Examination of Lipid-bound Conformation of Apolipoprotein E4 by Pyrene Excimer Fluorescence*

Apolipoprotein E (apoE) is a 34-kDa resident of lipoproteins that plays a key role in cholesterol homeostasis in plasma and in brain. It is composed of an N-terminal (NT) domain (residues 1–191) and a C-terminal (CT) domain (residues 201–299). Of the three major isoforms (apoE2, -E3, and -E4), apoE4 is considered a risk factor for both cardiovascular and Alzheimer disease. Compared with apoE3, domain interaction between NT and CT domains is believed to direct the lipoprotein distribution preference of apoE4 for very low density lipoprotein-sized particles. We examined the relative disposition of apoE4 NT and CT domains in lipid-free and lipid-bound forms by monitoring pyrene excimer fluorescence emission as a direct indicator of spatial proximity. Site-specific labeling of apoE4 by N-(1-pyrene)maleimide was accomplished after substitution of Cys residues for Arg-61 in NT domain and Glu-255 in CT domain. Pyrene labeling did not alter the lipoprotein distribution pattern of apoE4 in plasma. Pyrene excimer fluorescence was noted in lipid-free pyrene-R61C/E255C/apoE4 in mixtures containing excess wild-type apoE4, which was attributed to intramolecular spatial proximity between these specified sites. Upon disruption of tertiary interaction, a large decrease in excimer fluorescence emission was noted in pyrene-R61C/E255C/apoE4. In dimyristoylphosphatidylcholine/pyrene-R61C/E255C/apoE4 discoidal complexes, pyrene excimer fluorescence emission was retained. Taken together with fluorescence quenching and cross-linking analysis, a looped-back model of apoE4 is proposed in lipid-bound state, including spherical lipoprotein particles, wherein residues Arg-61 and Glu-255 are proximal to one another.

lipoproteins from blood (5,6). In addition, apoE in macrophages has a direct antiatherogenic effect independent of alterations in plasma lipoproteins (7) by promoting cholesterol efflux (8) from cell lining of arteries by a process termed reverse cholesterol transport (9 -11). The role of apoE in cholesterol homeostasis in the brain is emerging with evidence of its protective role against neuronal injury and neurodegeneration (12).
In humans, apoE displays polymorphism with three major genetic variants identified in the population, apo⑀2, -⑀3, and -⑀4, with allelic frequencies of 8%, 77%, and 14%, respectively, determining six apoE phenotypes, apoE2/E2, apoE3/E3, apoE4/ E4, apoE2/E3, apoE2/E4, and apoE3/E4 (13). The apoE isoforms determine the atherogenic fate of the lipoproteins on which they reside. Although the apoE2 isoform is associated with type III hyperlipoproteinemia, peripheral atherosclerosis, and accumulation of remnant lipoproteins, apoE4 is consistently associated with higher levels of low density lipoprotein (LDL) and cholesterol in plasma, and persons bearing the apoE4 isoform are more prone to develop atherosclerosis than those with the apoE3 isoform, which is generally considered antiatherogenic (5,14). In addition, the apo⑀4 allele is strongly linked to both sporadic and familial late-onset Alzheimer disease, although the precise mechanistic details of this association are unclear (15,16).
Understanding the structural basis of apoE isoforms provides insights into its physiological role in lipoprotein metabolism. ApoE is a 34-kDa, 299-residue protein composed of a 22-kDa N-terminal (NT) domain (residues 1-191) folded as a four-helix bundle (17) and a 10-kDa C-terminal (CT) domain (residues 201-299) that promotes apoE oligomerization (18 -22). The two domains are linked via a protease-sensitive linker loop region. The three isoforms differ at positions 112 and 158: apoE2 bears cysteines and apoE4 bears arginines at both sites, whereas apoE3 contains a cysteine and an arginine at positions 112 and 158, respectively. Thus, the differences between the three isoforms lie in the NT domain, which houses receptorbinding sites (residues 130 -150) for the LDL receptor family. Whereas apoE3 and -E4 display identical receptor binding activities, apoE2 seems to bear 1-60% activity compared with apoE3 (13). In turn, subtle differences exist in the lipoprotein binding preference of apoE3 and -E4; apoE3 displays a preference for high density lipoprotein (HDL)-sized particles, whereas apoE4 displays a preference for the larger very low density lipoprotein (VLDL)-sized particles compared with apoE3 (23)(24)(25)(26)(27).
Employing a combination of x-ray analysis of isolated apoE3 and E4 NT domains, site-directed mutagenesis, and plasma lipoprotein binding analysis of intact apoE, the presence of Arg at position 112 in apoE4 has been suggested to lead to salt bridge formation with Glu-109, which in turn displaces Arg-61 to potentially form a salt bridge with Glu-255 in the CT domain. ApoE3, however, bearing a Cys at position 112, accom-modates Arg-61 in a locale where it is unable to form a salt bridge with Glu-255. The salt bridge between Arg-61 and Glu-255 has been postulated to contribute to domain-domain interaction, which in turn directs the lipoprotein binding preference of apoE4 for VLDL (23,24). Indeed, replacement of Arg-61 in apoE4 with threonine (as noted in mouse apoE) shifts the lipoprotein binding preference from VLDL to HDL (28).
In the present study, we examined the relative disposition between apoE4 NT and CT domains in lipid-free and bound states, taking advantage of the unique fluorescence properties of N-(1-pyrene)maleimide as a probe of spatial proximity. We monitored the location of Arg-61 and Glu-255 with respect to each other in different environments and concluded that they remained proximal to each other after lipid interaction.
Site-directed Mutagenesis and Protein Expression-Cysteine residues were introduced into human apoE4 by site-directed mutagenesis as described previously, and the DNA sequences were verified (29). Arg-61 and Glu-255 were replaced by Cys to create a double-cysteinecontaining apoE4 variant R61C/E255C/apoE4. Recombinant human wild-type (WT) and R61C/E255C/apoE4 were expressed in Escherichia coli using the pTYB2 vector and purified using the IMPACT-CN system (New England BioLabs, Beverly, MA) as described earlier (30). The purified protein was present in 25 mM Tris-HCl, pH 7.4, containing 140 mM NaCl and 3 mM KCl (TBS), unless otherwise specified.
Pyrene Modification-R61C/E255C/apoE4 was initially incubated with 5-fold molar excess tris(2-cyanoethyl)-phosphine for 2 h at 37°C to maintain the sulfhydryl groups in a reduced state, followed by incubation with a 5-fold molar excess of N-(1-pyrene)maleimide at 37°C for 16 h (31). Unbound N-(1-pyrene)maleimide was removed by gel filtration chromatography. The stoichiometry of labeling was calculated to be 1.9 pyrene/apoE4 molecule in the pyrene-labeled variant (pyr-R61C/ E255C/apoE4), using the extinction coefficient for pyrene ⑀ 340 ϭ 40,000 M Ϫ1 cm Ϫ1 . Circular dichroism spectroscopy of the modified variant revealed that the labeling did not affect the overall secondary structural characteristics of apoE4.
Fluorescence Measurements-Fluorescence spectra were recorded on a Perkin-Elmer MPF-44B spectrofluorometer. A slit width of 5 nm was used for both excitation and emission monochromators. Emission spectra were collected from 350 -600 nm by setting the excitation wavelength at 340 nm; in some cases, excitation spectra were recorded between 250 and 350 nm by setting the emission wavelength at 375 nm. For all calculations, the area under the monomer emission peak was calculated from 370 to 410 nm, whereas the area under the excimer emission peak was calculated from 440 to 510 nm. To monitor the status of the excimer emission peak, data were normalized with respect to fluorescence emission intensity at 375 nm. Unless otherwise mentioned, all experiments were carried out with pyr-R61C/E255C/apoE4 mixed with a 10-fold molar excess wild-type apoE4 under denaturing conditions followed by extensive dialysis against 25 mM TBS (henceforth referred to as pyr-R61C/E255C/apoE4). This ensured that each labeled apoE4 was typically surrounded by unlabeled protein in lipid-free state or in the context of lipid particles.
Determination of Lipoprotein Binding Preference of pyr-R61C/E255C/ apoE4 -To verify that modified apoE4 still retained lipoprotein binding preference as reported by Weisgraber and colleagues (23,24), pyr-R61C/E255C/apoE4 (900 g of total protein) was incubated with 150 l of human plasma at 37°C for 2 h. The lipoprotein fractions were separated by size exclusion chromatography on a Superose 6 PC 3.2/30 column equilibrated with TBS using a fast performance liquid chromatography system (Amersham Biosciences) at a flow rate of 0.1 ml/min, with 50-l fractions. Each fraction was assayed for cholesterol (Bayer Health Care, Tarrytown, NY) and triglyceride (Sigma-Aldrich) content to confirm the nature of lipoprotein. In addition, fluorescence analysis of each fraction was carried out to determine the location of pyr-E255C/R61C/apoE4.
Effect of Trifluoroethanol, pH Alteration, and Micelle Formation on pyr-R61C/E255C/apoE4 Fluorescence-Fluorescence emission spectra of 12 g of pyr-R61C/E255C/apoE4 in the absence or presence of varying concentrations of trifluoroethanol (TFE) were recorded, after incubation of protein at room temperature for 16 h. The final TFE concentrations ranged from 0% to 50% (v/v) in the incubation mixtures. The effect of pH was evaluated by equilibrating 12 g of pyr-R61C/E255C/apoE4 with 50 mM sodium phosphate (pH 6.0, 7.0, and 8.0) and 50 mM sodium acetate (pH 3.0, 4.0, and 5.0) for 16 h, followed by fluorescence analysis. The effect of micelle formation was followed by incubating 12 g of pyr-R61C/E255C/ apoE4 with 0.4% lysophosphatidylcholine (lyso-PC).
Preparation and Characterization of DMPC/pyr-R61C/E255C/apoE4 Complexes-Discoidal complexes were prepared as described previously (30). Before disc formation, pyr-R61C/E255C/apoE4 was mixed with a 10-fold molar excess of WT apoE4 to enhance the probability of the presence of one pyr-R61C/E255C/apoE4 molecule surrounded by several WT apoE4 on a lipoprotein complex. Discoidal complexes were separated from unbound protein by density gradient ultracentrifugation, followed by extensive dialysis. The phospholipid and protein content were estimated using the phospholipids assay kit (Wako Chemicals GmbH, Neuss, Germany) and the bicinchoninic acid kit (Pierce Biotechnology), respectively, to calculate lipid/protein ratio of the reconstituted lipoprotein particles. Non-denaturing PAGE of the isolated lipoprotein complexes was carried out to evaluate the molecular mass and size of the particle on a 4 -20% gradient gel for 20 h at 150 V and stained with Amido Black. The particle sizes were calculated from a calibration curve using the following standards and their corresponding Stokes diameters: thyroglobulin, 17 nm; ferritin, 12.2 nm; catalase, 9.2 nm; and lactate dehydrogenase, 8.2 nm. For electron microscopic analysis, DMPC/pyr-R61C/E255C/apoE4 complexes were dialyzed against ammonium acetate buffer followed by negative staining with 2% sodium phosphotungstate as described previously (Zeiss 10; 80 kV) (32).
Cross-linking Studies-Cross-linking of DMPC/pyr-R61C/E255C/ apoE4 (10 g of protein) was carried out at DMS concentrations of 0.46, 4.6, and 23 mM for 2 h at 24°C (33,34). DMS was prepared in 1 M triethanolamine HCl, pH 9.7. The volume of 1 M triethanolamine HCl was maintained at one tenth of the final volume of the reaction mixture. DMS is a non-cleavable homobifunctional, amine-specific cross-linker with a spacer arm of 11.4-Å. The reaction was stopped by the addition of SDS-PAGE sample treatment buffer followed by electrophoresis on 4 -20% acrylamide gradient gel.
Fluorescence Emission Analysis of Lipid-associated pyr-R61C/ E255C/apoE4 -Fluorescence emission analysis of lipoproteinassociated pyr-R61C/E255C/apoE4 was carried out to evaluate lipidtriggered spatial repositioning of apoE4 domains on reconstituted discoidal lipoprotein particles and on large and small spherical lipoprotein particles. Fluorescence emission spectrum of DMPC/pyr-R61C/E255C/apoE4 complexes (12 g of protein) was recorded after excitation at 340 nm. Pyr-R61C/E255C/apoE4 was incubated with plasma and the various lipoprotein classes were separated by gel filtration as described above. Fluorescence emission spectra of VLDL-and HDLassociated pyr-R61C/E255C/apoE4 were recorded (fractions 7 and 21, respectively). Spectra of corresponding lipoprotein fractions collected from plasma incubations without any added pyr-R61C/E255C/apoE4 were subtracted to account for background fluorescence contribution.
Quenching Studies-Quenching of fluorescence emission of lipid-free pyr-R61C/E255C/apoE4 and DMPC/pyr-R61C/E255C/apoE4 complexes was carried out as described earlier (31,35) using aqueous quenchers such as potassium iodide (KI) to evaluate the degree of solvent exposure of the fluorophores. Increasing concentrations of KI were added in 50 mM potassium phosphate, pH 7.4, with 1 mM sodium thiosulfate to suppress free iodine formation. To assess the depth of location of the fluorophores with respect to the phospholipid bilayer in DMPC/pyr-R61C/E255C/apoE4 discoidal complexes, 5-DOXYL stearic acid (5-DSA) or 12-DOXYL stearic acid (12-DSA) were employed, where the DOXYL group (quenching moiety) is located at different depths along the fatty acyl chain. Aliquots of 5-DSA or 12-DSA (1.3 mM stock in ethanol) were added directly to DMPC/ pyr-R61C/E255C/apoE4 (keeping final concentration of ethanol Յ1% (v/ v)), and fluorescence intensities were measured at 375 nm after equilibration for 5 min. Effective quenching constants were calculated employing the Stern-Volmer equation, F 0 /F ϭ 1 ϩ K SV [Q], where F 0 and F are fluorescence intensities in the absence and the presence of varying quencher concentrations, respectively, K SV is the Stern-Volmer quenching constant, and [Q] is the quencher concentration (36). Quenching by KI and DSA cannot be compared directly by the classic Stern-Volmer equation because, in the case of DSA, the quencher is limited to the context of the bilayer and is not freely diffusing in solution. Therefore, apparent quenching constants were calculated for purposes of comparison between 5-DSA and 12-DSA.

RESULTS AND DISCUSSION
Pyrene Excimer Fluorescence: An Indicator of Spatial Proximity-When attached specifically to a single free sulfhydryl group, pyrene typically exhibits fluorescence emission maxima at 375 and 395 nm (excitation at 340 nm), attributed to a monomeric moiety. However, when there is a second bound pyrene within 10 Å in spatial proximity, it displays an additional broad and red-shifted fluorescence emission peak (ϳ460 nm), attributed to formation of an excited state dimer or 'excimer' (37). We exploit this unique spectral feature of pyrene to assess lipid-triggered repositioning of two specific sites, one in NT and the other in CT domain of apoE4, in an approach successfully employed to evaluate spatial proximity in other proteins (37)(38)(39)(40) including apolipoproteins such as apolipophorin III (31,41). Furthermore, when two pyrene molecules giving rise to excimer fluorescence move away from each other, the excimer fluorescence emission intensity is decreased or lost entirely. We employed site-directed mutagenesis to substitute unique cysteines in WT apoE4 (lacks endogenous cysteine residues) at positions 61 and 255, located in the NT and CT domains, respectively, yielding R61C/E255C/apoE4. R61C/ E255C/apoE4 was labeled with N-(1-pyrene)maleimide followed by secondary structural analysis using circular dichroism spectroscopy, which indicated that the global fold of the labeled variant was similar to that of WT apoE4 (data not shown).
Pyrene Excimer Fluorescence in Lipid-free pyr-R61C/E255C/ apoE4 -Fluorescence emission spectrum of pyr-R61C/E255C/ apoE4 alone (in the absence of excess WT apoE4) in aqueous buffer (Fig. 1) reveals the presence of a strong unstructured excimer peak centered at ϳ460 nm, in addition to the characteristic emission at 375, 385, and 395 nm attributed to vibronic transitions. Because excimer formation is an excited state event, the emission spectrum of pyr-R61C/E255C/apoE4 is not a mirror image of its excitation spectrum (data not shown) (42). Excimer emission formation is indicative of spatial proximity between positions 61 and 255 in lipid-free apoE4. However, apoE has been indicated to exist in solution as a tetramer by protein-protein interactions via its CT domain (18,21,22,43), leading us to examine whether the excimer emission originates from intermolecular spatial proximity between these specified sites. Pyr-R61C/E255C/apoE4 was mixed with a 10-fold molar excess of WT apoE4 under denaturing conditions, followed by slow refolding as described under "Experimental Procedures." This increases the likelihood that a given molecule of pyr-R61C/E255C/apoE4 is surrounded by unlabeled WT apoE4 in the native tetrameric organization. Fluorescence measurements reveal that excimer fluorescence emission was retained in these mixtures, indicative of intramolecular rather than intermolecular spatial proximity between pyrene moieties residing on R61C and E255C in apoE4.
Lipoprotein Binding Behavior of pyr-R61C/E255C/apoE4 -To ascertain that modification of apoE4 did not alter its lipoproteinbinding behavior, pyr-R61C/E255C/apoE4 was incubated with human plasma, followed by examination of its distribution profile among the various lipoprotein fractions by size-exclusion chromatography. The location of pyr-R61C/E255C/apoE4 was monitored by fluorescence analysis of each fraction (Fig. 2). Taking the sum of fluorescence intensities of all lipoprotein-bound fractions as 100%, the percentage distributions of pyr-R61C/E255C/apoE4 among the various lipoprotein fractions were calculated to be as follows: 38% in VLDL (fractions 5-9); 25% in intermediate density lipoprotein/LDL (fractions 11-15); 37% in HDL (fractions [17][18][19][20][21]. This distribution profile is largely consistent with that reported previously (24), indicating that the distinctive binding preference of apoE4 is retained in pyr-R61C/E255C/apoE4. A minor difference observed in the present case is the larger percentage of unbound pyr-R61C/E255C/apoE4 that is probably a result of the excess apoE/total plasma protein ratio used under our experimental conditions.

Effect of Alteration in Tertiary Contacts on Pyrene Excimer
Formation-To validate the use of pyrene excimer fluorescence as a sensitive monitor of spatial proximity in apoE4, we employed three independent tools with predictable features: effect of TFE, effect of altering pH, and effect of micelle formation with lyso-PC on pyrene excimer fluorescence. TFE is a solvent that destabilizes tertiary interactions in proteins while stabilizing and increasing ␣-helical structures. The effect of increasing TFE concentration on excimer fluorescence in pyr-R61C/ E255C/apoE4 was evaluated (Fig. 3A). A gradual decrease in excimer fluorescence intensity was noted with incremental increase in TFE (0 -50% v/v), with maximal decrease noted at 50% TFE. The decrease in excimer emission intensity in 50% TFE was accompanied by a decrease in the fluorescence intensity ratio of 395 nm (band III) to 375 nm (band I), indicative of an increase in local environment polarity (44 -46). We conclude that the decrease in excimer formation is the result of disruption of tertiary interaction between specified sites on NT and CT domains, leading to movement of R61C away from E255C. Next, the effect of altering pH on excimer fluorescence in pyr-R61C/E255C/apoE4 was examined (Fig. 3B). No alteration in excimer fluorescence occurred at pH 8.0 and pH 6.0, compared with that noted at pH 7.0. However, a substantial increase in excimer was observed at pH 5.0, followed by a trend of decrease at pH 4.0 and pH 3.0. The increase in excimer fluorescence emission may be attributed to: 1) movement of the pyrene moieties closer with respect to each other in a given population of pyr-R61C/E255C/apoE4 (pI 5.6 for R61C/E255C/apoE4) or 2) an increase in the fraction of apoE4 variant bearing proximal pyrene moieties. A similar movement of the two domains toward each other around pH 5 was also noted in earlier studies for apoE3 (pI 5.5), wherein sites Cys-112 and Trp-264 at the NT and CT domains, respectively, repositioned closer by 5 Å (30). It is conceivable that gradual protonation of the negatively charged residues draws the two domains toward each other (47,48). Finally, the effect of micelle formation on excimer fluorescence in pyr-R61C/E255C/apoE4 was examined (Fig. 3C). In the presence of 0.4% lyso-PC, a loss in excimer fluorescence was noted, consistent with movement of the two domains away from each other on a micellar surface. Taken together, these results validate the use of pyrene excimer fluorescence to monitor alterations in tertiary contacts in apoE4.
Effect of Lipid Interaction of pyr-R61C/E255C/apoE4 on Pyrene Excimer Formation-The relative spatial positioning of R61C and E255C in pyr-R61C/E255C/apoE4 was examined on three different lipid surfaces: reconstituted discoidal particles, plasma HDL, and VLDL particles. Because several apoE molecules are expected to be present on lipid particles, the presence of excess WT apoE4 reduced the likelihood of intermolecular excimer fluorescence formation. In the first case, DMPC-pyr-R61C/E255C/apoE4 particles were prepared as described previously (30), yielding complexes with a molecular mass of approximately 800,000 Da and Stokes diameter of ϳ19 nm as assessed by non-denaturing gel electrophoresis (data not shown). The lipid/protein molar ratio was calculated to be ϳ250:1 based on phospholipid and protein determination of the discoidal particles. Electron microscopic analysis revealed formation of discoidal particles (data not shown) with a diameter consistent with that obtained by gel electrophoresis. Finally, cross-linking of pyr-R61C/E255C/apoE4 was performed after formation of DMPC discoidal particles followed by SDS-PAGE analysis (Fig. 4). With increasing concentration of DMS, bands corresponding to a molecular mass of 68 kDa and higher appeared. This provides evidence of intermolecular cross-linking, suggesting the presence of approximately four molecules of apoE per lipoprotein particle. In addition, bands corresponding to monomeric apoE4 were present at all concentrations of DMS used, possibly representing intramolecular cross-linked apoE and apoE bearing covalently attached DMS.
Fluorescence emission analysis of DMPC-pyr-R61C/E255C/ apoE4 particles indicated that excimer fluorescence was retained (Fig. 5). We conclude that excimer fluorescence arose from intramolecular proximity between R61C and E255C in the context of reconstituted lipoprotein particles. In recent investigations, several lines of evidence indicated formation of discoidal phospholipid bilayer complexes of DMPC/apoE3, with the ␣-helices circumscribing the periphery of the particles (30,49,50). Furthermore, examination of VLDL-and HDL-bound pyr-R61C/E255C/apoE4 revealed excimer fluorescence (Fig. 6), consistent with spatial proximity between R61C and E255C in the context of spherical particles.
Fluorescence Quenching-To assess the relative location of pyrene fluorophores with respect to the lipid in representative DMPC/pyr-R61C/E255C/apoE4 complexes, two different types of fluorescence quenchers were employed: water-soluble and lipid-based quenchers. KI has been routinely employed as an aqueous collisional quencher that yields information about the microenvironment of the fluorophores. K SV was calculated to be 4.4 Ϯ 0.5 M Ϫ1 and 22.0 Ϯ 3.0 M Ϫ1 for KI quenching of lipid-free pyr-R61C/E255C/apoE4 and DMPC/pyr-R61C/E255C/apoE4, respectively. The 5-fold increase in K SV upon lipid binding indicates a dramatic alteration in the microenvironment of the fluorophores. The ability of KI to quench pyrene fluorescence emission in DMPC/pyr-R61C/E255C/apoE4 suggests localization of the pyrene moieties toward the lipid/water interface of the phospholipid bilayer. This approach was complemented with the second class of quenchers, spin-labeled fatty acids, which are lipid-based collisional quenchers. 5-DSA and 12-DSA act as molecular rulers, enabling assessment of the depth of location of the fluorophores with respect to the phospholipid bilayer (51). The apparent K SV values for 5-DSA and 12-DSA were calculated to be 20.2 Ϯ 5.8 ϫ 10 Ϫ3 M Ϫ1 and 2.1 Ϯ 0.2 ϫ 10 Ϫ3 M Ϫ1 , respectively. The 10-fold higher K SV for 5-DSA compared with 12-DSA indicates that the pyrene fluorophores are in a superficial location with respect to the membrane bilayer.
Taken together, we examined whether spatial proximity between positions 61 and 255 is retained upon interaction with a lipid surface by monitoring the signature spectral feature of pyrene. An apoE4 variant bearing Cys at these positions was constructed, which enabled us to selectively label with pyrene moiety employing the maleimide functional group. Secondary structural characterization indicated that the labeled apoE4 variant retained the overall fold of WT apoE4.
The Cys-to-Arg switch at position 112 in apoE4 compared with apoE3 has been proposed to direct its lipoprotein binding preference to VLDL-sized particles (23,24), with the positive charge at this position being an important factor (27) in forming a salt bridge with Glu-109 (23). The salt bridge between positions 109 and 112 in turn displaces Arg-61 from a neighboring helix, thereby making it available for salt bridge interaction with Glu-255 in the CT domain. This seems to be a unique case of salt bridge interaction because most salt bridges found in proteins (52), including those in apoE NT domain helix bundle (17), are formed between sequentially close residues. Human apoE is unique in that it bears Arg at position 61, whereas most other species have Thr at this position. Substitution of Arg-61 with Thr altered the lipoprotein preference of apoE4 from VLDL to HDL. Thus, the combination of positive charge at position 112 with Arg at position 61 seems to be responsible for mediating NT domain-CT domain interaction and the lipoprotein binding preference of apoE4. Therefore, in the present study, it was important for us to establish that substitution of Arg-61 and Glu-255 by Cys followed by covalent attachment with pyrene residues did not alter the lipoprotein- Lipid-free apoE4 is depicted as a monomer for the sake of simplicity, with the CT domain modeled as a series of helical structures (gray cylinders) linked to the NT domain (ribbon structure) (Dong et al., 1994) via a protease-sensitive loop (A). The lipid-bound conformation is probably valid for both discoidal HDL particles and spherical VLDL or HDL particles. The entire apoE molecule is depicted as a series of helical structures (gray cylinders) in lipid-bound state (B). In the case of discoidal particles, the helices are proposed to circumscribe a bilayer of phospholipid molecules with the helical axis oriented perpendicular to the plane of the bilayer. In the case of spherical particles, it is envisaged that the helices are embedded at the lipid/water interface near the phospholipid head groups as indicated previously for other exchangeable apolipoproteins (Sahoo et al., 1998). binding behavior of apoE4. Our studies indicate that we have effectively placed spectroscopic probes and retained NT domain-CT domain interaction in pyr-R61C/E255C/apoE4. We believe that we have maintained domain-domain interaction by pyrene-pyrene stacking interaction, as evidenced by the control experiment, which shows that the lipoprotein binding preference remains unaltered. The stabilization provided by pyrene excimer (dimer) formation is comparable with that contributed by surface or buried salt bridge (53,54). In other studies, an increase in stability was noted when well-packed hydrophobic residues replaced a salt bridge triad (55).
In summary, we observe that in the lipid-free state, positions Arg-61 and Glu-255, representing the NT and CT domains, respectively, are proximal to each other in apoE4, as proposed earlier (23,24). Upon interaction with phospholipid vesicles to form discoidal complexes, we propose that Arg-61 and Glu-255 lie spatially proximal to each other across the width of the phospholipid bilayer, such that each site is near the lipid/water interface (Fig. 7). In previous studies, it was suggested that apoE3 undergoes a dramatic conformational alteration upon lipid interaction, involving movement of the two domains away from each other (30,56), accompanied by further opening of the helix bundle (49,57) to yield a fully extended molecule circumscribing the periphery of discoidal bilayer of phospholipids (50). In the case of apoE4, we propose that a similar movement of the two domains about the hinge region occurs; however, spatial proximity between Arg-61 and Glu-255 is re-established upon lipid interaction, yielding a looped-back arrangement of helices. Based on an average particle diameter of 190 Å, average particle molecular mass of 800,000 Da and bilayer thickness of 40 Å, geometric calculations allow the presence of three or four apoE4 molecules in this configuration per discoidal particle. Taken together with native-PAGE, cross-linking, and fluorescence data, we postulate a looped-back conformation of apoE helices on discoidal particles. Proximity between the proposed sites seems to be maintained on spherical lipoprotein particles as well, whether VLDL or HDL, although in other studies, apoE was postulated not to accumulate on a lipoprotein particle (58). It is interesting that the lipid-bound forms of both isoforms display LDL receptor binding activity (13).
Despite the preference of apoE4 for VLDL in plasma, it was essential to examine apoE4 conformation on HDL-sized particles (discoidal and spherical), given the relevance of HDL in the brain, where no VLDL-sized lipoprotein particles have been detected (59). Furthermore, cholesterol delivery and synthesis are tightly regulated in an apoE-dependent manner, suggesting that lipid transport dysfunction in persons bearing the apoE4 isoform may be associated with Alzheimer disease pathogenesis (60). The pathophysiological behavior of apoE4 may be attributed to its distinctive structural arrangement in both lipid-free and lipid-bound states that probably distinguishes it from apoE3.