Structural features of apolipoprotein B synthetic peptides that inhibit lipoprotein(a) assembly

Lipoprotein(a) [Lp(a)] is assembled via an initial noncovalent interaction between apolipoprotein B100 (apoB) and apolipoprotein(a) [apo(a)] that facilitates the formation of a disulﬁde bond between the two proteins. We previously reported that a lysine-rich, (cid:2) -helical peptide spanning human apoB amino acids 4372–4392 was an effective inhibitor of Lp(a) assembly in vitro. To identify the important structural features required for inhibitory action, new vari-ants of the apoB 4372-4392 peptide were investigated. Introduction of a central leucine to proline substitution abolished the (cid:2) -helical structure of the peptide and disrupted apo(a) binding and inhibition of Lp(a) formation. Substitution of hydrophobic residues in the apoB 4372-4392 peptide disrupted apo(a) binding and inhibition of Lp(a) assembly without disrupting the (cid:2) -helical structure. Substitution of all four lysine residues in the peptide with arginine decreased the IC 50 from 40 (cid:3) M to 5 (cid:3) M. Complexing of the arginine-substituted peptide to dimyristoylphosphatidylcholine improved its activity further, yielding an IC 50 of 1 (cid:3) M. We that the (cid:2) -helical structure of apoB 4372-4392 , in combination with hydrophobic residues at the the interaction of apoB 4372-4392 with apo(a) is not lysine speciﬁc, because substitutions with arginine result in a more effective inhibitor. Structural features of apolipoprotein B synthetic peptides that inhibit lipoprotein(a) assembly.

Lipoprotein(a) [Lp(a)] is an atherogenic lipoprotein comprised of apolipoprotein(a) [apo(a)] attached to the apolipoprotein B100 (apoB) of a LDL (1). Large clinical trials have identified high plasma levels of Lp(a) as a risk factor for developing heart disease (2)(3)(4). The deposition of Lp(a) in vascular lesions has been well documented in humans (5,6) and in animal models (7). Unfortunately, Lp(a) has proven resistant to traditional lipid-lowering therapies (8,9), except high doses of niacin (10). Estrogen dramatically reduces Lp(a) levels (11), but a safe and effective agent for lowering Lp(a) levels has yet to be developed.
The majority of evidence suggests that Lp(a) assembly occurs in circulation (12) after secretion of apo(a) from the liver and metabolism of VLDL to LDL in the circulation. Lp(a) assembly is thought to be a two-step process (13,14) with noncovalent interactions between apo(a) and apoB preceding the formation of the disulfide bond between apo(a) Cys4057 (13,15) and apoB Cys4326 (16,17). The two-step model of Lp(a) assembly was confirmed in a study of a human apoB100 mutant lacking Cys4326, which still showed binding to apo(a), despite being unable to form the disulfide bond (18).
Recent research has focused on characterizing the apoB residues involved in the noncovalent interaction with apo(a). Lysine residues on apoB have been implicated in the noncovalent interaction with apo(a), because lysine analogues disrupt Lp(a) assembly in vitro (19)(20)(21) and deletion of lysine-binding sites in apo(a) kringle IV domains interfere with Lp(a) formation (22)(23)(24). A number of apoB sequences that noncovalently bind apo(a) have been reported. Becker et al. (25) identified an apoB lysine residue in the N terminus, apoBLys680, that mediates the noncovalent binding of an apoB18 fragment to apo(a); in their study, a synthetic peptide spanning the Lys680 residue was shown to inhibit Lp(a) formation at similar concentrations to lysine analogues.
Sequences in the carboxyl terminus of apoB have also been implicated in the noncovalent interaction with apo(a).
We recently reported that a synthetic peptide spanning apoB amino acids 4372-4392, bound apo(a) and inhibited Lp(a) assembly in vitro (26). Note that the apoB 4372-4392 peptide was 250-fold more effective than the lysine analogue, -amino-n -caproic acid, at inhibiting Lp(a) assembly (26). This work suggested that the apoB 4372-4392 sequence constitutes an apo(a) binding site that is important for Lp(a) assembly. The peptide sequence is highly conserved between species of apoB known to interact with apo(a) (27) and contains four highly conserved lysine residues surrounded by clusters of hydrophobic residues. Circular dichroism (CD) spectroscopy studies showed that apoB 4372-4392 is ␣ -helical (26). Binding of the peptide to dimyristoylphosphatidylcholine (DMPC) increased its ␣ -helical content and improved its capacity to inhibit Lp(a) assembly (26). To elucidate the key structural features responsible for the inhibitory properties of the apoB 4372-4392 peptide, a series of apoB peptides based on the apoB 4372-4392 sequence were designed and tested for their ability to bind apo(a) and inhibit Lp(a) assembly.

Peptides
Five new synthetic apoB peptides based on the human apo B4372-4392 sequence were synthesized by Chiron Mimotopes using an Applied Biosystems model 430A synthesizer. Peptides were purified by HPLC and stock solutions solubilized in Milli-Q water at a concentration of 1 mM, except for apoB 4372-92ARG , which was solubilized at 100 M. Peptides were dissolved in 10 mM Tris, pH 7.4, for CD spectroscopy studies and Milli-Q water for inhibitory studies. Table 1 gives the amino acid sequences and associated characteristics of the new series of peptides. The importance of the ␣ -helical structure for activity was tested with the apoB L4383P peptide, which contains a central leucine to proline substitution that was expected to perturb the ␣ -helical structure. The importance of the N-and C-terminal lysine residues was tested with the apoB K4372A and apoB K4392A peptides, which contain lysine to alanine substitutions at Lys4372 and Lys4392, respectively. The importance of a cluster of four conserved hydrophobic residues surrounding the Lys4392 residue was tested in the apoB 4372-92HYDRO peptide by substitution of these residues with alanines. To test whether the inhibitory capacity of the apoB 4372-4392 peptide is specific to lysines in the sequence, apoB 4372-92ARG was created, in which all four lysine residues were substituted with arginine. A murine sequence spanning apoB residues 4372-4392 (KYYEIEENMVELIK-TLLVSFR) was tested in Lp(a) formation assays. The murine sequence has conservation of two of the four lysine residues and a conservative substitution of the lysine at position 4392 to arginine.

Mice
Human apo(a) transgenic mice (19) were obtained from Dr. Robert Hammer (University of Texas Southwestern Medical Center, Dallas, TX) and human apoB transgenic mice (28) were provided by Dr. Stephen Young (Gladstone Institute of Cardiovascular Disease, San Francisco, CA). Wild-type mouse plasma was obtained from nontransgenic littermates.

Preparation of phospholipid vesicles
Small unilamellar vesicles of DMPC (Sigma-Aldrich, St. Louis, MO) were prepared using a method similar to that described by New (29). The DMPC was suspended in 10 mM Tris-HCl, pH 7.4, and sonicated using a 9.5 mm probe (Soniprep 150, MSE Scientific Instruments, Sussex, England) at 20 MHz for 10 ϫ 60 s with 30 s breaks in between. The temperature was maintained at 30 Њ C, above the phase transition temperature of DMPC ( ‫ف‬ 24 Њ C), using a jacketed vial with a circulating water bath (30). The vesicles were centrifuged at 2,250 ϫ g for 5 min to pellet titanium originating from the sonicator probe. Phospholipid concentrations were determined at 20 Њ C using an enzymatic phospholipid assay kit (Roche, Australia).

CD spectroscopy
The CD spectra of the apoB peptides in the absence and presence of DMPC were recorded at 25 Њ C on an Aviv 62DS CD spectrophotometer using a 1 mm pathlength quartz cuvette. Spectra were collected between the wavelengths of 195 nm and 250 nm, with a step size of 0.5 nm and a slit bandwidth of 1.5 nm. Signal averaging time was 1 s, and ellipticities were reported as mean residue ellipticity ([ ]) in Њ cm 2 dmol Ϫ 1 . The concentration of each peptide used in the CD studies (150 g/ml) was determined by absorption spectroscopy on a Cary 5 UV/Vis spectrophotometer at 280 nm. The peptides have a molar absorption or extinction coefficient of 2,980 M Ϫ 1 cm Ϫ 1 at 280 nm, except for apoB 4372-92HYDRO, for which is 1,490 M Ϫ 1 cm Ϫ 1 at 280 nm (31). Freshly prepared DMPC vesicles (3 mM) were added to aliquots of the peptides at a molar ratio of 50:1 DMPC-peptide to form discoidal complexes, which were characterized as previously described (26). The ␣ -helical content of each apoB peptide in the absence and presence of saturating amounts of DMPC was calculated from the molar ellipticity at 222 nm according to equation 1 by Aggerbeck et al. (32).

Apo(a) binding studies
Each peptide was evaluated for its ability to bind apo(a) using a solid-phase enzyme-linked ligand sorbent assay (26). Briefly, 96well polystyrene microtiter plates (Costar ® , Corning Inc., Corning, NY) were coated with 5 g per well of each apoB peptide. Apo(a) transgenic mouse plasma, and wild-type mouse plasma (negative control) were diluted 1 in 10 and 50 l aliquots and incubated on the peptide-coated plates for 90 min at room temperature to allow binding of apo(a) to the peptides. Bound apo(a) was detected with the HRP-labeled MAb-a5 antibody (33), washed, and then developed with substrate solution [ O -phenylenediamine dihydrochloride and hydrogen peroxide (Sigma-Aldrich). The level of background binding was determined from the wild-type plasma sample and this was subtracted from the apo(a) binding curves.

In vitro Lp(a) formation assays
The peptides were tested for their ability to act as inhibitors of Lp(a) assembly as described by Sharp et al. (26). Plasma (1 l) from an apo(a) transgenic mouse was incubated with plasma (2 l) from a human apoB transgenic mouse for 3 h at 37 Њ C in the presence of increasing amounts (1-400 M) of each apoB peptide. The amount of Lp(a) formed in each incubation was assessed after aliquots were size-fractionated on SDS/4% polyacrylamide gels under nonreducing conditions and the separated proteins transferred to nitrocellulose for Western blotting with the HRPlabeled MAb-a5 antibody. Bands were visualized using the ECL detection system (Amersham Pharmacia Biotech, Uppsala, Sweden). The amounts of Lp(a) formed in the incubations were quantified by an Lp(a)-specific ELISA, as detailed by Sharp et al. (26). Background absorbance was determined from incubations containing human apoB only and this was subtracted from all data points. The percent Lp(a) formed in each incubation was calculated relative to the incubation containing no inhibitor.

RESULTS
The WHEEL program (34) was used to produce a helical wheel diagram of the apoB 4372-4392 peptide ( Fig. 1 ). Amino acids targeted for substitution in the new series of peptides are highlighted in Fig. 1. The secondary structures of the five human apoB peptides were determined by CD spectroscopy both in the absence and in the presence of DMPC and each was compared with the control apoB 4372-4392 peptide ( Fig. 2 ). Using equation 1, the percent ␣ -helical structure for each peptide was calculated from the ellipticity value at 222 nm and the data are reported in Table 1. The spectra for the control apoB 4372-4392 peptide ( Fig. 2A) showed the characteristic double minima at 208 and 222 nm indicative of ␣ -helical structure. The ␣ -helicity of apoB 4372-4392 was further induced by the binding of the peptide to DMPC, as previously reported (26). As predicted, the proline-substituted peptide, apoB L4383P , had lost all secondary structure and adopted largely random structure, as indicated by a single minimum at 205 nm (Fig. 2B). Furthermore, the addition of DMPC had little effect on the structure of this peptide (Fig. 2B). Substitution of the N-terminal lysine residue to alanine resulted in a 19% decrease in ␣ helical content in the apoB K4372A peptide compared with apoB 4372-4392 (Fig. 2C). The addition of DMPC increased the helical content of the apoB K4372A peptide by 14%; however, the DMPC-complexed peptide was 25% less helical than apoB 4372-4392 in the presence of the same concentration of phospholipid. Substitution of the C-terminal lysine residue for alanine did not affect ␣-helical content; apoB K4392A showed a secondary structure similar to apoB 4372-4392 both in the absence and in the presence of DMPC (Fig. 2D). The hydrophobic replacement peptide, apoB 4372-92HYDRO , was slightly more helical than the original peptide, both in the absence (72% helical) and in the presence (94% helical) of DMPC (Fig. 2E). The apoB 4372-92ARG peptide showed ␣-helical structure (Fig.  2F), with the binding of DMPC significantly increasing its ␣-helical content from 57% to 100%.
The ability of each new human apoB peptide to bind apo(a) was assessed using a solid-phase binding assay (Fig. 3). The original apoB 4372-4392 peptide had previously been shown to bind apo(a) in a saturable manner (26). Both the apoB K4372A and apoB K4392A peptides showed minimal binding to apo(a) (Fig. 3). The apoB L4383P and apoB 4372-92HYDRO peptides did not demonstrate any apo(a) binding beyond background levels. The apoB 4372-92ARG peptide was the only peptide shown to effectively bind apo(a) from transgenic mouse plasma. ApoB 4372-92ARG bound apo(a) to a saturable level in a manner similar to that seen for the original apoB 4372-4392 peptide (26). Fig. 1. Helical wheel diagram of the apolipoprotein B100 apoB 4372-4392 peptide. A helical wheel diagram of the predicted amphipathic helix formed by apoB residues 4372-4392 was produced using the WHEEL program (33). Within the helix, the side chains of the four lysine residues are found at the lipid/aqueous interface projecting into the aqueous phase; the side chains of the hydrophobic residues (in boldface) are predicted to be buried in the lipid phase. The hydrophobic residues (shaded in gray) were substituted with alanine in the apoB 4372-92HYDRO peptide. The Leu4383 residue (marked with an asterisk) was targeted for substitution with proline in apoB L4383P . The four lysine residues on either side of the lipid/ water interface were targeted for substitution with arginines in apoB 4372-92ARG . The Lys4372 was targeted for substitution with alanine in apoB K4372A and Lys4392 was targeted for substitution in apoB K4392A .

Downloaded from
The capacity of each apoB peptide in Table 1 to inhibit Lp(a) assembly in vitro was investigated. Increasing amounts of each peptide were incubated with fixed amounts of apo(a) and human apoB in an Lp(a) formation assay. The amount of Lp(a) formed was visualized by Western blot analysis (Fig. 4A-F). The inhibitory effect of the apoB 4372-4392 control peptide is shown in Fig. 4A. Compared with the control (Fig. 4A), the apoB L4383P peptide (Fig. 4B) had little effect on Lp(a) assembly in vitro, suggesting that the loss in ␣-helical structure had significantly reduced its ability to inhibit Lp(a) assembly. The apoB K4372A peptide (Fig. 4C) showed only a slight disruption to Lp(a) assembly, and apoB K4392A (Fig. 4D) showed very little effect. The apoB 4372-92HYDRO peptide also showed no effect on Lp(a) assembly, despite having retained its ␣-helical structure (Fig. 4E). The apoB 4372-92ARG peptide proved to be a highly effective inhibitor of Lp(a) assembly. In the incubations containing apoB 4372-92ARG , Lp(a) levels decreased with increasing amounts of the peptide, and a corresponding increase in free apo(a) was seen (Fig. 4F). Indeed, the apoB 4372-92ARG peptide was more effective than the original apoB 4372-4392 peptide as an inhibitor, showing complete inhibition of Lp(a) assembly at a concentration of 25 M (compared with 100 M for apoB 4372-4392 ).
In addition to Western blot analyses, the amount of Lp(a) formed in each incubation was quantified by an Lp(a) ELISA and the results were used to determine IC 50 values for each of the peptides (Fig. 5). The original apoB 4372-4392 peptide showed an IC 50 value of 40 M, as previously reported (26). The apoB L4383P peptide had little inhibitory capacity, with an IC 50 value Ͼ 400 M. The hydrophobic replacement peptide, apoB 4372-92HYDRO , also showed very little inhibitory capacity, with an IC 50 Ͼ 400 M, whereas both apoB K4372A and apoB K4392A showed similar IC 50 values of approximately 100 M. By contrast, the argininesubstituted peptide, apoB 4372-92ARG , was the most potent inhibitor of Lp(a) formation, with an IC 50 of 5 M, 8-fold more effective than the apoB 4372-4392 peptide. The inhibi-   proved its inhibitory capacity by 5-fold, shifting the IC 50 value from 5 M to 1 M (Fig. 6). This is supported by Western blot analyses (Fig. 6, inset) showing that the DMPCbound apoB 4372-92ARG peptide caused complete inhibition of Lp(a) assembly at a concentration of 5 M. Complexing of the less effective apoB K4372A and apoB K4392A peptides with DMPC showed a similar decrease in IC 50 to that seen with the apoB 4372-4392 and apoB 4372-92ARG peptides (data not shown), suggesting that complexing to DMPC has a general stabilizing effect on peptide structure that results in a more effective inhibitor.

DISCUSSION
Although elevated levels of Lp(a) have been identified as an independent risk factor for developing vascular disease, there is currently no safe and effective Lp(a)-lowering therapy available. An understanding of the molecular interactions between apo(a) and apoB that initiate Lp(a) formation may lead to the design of novel Lp(a)-lowering agents. Previous research has established that apo(a) lysine-binding domains most likely interact with lysine residues on apoB to facilitate Lp(a) formation. A recent study indicates that the lysine-binding domains in apo(a) kringle IV 7 and IV 8 are the two most likely candidates to be involved in Lp(a) assembly (35). The apo(a) lysine-binding domains consist of an anionic center followed by a hydrophobic trough involving four hydrophobic residues. Lysine analogues interact with the pocket via the amino group of their side chain, making an ion-pair interaction with the anionic center; the aliphatic backbone interacts with the hydrophobic trough (36,37).
We are interested in defining the apoB residues involved in the noncovalent interaction with apo(a). The current study aimed to identify the key structural features of a synthetic apoB peptide spanning amino acids 4372-4392, which was recently reported to bind apo(a) and effectively inhibit Lp(a) assembly in vitro (26). The apoB 4372-4392 sequence is predicted to lie within a class A amphipathic helix with lipid-binding capacity (38). Indeed, CD spectroscopy studies confirmed that the apoB 4372-4392 peptide is ␣-helical in nature and binds phospholipid in a saturable manner (26). The importance of the ␣-helical structure for the inhibitory action of apoB 4372-4392 was tested in this study by replacing the central leucine at position 4383 with proline to create the apoB L4372P peptide. CD spectroscopy of apoB L4383P showed a complete loss of ␣-helix  with the L4383P substitution, which converted the peptide into a random coil structure (Fig. 2B). As a result, apo(a) binding studies showed that the peptide had lost the capacity to bind apo(a) (Fig. 3). Not surprisingly, the apoB L4383P had also lost the ability to inhibit Lp(a) assembly showing an IC 50 Ͼ 400 M. These data support the notion that the helical nature of the peptide is vital for its inhibitory capacity.
We also tested the importance of individual lysine residues in the apoB 4372-4392 sequence on Lp(a) assembly inhibition. We speculated that a lysine residue in close proximity to hydrophobic residues at the lipid/water interface could be important (Fig. 1). The Lys4372 and Lys4392 residues are highly conserved in species capable of binding apo(a) (27) and both lie close to hydrophobic residues at the lipid/aqueous interface. Relative to the control, CD spectroscopy studies showed a substantial reduction in the propensity for apoB K4372A to adopt an ␣-helical structure both in the absence and in the presence of DMPC (Fig. 2). In addition, only minimal binding of the apoB K4372A peptide to apo(a) was detected (Fig. 3); this was accompanied by a significant loss in the inhibition of Lp(a) assembly (Fig. 4). The apoB K4372A peptide had an IC 50 of 100 M (Fig.  5), which is 2.5-fold higher than the original apoB 4372-4392 peptide. As apoB K4372A had lost some ␣-helical structure (Fig. 2C), it is possible that this may account for its loss of activity. Substitution of the Lys4392 residue had no effect on secondary structure, but resulted in a similar loss of binding and inhibitory capacity to the apoB K4372A peptide, with an IC 50 of 100 M. As the secondary structure of apoB K4392A was unaffected, it is most likely the loss of positive charge at this position accounted for its reduced activity. We therefore propose that Lys4392 is a likely candidate for binding to the anionic lysine binding pocket of apo(a). Because the apoB K4392A peptide still shows inhibitory activity, however, other residues besides Lys4392 must be important in binding apo(a).
The importance of a cluster of hydrophobic residues adjacent to the Lys4392 residue was tested by creating a peptide, apoB 4372-92HYDRO, in which four hydrophobic residues located on one face of the helix were replaced with alanines ( Fig. 1; Table 1). Note that the apoB 4372-92HYDRO peptide showed a complete loss of apo(a) binding and inhibitory capacity, with an IC 50 value Ͼ 400 M. Because the secondary structure of this peptide has remained ␣-helical (Fig. 2E), the loss in activity is likely due to the loss of hydrophobic residues, indicating that both the primary and secondary structure of the apoB 4372-4392 peptide are important for its inhibitory activity. We conclude that the substituted hydrophobic residues form part of the apo(a) binding motif in the apoB 4372-4392 peptide. One or more of these hydrophobic residues presumably help to anchor the peptide into the binding pocket along with the Lys4392 residue.
Finally, to test whether the interaction of the apoB 4372-4392 peptide with apo(a) was specific to lysine residues, we replaced all four lysine residues in the apoB 4372-4392 peptide with arginines to create apoB 4372-92ARG . CD spectroscopy studies of apoB 4372-92ARG showed the peptide to be ␣-heli-cal, with a significant increase in helical content in the presence of phospholipid (Fig. 2F). Apo(a) binding studies showed that apoB 4372-92ARG specifically bound to apo(a) (Fig. 3). The ELISA analysis determined an IC 50 value for apoB 4372-92ARG of 5 M (Fig. 5) which is 8-fold more effective than the original apoB 4372-4392 peptide at inhibiting Lp(a) assembly. Furthermore, addition of the apoB 4372-92ARG peptide to DMPC increased the inhibitory capacity 5-fold (IC 50 ϭ 1 M), resulting in a highly effective inhibitor. The results with the apoB 4372-92ARG peptide suggest that the apo(a) lysine binding domains may have a stronger affinity for arginines than lysines. This is consistent with data showing that the addition of arginine can inhibit recombinant apo(a) derivatives from binding native LDL particles in vitro (39). We therefore propose that the guanidinium side chain of arginine may fit more tightly into the anionic binding pocket of apo(a). Note that the murine sequence contains an arginine at position 4392. A peptide spanning mouse apoB 4372-4392 was tested and found to inhibit Lp(a) formation, with an IC 50 similar to the human apoB 4372-4392 peptide (Fig. 7). Future studies investigating the importance of individual lysine to arginine substitutions in the context of the human apoB 4372-4392 sequence may help pinpoint the key lysine residue important for the interaction with the apo(a) lysine-binding pocket.
In summary, this study has determined several structural features that are responsible for the inhibitory action of the apoB 4372-4392 peptide on Lp(a) assembly. First, we show that the ␣-helical nature of this sequence is critical for inhibition, as demonstrated with the L4383P peptide, which shows a loss of helical structure coinciding with a loss in activity. This is also supported by data showing that the addition of phospholipid, which stabilizes the helical structure of the apoB 4372-4392 peptide, enhances the inhibition of Lp(a) assembly (26). Second, we show using the apoB K4392A and apoB 4372-92HYDRO peptides that a positive charge at position 4392, along with hydrophobic residues at the interface of the amphipathic helix, is important for the interaction with apo(a). Finally, we show that substitution of lysines for arginines in the sequence increases the inhibitory capacity of the peptide, suggesting a tighter fit of arginine in the anionic apo(a) binding cavity. The apoB 4372-4392 peptide has proved to be an excellent model for studying the nature of the noncovalent interactions between apoB and apo(a) in Lp(a) assembly and thus provides scope for the design of novel inhibitors of Lp(a) assembly.