Structural studies of a peptide activator of human lecithin-cholesterol acyltransferase.

The synthetic lipid-associating peptide, LAP-20 (VSSLLSSLKEYWSSLKESFS), activates lecithin-cholesterol acyltransferase (LCAT) despite its lack of sequence homology to apolipoprotein A-I, the primary in vivo activator of LCAT. Using SDS and dodecylphosphocholine (DPC) to model the lipoprotein environment, the structural features responsible for LAP-20's ability to activate LCAT were studied by optical and two-dimensional 1H NMR spectroscopy. A large blue shift in the intrinsic fluorescence of LAP-20 with the addition of detergent suggested that the peptide formed a complex with the micelles. Analysis of the CD data shows that LAP-20 lacks well defined structure in aqueous solution but adopts helical, ordered conformations upon the addition of SDS or DPC. The helical nature of the peptides in the presence of both lipids was confirmed by upfield H NMR secondary shifts relative to random coil values. Average structures for both peptides in aqueous solutions containing SDS and DPC were generated using distance geometry methods from 329 (SDS) and 309 (DPC) nuclear Overhauser effect-based distance restraints. The backbone (N, C, C=O) RMSD from the average structure of an ensemble of 17 out of 20 calculated structures was 0.41 ± 0.15 Å for LAP-20 in SDS and 0.41 ± 0.12 Å for an ensemble of 20 out of 20 calculated structures for LAP-20 in DPC. In the presence of SDS, the distance geometry and simulated annealing calculations show that LAP-20 adopts a well defined class A amphipathic helix with distinct hydrophobic and hydrophilic faces. A similar structure was obtained for LAP-20 in the presence of DPC, suggesting that both detergents may be used interchangeably to model the lipoprotein environment. Conformational features of the calculated structures for LAP-20 are discussed relative to models for apolipoprotein A-I activation of LCAT.

In human plasma, free and esterified cholesterol circulate as constituents of lipoproteins. The primary enzyme responsible for modulating plasma levels of cholesterol is lecithin-cholesterol acyltransferase (LCAT; EC 2.3.1.43) 1 (1,2), a 60-kDa glycoprotein that catalyzes transesterification of a fatty acid from the sn-2 position of lecithin (phosphatidylcholine) to cholesterol. The key in vivo metabolic activator of LCAT is apoA-I (3), a protein also believed to play an important role in the biogenesis of high density lipoprotein particles, which serve as acceptors of peripheral cell-associated free cholesterol in the reverse cholesterol transport pathway (4,5). High plasma high density lipoprotein-cholesterol levels have been inversely correlated with the risk of developing coronary artery disease (6).
A prominent feature of apoA-I and the other exchangeable apolipoproteins is repeating amino acid motifs of 11 or 22 residues, which, based upon predictions from primary sequences, may adopt amphipathic helical structures when associated with lipid (7). Such a motif is characterized by polar and nonpolar amino acid residues aligned on opposing faces of the long axis of an ␣-helix. The positively and negatively charged amino acid residues are distributed in the polar-nonpolar interface and along the center of the polar face, respectively. The lipid-associating properties of the apolipoproteins are believed to result from hydrophobic interactions between the nonpolar amino acid side chains and the phospholipid acyl chains (8). The model also allows for ionic interactions between (i) the positively charged protein side chains in the interface with the negatively charged phosphate groups of the phospholipid and (ii) the negatively charged protein side chains in the hydrophobic face with the positively charged quaternary amines of the phospholipid.
In order to more precisely define secondary structural features responsible for LCAT activation, synthetic model peptides have been studied. Synthetic fragments of apoA-I designed to localize the position of the major LCAT activating region indicate that residues 143-185 and 121-164, which both contain two predicted amphipathic regions (9,10), activate LCAT 24 and 30%, respectively, of the rate of native apoA-I. De novo peptides have shown that charge distribution plays a role in the peptides' affinity for lipid and ability to activate LCAT (11,12). For example, one acidic 30-residue peptide, GALA, was observed to activate LCAT almost as effectively as apoA-I (13). Studies using de novo amphipathic peptides of various lengths suggest that 10 -12 residues are enough for lipid association (14) but that longer sequences are necessary for LCAT activation (15). The critical role of the primary sequence is illustrated by the inability of a 16-residue peptide, LAP-16, to associate with DMPC or activate LCAT, while the addition of four residues to the amino terminus of LAP-16 (LAP-20) results in a dramatic increase in lipid association and LCAT activation (16). ApoA-I mimetic peptides have also illustrated that proline punctuation between contiguous pairs of predicted amphipathic ␣-helices increases lipid affinity and LCAT activation (17), although the absence of such a residue in LAP-20 and GALA suggests that intramolecular interaction between proline and LCAT is likely not involved in LCAT activation.
LAP-20 is unique in that, while it has no sequence homology with apoA-I, it activates transacylation by LCAT to 65% of that of apoA-I in the presence of DMPC and cholesterol (16). Because the latter value is greater than twice the value obtained with the most potent synthetic peptide from apoA-I, we chose to study in greater detail the conformational changes that occur upon the association of LAP-20 with lipid.
Sodium dodecyl sulfate and dodecylphosphocholine, agents commonly used to model membranes (18,19), were used to model the lipoprotein environment (20,21). The amount of helical structure adopted by LAP-20 in the absence and in the presence of increasing concentrations of SDS and DPC was estimated by circular dichroism spectroscopy and H ␣ NMR secondary shift analysis. Fluorescence Spectroscopy-Fluorescence measurements were obtained at 20°C on an SLM 4800 C spectrofluorometer. The samples (10 M LAP-20 Ϯ 400 M detergent) were irradiated at 280 nm, and the emission spectra were recorded between 300 and 450 nm. The emission maxima were reproducible to Ϯ 1 nm on replicate experiments. Circular Dichroism Spectroscopy-CD spectra were obtained on a Jasco J710 spectropolarimeter calibrated using ammonium d-(ϩ)-camphorsulfonate. The measurements were performed with 0.53 (SDS) and 0.64 (DPC) mM peptide (pH 5.0) and various concentrations of lipid (0 -25 mM) in a quartz cell of 0.02-cm path length at 25°C. Spectra were the average of two consecutive scans from 260 to 190 nm recorded with a bandwidth of 0.5 nm, a time constant of 0.25 s, and a scan rate of 10 nm/min. Following base-line correction and noise reduction, the observed ellipticities were converted to mean residue ellipticities, [], in units of degrees ϫ cm 2 /dmol. NMR experiments were run on a Bruker AMX spectrometer operating at a proton resonance frequency of 600.13 MHz as reported earlier (20). Standard phase-sensitive (TPPI) two-dimensional NOESY (23), TOCSY (24,25), and DQF-COSY (26) spectra were recorded at 25°C. Water suppression in the TOCSY and NOESY experiments was by WATERGATE (27) employing a 3-9-19 pulse sequence (28). NOESY data were recorded using mixing times of 75, 100, 150, and 225 ms. A 75-ms mixing time and 2.5-ms trim pulse were used in the MLEV-17 spinlocking sequence of the TOCSY experiments. Prior to Fourier transformation, the data were zero-filled to generate a 2K ϫ 2K matrix and apodized by a cosine function in D2 and a sine function in D1. A fifth-order polynomial function was applied to base-line correct all processed spectra in both dimensions.
Structure Calculations-Three-dimensional structures were calculated from the NOE distance data (FELIX) using the distance geometry program (DGII) of Insight II as described by Rozek et al. (20), except the initial distance restraints used to calculate structures for LAP-20 in DPC were obtained by classifying the peak volumes into strong (1.80 -3.0 Å), medium (2.80 -4.0 Å), and weak (3.80 -5.50 Å) ranges as opposed to ranges of 1.80 -2.50, 2.51-3.50 and 3.51-5.00 Å used to classify NOESY volumes for LAP-20 in SDS.

RESULTS
Optical Spectroscopy-The addition of 40 molar excess SDS and DPC to an aqueous solution of LAP-20 produced a 21-nm blue shift in the intrinsic tryptophan fluorescence. A blue shift of 16 nm was reported for LAP-20 upon the addition of 20 M excess DMPC (29). Such a shift in the maximum emission wavelength of the fluorescence of tryptophan suggests an association of the peptide with the lipid accompanied by a transfer of the tryptophan residue from a polar to nonpolar environment (16, 29 -31).
The CD spectra from a titration of an aqueous solution of LAP-20 with lipid are shown in Fig. 1. In the absence of detergent the LAP-20 spectrum shows a negative band around 200 nm and a very weak band near 220 nm, which characterize peptides that lack a well defined secondary structure (32). Spectra obtained without lipid did not change over the pH range 3-11 and, as observed by Pownall et al. (16), over the temperature range of 10 -50°C. The addition of SDS or DPC to LAP-20 effected changes in the CD spectra, which suggested an increase in ordered secondary structure. Above a peptide:detergent ratio of 1:4 (SDS) and 1:6 (DPC), no further changes were observed in the CD spectra, suggesting that LAP-20 was completely associated with lipid in a micelle-bound state (20,33). The lower molar ratio of SDS required to obtain such a condition may reflect a marginally greater affinity of LAP-20 for the negatively charged SDS than the zwitterionic DPC (31). In both lipid complexes the CD spectra of LAP-20 possess a double minimum at 222 and 208 -210 nm and a substantial maximum at 191-193 nm. Such features are indicative of a helical conformation (34) and were observed for LAP-20 in complexes with DMPC (16,29).
The helical content of LAP-20 in the absence and presence of SDS and DPC was estimated by deconvoluting the CD spectra using convex constraint analysis (35), and the results are pre- sented in Table I. While the absolute percentages of secondary structure obtained by such analyses will vary depending on the weighting of secondary structures contributing to the spectra used in the basis set, the convex constraint analysis results do illustrate trends. An increase in helical secondary structure of LAP-20 is observed as the lipid concentration is increased, with essentially identical values obtained at a peptide:lipid ratio of 1:40. Similar observations were made upon the titration of other peptides with SDS (20,33,36) and DPC (21,37). Because the calculated mean molar ellipticity values are sensitive to other factors (38,39), the CD data were also analyzed using two parameters, R1 and R2, which are independent of inaccuracies in determined peptide concentrations as well as those caused by small shifts in wavelength (40). Such parameters, defined and tabulated in Table I, also follow a trend characteristic of an increase in helical secondary structure as the lipid concentration is increased.
Proton Resonance Assignments and Secondary Shifts-In a non-lipid aqueous environment, the 1 H NMR spectrum of a saturated LAP-20 solution (approximately 1 mM) contains poorly resolved resonances characteristic of self-aggregation (41,42). Such observations agree with earlier Fourier transform infrared spectroscopy data, which suggested that, in the absence of lipid, LAP-20 self-aggregated (43). On the other hand, in the presence of either SDS or DPC, the solubility of LAP-20 increases severalfold and the 1 H NMR spectrum contains well resolved resonances with linewidths of approximately 10 Hz, and there is now a dispersion of the downfield amide resonances between 9 and 7.5 ppm and the side chain resonances between 4.5 and 0.5 ppm. Such observations suggest that LAP-20 forms a complex with the detergent micelles and that the conformation adopted by LAP-20 in the presence of lipid differs significantly from its structure in water alone. Fig. 2 illustrates the H ␣ -H N regions of 5 mM LAP-20 solutions in the presence of SDS-d 25 and DPC-d 38 . All of the proton resonances for LAP-20 in both detergents could be assigned to one unique species using TOCSY spectra to identify spin systems, NOESY spectra to obtain interresidue connectivities and to distinguish between degenerate spin systems, and DQF-COSY spectra to confirm side chain resonance assignments (44). The proton chemical shifts for LAP-20 in the presence of 40-fold molar excess SDS-d 25 and DPC-d 38 are summarized in Tables II and III, respectively. Unique proton assignments, and the preparation of clear LAP-20 solutions with similar NMR spectral features at 5 and 20 mM, implied that the peptide was tightly associated with the detergent micelles in a single conformation.
Using random coil proton chemical shifts determined for peptides in unstructured conformations (44), it is possible to calculate chemical shift changes that occur when a peptide or protein adopts an ordered conformation. With respect to random coil chemical shift values, the H ␣ resonances move upfield in an ␣-helical conformation and downfield in a ␤-sheet conformation (44,45). H ␣ secondary shifts (⌬H ␣ ) were calculated by subtracting the measured H ␣ chemical shifts of LAP-20 in detergent complexes (Tables II and III) from the corresponding random coil values obtained by Wü thrich (44). Most of the H ␣ secondary shifts are positive, which suggests that LAP-20 adopts a helical conformation in the presence of lipid. Furthermore, except for Trp 12 , Lys 16 , and two residues at the Nterminal, the difference in ⌬H ␣ for LAP-20 in the presence of both lipids is less than 0.04 ppm, suggesting a similar conformation in both lipid environments. A semiquantitative estimation of this helical content is obtained by dividing the average H ␣ secondary shift by 0.35, the average upfield H ␣ shift observed in the amino acid residues of proteins in an ␣-helical conformation (46). The results, tabulated in Table I, agree with the CD data and show that LAP-20 is highly helical in the presence of both detergents. A more detailed examination shows that the H ␣ secondary shifts of residues toward the N and C termini approach 0 in both lipids, suggesting some fraying at both ends.
Interresidue NOEs and Secondary Structure-An unambiguous indication of protein secondary structure is provided by the magnitude and pattern of the interresidue nuclear Overhauser effects (44). NOESY spectra highlighting the H ␣ -H N regions of LAP-20 (Fig. 2) show strong and medium, and in most cases well resolved, sequential H ␣ i -H N iϩ1 cross-peaks. As  ))/(no. of residues))/0.35. c Calculated as described by Rizo et al. (46). R1 is the ratio between the intensity of the maximum between 190 and 195 nm and the intensity of the minimum between 200 and 210 nm. R2 is the ratio between the intensity of the minimum near 222 nm and the intensity of the minimum between 200 and 210 nm.
, and H ␣ i -H ␤ iϩ3 NOE cross-peaks are also observed. The presence of strong to medium (i) Ϫ (iϩ1) and medium to weak (i) Ϫ (iϩ2), (i) Ϫ (iϩ3), and (i) Ϫ (iϩ4) NOE contacts throughout both structures suggests that they both adopt highly helical conformations when bound to detergent (44,47) in accord with the CD and secondary shift analyses.
Three-dimensional Structure Calculations-Based on the CD data, H ␣ secondary shifts, and NOE connectivity patterns for LAP-20, a right-handed ␣-helix was chosen as the starting structure for the distance geometry and simulated annealing calculations.  Fig. 4C the backbone of the average structure of the ensembles in Fig.  4, A and B, has been replaced by a ribbon and superimposed on residues 3-18. The quality of the final structures is apparent in Fig. 5, plots of the pairwise RMSDs to the mean structure for each residue. In general, the pairwise RMSD per residue is below 0.2 Å for the backbone atoms and never rises above 1.4 Å for all atoms. For instance, the backbone (N, C ␣ , CϭO) RMSD for residues 3-18 is 0.29 Ϯ 0.14 Å in SDS and 0.27 Ϯ 0.09 Å in DPC. If all atoms are included the RMSD for residues 3-18 is 1.13 Ϯ 0.14 Å in SDS and 1.02 Ϯ 0.13 Å in DPC. DISCUSSION LAP-20 was shown to associate with both SDS and DPC by optical spectroscopy and displayed optical properties similar to those observed for LAP-20 in the presence of DMPC (16,29), a lipid in which LAP-20 activates LCAT, suggesting that the peptide's conformation is similar in all three lipid environments. Previously, Pownall et al. (16) confirmed a physical association of LAP-20 with DMPC discoidal complexes using a variety of methods, including ultracentrifugation in a density gradient and size exclusion chromatography. Our 1 H NMR data indicate that LAP-20 associates with SDS and DPC, and all the evidence suggests it associates as a monomer. While the unique set of proton resonance linewidths for LAP-20 in the presence of detergent are sharp enough to allow unambiguous assignments, they are still too broad to obtain H ␣ -H N coupling constants, indicating the peptide is part of a large molecular weight micellar complex (19,20,46). The CD data, H ␣ secondary shifts (Table I), and the pattern of the interresidue NOEs (Figs. 2 and 3) (44,47) suggest that LAP-20 adopts a helical conformation when associated with SDS or DPC. Detailed three-dimensional structures generated for LAP-20 in the presence of SDS and DPC, using distance geometry calculations, verify such structures as illustrated by the superimposed backbone atoms of the calculated structures in Fig. 4, A (SDS) and B (DPC). Fig. 4C, which overlays ribbons drawn through the backbone atoms of the mean LAP-20 structure of the ensembles in Fig. 4, A and B, illustrates the formation of a similar, well defined helical structure over the length of the molecule in the presence of both detergents, apart from some dynamic fraying at the termini. Because the different head groups on SDS and DPC have little effect on the structure adopted by LAP-20 in the micelle, we suggest that both detergents may be equally suited to model the lipoprotein environment.  hydrophilic, and interfacial. Trp 12 is fixed in the interfacial region with the hydrophobic six-membered ring oriented toward the hydrophobic face and the polarizable imino group intruding into the hydrophilic face. Such an orientation (which generated a 21-nm blue shift in the maximum Trp fluorescence) is predicted to be energetically favored (48 -51) and is similar to the orientation observed for the Trp residue in a proposed lipid binding domain of apoC-I (35)(36)(37)(38)(39)(40)(41)(42)(43)(44)(45)(46)(47)(48)(49)(50)(51)(52)(53) (20). Glu 10 and Glu 17 , which are negatively charged in the SDS solution at pH 5.0, lie along the center of the hydrophilic face. The two positively charged residues, Lys 9 and Lys 16 (dark gray), are located in the polar-nonpolar interface. Such an overall orientation of the peptide side chains for LAP-20 in the presence of SDS and DPC fits the definition of a class A amphipathic helix (7,51) and was predicted for LAP-20 from the primary structure (16).
The nonpolar side chains that extend from the hydrophobic face of LAP-20 presumably interact with the hydrophobic acyl chains of the lipid. On the other hand, the polar and charged side chains, located at the polar-nonpolar interface or the hydrophilic face, presumably interact with the aqueous milieu that includes the negatively charged head groups of SDS or the zwitterionic head groups of DPC. Because the LAP-20 structures are similar regardless of the head group of the detergent, hydrophobic interactions between the nonpolar face of the peptide and the hydrophobic interior of the micelle are likely the primary force stabilizing the helix (50,52). The surface occupied by the hydrophobic face, as estimated by looking down the long axis of the helix, is a pie-shaped wedge that occupies ϳ30% of the total area. This proportion of peptide, which presumably penetrates into the micelle surface, is considered optimal for enzyme activation (9).
In addition to hydrophobic interactions, ␣-helical structures are stabilized by intramolecular hydrogen bond formation between backbone amide and backbone carbonyl groups four residues apart (5-7 kcal/bond). LAP-20, in a perfect ␣-helix, is predicted to form 16 backbone hydrogen bonds with a H N ⅐⅐⅐O distance of 2.06 Ϯ 0.16 Å and a N-H N ⅐⅐⅐O bond angle of 155 Ϯ 11° (53). The calculated structures for LAP-20 in SDS and DPC show that the N-H N ⅐⅐⅐O atoms are in a position to form 11 and 14 hydrogen bonds, respectively, that meet the following conditions: (i) Ϫ (i ϩ 4), N-H N ⅐⅐⅐O bond angle between 120 and 180°, H N ⅐⅐⅐O distance Ͻ 3.0 Å. Therefore, it is likely that the helical structure of LAP-20, in both detergents, is stabilized by the formation of intramolecular hydrogen bonds. Segrest (7,51) has proposed a stabilization of amphipathic structures by a "snorkeling" of basic amino acid side chains located at the polar-nonpolar interface. In such a model the Lys and Arg residues are oriented with the side chains aligned along the edge of the hydrophobic face, and the positively charged ends are extended into the hydrophilic face. It is evident from Figs. 6 and 7 that, while the alkyl groups of Lys 9 and Lys 16 align along the edge of the hydrophobic face, the positively charged ends extend into the polar-nonpolar interface and not into the hydrophilic face, i.e. hydrophobic interactions dominate. We noted previously that snorkeling was not exten- sive in apoC-I fragments bound to SDS, and consequently, this study reinforces our contention that snorkeling is likely not a general characteristic of amphipathic helices (20).
Intraresidue charge interactions, in the form of salt bridges between oppositely charged side chains, may stabilize ␣-helical structures by up to 6 kcal/mol/salt bridge (54 -56). While LAP-20 contains two positively and two negatively charged side chains at pH 5.0, they are either too close (1 residue) or too distant (Ͼ5 residues) to stabilize the ␣-helix through intramolecular salt bridge formation. It is therefore not surprising to find that the structure of LAP-20 at pH 4.0 in DPC, where the Glu residues are protonated, does not differ significantly from the structure obtained at pH 5.0 in SDS, where the Glu residues are negatively charged. Furthermore, detailed structural studies of the proposed lipid binding domains of apoC-I, which adopted helical structures in the presence of SDS, showed that salt bridges did not form (20).
Features associated with peptides that activate LCAT are affinity for lipid (57) and helical secondary structure (58). Our NOE-derived structures for LAP-20 confirm a helical conformation and show that ionic interactions (intermolecular salt bridges and snorkeling) play a minor role in stabilizing the lipid-bound state. We find, instead, that hydrophobic interactions between the nonpolar amino acid side chains and the phospholipid acyl chains (8) play the most important role in stabilizing the complex, followed by intermolecular backbone hydrogen bonding. The hydrophobic interactions are likely crucial for LCAT activation because LCAT is surface-active, i.e. it binds to the lipid/water interface (3). Indeed, it has been suggested that the primary role of the amphipathic helix in activating LCAT is to disrupt the water-phospholipid interface and expose buried substrate to LCAT (15,59). Support for this hypothesis is the observation that LCAT does not require a co-factor to hydrolyze water soluble substrates such as the p-nitrophenyl esters of fatty acid (59). However, while lipid binding is necessary for LCAT activation, many amphipathic peptides that bind to lipid do not activate LCAT (51). Consequently, there must be one or more key topological features of these bound peptides responsible for their ability to effectively activate LCAT.
Segrest et al. (51) have proposed that the unique position of two negatively charged Glu residues, located in the nonpolar face of apoA-I regions 66 -87 and 99 -120, play a major role in the apoA-I activation of LCAT. In each of these two apoA-I 22-residue regions, the Glu is located at the 13th position. While a consensus 22-residue sequence containing a Glu at the 13th position is a poor activator of LCAT, a 44-residue dimer, obtained by linking two 22-mers together head-to-tail, activates LCAT as well as apoA-I (57). Because the two Glu residues of LAP-20, a potent activator of LCAT, are located in the center of the hydrophilic face when bound to lipid, it is unlikely that negatively charged Glu residues in the nonpolar face are directly involved in the intermolecular activation of LCAT.