Effect of Leucine to Phenylalanine Substitution on the Nonpolar Face of a Class A Amphipathic Helical Peptide on Its Interaction with Lipid

Model class A amphipathic helical peptides mimic several properties of apolipoprotein A-I (apoA-I), the major protein component of high density lipoproteins. Previously, we reported the NMR structures of Ac-18A-NH2 (renamed as 2F because of two phenylalanines), the base-line model class A amphipathic helical peptide in the presence of lipid ( Mishra, V. K., Anantharamaiah, G. M., Segrest, J. P., Palgunachari, M. N., Chaddha, M., Simon Sham, S. W., and Krishna, N. R. (2006) J Biol. Chem. 281, 6511-6519 ). Substitution of two Leu residues on the nonpolar face (Leu3 and Leu14) with Phe residues produced the peptide 4F (so named because of four phenylalanines), which has been extensively studied for its anti-inflammatory and antiatherogenic properties. Like 2F, 4F also forms discoidal nascent high density lipoprotein-like particles with 1,2-dimyristoyl-sn-glycero-3-phosphocholine (DMPC). Since subtle structural changes in the peptide-lipid complexes have been shown to be responsible for their antiatherogenic properties, we undertook high resolution NMR studies to deduce detailed structure of 4F in 4F·DMPC discs. Like 2F, 4F adopts a well defined amphipathic α-helical structure in association with the lipid at a 1:1 peptide/lipid weight ratio. Nuclear Overhauser effect (NOE) spectroscopy revealed a number of intermolecular close contacts between the aromatic residues in the hydrophobic face of the helix and the lipid acyl chain protons. Similar to 2F, the pattern of observed peptide-lipid NOEs is consistent with a parallel orientation of the amphipathic α helix, with respect to the plane of the lipid bilayer, on the edge of the disc (the belt model). However, in contrast to 2F in 2F·DMPC, 4F in the 4F·DMPC complex is located closer to the lipid headgroup as evidenced by a number of NOEs between 4F and DMPC headgroup protons. These NOEs are absent in the 2F·DMPC complex. In addition, the conformation of the DMPC sn-3 chain in 4F·DMPC complex is different than in the 2F·DMPC complex as evidenced by the NOE between lipid 2.CH and βCH2 protons in 4F·DMPC, but not in 2F·DMPC, complex. Based on the results of this study, we infer that the antiatherogenic properties of 4F may result from its preferential interaction with lipid headgroups.

bic face of the helix and the lipid acyl chain protons. Similar to 2F, the pattern of observed peptide-lipid NOEs is consistent with a parallel orientation of the amphipathic ␣ helix, with respect to the plane of the lipid bilayer, on the edge of the disc (the belt model). However, in contrast to 2F in 2F⅐DMPC, 4F in the 4F⅐DMPC complex is located closer to the lipid headgroup as evidenced by a number of NOEs between 4F and DMPC headgroup protons. These NOEs are absent in the 2F⅐DMPC complex. In addition, the conformation of the DMPC sn-3 chain in 4F⅐DMPC complex is different than in the 2F⅐DMPC complex as evidenced by the NOE between lipid 2.CH and ␤CH 2 protons in 4F⅐DMPC, but not in 2F⅐DMPC, complex. Based on the results of this study, we infer that the antiatherogenic properties of 4F may result from its preferential interaction with lipid headgroups.
Class A amphipathic helical peptides have been shown to mimic apolipoprotein A-I (apoA-I), 3 the major protein accounting for 70% of the total protein present in high density lipoproteins (HDL). Epidemiological studies have established an inverse correlation between the plasma levels of HDL cholesterol and the risk of coronary artery disease. Reconstituted HDL of human apoA-I in combination with phospholipid has been shown to inhibit atherosclerosis in animal models of atherosclerosis (1-6) as well as in humans (7).
ApoA-I mimetic peptides may represent an alternative to apoA-I for large scale production of synthetic HDL as a therapeutic agent. The motif that is responsible for the association of apoA-I with lipid was determined as tandem repeating amphipathic ␣-helical domains that are present throughout the sequence of apoA-I. The observation that the amphipathic ␣-helical domains in exchangeable apolipoproteins possess a class A amphipathic motif with basic amino acid residues at the polar-nonpolar interface and negatively charged residues at the center of the polar face (8 -10) led us to design the first model class A amphipathic helical peptide 18A with an amino acid sequence DWLKAFYDKVAEKLKEAF (11). This sequence does not possess any sequence homology with any of the exchangeable apolipoproteins, but the synthetic peptide 18A was able to form peptide-lipid complexes similar to apoA-I⅐lipid complexes (11,12). Adding an acyl group to the amino terminus of this peptide and an amide to the C-terminal end resulted in Ac-18A-NH 2 (also known as 2F, since the nonpolar face possesses two Phe residues) with increased ␣ helicity and affinity for lipids (13). Recently, we determined the structure of this peptide in 50% (v/v) trifluoroethanol (14) and complexed to lipid (15) using high resolution solution NMR methods. A number of variations of 2F peptide were synthesized by replacing existing nonpolar amino acids with Phe residues on the nonpolar face. Thus, the peptides synthesized were 3F 3 (Ac-F 3 18A-NH 2 ), 3F 14 (Ac-F 14 18A-NH 2 ), 4F(Ac-F 3,14 18A-NH 2 ), 5F(Ac-F 11,14,17 18A-NH 2 ), 6F(Ac-F 10,11,14,17 18A-NH 2 ), and 7F(Ac-F 3,10,11,14,17 18A-NH 2 ) (16). All of the peptides were able to solubilize multilamellar vesicles of egg yolk phosphatidylcholine and thus possessed increased lipid affinity (16). However, peptides 4F, 5F, and 6F were significantly more effective than the homologues 2F and 7F in their ability to inhibit low density lipoprotein-induced monocyte chemotaxis (16). Since these observations, apoA-I mimetic peptide 4F has been studied extensively as an anti-inflammatory agent in inhibiting atherosclerosis in various animal models of atherosclerosis (17)(18)(19)(20), improving vascular function in streptozotokine-induced diabetes (21)(22)(23), influenza A-mediated inflammation (24,25), and lipopolysaccharide-mediated inflammation (26) and even in human patients in the Phase I clinical trial demonstrating an improvement in HDL quality (27). Although the lipid-associated structure of 2F has been determined (15), the lipid-associated structure of 4F is unknown. In this paper, we describe the lipid-associated structure of 4F in an effort to understand unique properties of this peptide that may be responsible for its anti-inflammatory properties. The helical wheel and helical net representations of 4F are shown in Fig. 1. The location of Phe residues in 4F that were substituted for Leu in 2F are indicated by arrows in Fig. 1.

MATERIALS AND METHODS
Peptide Synthesis, Purification, and NMR Sample Preparation-The peptide 4F (Ac-DWFKAFYDKVAEKFKEAF-NH 2 ) was synthesized and purified as described earlier (16). Two separate sets of samples were prepared for the NMR studies: (a) perdeuterated 1,2-dimyristoyl-d 54 -sn-glycero-3-phosphocholine-1,1,2,2-d 4 -N,N,N-trimethyl-d 9 (DMPC) (Avanti Polar Lipids, Inc., Alabaster, AL) and peptide in 90% H 2 O, 10% D 2 O (100% deuterium oxide; Cambridge Isotope Laboratories, Inc.) and (b) protonated DMPC and peptide in 90% H 2 O, 10% D 2 O. To prepare these samples, lyophilized dry peptide was dissolved in methanol. DMPC (either protonated or perdeuterated) was dissolved separately in methanol. Appropriate amounts of peptide and DMPC solutions were mixed together to obtain a peptide/total lipid ratio of 1:1 (w/w). Organic solvent was evaporated under a stream of nitrogen gas, and the mixture was lyophilized to remove any residual solvent. To the dry peptide and lipid film, 0.6 ml of 5 mM KH 2 PO 4 was added, and the sample was incubated at 37°C overnight. The pH of the final solution was 5.5. The final peptide concentration in solution was ϳ3 mM.
Size of the Discoidal Peptide-Lipid Complex-Apparent size of the discoidal peptide-lipid complex was determined using a fast protein liquid chromatography system (BioLogic, Duo-Flow; Bio-Rad) and Superose 6 10/300 GL and Superdex 200 10/300 GL columns (Amersham Biosciences) linked in tandem and run at a flow rate of 0.4 ml/min in phosphate-buffered saline containing 0.02% sodium azide (pH 7.4). Complex elution was monitored using absorbance at 280 nm. The apparent Stokes diameter of the complex was estimated using high and low molecular weight calibration kits (Amersham Biosciences). The void and total volume of the columns were determined using blue dextran and K 3 (Fe(CN) 6 ), respectively. The amounts of lipid and peptide in the eluted peak fraction were measured using an enzymatic colorimetric method (Phospholipids B; Wako Chemicals) and absorbance at 280 nm in the presence of 6 M guanidinium hydrochloride, respectively.
NMR Measurements and Structure Calculations-One-and two-dimensional NMR experiments were performed on a Bruker Avance-600 NMR system with a TCI CryoProbe and an Avance-500 NMR spectrometer at 37°C. One-dimensional 31 P experiments were run at 202.45 MHz on an Avance-500 spectrometer with sodium phosphate buffer as an external reference set at 0.0 ppm. 50-, 100-, 150-, and 200-ms mixing times were used in NOESY, and a 100-ms spin-lock time was used in total correlation spectroscopy (TOCSY) measurements. Twodimensional data sets were collected with 2048 complex t 2 points with 512 t 1 increments. States-time-proportional phase incrementation was employed for frequency discrimination in the indirect dimension. Chemical shifts were referenced with respect to sodium 2,2-dimethyl-2-silapentane-5-sulfonate (0.0 ppm) used as an internal standard. To improve spectrum resolution, additional spectra were obtained at 37°C on a Varian Inova 900 MHz spectrometer equipped with a cryogenic probe. Data were acquired using conventional sequences from the The two Phe residues that were substituted in 4F for Leu residues in 2F are indicated by arrows. The primary amino acid sequences of 2F and 4F are shown above the helical wheel and helical net diagrams. Red, acidic; blue, basic; thick black, hydrophobic; thin black, Ala.
Varian library. In all experiments, the spectral width was 14 kHz, and a 3919 watergate scheme was used to suppress water in all experiments. The two-dimensional TOCSY data set was collected with a 60-ms DIPSI spin-lock of 10 kHz strength and an acquisition time of 400 ms in t2 and 28 ms in t1. The NOESY data set was collected with the same general parameters and a mixing time of 200 ms. These parameters provided a digital resolution in F2 of 2.5 Hz/point prior to zero filling.
The NMR data were transferred to a Silicon Graphics IRIS Indigo work station and processed using the program FELIX (version 2007) (Felix NMR Inc., San Diego, CA). NMR structures were calculated on a Silicon Graphics IRIS Indigo work station using X-PLOR (online) version 3.851. The structure calculation protocol involved (a) generation of a "template" coordinate set; (b) bound smoothing, full structure embedding, and regularization to produce a family of 100 distance geometry structures; (c) simulated annealing regularization and refinement for embedded distance geometry structures; and (d) simulated annealing refinement, as described previously (15). Average coordinates of the accepted structures were energyminimized using 200 cycles of conjugate gradient energy minimization. All of the NMR constraints were enforced during energy minimization.
Distance and Dihedral Angle Constraints-On the basis of cross-peak intensities in the NOESY spectra, recorded with mixing times of 100 and 200 ms, the NOE distance constraints were classified as follows. For NOEs involving intraresidue and sequential NH, ␣H, and ␤H protons, a distance constraint of 2.0 -2.5 Å for strong, 2.0 -3.0 Å for medium, and 2.0 -4.0 Å for weak NOEs was used; for medium or long range NOEs involving NH and ␣H protons, a distance constraint of 2.0 -4.0 Å was used; for NOEs involving medium or long range side chain pro-tons, a distance constraint of 2.0 -5.0 Å was used. Pseudoatom corrections, where appropriate, were added to the upper distance bounds, as described previously (28). Based on the difference (⌬␦) in the observed and "coil" NH proton chemical shifts, hydrogen bond (H-O) distances were calculated as described earlier (29). These hydrogen bond distances were included as additional constraints (30) during structure calculation. Dihedral angles ⌽ and were based on the difference in the observed and "coil" C ␣ H proton chemical shifts (29); a minimum deviation of 40°was allowed in the dihedral constraints.
Molecular modeling was performed on a Silicon Graphics IRIS Indigo work station using the program SYBYL (version 7.2) (Tripos, Inc., St. Louis, MO).

RESULTS
The size exclusion profiles of 2F⅐DMPC (15) and 4F⅐DMPC complexes are compared in Fig.  2. The peptides elute predominantly as a lipid complex at the peptide/lipid ratio used (1:1, w/w) (Fig. 2). The apparent Stokes diameter of the complexes, estimated using high and low molecular weight calibration kits (Amersham Biosciences), are 69 Å (15) and 67 Å for 2F⅐DMPC and 4F⅐DMPC, respectively (Fig. 2). The lipid/peptide ratio in the eluted fractions was determined to be nearly identical.
A combination of TOCSY and NOESY spectrum was used to make sequence-specific resonance assignments of the individual amino acids of 4F in the 4F⅐DMPC complex (31). Fig. 3, A and B, show the amide/aromatic proton region and the fingerprint region, respectively, of the NOESY spectrum of 4F⅐DMPC complex obtained using a mixing time of 200 ms. TOCSY crosspeaks in Fig. 3A are shown in red. Sequential d NN (i, i ϩ 1) NOESY cross-peaks for all of the residues in 4F are identified in Fig. 3A. In addition, d NN (i, i ϩ 2) cross-peaks were observed for Asp 1 /Phe 3 , Phe 3 /Ala 5 , Phe 6 /Asp 8 , and Lys 13 /Lys 15 pairs and are indicated in Fig. 3A. In Fig. 3B, intraresidue HN/C ␣ H NOEs for all of the residues in 4F are labeled. A number of other long range (i, i ϩ 3, and i, i ϩ 4) NOEs involving side chain protons of the nonpolar amino acid residues (Trp 2 /Ala 5 , Trp 2 /Phe 6 , Phe 3 /Phe 6 , Phe 3 /Tyr 7 , Phe 6 /Val 10 , Tyr 7 /Val 10 , Tyr 7 /Ala 11 , Val 10 /Phe 14 , Ala 11 /Phe 14 , Phe 14 /Ala 17 , and Phe 14 /Phe 18 ) were observed (Fig. 3, A and B). The methyl protons of the N terminus acetyl protecting group showed NOEs to Asp 1 , Trp 2 , Phe 3 , and Lys 4 backbone NH protons as well as to the Trp 2 aromatic ring protons (Fig. 3B). In addition, NOEs were observed between the aromatic ring protons of Trp 2 /Phe 3 and Phe 3 /Tyr 7 (Fig. 3A). NOEs between Trp 2 /Phe 6 and Phe 6 /Tyr 7 ring protons, although probably present, could not be assigned unambiguously because of resonance overlap. The observed (i, i ϩ 1), FIGURE 2. Size exclusion chromatography of 2F⅐DMPC (open circles) and 4F⅐DMPC (filled circles) complexes at a 1:1 (w/w) peptide/lipid ratio. Two columns, Superose 6 10/300 GL and Superdex 200 10/300 GL (Amersham Biosciences), were linked in tandem and run at a flow rate of 0.4 ml/min in phosphate-buffered saline containing 0.02% sodium azide (pH 7.4) using a fast protein liquid chromatography system (BioLogic, DuoFlow). Complex elution was monitored using absorbance at 280 nm. The apparent Stokes diameter of the complex was determined using high and low molecular weight calibration kits (Amersham Biosciences). Inset, calibration curve obtained using these standards.
(i, i ϩ 2), (i, i ϩ 3), and (i, i ϩ 4) NOEs are schematically shown in Fig. 4. For the sequential NOEs, thickness of the line is in proportion to their intensity (Fig. 4). Proton chemical shifts are related to the secondary structures of peptides and proteins (29,32). The plots of the difference between the observed chemical shift and the "coil" chemical shift (29) (⌬␦ ϭ ␦ obs Ϫ ␦ c ) for the C ␣ H and the NH protons in 4F are shown in Fig. 5, A and B, respectively. A periodic (3-4 residues) variation is evident in the case of NH proton chemical shifts (⌬␦) (Fig. 5B).
The distance and dihedral constraints were used to calculate NMR structures of the peptide associated with DMPC disc. Starting from a set of 100 initial extended structures of 4F, 70 structures were accepted based on the following criteria: (a) no NOE violations Ͼ0.5 Å, (b) no dihedral angle violations Ͼ5°, (c) root mean square (r.m.s.) difference for bond deviations from ideality Ͻ0.01 Å, and (d) r.m.s. difference for angle deviations from ideality Ͻ2°. To obtain the average NMR structure, average coordinates of the 70 accepted structures were energy-minimized. All of the distance and the dihedral angle constraints were enforced during the energy minimization of the average coordinates. The structural statistics for the 70 accepted NMR structures and the energy-minimized average structure of 4F are summarized in Table 1. The 70 accepted structures were superimposed onto the energy-minimized average structure using MATCH (Sybyl, version 7.2). The program MATCH performs an automatic least squares fit of two molecules that differ only in the coordinates of their atoms. Fig.  6 shows superposition of the C␣ atoms of all the 70 accepted structures on the energy-minimized average NMR structure of 4F. The energy-minimized average structure is shown as a ribbon/tube (Fig. 6). The side chains of energy-minimized average structure are also shown (Fig. 6). Fig. 7 shows a DMPC molecule with different protons labeled. To observe intermolecular NOEs between peptide and lipid protons, NOESY spectra were obtained with the peptide and protonated lipid and peptide and perdeuterated lipid in 90% H 2 O and 10% D 2 O. Fig. 8 shows an overlay view of the two NOESY spectra obtained using a mixing time of 200 ms. The spectrum obtained with protonated DMPC is shown in black, and that with perdeuterated DMPC is shown in red. The assignments of the observed lipid protons are marked beside the observed NOESY cross-peaks (Fig. 8). Fig. 9A shows the onedimensional horizontal slices of the NOESY spectrum of the 4F⅐DMPC complex in the presence of protonated DMPC. The corresponding lipid protons are indicated beside the horizontal slices (Fig. 9A). Fig. 9B shows the one-dimensional horizontal slice of the NOESY spectrum corresponding to the (CH 2 ) n protons of the DMPC alkyl chain and the observed peptide aromatic proton signals. To compare intermolecular peptide-lipid NOEs, NOESY spectra of 2F⅐DMPC and 4F⅐DMPC complexes were recorded under identical conditions. Fingerprint regions of the spectra obtained using a 200-ms mixing time and pro-  . Schematic representation of the observed sequential and long range NOEs. Line thickness for the sequential NOEs is proportional to their measured intensity. Medium and long range NOEs were not classified as strong, medium, or weak. Other observed long range NOEs, especially those observed between side chain protons of the nonpolar residues, are described under "Materials and Methods." cessed using identical window functions are shown in Fig. 10, A  and B, for 2F⅐DMPC and 4F⅐DMPC, respectively. It is evident that, compared with 2F⅐DMPC, stronger peptide-lipid NOEs are observed in the 4F⅐DMPC complex (Fig. 10). This is further shown in Fig. 11, which compares one-dimensional horizontal slices of NOESY cross-peaks between 2F and 4F and methyl protons of the choline headgroup in 2F⅐DMPC and 4F⅐DMPC, respectively. It is also important to note that intensities of the NOE cross-peaks between DMPC protons are also different between 2F⅐DMPC and 4F⅐DMPC complexes (Fig. 10). For example, although there is a relatively strong NOE cross-peak between DMPC 2.CH and ␤CH 2 protons in 4F⅐DMPC complex, this NOE is not seen in the case of the 2F⅐DMPC complex (Fig. 10).

DISCUSSION
We have shown that in a set of class A amphipathic helical peptides, despite similar physical-chemical properties, only some of the peptide-lipid complexes are antiatherogenic (33). Such complexes are also involved in improving the antiatherogenic properties of HDL in atherosclerosis-sensitive animal models (33). We undertook the present studies to understand subtle differences in the peptide-lipid complex properties that lead to differences in the biological properties.
Peptide 4F Forms a Well Defined Class A Amphipathic ␣-Helix-The NMR structure of 2F has been reported by us previously (15). It has been shown previously that the end group blockage increases the helical content of the unblocked peptide 18A by removing the destabilizing interactions of the helix macrodipole with the charged termini (13). It is worthy of note that in an ␣-helix the N terminus acetyl and the C terminus amide blocking groups also provide for the formation of two additional hydrogen bonds (one at each end) between residues i (acceptor) and i ϩ 4 (donor), compared with the unblocked peptide. The N-terminal acetyl group shows NOEs to Asp 1 , Trp 2 , Phe 3 , and Lys 4 backbone NH protons as well as to Trp 2 ring protons (Fig. 3B), suggesting its involvement in helix stabilization.
The observation of NOEs between hydrophobic side chains spaced at (i, i ϩ 3) and (i, i ϩ 4) positions in the linear sequence (Trp 2 /Phe 6 , Trp 2 /Ala 5 , Phe 3 /Tyr 7 , Phe 6 /Val 10 , Tyr 7 /Ala 11 , and Phe 14 /Phe 18 ) (Fig. 3, A and B) is also consistent with an ␣-helical structure of the peptide. In addition, the aromatic rings of Trp 2 , Phe 3 , Phe 6 , and Tyr 7 are sufficiently close in space to give rise to the observed NOEs between their ring protons (Fig. 3A).
The pattern of stronger d NN (i, i ϩ 1) and weaker d ␣N (i, i ϩ 1) NOEs as well as the observation of several well resolved d ␣N (i, i ϩ 4) and d ␣␤ (i, i ϩ 3) NOEs (Fig. 4) confirms that 4F adopts an ␣-helical structure in DMPC discs. A few of the d ␣N (i, i ϩ 4) and d ␣␤ (i, i ϩ 3) NOEs could not be confirmed due to resonance overlap but are likely to be present as well (Fig. 4). The 1 H NMR chemical shifts in peptides and proteins are dependent on the secondary structure (29, 32) (e.g. C ␣ H protons move upfield (relative to their random coil value) upon helix formation). An inspection of Fig. 5A reveals that all of the C ␣ H protons in 4F experience an upfield chemical shift relative to their "coil" values (29). Thus, the chemical shifts of C ␣ H protons are consistent with an ␣-helical structure of 4F.  The NH protons in 4F show a periodic 3-4-residue variation in the (⌬␦) chemical shifts (Fig. 5B). The periodicity in the NH proton chemical shifts has been related to the amphipathicity of the helix (15,32). The periodicity in the NH proton chemical shift is thought to result from a shortening (strengthening) of the hydrogen bonds on the hydrophobic side (a medium of low dielectric) and a lengthening (weakening) of the hydrogen bonds on the hydrophilic side (a medium of high dielectric) of the helix. The NMR structures obtained based on the distance and dihedral angle constraints show that the peptide 4F forms a well defined class A amphipathic ␣ helix with hydrophobic residues clustered on one side and hydrophilic residues clustered on the opposite side of the long axis of the helix (Fig. 6).

Peptide 4F Interacts with the Lipid Acyl Chains and Oriented
Parallel to the Plane of the Bilayer-To determine intermolecular peptide/ lipid NOEs, NOESY spectra of the peptide 4F obtained with protonated DMPC and deuterated DMPC in H 2 O were compared. As in the case of the 2F⅐DMPC complex reported by us previously (15), a number of intermolecular peptidelipid NOEs were identified in the case of protonated DMPC (Fig. 8). These NOEs were further identified by extracting one-dimensional horizontal slices of the NOESY spectrum obtained with protonated DMPC. Like 2F (15), a number of NOEs were seen between lipid alkyl chain methylene protons, (CH 2 ) n , and the peptide aromatic protons (Fig. 9A). It is important to note that these intermolecular NOEs were seen with aromatic ring protons of the residues that are distributed throughout in the linear sequence (e.g. Trp 2 , Phe 3 , Phe 6 , Tyr 7 , Phe 14 , and Phe 18 ) (Fig. 9B). Similar patterns of intermolecular NOEs between DMPC methylene protons, (CH 2 ) n , and 4F aromatic ring protons were seen at different mixing times (50, 100, 150, and 200 ms) used in the NOESY experiments (supplemental Figs. 1 and 2). This observation rules out the possibility of spin diffusion and strongly suggests that, similar to 2F (15), 4F is oriented parallel to the plane of the membrane (and, therefore, perpendicular to the lipid acyl chains; "belt orientation") in the DMPC discoidal particles.
To probe and compare the microenvironments of Lys residues in 2F⅐DMPC (15,34) and 4F⅐DMPC complexes, labeling of the Lys residues of 2F and 4F with 13 CH 3 groups was performed by the reductive methylation procedure of Jentoft and Dearborn (35), as described earlier for apoA-I discoidal complexes (36). The chemical shifts of 13 C-labeled Lys residues in 2F⅐DMPC and 4F⅐DMPC complexes were almost identical, indicating that the microenvironments of Lys residues in the two complexes are similar ( 13 C NMR spectra not shown). A similar pattern of bis(sulfosuccinimidyl) suberate cross-linked products resolved on the SDS-PAGE (results not shown) further indicates that Lys side chains are oriented similarly in 2F⅐DMPC and 4F⅐DMPC discs.
Peptide 4F Is Located Closer to the DMPC Headgroup in 4F⅐DMPC Complex than 2F in 2F⅐DMPC Complex-Compared with 2F⅐DMPC, the observation of stronger NOEs between 4F and DMPC methyl protons of the choline (-N-(CH 3 ) 3 ) headgroup (Fig. 11) as well as ␤CH 2 protons (Fig. 10B), indicates that 4F is located closer to the DMPC headgroup than 2F in DMPC discoidal particles. Both 2F and 4F contain a single Trp residue  as the second amino acid residue in the primary sequence (Fig.  1). It is interesting to note that methyl protons of the DMPC choline headgroup show a much stronger NOE to Trp indole ring proton in the 4F⅐DMPC complex than in the 2F⅐DMPC complex (Fig. 11). Based on the steady-state fluorescence quenching experiments with water-soluble quencher acrylamide, we observed that Trp residues in both 2F and 4F DMPC discoidal particles are equally inaccessible (data not shown) to the quencher compared with peptide free in solution. The Trp indole ring possesses hydrogen bonding ability, with the NH group as hydrogen donor. In addition, the aromatic rings of Trp, Phe, and Tyr can participate in cationinteractions (37). It is important to note that, compared with Leu, Phe has an about 12% larger hydrophobic surface area (38). In addition, compared with Leu, Phe shows a preference for the water-lipid interfacial region (39). Since, compared with 2F, 4F has two additional Phe residues located on the nonpolar face of the helix, we propose that compared with 2F⅐DMPC, additional cationinteractions between choline headgroup and Phe residues of 4F in the 4F⅐DMPC complex, facilitate the location of 4F closer to the headgroup region. Such a preferential interaction of 4F with the lipid headgroup also results in changes in the lipid headgroup as evidenced by the NOE between lipid 2.CH and ␤CH 2 protons in 4F⅐DMPC but not in the case of 2F⅐DMPC (Fig. 10).
It has been shown earlier that 4F exerts a greater degree of penetration into vesicles of pure phosphatidylcholine in the absence of cholesterol than into vesicles of phosphatidylcholine and cholesterol (40). 4F promotes the separation of cholesterol from the phospholipid, resulting in the formation of cholesterol crystallites, even at mol fractions of cholesterol as low as 0.3 (40). Preferential interaction of 4F with the lipid headgroup in the discoidal complexes observed in this study helps explain these earlier observations in lipid vesicles (40).
Proposed Model of 4F⅐DMPC Discoidal Structure-Based on the similar size and similar lipid/peptide ratio in 2F⅐DMPC and 4F⅐DMPC discoidal particles, the overall structure of the two particles is likely to be quite similar. The discoidal morphology of both 2F⅐DMPC and 4F⅐DMPC complexes is confirmed by negative stain transmission electron microscopy (results not shown). However, as discussed above, 4F is located closer to the DMPC headgroup than 2F in the peptide-DMPC discs. Also, the conformation of the DMPC headgroup is different in 4F⅐DMPC than in 2F⅐DMPC, as evidenced by the presence of stronger NOE between 2.CH and ␤CH 2 protons seen in the case of 4F⅐DMPC complex and the absence of this NOE in the case of 2F⅐DMPC complex (Fig. 10). This is presumably because of overall reduction in the thickness of the DMPC bilayer in the case of 4F⅐DMPC discoidal complex compared with 2F⅐ DMPC complex (41). It is important to note that the interfacial region contributes equally to the   DECEMBER 5, 2008 • VOLUME 283 • NUMBER 49 overall thickness of a lipid bilayer as the hydrocarbon region (41). Our proposed model for the 4F⅐DMPC discoidal particles is shown in Fig. 12.

NMR Studies of 4F⅐DMPC Complex
Relevance to Biological Activities of 4F-Dyslipidemia is associated with oxidative stress and the generation of biologically active oxidized lipids. Biologically active oxidized phospholipids can initiate and modulate many of the cellular events attributed to the pathogenesis of atherosclerosis (42). A novel family of atherogenic oxidized choline glycerophospholipids (oxPC CD36 ) that are formed during the oxidization of low density lipoprotein by multiple pathways and are present in vivo at sites of enhanced oxidative stress have been isolated and structurally characterized (43,44). oxPC CD36 serve as high affinity ligands for the macrophage scavenger receptor CD36 (43) and facilitate macrophage foam cell formation through recognition and uptake of oxidized low density lipoprotein by CD36 (45). It has been suggested that the selectively blocking the interaction between oxPC CD36 and CD36 might be an useful therapeutic approach against atherothrombotic diseases (46). The conformation of a prototypic high affinity CD36 ligand, 1-palmitoyl-2-(5-keto-6-octene-dioyl)phosphatidylcholine (KOdiA-PC), near the hydrophobic-hydrophilic interface within membrane bilayers was probed by determining multiple critical internuclear distances using NOESY (47). It was shown that the truncated oxidized sn-2 fatty acid chain of KOdiA-PC within membranes protrudes into the aqueous phase (47). It was hypothesized that this allows a direct physical access of the KOdiA-PC to the cell surface macrophage CD36 receptor (47). It has been shown recently that this unusual conformation for an oxidized phospholipid within a membrane may not be unique to KOdiA-PC but rather might represent a more global phenomenon of oxidized phospholipids, enabling direct physical contact between pattern recognition receptor and molecular pattern ligand (48,49).
We have shown previously that the two homologous apoA-I mimetic peptides, 3F-2 and 3F 14 , differ in their in vitro antiatherogenic properties (50), and the peptide 3F-2 inhibits atherosclerosis in female apoE null mice, whereas 3F 14 does not (51). Using several NMR methods, we have shown that these two peptides insert to different extents into lipid vesicles (51). 3F-2 with aromatic residues at the center of the nonpolar face partitions closer to the phospholipid headgroup in lipid vesicles compared with 3F 14 (51). We have shown that compared with 2F, 4F exhibits more potent in vitro antiatherogenic properties (16). The present study provides further support to our hypothesis that peptides located closer to the lipid headgroup exhibit potent antiatherogenic properties.
It is likely that a potential mechanism of antiatherogenic property of 4F is due to its preferential association with the lipid headgroups, as demonstrated in the present study for the first time. This may allow the shielding of the oxidized phospholipids at the cell surface by 4F, thereby preventing interaction of oxidized phospholipids with their receptor (e.g. CD36).  It has been shown recently using surface plasmon resonance that 4F bound oxidized lipids with much higher affinity than human apoA-I (52). The extraordinary ability of 4F to bind proinflammatory oxidized lipids is suggested to account for its remarkable anti-inflammatory properties (52). It is likely that 2F and 4F bind to membranes containing oxidized lipids differently (53). It is interesting to note that the peptide 3F-2, which has been shown to inhibit atherosclerosis in apoE null mice (51), also bound oxidized phospholipids with very high affinity similar to 4F (52). 3F-2 has been shown to associate with the lipid headgroup (51). The results obtained in the present study further support our hypothesis that antiatherogenic peptides associate with the lipid headgroup and bind oxidized lipids with high affinity.