/B16 x SJL F, zygotes; one of these animals expressed apoB9O. The apoB9O concentration in the plasma of founder animals ranged from 5 to 60 mg/dl, similar to the range of apoBlOO concentrations in the apoBlOO transgenic mice (20).
DNA sequencing of p158/neo8 predicted that the construct should code for a truncated apoB protein containing the aminoterminal 4084 amino acids of apoB100. To confirm that the apoB protein in the plasma of the mice was of the predicted size, we performed Western blots of plasma samples from a human apoBlOO transgenic mouse and an apoB9O transgenic mouse, using selected apoB-specific monoclonal antibodies (Fig. 1). A Western blot using antibody 1D1 revealed that the plasma of both the apoB9O and the apoBlOO mice contained human apoB48. As expected, apoB9O was bound by antibody MB43 (the epitope for antibody MB43 is located between amino acids 4027 and 4081) but was not bound by antibody Bsol 16 (the epitope for antibody Bsoll6 is located between amino acids 4157 and 4189). Of note, these Western blots demonstrated that the concentration of apoB9O in the plasma of the apoB9O transgenic animal was equal to that observed in the "high expressing" apoBlOO animals.
To assess the distribution of cholesterol, triglycerides, and human apoB in the plasma of the transgenic animals, the plasma from one apoB9O transgenic mouse and one apoBlOO transgenic mouse was fractionated by Superose 6 chromatography. Compared with the plasma of the male apoBlOO transgenic animal, the plasma of the male apoB9O transgenic animal had a slightly lower LDL cholesterol peak (Fig. 2 A ) but a slightly higher LDL triglycerides peak (Fig. 2B 1. The amounts of apoB in the plasma of the apoB9O and apoBlOO mice were similar (Fig. 2C). Almost all of the human apoB9O and human apoBlOO was found in LDL-sized particles. The apoB9O lipoproteins were, as expected (35)(36)(37)(38), slightly smaller than the lipoproteins containing apoB100. One could argue that the smaller size distribution of the apoB9O particles might reflect more extensive metabolism of these particles in the plasma. However, we have used p158 and p158Ineo8 to generate rat hepatoma cell lines that secrete human apoBlOO and apoB9O. When the cell culture medium from these cells was subjected to Superose 6 chromatography, we found that the apoB90-containing lipoproteins were slightly smaller than the apoB100containing lipoproteins.' Of note, Chiesa and co-workers (9)   To assess the capacity of apoB9O to form a complex with recombinant human apo(a), the plasma of the apoB9O transgenic mice was incubated with the plasma of human apo(a) transgenic mice at 37 "C for 4 h. Following the incubation, the mixtures were size-fractionated on SDS-polyacrylamide gels, and Western blotting was performed with the apo(a)-specific antibody, IgG-1A2. In this assay system, Lp(a) was distinguished from apo(a) by its size. No Lp(a) formation was detected after incubation of the apoB9O transgenic mouse plasma with apo(a) (Fig. 3A, lane 4 ) . Similarly, no Lp(a) was formed during the incubation of apo(a) with the plasma of a human subject who synthesized only apoB45.2 (Fig. 3A, lane 5). In contrast, the incubation of apo(a) with human d < 1.063 g/ml lipoproteins or with the plasma of a human apoBlOO transgenic mouse (Fig. 3A, lanes 2 and 3, respectively) resulted in abundant Lp(a) formation. In fact, virtually all of the free apo(a) in these incubations was converted to Lp(a). The formation of Lp(a) was also eliminated when a reducing agent, 3% 2-mercaptoethanol, was added to the incubation mixtures before electrophoresis on the SDS-polyacrylamide gels (Fig. 3B). We also found that the formation of Lp(a) could be prevented by including 100 mM eaminocaproic acid in the in vitro incubations (Fig.  3C). It is important to note that the amount of human apoB in each of the in vitro incubations was the same, as judged by a Western blot using monoclonal antibody MB43 (Fig. 3 0 ) or by a human apoB sandwich RIA (Fig. 4, shaded bars). A monoclonal antibody-based sandwich RIA was also used to assess the formation of Lp(a) in the in vitro incubations (Fig. 4, solid bars). In this assay, antibody LPA6 was used as the "capture antibody," and '251-C1.4 was used to quantify the bound Lp(a). Substantial lZ5I-C1.4 binding to the Lp(a) was observed upon incubation of apo(a) with human d < 1.063 g/ml lipoproteins or with the plasma of the apoBlOO transgenic mice, whereas little lZ5I-C1.4 binding was observed upon incubation of mixtures consisting of apo(a) and plasma samples containing either apoB45.2 or apoB9O.
We previously demonstrated that mice expressing both human apoBlOO and apo(a) have bona fide Lp(a) in their plasma (20). To test the possibility that apoB9O might associate with apo(a) in vivo, we established a mating between a female apoB9O transgenic animal and a male that was hemizygous for the apo(a) transgene. One of the four offspring expressed both transgenes. In that animal, all of the apo(a) circulated free of the lipoproteins, and no Lp(a) was detectable, even though the animal had high plasma levels of apoB90, -50 mg/dl (Fig. 5A). In parallel, we established a mating between an apo(a) male and an apoBlOO female (from human apoBlOO transgenic line 1095, which expresses relatively low levels of apoB100). One of the four offspring expressed both transgenes. Even though this animal had an apoBlOO level of -8 mg/dl, virtually all of the apo(a) was in the form of Lp(a) (Fig. 5B).  (a) with h u m a n d < 1.063 g/ml lipoproteins, h u m a n apoBlOO transgenic mouse plasma, human apoB9O transgenic mouse plasma, and the plasma of the h u m a n apoB45.2 homozygote. The amount of human apoB in each incubation mixture (shaded bars) and the amount of Lp(a) formed during the incubations (solid bars) were assessed by monoclonal antibody-based sandwich RIAs as described under "Materials and Methods." Each bar shows the specific counts bound per well. Identical data were obtained from four independent experiments with different apo(a)-apoB incubations. Equivalent results for Lp(a) formation were obtained using a different sandwich RIA, in which antibody MB47 was used as the capture antibody and ""I-LPA6 (iodinated by Bolton-Hunter reagent) was the detecting antibody. This study, which involved both in vitro incubation studies and in vivo studies with apoB90/apo(a) transgenic mice, demonstrated that human apoB9O cannot form Lp(a). "he failure of apoB9O to form Lp(a) very strongly suggests that the carboxylterminal 10% of apoBlOO is required for Lp(a) formation. The results do not, however, completely exclude the possibility that all of the important structural features for Lp(a) formation exist within apoB9O but that these structures are not in the proper conformation because of secondary factors related to the length of apoB9O. For example, altered lipid composition of apoB90-containing lipoproteins might alter the conformation of apoB9O on the surface of the lipoprotein, affecting its ability to bind to apo(a). In any case, our data indicating that apoB9O cannot form Lp(a) are consistent with the preliminary clinical observations on the Lp(a) of a human subject who is a compound heterozygote for familial hypobetalipoproteinemia, H. J. B. H. J. B. had two mutant apoB alleles, one yielding apoB37 (1728 amino acids) (39) and another yielding apoB86 (3896 amino acids) and apoBlOO (40). The d < 1.063 g/ml lipoproteins of H. J. B. contained a small amount of both apoB86 and a larger amount of apoB100; a nonreduced SDS-gel of the high density lipoprotein fraction (d = 1.075-1.21 g/ml) revealed a large amount of Lp(a) but no free apoB86 or apoB100. When the Lp(a) was subjected to a disulfide reducing agent, it dissociated into apo(a) and apoB100, but we were never able to detect any apoB86 resulting from the dissociation of the Lp(a) (41). The carboxyl-terminal 10% of apoBlOO contains two cysteine residues, CYS~'~' and Cys4"'. A possibility consistent with our results is that one of these carboxyl-terminal cysteines is involved in the disulfide linkage. Of these 2 cysteines, the studies of Coleman et al. (19) would favor CYS~'~'', because that residue, unlike Cys4325, is accessible to labeling with a fluorescent sulfhydryl probe. However, our data do not permit us to exclude the possibility that a more remote cysteine is involved in the disulfide linkage. Molecular modeling studies (17) and immunochemical studies (18) have suggested the possibility that apoB Cys3734 might be involved in the disulfide linkage. It is conceivable that the carboxyl-terminal 10% of apoBlOO contains sequences that are required for the noncovalent association of apo(a) and apoBlOO and that this noncovalent association facilitates the formation of a disulfide bond involving a more remote cysteine, such as Cys3734. We believe that site-directed mutagenesis of each of the 4 carboxyl-terminal cysteine residues of apoBlOO will ultimately be required to establish the identity of the cysteine residue that is involved in the disulfide linkage between apoB and apo(a). An equally important topic of investigation, however, will be the identification of the structural features within the carboxyl-terminal 10% of apoBlOO that facilitate its interaction with apo(a). Advances in molecular techniques (42,43) will make the generation of subtle mutations within the full-length apoB gene clone practical and will allow us in the future to undertake more detailed studies of the structural requirements for Lp(a) formation.

A. Apo(a)
In 1992, Sternberg (44) pointed out that it should be possible to interrupt P1 clones with transposons to generate truncated proteins, thereby facilitating the investigation of protein structure/function relationships. To our knowledge, our study is the first application of Sternberg's suggestion, and it represents the first example of the expression of a mutant form of human apoB in transgenic mice.