Only the Two End Helixes of Eight Tandem Amphipathic Helical Domains of Human Apo A-I Have Significant Lipid Affinity
Implications for HDL Assembly
Abstract Human apolipoprotein A-I (apo A-I) possesses multiple tandem repeating 22-mer amphipathic α-helixes. Computer analysis and studies of model synthetic peptides and recombinant protein-lipid complexes of phospholipids have suggested that apo A-I interacts with HDL surface lipids through cooperation among its individual amphipathic helical domains. To delineate the overall lipid-associating properties of apo A-I, the first step is to understand the lipid-associating properties of individual amphipathic helical domains. To this end, we synthesized and studied each of the eight tandem repeating 22-mer domains of apo A-I: residues 44-65, 66-87, 99-120, 121-142, 143-164, 165-186, 187-208, and 220-241. Among the 22-mers, only the N- and C-terminal peptides (44-65 and 220-241) were effective in clarifying multilamellar vesicles (MLVs) of dimyristoylphosphatidylcholine (DMPC). These two peptides also exhibited the highest partition coefficient into 1-palmitoyl-2-oleoyl-sn-glycero-3-phosphatidylcholine liposomes, the highest exclusion pressure for penetration into an egg yolk phosphatidylcholine monolayer, and the greatest reduction in the enthalpy of the gel-to–liquid crystalline phase transition of DMPC MLVs. These results suggest that the strong, lipid-associating properties of apo A-I are localized to the N- and C-terminal amphipathic domains. Although each of the eight peptides studied has an amphipathic structure, models based on changes in residual effective amino acid hydrophobicity resulting from differing depths of helix penetration into the lipid are best able to explain the high lipid affinity possessed by the two terminal domains. Differential scanning calorimetry (DSC) studies showed that on a molar basis, apo A-I is about 10 times more effective than the most effective peptide analyzed in reducing the enthalpy of the gel-to–liquid crystalline phase transition of DMPC MLVs. Because previous proteolysis experiments coupled with the present DSC results suggest that the lipid-associating domains of apo A-I are distributed throughout the length of the 243 amino acid residues, we propose that the terminal amphipathic helical domains are involved in the initial binding of apo A-I to the lipid surface to form HDL particles, followed by cooperative binding of the middle six amphipathic helical domains, perhaps aided by salt-bridge formation between adjacent helixes arranged in an antiparallel orientation.
- protein-lipid interactions
- amphipathic helical peptides
- lipid affinity
- helix-helix interactions
- cooperative lipid association
Reprint requests to Dr G.M. Anantharamaiah, Departments of Medicine, Biochemistry, and Molecular Genetics, UAB Medical Center, Birmingham, AL 35294.
- Received June 14, 1995.
- Accepted October 27, 1995.
Apo A-I is the major protein component of human HDL. Apo A-I has significant clinical relevance, because higher levels of HDL are a negative risk factor for coronary artery disease. Recently, it has been shown that expression of human apo A-I in transgenic mice that are normally susceptible to atherosclerosis protects these animals from this disease.1 2 The protective effect of HDL is associated exclusively with those particles that contain human apo A-I alone,3 4 thus providing compelling in vivo evidence for the antiatherogenicity of human apo A-I.
Because this protein has not yet been studied by x-ray crystallography and particularly because the lipid-associated structure of this protein is physiologically important, many approaches have been used to understand its structure and function. On the basis of the human apo A-I genomic structure, it is known that the fourth exon comprises eight tandem repeating 22-mer amphipathic helical domains. From fragments obtained by cyanogen bromide degradation, proteolysis, and synthesis of peptides corresponding to the apo A-I sequence, these tandem repeating amphipathic helical domains appear to be responsible for the lipid-associating properties of this protein.5 6 7 It has also been reported that the domains responsible for functional properties, such as activation of lecithin:cholesterol acyltransferase and cholesterol efflux, reside in the midportion of apo A-I.8 9 10 11 12 Investigations using monoclonal antibodies to apo A-I and limited proteolysis also suggest that the lipid-associating properties are located toward the C-terminal region and the functional domains in the midportion.8 Proteolysis experiments further suggest that when apo A-I is associated with lipid, the N terminus is susceptible to proteolysis and therefore not involved in lipid association.8 13
Based on these studies, many structures have been proposed for lipid-associated apo A-I.5 6 10 11 12 14 Calculation of the size and stoichiometry of the recombinant lipoprotein complexes and the assumption that there are eight 22-mer amphipathic helixes (most of them punctuated by a Pro) have yielded models that show all 22-mer amphipathic helical domains of apo A-I to be associated with phospholipids in an antiparallel arrangement, with the helix axes parallel to the phospholipid acyl chains. Most of these proposed structures also show that the N-terminal domain (≈residues 1 through 43) is not helical and not involved in the lipid association.5 9 14 15 16 Such structures have also been used to explain subspeciation in discoidal apo A-I–phospholipid complexes.
Although studies of synthetic peptides corresponding to the apo A-I sequence have been reported,17 18 none are yet capable of determining the ability of individual helixes to associate with phospholipid. Because it has been suggested that cooperation among amphipathic helical domains is involved in lipid association,7 19 a detailed analysis of the lipid-associating ability of individual helixes is essential for understanding the lipid-associating and other functional properties of apo A-I. In this article, we report the lipid-associating properties of eight candidate amphipathic helical domains of apo A-I that have been suggested by Segrest et al5 and others.9 10 11 12 14 15 16 Peptides were synthesized as Ac-peptide-NH2. The 22-mer peptides studied were Ac–44-65–NH2, Ac–66-87–NH2, Ac–99-120–NH2, Ac–121-142–NH2, Ac–143-164–NH2, Ac–165-186–NH2, Ac–187-208–NH2, and Ac–220-241–NH2.
DMPC, POPC, lyso PC, and egg PC were purchased from Avanti Polar Lipids and used without further purification. All other chemicals were of the highest purity available commercially.
The peptides were synthesized by the solid-phase method using an automated solid-phase peptide synthesizer (Advanced ChemTech). To a benzhydrylamine resin support (cross-linked with 1% divinylbenzene [0.536 mEq/g], Peninsula Laboratories, Inc), FMOC–amino acids were coupled in the presence of DCC and 1-hydroxybenzotriazole. The other protecting groups used for the side-chain functional amino acids were tert-butyl for Tyr, Ser, Thr, Asp, and Glu; N-t-BOC for Lys; trityl for His; and 4-methoxy-2,3,6-trimethylbenzenesulfonyl for Arg. Three equivalents of the appropriate first FMOC–amino acids were coupled to the resin after neutralization of the resin hydrochloride with 5% N,N-diisopropylethylamine in N,N-dimethylformamide. The remaining free amino groups in the resin were blocked by treatment with DCC–acetic acid in dichloromethane for 30 minutes. The FMOC group that had been used for temporary protection of the N terminal was removed at each stage by two treatments with 50% piperidine in N,N-dimethylformamide (10 mL/g of resin), the first for 5 minutes and the second for 15 minutes throughout. Coupling and deprotection steps during synthesis were monitored by the Kaiser test.20 Whenever the couplings were incomplete, the amino acids were recoupled. After the completion of the addition of amino acids in sequence and deprotection of the N-terminal α-amino acid, the N terminal was acetylated by treatment with DCC–acetic acid as described above.
Peptides were cleaved from the resin by addition of anhydrous HF (10 mL/g of peptide resin) containing 10% anisole or p-cresol. For peptides with Trp or Met, 10% dimethylsulfide, p-cresol, and p-thiocresol were used. The peptides were extracted with 6 mol/L Gdn HCl (3× 10 mL/g of peptide resin), dialyzed against water (Spectrapor dialysis membrane, 1000–molecular weight cutoff), and lyophilized. The peptides were purified on a preparative C-4 reversed-phase HPLC column (VYDAC: 22-mm inner diameter×25 cm; particle size, 10 μm) on a Beckman HPLC system using a gradient of 10% acetonitrile to 80% acetonitrile (containing 0.1% trifluoroacetic acid, vol/vol) at a flow rate of 4.8 mL/min (total, 66 minutes). Peptide purity was checked on an analytical HPLC (Beckman System Gold 166) using a C-18 reversed-phase column (4.6-mm inner diameter×25 cm; particle size, 5 μm). The purity and authenticity, respectively, of each synthetic peptide were confirmed by amino acid analysis using the phenylisothiocyanate method21 and mass spectral analysis using a PE-Sciox APT-III triple-quadrupole ion-spray mass spectrometer.
Preparation of Peptide Solutions
Peptide solutions were prepared by dissolving the solid in PBS (pH 7.4; 6.45 mmol/L Na2HPO4·7H2O, 1.47 mmol/L KH2PO4, 136.89 mmol/L NaCl, and 2.68 mmol/L KCl). Peptide concentrations were determined in 6 mol/L Gdn HCl solution by measuring the absorbance at either 280 (ε280=[5500 mol/L]−1·cm−1 for Trp and ε280=[1800 mol/L]−1·cm−1 for Tyr) or 275 (ε275=[1400 mol/L]−1·cm−1 for Tyr) nm. For those peptides with neither Tyr nor Trp, quantitative amino acid analysis was used to determine the concentration.
Interaction of Peptides With MLVs of DMPC
Association of the peptides with DMPC MLVs was assessed by monitoring the rate of clarification of turbidity due to solubilization of the MLVs. In brief, a chloroform solution of DMPC was transferred to a test tube and the solvent removed under a stream of N2. Residual solvent was removed by storing the tube overnight under high vacuum in a vacuum oven at room temperature. The lipid film thus deposited onto the walls of the test tube was hydrated overnight by adding appropriate amounts of PBS, and the test tube was vortexed to obtain MLVs. Peptides at a 1:1 ratio (wt/wt) were mixed with the MLVs, and turbidity clarification was followed by measuring the scattered light intensity on an SLM 8100 photon-counting spectrofluorometer with both excitation and emission monochromators at 400 nm. The sample, containing 50 μg each of DMPC and peptide, was maintained with stirring at 25°C. The protein-to-lipid ratio of apo A-I/DMPC was 1:2 (wt/wt). As a positive control, completely dissolved DMPC was achieved by adding Triton X-100 to the MLV suspension at a final concentration of 1 mmol/L.
Negative-stain (potassium phosphotungstate) electron microscopy was performed as described by Forte and Nordhausen.22 Micrographs were obtained at an instrument magnification of 80 000, and dimensions of 100 particles were measured and averaged.
High-sensitivity DSC studies were performed in a Microcal MC-2 scanning calorimeter (MicroCal, Inc) at a scan rate of 20°·h−1. The lipid MLVs and the peptide-lipid mixtures for DSC were prepared as follows. DMPC (≈2 mg) was dissolved in chloroform in a test tube and dried by slow evaporation of the solvent under a stream of dry N2. Residual solvent was removed under high vacuum in a vacuum oven as described above. To the dry lipid film, either buffer alone or buffer with peptide was added to obtain a lipid-to-peptide molar ratio of 20:1. The lipid was hydrated by vortexing at room temperature for 30 minutes. The suspension was scanned with buffer alone in a reference cell. Four consecutive scans with a 60-minute equilibration period between each were run for each sample. No significant changes were observed between the first and the last scan. To obtain the ΔHT of the gel-to–liquid crystalline phase transition of the lipid, the thermograms were analyzed with software (DA2, V2.1) provided by MicroCal Inc.
CD spectra were recorded on an AVIV 62DS spectropolarimeter. Details of these measurements have been described previously.23 The helical contents of the peptides were estimated from the mean residue ellipticity at 222 nm as described earlier.24
Binding of Peptides With POPC Vesicles
Binding studies were performed according to procedures described elsewhere.25 In brief, ≈7 mg POPC in 10 mmol/L Tris HCl and 100 mmol/L NaCl buffer at pH 7.25 was used in each experiment. The dry lipid was reconstituted in the Tris buffer to yield a suspension that contained 7 mg lipid per 50 μL. Various concentrations (5 to 25 μmol/L) of peptide were prepared, and 7 mg lipid was added in a final volume of 1.1 mL. The lipid-peptide mixture was vortexed, followed by six freeze-thaw cycles (−70°C) and further vortexing to produce MLVs and to ensure homogeneous equilibration of the peptide. The mixtures were left at room temperature for 4 hours. They were then centrifuged at 300 000g in a Beckman table-top ultracentrifuge (Beckman model TL 100), which produced a clear, almost lipid-free supernatant. The amount of lipid-bound peptide was calculated as the difference in the peptide concentration before and after equilibration with the lipid. The amount of lipid-bound peptide per mole of total lipid was calculated and represented by Xb (mmol/mol). A correction factor was applied to the OD observed at 280 nm. This factor was calculated by measuring the OD of both the supernatant and solutions at 280 and 328 nm. The OD at 328 nm was attributed to light scattering, and the small residual reading was subtracted from the OD at 280 nm before calculation of Xb. For peptides that did not contain either Tyr or Trp, concentrations were determined by either quantitative amino acid analysis or reversed-phase HPLC.
Interaction of Peptides With Phospholipid Monolayers
The relative affinities of peptides for the lipid-water interface were investigated with a surface balance technique. As per published procedures,26 27 an insoluble monolayer of egg PC was spread at the air-water interface (85 cm2) in a circular polytetrafluoroethylene (Teflon) dish containing 80 mL PBS (pH 7.0) at room temperature. The surface pressure (π) was monitored by the Wilhelmy plate technique using a mica plate connected to a Cahn RTL recording electrobalance. Sufficient egg PC was spread from a 9:1 (vol/vol) hexane/ethanol solution to give an initial surface pressure (πi) in the range 5 to 35 dyn/cm. Peptides that were dissolved in the aforementioned buffer (with 1.5 mol/L Gdn HCl) were injected into the subphase to give an initial concentration of 5×10−5 g/dL. A small polytetrafluoroethylene tube, which projected downward through the monolayer into the aqueous subphase, was used for this injection so that the egg PC monolayer was not disrupted. Gdn HCl in the buffer solution ensured that the peptide molecules were initially present as random-coil monomers. The peptide molecules renatured in the subphase as Gdn HCl was diluted to a final concentration of ≤1 mmol/L. The solution was stirred continuously with a magnetic stirrer, and the surface pressure (Δπi) was recorded until a steady-state value was obtained. Steady-state values for Δπi were plotted as a function of πi. Linear extrapolation of the πi versus Δπi curve to Δπi=0 dyn/cm gave the exclusion pressure, ie, the value of πi at which the peptides were no longer able to penetrate the egg PC monolayer.
Synthetic peptides for the present studies are summarized in Table 1⇓. Although most peptides displayed one major HPLC peak after synthesis and cleavage of the peptide from the resin, synthetic peptides Ac–66-87–NH2, Ac–99-120–NH2, and Ac–143-164–NH2, which contain Met on the nonpolar faces of the amphipathic helixes at positions 86, 113, and 148, respectively, displayed two peaks. Each pair of peaks was subjected to mass spectral analysis using ion-spray mass spectroscopy. In every case, the peak with the longer retention time corresponded to the theoretical mass, whereas the peak with the shorter retention time corresponded to a mass compatible with an oxidized Met (Fig 1B⇓).
We have previously shown that oxidation of Met alters the retention time28 29 of apo A-I and apo A-II because the otherwise nonpolar Met becomes polar upon oxidation. An example of such an altered profile in comparison with that of the crude peptide is shown in Fig 1⇑. Fig 1A⇑ is the HPLC profile of the crude peptide that does not contain Met (Ac–44-65–NH2) and Fig 1B⇑ is the HPLC profile of the crude peptide that contains Met (Ac–66-87–NH2). HPLC profiles for all of the purified peptides under the same conditions are shown in Fig 1C⇑, and Table 1⇑ gives the results of mass spectral analysis.
The secondary structures of the peptides in different environments were determined by CD spectroscopy. In PBS, all peptides except Ac–99-120–NH2 showed a predominantly random structure (Fig 2A⇓); the helical content of Ac–99-120–NH2 was estimated to be 38%. Because most of the peptides did not spontaneously “micellize” the DMPC multilamellar vesicles (Fig 3⇓), we attempted to reconstitute the peptide-DMPC complexes by cholate dialysis.30 With this method, the CD spectra of three of the peptides could be measured in the presence of DMPC because the solution remained clear after cholate dialysis; however, the reaction mixtures of other peptides became turbid upon overnight dialysis, with four changes of buffer. In Fig 2B⇓, two of the three aforementioned peptides that clarified the DMPC solution showed significant helical structure in the presence of DMPC. For the other peptides (which could not be reconstituted in DMPC), CD spectra were recorded in the presence of 0.4% lyso PC micelles.31 As controls, CD spectra of the three peptides that had been reconstituted in DMPC were also recorded and are shown in Fig 2C⇓. In the presence of lyso PC micelles, all peptides except Ac–66-87–NH2 showed significant helical structure. The results of these experiments are summarized in Table 2⇓.
Right-Angle Light-Scattering Measurements
Interactions between peptides and DMPC MLVs were studied by monitoring the rate of clarification of turbidity due to addition of the peptide to the lipid vesicles. The results of this study are shown in Fig 3⇑. Only two (Ac–44-65–NH2 and Ac–220-241–NH2) of the eight peptides tested had the ability to micellize the DMPC MLVs. The rate of micellization was much faster with these two peptides compared with that of apo A-I that was tested at a lipid-to-protein weight ratio of 1:2. Furthermore, peptide Ac–220-241–NH2 showed the fastest rate of micellization.
Lipid Binding Assays
Because of problems discussed elsewhere,25 binding studies were performed at low peptide-lipid ratios. Under these conditions, only the two peptides that clarified the DMPC MLVs (ie, Ac–44-65–NH2 and Ac–220-241–NH2) showed a significant Kp. The natural log of Kp for each peptide (proportional to ΔG, the free energy of association) is shown in Table 2⇑.
Measurements of π
The relative abilities of the peptides to penetrate an egg PC monolayer were determined by measurements of π. The πe for apo A-I is 34 dyn/cm. Among the 22-mers, only two (Ac–44-65–NH2 and Ac–220-241–NH2) exhibited π values >25 dyn/cm (Table 2⇑). Peptide Ac–187-208–NH2 exhibited the next highest πe (23 dyn/cm). The other five peptides exhibited significantly lower πe values (20 dyn/cm or less). The two peptides that exhibited the highest πe values (Ac–44-65–NH2 and Ac–220-241–NH2) were also the two that spontaneously interacted with DMPC MLVs to produce discoidal complexes and to show significant Kp values (Table 2⇑).
The effects of peptides on the thermotropic phase-transition properties of the DMPC MLVs were investigated by DSC. DMPC vesicles alone showed a sharp gel-to–liquid crystalline phase transition (Fig 4⇓). It is evident from Table 2⇑ that only peptides Ac–44-65–NH2 and Ac–220-241–NH2 caused large reductions in the ΔHT of the gel-to–liquid crystalline phase transition of the DMPC vesicles. Negative-stain electron microscopy showed these were the only two peptides that produced discoidal structures (results not shown). It is interesting to note that human apo A-I at a lipid-to-protein molar ratio of 200:1 reduces the ΔHT more than the most effective peptide, Ac–220-241–NH2, at a lipid-to-peptide molar ratio of 20:1 (Fig 4⇓; ΔHT for apo A-I–DMPC is 3.1 kcal/mol).
Determinants of Λ
Table 2⇑ provides a summary of measured and calculated properties of apo A-I 22-mer peptides. The KP values of Ac–44-65–NH2 and Ac–220-241–NH2 indicate that these two peptides associate with phospholipids more avidly than do the others. These two peptides were also able to reduce the ΔHT of DMPC and exhibited the highest monolayer πe. These results suggest that the first and last of the eight tandem 22-mer amphipathic helical repeats in apo A-I (44-65 and 220-241) have substantial values for Λ and the other six have much lower ones.
For comparison, we obtained πe and KP values for two model amphipathic helical peptide analogues: 18A (a model class A amphipathic helical peptide) and 18R (18A with charged residue positions reversed). 18A has both a greater exclusion pressure (πe=30 dyn/cm) and a higher KP (lnKP=5.1) than does 18R (πe=23 dyn/cm and lnKP=3.7, respectively). The lower limit for πe can be determined from studies with a randomly scrambled amino acid sequence derived from 18A, 18S, that is a nonamphipathic helix32 with a πe of 19 dyn/cm.
By using lnKP for correlation analysis of the calculated properties of the 22-mers, neither hydrophobicity per residue of the hydrophobic face (r=.03), hydrophobic moment (r=.28), nor percent α-helix content (r=.38) can explain the results of Table 2⇑. Two variables in Table 2⇑ have a reasonable degree of correlation with the lnKP of the eight 22-mers: (1) total hydrophobicity of the nonpolar face (r=.68, Fig 5A⇓)—defined as the total hydrophobicity (on the GES* scale; see the footnote to Tables 2⇑ and 3⇓) of all residues between the angle (θΔ) formed by the charged residues nearest to and located on opposite sides of the center of the hydrophobic face (on a helical wheel projection) and (2) total 22-mer hydrophobicity (r=.61).
To derive a more mechanistic explanation for the results of Table 2⇑, we explored the implications of the “snorkel” hypothesis.23 33 Lys and Arg residues represent flexible, rod-shaped, amino acid side chains that are also markedly amphipathic. Because of these unique physical-chemical properties, class A amphipathic helixes associated with the outer surface of hydrated phospholipids should be able to extend (snorkel) their basic residues toward the polar face of the helix, allowing their charged moieties to come in contact with the aqueous milieu. This hypothesis is supported by a body of experimental evidence from our laboratory.23
The snorkeling of basic residues allows greater penetration of class A amphipathic helixes into the hydrophobic interior of phospholipid monolayers than would otherwise be possible23 ; the greater the angle of the snorkel wedge, the greater the lipid penetration (see Fig 6⇓). Using neutron diffraction, Jacobs and White 34 measured the gradient that water forms from the outside to the inside of a phospholipid monolayer. The hydrocarbon core starts ≈7 Å beneath the center of the PC head group (see Fig 6⇓); at this depth, water has a molar concentration ≈15% of that at the level of the PC head groups.34 Because the free energy of the hydrophobic effect decreases with decreasing water concentration,34 the deeper the penetration of an amphipathic helix into the interior of a phospholipid monolayer, the more effective is the hydrophobicity (ie, the lower the free energy) of the helical nonpolar face. Therefore, the overall lipid affinity of an amphipathic helix will partially depend on its depth of lipid penetration. Nolte and Atkinson16 also emphasized the importance of depth of penetration for determining lipid affinity of amphipathic helixes.
The depth of lipid penetration by an amphipathic helix is related to the angle of the snorkel wedge as follows. Lipid affinity (Λ)=(ΣΔGΨ×d), where ΣΔGΨ is the total hydrophobicity of the nonpolar face, and d, the relative depth of penetration, =1−cosθd/2, where θd is the snorkel wedge angle (Fig 6A⇑). Because depth of penetration will vary for each residue on the hydrophobic face, then ΛΣ(ΔGΨi×di), where ΔGΨi is the hydrophobicity of each residue on the hydrophobic face and di is the relative depth of penetration of each residue. As shown in Fig 6A⇑, di=[cosθi/2−cosθd/2]×r. Based on computer-generated models from the model peptide 18A,35 the maximum depth of penetration of a class A amphipathic helix into a phospholipid monolayer is ≈14 Å35 ; therefore, r=7 Å and di (expressed in Å) =di×7 Å=[cosθi/2−cosθd/2]×7 Å. From the neutron diffraction studies of Jacobs and White,34 a level 14 Å beneath the PC head group region of a monolayer water has a molar concentration <1% of that at the level of the PC head groups.
The angle of the hydrophobic face, θΔ, is used for calculation of ΣΔGΨ (total hydrophobicity of the nonpolar face). The numerical value of θΔ may not necessarily be the same as that of the snorkel wedge angle, θd (see below), that is used to calculate d. After trying a number of definitions for θΔ, some of which included hydrophobic residues on the polar face but did not account for the results of Table 2⇑, we selected ΣΔGΨ to represent the total hydrophobicity of all residues between the angle (θΔ) formed by the charged residues nearest to and located on opposite sides of the center of the hydrophobic face on a helical wheel projection.
We examined three ways of defining θd: (1) θd=θΔ, (2) θb=the Brasseur hydrophobic angle as defined elsewhere,36 and (3) θd(Ψ)=snorkeled θd=θΔ+20° or 40° for each Lys and Arg used to define θΔ, and 20° represents the angle between adjacent residues on a helical wheel projection. Neither θd=θΔ nor θb=Brasseur angle accounted for the results of Table 2⇑. A further consideration in defining θd(Ψ) was that Lys is flexible and presumably can snorkel well, whereas Arg is much less so and presumably can snorkel poorly, if at all.37 Perhaps this explains why class A2 amphipathic helixes have 5 Lys for every Arg, whereas class A1 helixes with a lower lipid affinity have 2 Arg for every Lys.38 On the basis of this argument, we developed two additional definitions for θd(Ψ). Because Lys residues in the snorkeled conformation can contribute to the hydrophobicity of the hydrophobic face and thus increase the hydrophobic face angle, we used 20° or 40° for Lys but not for Arg that appeared immediately after the hydrophobic face; ie, θd(Ψ)=θΔ+20° or 40° for each Lys used to define θΔ. We selected θd(Ψ)=θΔ+40° for each Lys used to define θΔ, because this definition was best able to account for the results of Table 2⇑. Therefore, di(Å)=[cosθi/2−cosθd(Ψ)/2]×7 Å (Fig 6B⇑).
Jacobs and White34 determined a water gradient from the head group position of a phospholipid bilayer to the acyl chains of the phospholipid. Using this information and plotting the free energy of transfer of an amino acid residue from water to varying water–organic solvent mixtures39 onto the water gradient from the outside to the inside of a phospholipid monolayer determined by Jacobs and White,34 we derived the free energy gradient δi(Å) as a function of depth of penetration in Å, di(Å). Therefore, Λ=Σ(ΔGΨi×δi)(Å).
Table 3⇑ shows the results of calculated values for ΣΔGΨ and Λ for the eight apo A-I 22-mers and for the two well-characterized model amphipathic helical peptides, 18A and 18R.40 Calculated Λ values for the eight apo A-I 22-mers are strongly correlated with measurements of ΔHT (r=.92, Fig 5B⇑). Combining the results for the eight apo A-I 22-mers with 18A and 18R (for which KP and πe values are known), the calculated Λ correlates very well with measured values for lnKP (r=.94) and fairly well with πe (r=.86) (Fig 5C⇑ and 5D⇑, respectively).
The rationale for synthesizing N- and C-terminal–protected 22-mer peptides is the fact that all of these are in the interior of the apo A-I sequence. Given the rather compelling experimental evidence that most of the apo A-I sequence is involved in lipid association,6 8 9 10 11 12 13 14 15 16 40 it is striking that only two of the eight amphipathic helical domains of this protein, apo A-I–44-65 and apo A-I–220-241, have demonstrated an affinity for lipid by all of our study methods. These results strongly support the notion that only the two terminal 22-mer domains of apo A-I (44-65 and 220-241) can directly interact with lipid as individual 22-mers, thus suggesting the likelihood that cooperative helix-helix interactions are required for the remaining six 22-mers to associate with lipid.
The DSC results indicate that the peptide with the highest lipid affinity, Ac–220-241–NH2 at a 1:20 peptide-to-lipid molar ratio, produced a decrease in ΔHT equaled by that of apo A-I at a 1:200 protein-to-lipid molar ratio (Fig 4⇑). Thus, apo A-I appears to be 10 times more effective than any single 22-mer, suggesting cooperation among the middle six 22-mer amphipathic helical domains of apo A-I.
Such cooperation is supported by earlier studies from our laboratory, which suggest that a tandem dimer of class A amphipathic helix connected by a Pro is more effective as a lipid-associating peptide than is one connected by an Ala or a tandem dimer with no connecting amino acid.19 41 Additional evidence for cooperation comes from studies of synthetic analogues of consensus sequences of the 22-mer tandem repeating sequences in apo A-I.42
The turbidity clarification studies indicate slow kinetics for apo A-I compared with that of its 22-mer constituent peptides with high lipid affinity. This is consistent with the assumption that apo A-I must undergo a major “cooperative” structural change when it binds to phospholipid.
On the basis of the present results and other data for apo A-I–phospholipid recombinant studies, we propose a simple two-step model for the association of apo A-I with phospholipid. In step 1, the two terminal 22-mer domains, 44-65 and 220-241 (helixes 1 and 8), associate with phospholipid (Fig 7A⇓). This association then “triggers” step 2 (Fig 7B⇓), in which the intervening 22-mer domains with little individual lipid affinity create a ladderlike, continuous unit with significantly increased lipid affinity via cooperative antiparallel amphipathic helix salt-bridge formation.16 43 44 This amphipathic protein sheet then joins the terminal amphipathic helixes in associating with phospholipid (Fig 7C⇓).
Because the two-step model in Fig 7⇑ is based on studies of the 22-mer fragments of apo A-I, the possible role of intramolecular interactions in stabilizing apo A-I–lipid interactions cannot be addressed by the experiments reported here. However, studies of larger apo A-I fragments show that only those that contain the 44-65 and 220-241 sequences associate with phospholipid, a finding that supports our model (V.K. Mishra et al, unpublished observations, 1995). Support for the two-step model is also provided by our studies of a deletion mutant of apo A-I in which the first 43 residues of the N-terminal segment were removed. The resulting Δ1-43–apo A-I has an affinity for lipid equal to that of intact plasma apo A-I.6
After the initial lipid association by amphipathic helical domains 1 and 8 and depending on the properties of the local protein-lipid microenvironment, the interaction among domains 2 through 7 (Fig 7⇑) can vary, thereby allowing HDL subspeciation.5 6 9 15 44 45 46 47 We propose that portions of the middle domains in apo A-I (the hinged domain45 47 ) interact either intramolecularly with other amphipathic helical domains of apo A-I (eg, the G* amphipathic helical domain 1-33 located at the N terminus of apo A-I) or intermolecularly with lipids and/or other apolipoprotein molecules, such as apo A-II amphipathic helical domains, thus allowing HDL subspeciation.47
Although each of the eight peptides studied has a well-defined amphipathic structure, only those models based on changes in effective amino acid residue hydrophobicity resulting from differing depths of lipid penetration were able to explain the high lipid affinity possessed by the two terminal domains. Use of the depth of penetration concept to derive the equation Λ=Σ(ΔGΨi×δi)(Å) provides a calculated lipid affinity that has an average correlation of r=.88 with the three methods for determining lipid affinity: lnKP, ΔHT, and monolayer πe. By comparison, two other calculated properties of the 22-mer peptides, total hydrophobicity and total hydrophobicity of the nonpolar face, have considerably lower average correlations with the same three methods for determining lipid affinity (r=.64 and .48, respectively).
From Table 3⇑, θd for five of the six middle 22-mer amphipathic helixes of apo A-I with little or no measurable lipid affinity ranges from 120° to 160° (corresponding to a maximum depth of penetration of 3.5 to 6 Å; Fig 6B⇑), whereas the two 22-mers with appreciable lipid affinity have a θd of 200° and 220° (corresponding to a maximum depth of penetration of 8 and 9.5 Å; Fig 6B⇑). (Note again that the hydrocarbon region starts at a depth of ≈7 Å,33 for a value of θd=180°.) It is interesting to note that peptide 187-208 has a large θd (240° and d=10.5 Å) but low total hydrophobicity of the nonpolar face (ΣΔGΨ=2.8 kcal/mol). The surface pressure experiments show that this peptide has a πe of 23 dyn/cm, the next highest value from that of the two peptides that can spontaneously solubilize DMPC. Peptide 187-208, however, does not spontaneously solubilize DMPC and has both a low lnKP and a high ΔHT.
All of the other exchangeable apolipoproteins with known high lipid affinities (apo A-II, C-II, and C-III) have amphipathic helixes with a θd that ranges from 200° to 360° (corresponding to a maximum depth of penetration of 8 to 14 Å; J.P. Segrest et al, unpublished observations, 1995). The calculated Λ for helix model 18A is considerably greater than that for apo A-I–44-65 and somewhat greater than that for apo A-I–220-241. These results are compatible with studies that have shown that apo A-I can be displaced from HDL by peptide 18A (G.M. Anantharamaiah et al, unpublished results, 1990).
Recent deletion mutagenesis studies have shown that the C-terminal region of apo A-I (residues 227-243) is critical in modulating its lipoprotein association.48 It has also been shown that mutation of a single Leu residue in helix 8 (Fig 6⇑) to an Asn (Leu240→Asn) in apo A-I drastically reduces its lipid-associating properties.49 This finding agrees with our present results, that of the two amphipathic helical domains that spontaneously associate with lipid, 220-241 has a higher lipid affinity and a mutation like Leu240→Asn is predicted to reduce its affinity for lipid. The lipid affinity for 220-241 (Leu240→Asn) is calculated to be 6.3, compared with a value of 11.9 kcal/mole for the native sequence (Table 3⇑).
Despite the strong predictive power of the equation Λ=Σ (ΔGΨi×δi)(Å) demonstrated in this article, additional improvements are possible. For example, the present algorithm is based on a two-dimensional helical wheel representation; extrapolation to three dimensions and ultimately to molecular models is contemplated. Furthermore, water penetration into a lipid bilayer is increased by tripeptide association34 and thus, δi(Å) will very likely be a function of the type of peptide-lipid interaction, eg, a tripeptide versus an amphipathic helix. We have begun preliminary neutron diffraction studies to address the latter question.
In summary, in this report we have determined the amphipathic helical domains of apo A-I responsible for its strong lipid association and suggest that models of protein-lipid interactions based on variations in the depth of peptide penetration into lipid best account for the variations in lipid affinity among the different amphipathic helical 22-mers. Furthermore, we propose that the two strong lipid-associating terminal amphipathic helical domains (1 and 8) are involved in the initial binding of apo A-I to the HDL surface to form HDL particles, followed by cooperative binding of the middle six weakly lipid-associating amphipathic helical domains, which may be aided by antiparallel helix-helix salt-bridge formation. Finally, we hypothesize that microenvironment-driven variations in the helix-helix interactions of amphipathic helixes 2 through 7 of apo A-I are important in the creation of HDL subspecies.47
Selected Abbreviations and Acronyms
|DSC||=||differential scanning calorimetry|
|egg PC||=||egg yolk phosphatidylcholine|
|HPLC||=||high-performance liquid chromatography|
|K p||=||partition coefficient|
|ΔHT||=||enthalpy of the gel-to–liquid crystalline phase transition|
|π(i)||=||(initial) surface pressure|
This research was supported in part by National Institutes of Health (Bethesda, Md) Program Project grants HL34343 and HL 22633. We thank Dr Donald D. Muccio for allowing us to use the spectropolarimeter. Mass spectral analyses were performed at the Cancer Core Facility. We thank Dr Steve Barnes and Marian Kirk for their help in the mass spectral analysis and Sheila Benowitz for technical assistance.
Chajek-Shaul T, Hayek T, Walsh A, Breslow JL. Expression of the human apolipoprotein A-I in transgenic mice alters high density lipoprotein (HDL) particle size distribution and diminishes selective uptake of HDL cholesterol esters. Proc Natl Acad Sci U S A. 1991;88:6731-6735.
Warden C, Hedrick CC, Qiao J-H, Castellani LW, Lusis AJ. Atherosclerosis in transgenic mice overexpressing apolipoprotein A-II. Science. 1993;261:469-471.
Ji Y, Jonas A. Properties of an N-terminal proteolytic fragment of apolipoprotein A-I in solution and reconstituted high density lipoprotein. J Biol Chem. 1995;270:11290-11297.
Banka CL, Bonnet DJ, Black AS, Smith RS, Curtis LK. Localization of an apolipoprotein A-I epitope critical for activation of lecithin:cholesterol acyltransferase. J Biol Chem. 1991;266:23886-23892.
Collet X, Perret B, Simard G, Raffai E, Marcel YL. Differential effects of cholesterol in the immunoreactivities and conformation of apolipoprotein A-I in high density lipoproteins. J Biol Chem. 1991;266:9145-9152.
Sparks DL, Phillips MC, Lund-Katz S. The conformation of apolipoprotein A-I in discoidal and spherical recombinant high density lipoprotein particles. J Biol Chem. 1992;267:25830-25838.
Marcel YL, Provost PR, Koa H, Raffai E, Dac NV, Fruchart JC, Rassart E. The epitopes of apolipoprotein A-I define distinct structural domains including a mobile middle region. J Biol Chem. 1991;266:3644-3656.
Ponsin G, Hester L, Gotto AM Jr, Pownall H, Sparrow JT. Lipid:peptide association and activation of lecithin:cholesterol acyltransferase: effect of α helicity. J Biol Chem. 1986;261:9202-9205.
Anantharamaiah GM, Jones JL, Brouillette CG, Schmidt CF, Chung BH, Hughes TA, Bhown AS, Segrest JP. Studies of synthetic peptide analogs of the amphipathic helix. J Biol Chem. 1985;260:10248-10262.
Kaiser E, Colescott RL, Bossinger CD, Cook PI. Color test for the detection of free terminal amino groups in the solid phase synthesis of peptides. Anal Chem. 1970;34:595-598.
Mishra VK, Palgunachari MN, Segrest JP, Anantharamaiah GM. Interaction of synthetic peptide analogs of the class A amphipathic helix with lipids: evidence for snorkel hypothesis. J Biol Chem. 1994;269:7185-7191.
Spuhler P, Anantharamaiah GM, Segrest JP, Seelig J. Binding of apolipoprotein A-I model peptides to lipid bilayers: measurement of binding isotherms and peptide-lipid head group interactions. J Biol Chem. 1994;269:23904-23910.
Anantharamaiah GM, Hughes TA, Iqbal M, Gawish A, Neame PJ, Meadley MF, Segrest JP. Effect of oxidation on the properties of apolipoproteins A-I and A-II. J Lipid Res. 1988;29:309-318.
Hughes TA, Moore MA, Neame P, Meadley MF, Chung BH. Rapid quantitative analysis of apolipoproteins by gradient ultracentrifugation and reversed-phase high performance liquid chromatography. J Lipid Res. 1988;229:363-376.
Owens RJ, Anantharamaiah GM, Kalhone JB, Srinivas RV, Compans RW, Segrest JP. Apolipoprotein A-I and its amphipathic helical peptide analogs inhibit human immunodeficiency virus-induced syncytium formation. J Clin Invest. 1990;86:1142-1150.
Tytler EM, Segrest JP, Epand RM, Nie S-Q, Epand RF, Mishra VK, Venkatachalapathi YV, Anantharamaiah GM. Reciprocal effects of apolipoprotein and lytic peptide analogs on membrane: cross-sectional molecular shapes of amphipathic α helices control membrane stability. J Biol Chem. 1993;268:22112-22118.
Brasseur R. Differentiation of lipid-associating helices by the use of three dimensional molecular hydrophobicity potential calculations. J Biol Chem. 1991;266:16120-16127.
Richardson JS, Richardson DC. In: Fasman GD, ed. Prediction of Protein Structure and the Principles of Protein Conformation. New York, NY: Plenum Press; 1989:1-98.
Segrest JP, Jones MK, de Loof H, Brouillette CG, Venkatachalapathi YV, Anantharamaiah GM. The amphipathic helix in exchangeable apolipoproteins: a review of secondary structure and function. J Lipid Res. 1992;33:141-166.
Cohn EJ, Edsall JT. Amino acids and peptides as ions and dipolar ions. In: Proteins, Amino Acids and Peptides. New York, NY: Reinhold Publishing Co; 1943:196-216.
Lins L, Rosseneu M, Ruysschaert JM, Brausseur R. Mode of assembly of apolipoproteins with lipids: structural determinants of the complexes. In: Rosseneu M, ed. Structure and Function of Apolipoproteins. Boca Raton, Fla: CRC Press; 1992:251-268.
Mishra VK, Palgunachari MN, Lund-Katz S, Phillips MC, Segrest JP, Anantharamaiah GM. Effect of the arrangement of tandem repeating units of class A amphipathic helixes on lipid interaction. J Biol Chem. 1995;270:1602-1611.
Anantharamaiah GM, Venkatachalapathi YV, Brouillette CG, Segrest JP. Use of synthetic peptide analogues to localize lecithin:cholesterol acyltransferase activating domain in apolipoprotein A-I. Arteriosclerosis. 1990;10:95-105.
Rosseneu M, Vanloo B, Lins L, DePawn M, Ruysschaert J-M, Brasseur R. Analysis of the primary and of the secondary structure of the apolipoproteins. In: Rosseneu M, ed. Structure and Function of Apolipoproteins. Boca Raton, Fla: CRC Press; 1992:159-183.
Dalton MB, Swaney JB. Structural and functional domains of apolipoprotein A-I within high density lipoproteins. J Biol Chem. 1993;268:19274-19283.
Chung MC, Segrest JP, Albers JJ, Cone JT, Brouillette CG, Chung BH, Kashyap MA, Glasscock MA, Anantharamaiah GM. Characterization of high density lipoprotein subspecies: structural studies by single vertical spin ultracentrifugation and immunoaffinity chromatography. J Lipid Res. 1987;28:913-929.
Schmidt HH-J, Remaley AT, Stonik JA, Ronan R, Wellmann A, Thomas F, Zech LA, Brewer HB Jr, Hoeg JM. Carboxyl-terminal domain truncation alters apolipoprotein A-I in vivo catabolism. J Biol Chem. 1995;270:5469-5475.
Hiebsch RR, McGuire KA, Jonas A, Wattenberg BW. A glycosylated apolipoprotein A-I displays alterations in lipoprotein formation: use of glycosylated form to follow intracellular trafficking of apolipoprotein A-I. Personal Communication. 1995.