Anti-inflammatory apoA-I-mimetic peptides bind oxidized lipids with much higher affinity than human apoA-I^*

MATERIALS AND METHODS

RESULTS

We previously reported that adding 50 μg/ml of human apoA-I to human artery wall cells in culture did not prevent LDL oxidation and induction of MCP-1 as measured by a bioassay if the apoA-I was added at the same time as LDL (25). As shown in Fig. 1, adding 1 mg/ml (∼35 μM) of human apoA-I to HAECs did not decrease LDL-induced monocyte chemotactic activity and actually slightly increased it. In the experiments shown in Fig. 1, L-4F was added at concentrations ranging from 0.43 nM (0.001 μg/ml) to 0.43 μM (1 μg/ml) in the presence of 35 μM (1 mg/ml) of human apoA-I. At a concentration of 4.3 nM (0.01 μg/ml), L-4F significantly (P < 0.001) decreased the LDL-induced monocyte chemotactic activity, and 0.43 μM of L-4F reduced the levels to those achieved with normal human HDL (Fig. 1). The effect of L-4F on the lipoprotein inflammatory index in the absence of apoA-I was not significantly different at any concentration tested, compared with L-4F in the presence of apoA-I (data not shown). Thus, the effect of L-4F did not depend on the presence of full-length apoA-I. These results confirm and extend our previous results (25) and directly prove that 4F is effective at nanomolar concentrations even in the presence of a very large excess of human apoA-I.

Figure 2 shows the effects of oxidizing PAPC on its binding to apoA-I and the 4F peptides. The total duration of oxidation was 48 h, and during that time, PAPC was converted to multiple different components, collectively designated “oxPAPC.” At the conclusion of oxidation, no intact PAPC was observed by mass spectrometry (data not shown). Samples of oxPAPC were removed and tested at intervals, and the “apparent K” (K_Dapp) values were calculated. The K_Dapp of oxidized PAPC for binding apoA-I changed very little, whereas for D-4F and L-4F, it decreased by several orders of magnitude (Fig. 2A). Even at the zero-hour time point, the binding isotherms of apoA-I and the 4F peptides differed substantially when concentrations well below the K_Dapp were tested (Fig. 2B). Similar behavior was also seen when subsequent oxPAPC samples were tested (Fig. 2C, D).

When comparing K_D data with a binding isotherm, it must be recalled that the K_D value is derived from the equation: K_D = k_d/k_a. In this equation, k_d represents the dissociation rate constant and has units of 1/sec, and k_a represents the association rate constant and has units of 1/Msec. The high affinity (low K_D value) of the oxPAPC preparations derives principally from their much higher association rates, which are reflected in the isotherms by the much steeper upward slope of the binding curves for the 4F peptides. In Fig. 2, the k_a values for oxPAPC were 9.92 × 10⁶/Msec, 9.04 × 10⁶/Msec, and 6.41 × 10³/Msec for L-4F, D-4F, and apoA-I, respectively, whereas for PAPC, the k_a values were 1.22 × 10³/Msec, 1.12 × 10³/Msec, and 319/Msec for L-4F, D-4F, and apoA-I, respectively.

We used homogeneous lipid preparations to further evaluate the binding of oxidized lipids by the 4F peptides. Table 1 shows the affinity of four pro-inflammatory oxidized phospholipids for D-4F, L-4F, and apoA-I. On average, the 4F peptides had K_D values that were about 3,000-fold lower than those of apoA-I. For PEIPC, one of the most bioactive pro-inflammatory phospholipids (26), D-4F had a K_D of 0.064 nM, L-4F had a K_D of 0.01 nM, and apoA-I had a K_D of 50 μM, a difference of about six orders of magnitude.

Several nonoxidized FFAs that varied in chain length and saturation were bound with equal affinity by apoA-I, D-4F, and L-4F (Table 2). However, oxidized species such as HPETE (hydroperoxyeicosatetraenoic acid), HPODE (hydroperoxyoctadecadienoic acid), HETE (hydroxyeicosatetraenoic acid), and HODE (hydroxyoctadecadienoic acid) were bound with K_D values that were often five orders of magnitude lower for 4F peptides than for apoA-I (Table 2). This difference resulted because the K_D values for binding to the peptides were similar for the oxidized FAs and their nonoxidized precursors. In contrast, these oxidized FAs bound apoA-I with much lower affinity than their nonoxidized precursors. Figure 3 shows binding curves for eight concentrations of one of these oxidized lipids, 13(S)-HPODE: 1.5, 2.0, 3.0, and 5.0 μg/ml (∼4.8, 6.4, 9.7, and 16.1 μM) and 30, 40, 50 and 90 ng/ml (∼97, 129, 161, and 290 nM). At the higher concentrations, L-4F had a K_D of 16.8 ± 4.3 nM. The corresponding K_D for apoA-I was 1.23 ± 0.57 mM. Native and oxidized cholesterol also bound to the mimetic peptides with substantially higher affinity than to apoA-I (Table 3). Neither cholesterol nor its oxidized analogs bound sufficiently to apoA-I for us to estimate a K_D.

In the experiments described above, the lipids were not associated with lipoproteins. To determine whether lipids associated with lipoproteins would behave similarly, the binding of LDL that had been incubated overnight with HAECs was compared with the binding of LDL that was not incubated with HAECs. We have previously reported that incubating LDL with HAECs results in the transfer of hydroperoxy FAs such as HPODE from the endothelial cells to LDL (25). The binding affinity (K_D) of LDL that had been incubated with HAECs was not different from the binding affinity of native LDL for human apoA-I (K_D = 167 ± 27 nM vs. 137 ± 16 nM). In contrast, the binding affinity of LDL that had been incubated with HAECs was increased by 6.1-fold, compared with the binding affinity of native LDL for L-4F (K_D = 141 ± 32 nM vs. 23 ± 9 nM; P < 0.001). Adding 13(S)-HPODE to native LDL (500 ng 13(S)-HPODE incubated with 50 μg LDL/ml for 2 h at 37°C) resulted in a dramatic increase in the binding affinity of LDL for L-4F (K_D = 0.42 nM; P < 0.0079) but did not result in a significant change in the binding affinity of LDL for apoA-I.

In the examples shown up to this point, apoA-I and the N-terminally blocked apoA-I-mimetic 4F peptides were immobilized to the sensor chip by an amine-coupling procedure that converted one or more of their lysine-ɛ-amino group(s) into one or more amides. To assure ourselves that this immobilization procedure had not somehow enhanced the ability of the 4F peptides to bind oxidized lipids, binding studies were also performed by attaching N-biotinylated L-4F to a streptavidin-coated CM5 chip. Using results obtained with a series of phospholipid concentrations, we calculated a K_Dapp value of 3.1 × 10⁻⁸ M for oxPAPC and a K_D of 2.2 × 10⁻⁴ M for native PAPC. Figure 4A shows studies done with two concentrations of native PAPC and oxPAPC: 1 μM and 1 pM. Immobilized biotinylated L-4F bound both at the 1 μM concentration, but bound only oxPAPC at the 1 pM level. Figure 4B shows that immobilized biotinylated L-4F could even bind subpicomolar (attomolar) concentrations of KOdiA-PC. In other studies, it was determined that the biotinylation of L-4F did not alter its biologic activity (data not shown).

As a further control, we reversed the binding conditions by immobilizing the lipid on the sensor chip and introducing the 4F peptides in solution. To do so, we biotinylated 1-palmitoyl-2-arachidonoyl-sn-glycero-3-phosphatidylethanolamine (PAPE) to prepare PAPE-N-Biotin. Whereas one aliquot of this reagent was stored under argon at −80°C, another was allowed to oxidize for 48–72 h (as monitored by electrospray ionization-mass spectrometry) to prepare oxPAPE-N-Biotin as described in Materials and Methods, and then both were coupled to streptavidin-coated sensor chips. Where as little as 60 femtograms of L-4F/ml (∼26 fM) gave a good binding signal with immobilized N-biotin-oxPAPE (Fig. 5), even 100 μg/ml of L-4F (43.3 μM) failed to bind immobilized PAPE-N-Biotin that had not been oxidized. The apparent affinity of soluble L-4F for immobilized oxPAPE-N-Biotin (K_Dapp = 1.36 × 10⁻⁹ M) was similar, and perhaps greater than its affinity when the L-4F was biotinylated and immobilized to the biosensor and the unmodified lipids were in solution.

Whereas apoA-I contains 243 amino acid residues, the apoA-I-mimetic peptides contain only 18 amino acids. Accordingly, we tested two additional 18-residue apoA-I-mimetic peptide analogs, 3F-2 and 3F¹⁴. Although both have class A, amphipathic helical structures and they bind nonoxidized lipids similarly, 3F-2 inhibited atherosclerotic lesion formation in a mouse model and 3F¹⁴ did not (30). Figure 6A shows that immobilized 3F-2 and L-4F bound oxPAPC identically; however its binding by 3F¹⁴ was significantly less, resembling its binding by apoA-I (Fig. 2A). Moreover, when 3F-2 and 3F¹⁴ were compared for their ability to bind specific oxPAPC components (POVPC, PGPC, and PEIPC) as well as the oxidized lipid 15(S)-HPETE, 3F-2 was more effective (Fig. 6B). Thus, the anti-inflammatory peptides 3F-2 and 4F bind pro-inflammatory oxidized lipids and phospholipids with an affinity that is much higher than that shown by apoA-I or by 3F¹⁴.

DISCUSSION

We previously reported that human apoA-I was significantly more potent than mouse apoA-I in rendering HDL anti-inflammatory in a mouse model of atherosclerosis (35). In other studies, we reported that the apoA-I-mimetic peptide D-4F was much more anti-inflammatory than human apoA-I in LDL receptor-null mice with influenza A viral pneumonia (3). The data reported here indicate that the binding affinities for nonoxidized phospholipids are similar for human apoA-I and the mimetic peptides D-4F, L-4F, 3F-2, and 3F¹⁴. Oxidation of PAPC only minimally changed the binding affinity for apoA-I and 3F¹⁴ (a biologically inactive peptide). In contrast, the binding affinities of oxPAPC and defined oxidized phospholipids for D-4F, L-4F, and 3F-2 (peptides that inhibit lesions in mouse models of atherosclerosis), compared with the nonoxidized phospholipids, were dramatically increased. Interestingly, the binding of oxidized and nonoxidized FFAs to the 4F peptides was of equally high affinity, with K_D values ranging between 10⁻⁵ and 10⁻¹⁰ M (Table 2). Thus, the binding affinity of the phospholipids for the 4F peptides only approached that of the FAs after the phospholipids were oxidized.

As noted above, in the case of apoA-I, the binding of phospholipids was similar whether oxidized or not oxidized. However, the binding of the FFAs (arachidonic and linoleic acids) to apoA-I decreased dramatically when they were oxidized (Table 2).

Epand et al. (9) studied the interaction of L-4F with a synthetic membrane containing phospholipid with and without cholesterol. They found that the peptide penetrated more deeply into the membrane in the absence of cholesterol and in the presence of cholesterol, resulted in the separation of cholesterol from the phospholipid. These results might suggest a higher binding affinity of L-4F for the phospholipid, compared with the cholesterol. As shown in Fig. 2, before oxidation, the apparent K_D (K_Dapp) value for the binding of the phospholipid to the 4F peptides was ∼10⁻⁴ M. As shown in Table 3 the K_D value for the binding of cholesterol to the 4F peptides was ∼10⁻⁷ M. These differences may be due to several factors: i) differences in the phospholipids studied (SOPC in the case of Epand et al. and PAPC here); ii) the mechanism promoting the separation of cholesterol from the membrane phospholipid may not be related to differences in the binding affinities; or iii) the differences in studying these interactions in a membrane versus studying the interactions in a nonmembrane environment.

Both oxidized phospholipids and oxidized FAs bound with much higher affinity to the peptides than to apoA-I (Tables 1 and 2). Interestingly, the binding affinity of oxidized phospholipids for the peptides was even higher than the binding affinity of the oxidized FAs for the peptides (Tables 1 and 2). Differences in relative polarity or differences in tertiary structure between the two classes of oxidized lipids may have contributed to these differences in binding affinity for the peptides.

The results reported here cannot be attributable to the amine-coupling procedure used to attach the peptides to the biosensor chip surface. The remarkably high affinity of the 4F mimetic peptides for oxidized phospholipids was seen both when L-4F was immobilized to the sensor chip by a biotin-streptavidin linkage and the phospholipids were in solution (Fig. 4) and when biotinylated phospholipids were immobilized on the sensor chip and the 4F peptides remained in solution (Fig. 5).

The activity of class A amphipathic helical peptides in biological systems was suggested to depend on hydrophobic face-lipid acyl chain interactions (10, 29, 36). Based on the imputed molecular shapes of 3F-2 and 3F¹⁴, it was predicted that the entry of water and oxidized lipids into the surrounding hydrophobic milieu of the peptide would be favored for the 3F-2 molecule but not for 3F¹⁴ (29). Similar considerations may explain why the apoA-I-mimetic peptide 4F inhibits lesion formation in mouse models of atherosclerosis, whereas 2F (another class A amphipathic helical peptide) does not (36). Consistent with the present data, the mimetic peptides 3F-2 and 4F removed lipid hydroperoxides from the LDL of Watanabe rabbits, whereas 3F¹⁴ did not (10). 3F-2 and 4F also prevented oxPAPC from inducing MCP-1 activity in cocultured human aortic endothelial and smooth muscle cells, whereas 3F¹⁴ did not (10).

The enhanced binding of the 4F peptides has been attributed to their ability to interact with lipid membranes, creating microdomains for the sequestration of polar lipids derived from an aqueous environment (36). In the experiments utilizing sensor chips, this mechanism would not be operative. However, it is possible and, we think, likely that microdomains in the water layer surrounding the immobilized peptides provide a similar thermodynamically favored environment for oxidized lipids. We believe that the structural basis of the binding of oxidized lipids to the peptides relates to the fact that the oxidized lipids are more polar than the nonoxidized lipids from which they are derived and the peptides provide a microenvironment that thermodynamically is more favorable than either the pure lipid or the aqueous environment in which they were formed.

Recently Wool, Reardon, and Getz (37) reported in vitro studies showing that dimers of the 4F peptide linked by proline or alanine or the sequence KVEPLRA produced more remodeling of HDL and more cholesterol efflux compared with 4F alone. However, 4F alone was more effective in preventing LDL oxidation than was the case for the dimers. The physiologic relevance of the differences in the affinities of apoA-I and apoA-I-mimetic peptides will be determined in vivo by the amounts or fluxes of the oxidized lipids and the amounts or fluxes of peptide versus apoA-I in the plasma at a steady state. The concentration of cholesterol in the plasma of apoE-null mice is measured in the range of mg/ml. The concentration of oxidized phospholipids in the plasma of apoE-null mice is measured in the range of μg/ml (38). The four to six orders of magnitude difference in the affinity of the apoA-I-mimetic peptides for binding these oxidized phospholipids compared with apoA-I (Table 1) appears sufficient to favor the peptides over apoA-I, as shown by the experiment in presented in Fig. 1.

Recently, we demonstrated that feeding a Western diet to LDL receptor-null mice caused an accumulation of oxidized phospholipids in the kidneys associated with renal inflammation. Consistent with in vitro studies indicating that oxPAPC induces cytokine production through a sterol-regulatory element binding protein (SREBP)-mediated mechanism (39, 40), renal SREBP-1c mRNA levels increased dramatically in mice fed the Western diet, compared with the same mice on a chow diet (22). The increased SREBP-1c mRNA levels were associated with a significant increase in renal triglyceride levels. Adding D-4F to the drinking water of the LDL receptor-null mice on the Western diet did not alter plasma triglycerides, cholesterol, or LDL + VLDL cholesterol levels, nor did it significantly change renal cholesterol levels. However, the addition of D-4F to the drinking water reduced the levels of oxidized phospholipids in the kidneys to those seen in mice on chow and similarly reduced SREBP-1c mRNA levels and renal triglyceride levels without altering the levels of apoB retained in the kidneys. Thus, the action of the apoA-I-mimetic peptide appeared to be specifically related to the levels and known mechanisms of action of oxPAPC in inducing inflammatory cytokines (22).

Because the maximum plasma levels of D-4F achieved in mice after oral administration are only about ∼130 nM (7), and plasma levels of apoA-I in mice are ∼35 μM, it has been difficult to understand how 4F might achieve its beneficial effects in vivo. The dramatic difference in binding affinities for oxidized phospholipids between the mimetic peptides and apoA-I could explain how these apoA-I-mimetic peptides exert their potent biological activities, even when surrounded by a “sea” of apoA-I. The results of the experiments shown in Fig. 1, where concentrations of full-length human apoA-I were similar to those in plasma and did not inhibit LDL-induced monocyte chemotactic activity, but the addition of 4.3 nM of L-4F significantly reduced this activity and 0.43 μM was as effective as normal human HDL, are consistent with this explanation.

We previously reported that the peptide 3F¹⁴ was unable to inhibit LDL-induced monocyte chemotactic activity, in contrast to the potent inhibition seen with L-4F (see Fig. 5 in Ref. 1). The binding data reported here (Fig. 6), coupled with the data previously reported on 3F¹⁴ (see Fig. 5 in Ref. 1; see Ref. 30 for studies in a mouse model of atherosclerosis), suggest that a strategy seeking to design anti-inflammatory apoA-I-mimetic peptides should not focus on the binding properties of the peptides for nonoxidized lipids, but rather should focus on developing peptides that bind oxidized lipids with very high affinity.