NMR Structure and Dynamics of a Receptor-active Apolipoprotein E Peptide*

Apolipoprotein E (apoE) is important in lipid metabolism due to its interaction with members of the low density lipoprotein (LDL) receptor family. ApoE is able to interact with the LDL receptor only when it is bound to lipid particles. To address structural aspects of this phenomenon, a receptor-active apoE peptide, encompassing the receptor-binding region of the protein, was studied by NMR in the presence of the lipid-mimicking agent trifluoroethanol. In 50% trifluoroethanol, apoE-(126–183) forms a continuous amphipathic α-helix over residues Thr130–Glu179. Detailed NMR relaxation analysis indicates a high degree of plasticity for the residues surrounding 149–159. This intrinsic flexibility imposes a curvature to the peptide that may be important in terms of interaction of apoE with various sized lipid particles and the LDL receptor. Residues 165–179 of apoE may act as a molecular switch whereby these residues are unstructured in the absence of lipids and prevent interaction with the LDL receptor. In the presence of lipids, these residues become helical resulting in a receptor-active conformation of the protein. Furthermore, the electrostatic characteristics and geometric features of apoE-(126–183) suggest that apoE binds to the LDL receptor by interacting with more than one of the receptor ligand-binding repeats.

olism due to its interaction with members of the low density lipoprotein (LDL) receptor family. ApoE is able to interact with the LDL receptor only when it is bound to lipid particles. To address structural aspects of this phenomenon, a receptor-active apoE peptide, encompassing the receptor-binding region of the protein, was studied by NMR in the presence of the lipid-mimicking agent trifluoroethanol. In 50% trifluoroethanol, apoE-(126 -183) forms a continuous amphipathic ␣-helix over residues Thr 130 -Glu 179 . Detailed NMR relaxation analysis indicates a high degree of plasticity for the residues surrounding 149 -159. This intrinsic flexibility imposes a curvature to the peptide that may be important in terms of interaction of apoE with various sized lipid particles and the LDL receptor. Residues 165-179 of apoE may act as a molecular switch whereby these residues are unstructured in the absence of lipids and prevent interaction with the LDL receptor. In the presence of lipids, these residues become helical resulting in a receptoractive conformation of the protein. Furthermore, the electrostatic characteristics and geometric features of apoE-(126 -183) suggest that apoE binds to the LDL receptor by interacting with more than one of the receptor ligand-binding repeats.
Human apolipoprotein E (apoE) 1 plays an important role in lipid metabolism by stabilizing lipoprotein particles and regulating plasma triglyceride clearance and cholesterol homeostasis (1, 2) through its interaction with members of the lipoprotein receptor family, such as the low density lipoprotein (LDL) receptor, the LDL receptor-related protein (3), and the very low density lipoprotein receptor (4). ApoE is also found associated with lipoproteins in cerebrospinal fluid (5) and is involved in nerve regeneration (6). The E4 isoform has been demonstrated as a major genetic risk in the predisposition of Alzheimer's disease (7,8).
ApoE is composed of two structurally and functionally independent domains (9,10). The 10-kDa C-terminal domain exhibits a high affinity for lipids and is thought to be the primary site for binding of apoE on spherical lipoprotein particles (11). The 22-kDa N-terminal domain contains the receptor-binding region of the protein. This region has been localized between residues 136 and 150 and is essentially made of basic amino acids whose chemical modifications render apoE non-receptorcompetent (12,13). In the absence of lipids, neither the Nterminal domain nor the full-length apoE protein are recognized by the LDL receptor (14). The N-terminal domain has a lower affinity for plasma lipoproteins than the C-terminal domain; however, the N-terminal domain is able to bind phospholipid vesicles and transform them into discoidal complexes that bind the LDL receptor with a similar efficiency as the entire protein (15). The two domains are linked by a protease-sensitive hinge region that has had no function associated to date.
The structure of the N-terminal domain in lipid-free solution was solved at high resolution by x-ray crystallography. This domain is an elongated four-helix bundle made of amphipathic helices with hydrophobic faces oriented toward the interior of the bundle (16). It has been postulated that the N-terminal four-helix bundle undergoes a conformational change upon lipid binding to expose the hydrophobic side-chains of the protein that interact with the lipid surface (2,17,18).
We have described previously (19) the purification and characterization of a peptide encompassing the portion of apoE that contains the receptor-binding region, apoE-(126 -183). In this report, the structure and dynamics of the receptor-active apoE peptide, as determined by NMR spectroscopy in presence of TFE, are shown. Results indicate that apoE-(126 -183) consists of an extended amphipathic helix that runs almost the entire length of the sequence. Structural calculations, in addition to relaxation data, indicate that the helix shows a high degree of plasticity. The dynamics study reported here marks the first such study of an extended helix in solution. The implications of the flexibility and helical structure of apoE-(126 -183) are discussed in terms of whole apoE in the presence of a wide range of lipoprotein particles in the bloodstream. Furthermore, this study emphasizes the requirement of lipids bound to apoE for proper LDL receptor recognition. purified recombinant apoE3-(1-183) by CNBr as described previously (19).
NMR Spectroscopy-NMR experiments were performed on ϳ2 mM unlabeled or 15 N-labeled apoE-(126 -183) in 500 l of 50% TFE-d 3 , 40% H 2 O, 10% D 2 O, pH 3.3, containing 0.01% (w/v) NaN 3 and 0.25 mM of 2,2-dimethyl-2-silapentane-5-sulfonate as an internal chemical shift reference. NMR experiments were carried out at 30°C on Varian INOVA 500-MHz and Unity 600-MHz NMR spectrometers. Data were processed using NMRPIPE (22) and analyzed using NMRVIEW (23). Complete 1 H and 15 N spectral assignments of apoE-(126 -183) were obtained using gradient-enhanced three-dimensional 15 N-edited total correlation spectroscopy ( mix 51.4 ms) and NOESY ( mix 50 ms) experiments to identify spin systems and inter-residue connectivities as described by Wü thrich (24). Confirmation of side-chain assignment was obtained through the use of three-dimensional HNHB and two-dimensional natural abundance 13 C HSQC spectra. 15 N T 1 , T 2 , and heteronuclear NOE relaxation data were recorded at 30°C on both 500-and 600-MHz spectrometers using the enhanced sensitivity gradient pulse sequences developed by Farrow et al. (25) 6, and 162.9 ms on the 600-MHz spectrometer. The exponential decay curves for T 1 and T 2 peak intensities were fit using the in-house written program xcrvfit. 2 { 1 H}-15 N NOE values were obtained from the ratio of the peak intensity from proton-saturated and -unsaturated spectra. Reduced spectral density mapping was carried out as described by Farrow et al. (26).
Structure Calculations-An ensemble of 388 apoE-(126 -183) structures was computed from 525 distance restraints (203 intraresidue, 202 sequential, and 120 medium range (defined as 2 Ͼϭ i Ϫ j Ͻϭ 4)), and 76 dihedral angle restraints (46 , 30 ) starting with an extended chain using a simulated annealing protocol (27,28) in X-PLOR version 3.851 (29). Interproton distance restraints were derived from threedimensional 15 N-edited NOESY experiments recorded with a mix of 50 and 100 ms, as well as a two-dimensional homonuclear NOESY experiment in D 2 O recorded with a mix of 150 ms. Distances were calibrated according to Slupsky and Sykes (30). backbone dihedral angles were calculated based on measured 3 J HN-H␣ coupling constants in an HNHA experiment (31) and the Karplus equation (32). dihedral angle restraints were obtained from the ratio of the d N␣ (i,i)/d ␣N (i Ϫ 1,i) in the three-dimensional 15 N-edited NOESY spectrum (33).
Families of structures were extracted from the ensemble of structures by superimposing the backbone of residues 149 -159 of apoE-(126 -183) using the program NMRCLUST (34).

RESULTS
Characterization of ApoE-(126 -183)-ApoE-(126 -183) is a 58-residue peptide that includes a lipid-binding region as well as the LDL receptor-binding moiety of apoE. ApoE-(126 -183) is insoluble in aqueous solution above pH 4. Below pH 4, the peptide does not appear structured in water; however, in the presence of either lipids, such as dodecylphosphocholine (DPC), or lipid-mimicking agents, such as trifluoroethanol (TFE), it displays a high (70 -80%) helical content (19). Previously, it was determined that 50% TFE or 13 mM DPC is sufficient to induce full structuring of this peptide (19). Furthermore, functional characterization has shown that, in the presence of lipids, this peptide is able to interact with the LDL receptor (19).
Structure of ApoE-(126 -183)- Fig. 1 exhibits a two-dimensional 1 H-15 N HSQC NMR spectrum of apoE-(126 -183) illustrating the quality of the data obtained for this peptide at an NMR frequency of 500 MHz. For the 58-residue peptide, most resonances are clearly resolved. Four resonances overlap (Lys 157 with Arg 158 and Ile 177 with Arg 180 ); one resonance is very weak and close to another peak (Leu 148 ), and three residues are missing (the two N-terminal residues, Leu 126 , Gly 127 , and the residue preceding the C-terminal proline, Gly 182 ). De-2 R. Boyko, unpublished data. spite having a well resolved spectrum, normally indicative of a protein with a well defined tertiary structure, no long range distance restraints (i Ϫ j Ͼϭ 5) could be found in an 15 N-edited NOESY spectrum with a mixing time of 150 ms.
The secondary structure of apoE-(126 -183) was determined using NMR spectroscopy based upon NOE connectivities, the H ␣ NMR chemical shift index (CSI) (35,36), and the ratio of the d N␣ /d ␣N NOEs (33). A summary of these data are illustrated in Fig. 2. In general, helical secondary structure was defined as follows: 1) the presence of a d ␣N(i,iϩ3) NOE; 2) a d N␣ /d ␣N ratio Ͼ1; 3) a 3 J HNH␣ Ͻ6 Hz; and 4) upfield shifted H ␣ NMR chemical shifts relative to random coil chemical shifts (negative CSI). When more than half of the available criteria were met for a particular residue and the flanking residues, the secondary structure was assigned as helical. Fig. 2 shows that apoE-(126 -183) appears to be composed of a long ␣-helix spanning the sequence from Thr 130 to Glu 179 , with the first and last four residues appearing unstructured. For residues Ala 176 to Glu 179 , the d ␣N(i,iϩ3) NOEs could not be assigned due to ambiguities arising from an almost complete overlap of Ile 177 and Arg 180 . Nevertheless, the 3 J HNH␣ , d N␣ /d ␣N , and CSI values for these residues favor a helical conformation. Fig. 2 also shows evidence of the simultaneous presence of d ␣N(i,iϩ2) and d ␣N(i,iϩ4) NOEs for some residues along the helix. These data were collected from a short mixing time NOESY (50 ms) and thus are not likely due to spin diffusion. According to Wü thrich (24), d ␣N(i,iϩ4) is characteristic of an ␣-helix, whereas d ␣N(i,iϩ2) is characteristic of a 3 10 helix. The simultaneous presence of both NOEs for the same residue may reflect some sort of internal flexibility.
The three-dimensional structure of apoE-(126 -183) was calculated from the NMR data as described under "Experimental Procedures." An ensemble of 388 structures was computed, none of which contained distance restraint violations greater than 0.2 Å nor dihedral angle violations greater than 2°. The 100 structures with the lowest calculated total energy were subsequently selected for further consideration. According to PROCHECK-NMR (37), 99.90% of the non-glycine residues have (, ) angles in the most favored or the additionally allowed regions of the Ramachandran plot for these 100 structures (data not shown).
Attempts to superimpose the final 100 structures revealed a rather poor convergence. In general, it is well known that proteins, nucleic acids, and polysaccharides are flexible in solution and can adopt multiple conformations. Such conformational flexibility has been shown to be important for biological function. Most structures solved by NMR are usually reported as ensembles of structures fulfilling the experimental restraints to a similar and high extent (38). However, when the dynamics or motion in a protein is more extensive, conformational families are believed to be a more realistic representation of the protein structure (34,38,39,40). In the case of apoE-(126 -183), it is apparent that there are several conformationally related subfamilies that all fulfill the experimental restraints after the simulated annealing calculations. To visualize the structural diversity of apoE-(126 -183), conformationally related subfamilies of structures were extracted using the program NMRCLUST (34). For clustering, the backbone heavy atoms of residues 149 -159, the central core region of the peptide displaying similar R 2 /R 1 ratios (see "Discussion" below), were superimposed, and clustering was done on the backbone heavy atoms of residues 132-170. In total, 13 subfamilies of structures were found. Among these, the first five contain 68 of the 100 structures. Each of the other classes contained 5 or fewer structures. The subfamilies could be further divided in two categories as follows: those with their hydrophobic residues on the convex face of the helix, and those with their hydrophobic residues on the concave face of the helix. The first three subfamilies (Fig. 3A) have their hydrophobic residues located on the concave portion of the helix and represent 45 of the 68 structures found in the first 5 subfamilies.
The structures in each of the first 3 subfamilies are moderately well defined over the entire helical length with root mean square distributions about the mean coordinate positions of 1.2 Å for subfamily 1, 1.3 Å for subfamily 2, and 1.6 Å for subfamily 3 for backbone atoms of residues 134 -168. However, these subfamilies have quite differing degrees of helix curvature. With the hydrophobic face of the helix pointing toward the reader, a van der Waals surface of the representative structure of subfamily 1 is shown in Fig. 3B. A continuous and almost perfectly aligned surface of hydrophobic residues appears to cover the inside face of the apoE-(126 -183) peptide in this conformation. Such a disposition of the hydrophobic residues is well suited for an interaction of this peptide with another hydrophobic surface such as a lipid particle. Arg 134 , close to the hydrophobic-hydrophilic interface, is the only non-hydrophobic residue in this face of the peptide.
Relaxation Measurements of ApoE-(126 -183)-To gain insight into the motions of apoE-(126 -183) in solution, longitudinal (T 1 ) and transverse (T 2 ) 15 N NMR relaxation times as well as { 1 H}-15 N heteronuclear NOEs were measured. The R 1 (1/T 1 ) and R 2 (1/T 2 ) relaxation rates as well as the heteronuclear NOEs at field strengths of 500 and 600 MHz are shown in Fig. 4. R 1 , R 2 , and the heteronuclear NOE are useful NMR parameters for probing backbone dynamics on several time scales. Picosecond to nanosecond time scale motions are reflected in all three parameters. Slower millisecond to microsecond time scale motions are manifested in the parameter R 2 . Of the 58 residues in apoE-(126 -183), 50 were used in the backbone dynamics analysis excluding the four overlapping resonances, the three missing resonances (the two N-terminal residues, and the penultimate C-terminal residue) and the C-terminal proline.
The 1st panel in Fig. 4 illustrates the R 1 relaxation rate of apoE-(126 -183). For a rigid globular (and hence isotopically tumbling) protein with flexible ends, it is typically observed that the R 1 relaxation rates decrease sharply from the flexible N terminus to a plateau value that is maintained for the majority of the protein, ending with a sharp increase for the flexible C-terminal residues. For apoE-(126 -183), there appears to be a gradual decrease in the R 1 relaxation rate up to approximately residue 149 followed by a gradual increase after approximately residue 159. The 2nd panel illustrates the R 2 relaxation rates. For a globular protein, it is typically observed that the R 2 relaxation rate will follow the opposite behavior of R 1 ; R 2 sharply rises followed by a plateau region and subsequently sharply falls due to flexibility at the N and C termini. For apoE-(126 -183), the first and last four residues follow this pattern of the sharp rise and fall. However, for residues 130 -148, R 2 appears to rise gradually followed by a plateau region from 149 -159 followed by a more rapid decrease for residues 160 -178. The 3rd panel in Fig. 4 shows the { 1 H}-15 N heteronuclear NOE for apoE-(126 -183). For a rigid globular protein, the heteronuclear NOE should follow a similar pattern to R 2 . For apoE-(126 -183), the NOE rises sharply for the first few residues, followed by a gradual increase up to residue 136. The NOEs stay constant up to residue 156 followed by a slight decreasing trend to residue 179. Often there is a region such as a loop that exhibits higher R 1 and lower R 2 relaxation rates, which is characteristic of a greater mobility in that portion of the polypeptide chain. For apoE-(126 -183), no such regions are observed in the measured relaxation data.
In general, relaxation rates are field-dependent: as magnetic field strength increases, R 1 decreases, whereas the { 1 H}-15 N heteronuclear NOE and to a much lesser extent R 2 increase. The only parameter sensitive to millisecond to microsecond time scale motions is R 2 . Large differences in R 2 at different field strengths imply that motion is occurring in the millisecond to microsecond time scale. This motion is termed exchange and can arise from structural interconversions or aggregation. It is unlikely that apoE-(126 -183) forms an oligomer as these studies were performed in 50% TFE. TFE has a tendency to weaken hydrophobic interactions and has been shown to be a denaturant of quaternary structure (41,42). Fig. 4 shows typical behavior for a monomeric protein. R 1 and NOE show the proper field dependence, whereas R 2 shows very little field dependence, ruling out exchange as a mechanism for describing the motions present in apoE-(126 -183). In the upper panel, the peptide was oriented with its hydrophobic face (i.e. the concave part of the curvature) out of the page. In the lower panel, the peptide was rotated by 180°around the x axis. Residues known to be important for receptor binding are indicated. The color code (for the side-chains and the molecular surface) is as follows: green, Ala, Gly, Ile, Leu, Pro, and Val; light green, Tyr; blue, Arg and Lys; light blue, His, Ser, and Thr; red, Asp and Glu; pink, Asn and Gln. The molecular surface was calculated using MSMS (84), and the figure was made using DINO (www.bioz.unibas.ch/ϳx-ray/dino).
The bottom panel in Fig. 4 shows the R 2 /R 1 ratio. In general, R 2 /R 1 ratios are determined by the overall rotational correlation time, assuming the value of R 2 is not dominated by exchange (43,44), and the position of the N-H bond vector relative to the diffusion tensor of the protein. ApoE-(126 -183) exhibits much larger R 2 /R 1 ratios for the central residues (residues 149 -159) than for the termini. Because the lack of field dependence of R 2 indicates there are no significant exchange contributions, anisotropic motion was investigated as a source for the large R 2 /R 1 values. Assuming that the protein is an extended helix and can be represented as a prolate ellipsoid, the motion will be anisotropic, and the rotational correlation times along the long axis (ʈ) and short axis (Ќ) will be different. Assuming a diffusion tensor ratio Dʈ:D Ќ ϳ7, expected for an extended helix of 58 residues, values of 0.7 s Ϫ1 are calculated for R 1 , 25 s Ϫ1 for R 2 , and 0.77 for the NOE (45)(46)(47). For residues 149 -159, the average R 1 is 0.95 s Ϫ1 , R 2 is 25 s Ϫ1 , and the heteronuclear NOE is 0.56. The similarity of calculated versus experimental values suggests that the motion of apoE-(126 -183) approximates that of a prolate ellipsoid, at least for the central residues.
The decrease in R 2 /R 1 ratios on either side of the central region (residues 149 -159) of the peptide implies that the helix is "fraying" or "unfolding" as one approaches the N or C termi-nus. To determine the time scales of the motions involved, we undertook a reduced spectral density approach to study the motions along the helix. The spectral densities quantitate the contributions of motions at various frequencies obtainable from the NMR experiment. Fig. 5 illustrates the reduced spectral density functions J(0), J( N ), and J(0.87 H ), which describe the motion of the H-N bonds, derived from the relaxation parameters R 1 , R 2 , and { 1 H}-15 N NOE. The spectral density function at zero frequency, J(0), is sensitive to motions on all time scales (48). The high frequency spectral density functions, J( N ) and J(0.87 H ), are sensitive to fast internal motions on the time scales of 1/ N and 1/w H , respectively. Large J(0.87 H ) (greater than 15 ps/rad) indicate fast internal motions, whereas spectral densities less than 7.5 ps/rad indicate a lack of internal flexibility (48). J(0) values suggesting that they have slow motions. These motions are not due to exchange between stable conformational states, as indicated by the lack of field dependence of R 2 . Rather, the motions are better described as "plastic" in the sense that small variations in local structure can accumulate to produce the large scale structural plasticity that is evidenced in this dynamics study and the range of subfamilies of curved helices calculated from the NMR restraints. This sort of cumulative dynamical behavior has been reported elsewhere (44,49). Interestingly, the J(0) (as well as R 1 and R 2 ) values for residues 131-143 appear to have a disproportionate change relative to residues 163-176 such that the residues at the C-terminal end of the peptide appear to exhibit a higher degree of motion. This is also observed in the { 1 H}-15 N heteronuclear NOE plot (Fig. 4) where the NOEs appear to be constant from residues 136 to 162 and then appear to gradually decrease. This may be explained by more stability for the N-terminal portion of the peptide.

DISCUSSION
The three-dimensional structure of the N-terminal portion (residues 1-191) of three major isoforms of apoE was determined previously (16, 50 -52) in aqueous solvent using x-ray crystallography. These structures revealed the presence of four amphipathic ␣-helices arranged in a 4-helix bundle. The lack of electron density for residues 1-22 and 166 -191 implied disorder for these two regions. The interaction of apoE with lipid molecules is believed to arise by the opening of the 4-helix bundle through a hinge region between the helices exposing hydrophobic residues that bind to the acyl chains of the lipid molecules (2). It appears that helices 1 and 2 as well as helices 3 and 4 remain preferentially paired during the first stage of the opening of the bundle (53) with possible rearrangement of the helices occurring subsequently (53)(54)(55). Other models have also been proposed (56).
The interaction of apoE with the LDL receptor involves a cluster of basic amino acids (residues 136 -150) located on helix 4 (2). As well, Arg 172 appears important for receptor binding activity (57). It was also found that residues 171-183 contain residues essential for receptor binding as truncated variants of the 22-kDa fragment of apoE illustrated that apoE3-(1-183) had nearly full receptor binding activity, whereas apoE3-(1-174) had only 19% activity, and further truncations reduced the activity to 1% (58).
In order to understand how residues 126 -183 are able to bind to and interact with lipids and the LDL receptor, we undertook a structural study of the receptor-active portion of apoE, apoE-(126 -183), using nuclear magnetic resonance spectroscopy in the presence of the lipid-mimicking agent TFE. This portion of apoE encompasses helix 4 in the crystal structure as well as the unstructured residues that appear important for receptor binding. TFE was used as a co-solvent because of its relatively small size compared with lipids, as well as its lipidlike characteristics. Historically, TFE has been used to study amphipathic and hydrophobic peptides and proteins. TFE is known to disrupt quaternary interactions and stabilize nativelike helical structure in proteins (41,59,60). Its use has been criticized because of its tendency to promote the formation of nonspecific helical structures in some cases (61)(62)(63). However, it has been shown that TFE only induces helical structures in regions with predicted helical propensity. Without such a propensity, helical structures are not found even at high TFE concentration (42,61,64). In the presence of 50% TFE, apoE-(126 -183) was shown to adopt 70% helical conformation as assessed by CD spectroscopy, which is in good agreement with values found on lipidic discoidal particles (19).
We show here that in the presence of 50% TFE, apoE-(126 -183) is composed of a helix that spans 50 residues from residue Thr 130 to Glu 179 , in contrast to the crystal structure in aqueous solvent where the fourth helix only spans residues Thr 130 to Ala 164 (19). The helix is relatively rigid from residues 149 to 159; however, residues on either side of this region exhibit a certain flexibility or plasticity such that the helix is able to adopt different curvatures. A calculated helical content of 86% for the lipid-bound species is slightly more than that obtained experimentally by CD spectroscopy (19). The flexible character of the peptide could be responsible for the slightly lower percentage (70%) obtained by CD. The helix length determined by NMR is 15 residues longer than that determined by x-ray crystallography. Compared with the 191 residues of the Nterminal domain, this 8% increase in helical content upon lipid binding is in good agreement with the 10 -15% increase found by CD spectroscopy (65). Interestingly, this region shows resistance to proteolysis when the N-terminal domain is lipidbound, whereas in the absence of lipid, this linker region is completely accessible to protease (66). Because apoE and/or its N-terminal domain only binds to the LDL receptor when complexed to lipid (14,15), the results presented here represent the first experimental description, at a molecular level, of how lipid binding may affect this region of apoE to modulate receptor binding activity.
ApoE-(126 -183) appears to be amphipathic (Fig. 3B), with one face composed almost entirely of hydrophobic residues. This surface is ideally suited to interact with the acyl chains of lipids on a lipoprotein surface or lipidic discoidal particles. An amphipathic helix seems to be a common feature of exchange- able apolipoproteins and likely provides a structural basis for their lipid-binding properties (67,68). A similar arrangement of hydrophobic residues was observed for a deletion mutant of human apoA-I, (⌬1-43) apoA-I. A continuously curved series of 10 amphipathic helices arranged as pairs of antiparallel dimers with one strongly hydrophobic face was observed (69). Other exchangeable apolipoproteins containing amphipathic helices that have been studied by NMR spectroscopy in lipid mimicking environments include apoC-I (58 residues) and apoC-II (79 residues) (70,71).
ApoE-(126 -183) shows a high flexibility and curvature not unexpected for a lipid-binding protein. Less recognized than amphipathicity, flexibility and curvature are common features of helical exchangeable apolipoproteins. ApoE is known to bind different sized lipoprotein particles from small (e.g. HDL, diameter 5-12 nm) to very large (e.g. chylomicron, diameter 75-1200 nm). Thus, apoE needs a certain flexibility or plasticity to accommodate these different surfaces and wrap itself around these lipoprotein particles. Other exchangeable apolipoproteins, known to bind different lipoproteins, are also curved helical structures. As already stated above, the (⌬1-43)apoA-I deletion mutant x-ray structure shows a high curvature giving a horseshoe shape to the molecule (69). ApoA-I is known to bind to small HDL particles. Measurement of the helix curvature shows that apoA-I is compatible with the size of these particles. ApoC-I is made of two helices connected by a flexible linker region giving an overall curved structure for the protein. Both helices contain poorly defined regions that appear to have distributions of more or less bending. ApoC-I also appears to assume a slightly bent flexible conformation on average, and the flexibility may be an asset in binding to lipoprotein particles (70). The apoC-II structure, as determined by NMR spectroscopy in the presence of SDS micelles, shows three well defined helical regions with connecting regions showing a loosely defined helical conformation. The first and longest helical region displays a marked curvature. As in our case, the authors found d ␣N(i,iϩ4) and d ␣N(i,iϩ2) NOEs to the same residues in the helical regions (71).
In the case of apoE, flexibility and curvature might also have important implications. It was recently demonstrated (65) that in presence of lipids, the two lysines in the receptor-binding region (Lys 143 and Lys 147 ) become more accessible to solvent and that their respective pK a values are decreased due to a local increase of positive electrostatic potential. This increase in positive potential and solvent exposure upon lipid binding could improve the electrostatic interaction of apoE with the ligand-binding repeat(s) of the LDL receptor and, therefore, may explain why lipid association is required for high affinity binding of apoE to the LDL receptor. We suggest that this increase in solvent accessibility might be a direct consequence of the curvature imposed on the receptor-binding region of apoE upon lipid binding or in a lipid-mimicking environment. A superposition of the three most populated subfamilies of structures with the fourth helix of the crystal structure of apoE (Protein Data Bank code 1BZ4) shows that our structures have significantly different curvatures than in lipid-free solution (data not shown).
Our results on the flexibility of the fourth helix of the apoE N-terminal domain may also be related to the conformational flexibility observed recently in three new crystal forms of apoE in lipid-free solution (56). Superposition of the three structures illustrate flexibility and kinks in helices near one end of the four-helix bundle, namely the loop connecting helix 2 to helix 3, the beginning of helix 3, and the end of helix 4. The flexible end of the bundle was associated with a potential initial site for lipid binding that could initiate the opening of the bundle upon lipid association.
Potential Interactions with the LDL Receptor-The structuring of residues 165-179 in the presence of a lipid-mimicking agent may only indirectly affect the interaction between apoE and the LDL receptor, perhaps by anchoring the receptorbinding region of apoE (residues 136 -150) at the lipid interface. In turn, this could increase the accessibility of apoE to the receptor or orient apoE in a favorable way. It is more likely, however, that a direct interaction of the region spanning residues 165-179 with the LDL receptor occurs due to the absolute requirement of Arg 172 for correct association of apoE to the receptor (57).
The LDL receptor (LDLR) binds different types of lipoproteins having at their surface either apoB (LDL) or apoE (␤-VLDL). LDLR is the principal mechanism for cell cholesterol uptake. It is part, and is considered the prototype, of a large family of cell surface receptors (72). These receptors share a common architecture composed of different structural units including a group of ligand-binding repeats (LR) followed by epidermal growth factor and YWTD domains, a single transmembrane segment, and a cytoplasmic tail required for targeting of the receptor to coated pits as well as internalization. The LDLR contains seven ligand-binding repeats. The three-dimensional structures of several of the ligand-binding repeats have been studied. LR1 and LR2 (73,74), LR5 (75), LR6 (76,77), and one repeat (CR8) from the LDLR-related protein (78) have been solved. Each of these ligand-binding repeats show similar features including three disulfide bonds, a short antiparallel ␤-sheet, a single turn of 3 10 helix, and a high affinity calciumbinding site. The presence of highly conserved acidic residues within the LR modules and the positively charged region of helix 4 (residues 136 -150) of apoE have led to the hypothesis that ligand-receptor recognition may be due to electrostatic interactions.
It has been shown that ligand-binding repeats 3 through 7 (LR3-LR7) and epidermal growth factor repeat A are required for the binding of apoB (LDL), whereas the binding of apoE (␤-VLDL) is primarily mediated by LR5 (79). Indeed, a deletion of LR5 produced a 50% decrease in the receptor capability to bind ␤-VLDL. No other LR deletion showed such a decrease. Nevertheless, even if less spectacular, the LR4 deletion decreased the binding by ϳ20% (79), and the deletion of the 3 first repeats at once showed an ϳ30% decrease (80). These data are in favor of a role of some other repeats, located upstream of LR5, in apoE binding to the receptor. It was noticed that LR5 contained extra negatively charged residues with regard to other repeats that probably contribute to the importance of LR5 (79). Nonetheless, a comparison of 3 different binding repeats has shown a region in each of negative electrostatic potentials surrounding the coordinated calcium ion (76). The rest of the surface, although mainly negative, is unique for each repeat. Ligand binding specificity may result from the rearrangement of 2 or more repeat modules (76).
As the counterpart, the well characterized receptor-binding region of apoE-(136 -150) displays a highly positive electrostatic surface potential (Fig. 6) and can thus interact electrostatically with the most negatively charged ligand-binding repeats (LR5). The other part of the peptide (residues 165-179), structured in the presence of a lipid environment, also shows a mainly positively charged surface (Fig. 6), and could conceivably interact with another ligand repeat from the LDL receptor. The glutamic acid residues punctuating the largely positive surface might be important for proper orientation of the peptide at the surface of the receptor, interacting with the few arginine and lysine residues present in almost every repeat. Other arguments seem to reinforce the idea that apoE is interacting with more than just LR5 at the receptor surface. Little is known of the respective alignment and orientation of these repeats in the whole receptor. Electron microscopy on LDLR reconstituted into lipidic vesicles suggests it forms extended structures (81). In LDLR, ligand-binding repeats are separated from each other by a 4-residue linker (except for LR4 and LR5 which are separated by 12 residues). The study of ligandbinding repeat pairs (LR1-LR2 and LR5-LR6) has shown that the repeats appear to act as isolated modules, independently folded with few (if any) interactions between adjacent modules (72,82). Furthermore, analysis of backbone dynamics of the LR5-LR6 pair suggests that the linker region between the two modules is highly flexible and that the position of the two modules with respect to each other is not fixed (83). The flexibility of the LDLR in the ligand-binding repeat domain might be seen as the perfect counterpart to the observed flexibility of apoE-(126 -183). Indeed, as discussed previously, if the peptide and, by extension, apoE need flexibility to bind to lipoprotein particle surfaces of different sizes, it might seem logical that the LDLR that recognize and bind apoE on these different lipoproteins needs the same flexibility. Once a ligand-binding repeat (presumably LR5) has recognized the most basic region of apoE (residues 136 -150) through electrostatic interactions, the adjacent repeat(s) might interact with other region(s) of apoE (notably around Arg 172 ). This interaction may orient the apolipoprotein and subsequently the entire lipoprotein in a favorable manner with respect to the receptor. The correct interaction with the lipoprotein might be an important step in the transduction of the internalization signal to the cytoplasmic side of the receptor. Interestingly, the apoE-(126 -183) peptide, containing all the required regions of apoE that bind to the receptor, is an elongated structure ϳ70 Å long, whereas a single ligand-binding repeat from the receptor has a maximum diameter of 25 Å (Fig. 6). These distances suggest that apoE could be interacting with more than one repeat at the surface of the receptor. When modeled in an extended conformation, the linker region between LR5 and LR6 separates the two repeats by ϳ15 Å (83). One can expect linker regions between other repeats having the same approximate length. This distance, which corresponds to ϳ65 Å, matches well with the ϳ70 Å separating the basic cluster region (residues 136 -150) from Arg 172 of apoE (Fig. 6).
We have shown that the lipid-and receptor-binding portion of apoE in a lipid-like environment is a single continuous ␣-helix that exhibits considerable flexibility. Preliminary results indicate that apoE in the presence of DPC is similar to apoE in the presence of TFE. Our results suggest that residues 165-179 of apoE may act as a molecular switch, whereas in the absence of lipids these residues are unstructured and prevent interac-tion with the LDL receptor. In the presence of lipids, these residues become helical resulting in a receptor-active conformation of the protein. Furthermore, it is likely that the inherent flexibility of apoE allows it to bind to a variety of sizes of lipid particles. Finally, it appears as if the basic portion of the helix would certainly be exposed to interact with at least two of the acidic ligand repeats of the LDL receptor, each of which would interact with apoE in any curved conformation.