Address correspondence and reprint requests to Dr. T. Pillot at CJF INSERM 9508, Institut des Cordeliers, 15 Rue de l'Ecole de Médecine, 75006 Paris, France.
Abstract : Growing evidence indicates the involvement of apolipoprotein E (apoE) in the development of late-onset and sporadic forms of Alzheimer's disease, although its exact role remains unclear. We previously demonstrated that β-amyloid peptide (Aβ) displays membrane-destabilizing properties and that only apoE2 and E3 isoforms inhibit these properties. In this study, we clearly demonstrate that the carboxy-terminal lipid-binding domain of apoE (e.g., residues 200-299) is responsible for the Aβ-binding activity of apoE and that this interaction involves pairs of apoE amphipathic α-helices. We further demonstrate that Aβ is able to inhibit the association of the C-terminal domain of apoE with lipids due to the formation of Aβ/apoE complexes resistant to sodium dodecyl sulfate-polyacrylamide gel electrophoresis. On the contrary, the amino-terminal receptor-binding domain of apoE (e.g., residues 129-169) is not able to form stable complexes with Aβ. These data extend our understanding of human apoE-dependent binding of Aβ by involving the C-terminal domain of apoE in the efficient formation of apoE/Aβ complex.
Among the different events leading to neuronal death and finally to dementia, the apolipoprotein E (apoE) polymorphism is one of the major risk factors identified so far that influences the progression and the age at onset of Alzheimer's disease (AD) (Wisniewski et al., 1995 ; Weisgraber and Mahley, 1996). apoE (299 residues) is a highly α-helical protein, primarily involved in cholesterol metabolism and cellular uptake of lipoproteins via its receptor-binding domain (for review, see Weisgraber, 1994). The crystal structure of full-length apoE is not known, but apoE is organized into two structural domains that are responsible for different functions of the protein (Aggerbeck et al., 1988 ; Wetterau et al., 1988). The N-terminal domain (residues 1-191) contains the lipoprotein receptor-binding region (residues 140-160) (Innerarity et al., 1983) and is structured as an extended four-helix bundle, arranged in an antiparallel manner (Wilson et al., 1991). The C-terminal domain (residues 216-299), the structure of which is not yet determined, contains the major lipid-binding determinants (Sparrow et al., 1992 ; Wang et al., 1996). apoE occurs as three isoforms in humans, differing by a single amino acid substitution : apoE3 (Cys112, Arg158) is the major isoform (Mahley et al., 1990), whereas isoforms apoE2 (Cys112, Cys158) and E4 (Arg112, Arg158) may be associated with disorders of cholesterol metabolism (Rosseneu and Labeur, 1995).
apoE4 has been linked to AD owing to the increased frequency of the ε4 allele in AD (Corder et al., 1993 ; Strittmatter et al., 1993a). Moreover, the cerebral amyloid plaque core, a major neuropathological feature of AD that contains fibrils of β-amyloid protein (Aβ), is also immunostained for apoE (Namba et al., 1991 ; Wisniewski and Frangione, 1992). We have demonstrated that the C-terminal domain of Aβ displays fusogenic properties, which might account, at least partly, for the cytotoxicity of Aβ through a direct destabilization of neuronal membranes (Pillot et al., 1996a). Conflicting results have been published concerning the in vitro apoE isoform-specific binding to Aβ (Strittmatter et al., 1993b ; LaDu et al., 1994). We recently demonstrated that the apoE2 and E3 isoforms specifically interact with Aβ, whereas apoE4 exhibits defective binding activity (Pillot et al., 1997a). These results have been confirmed by other groups (Aleshkov et al., 1997 ; LaDu et al., 1997 ; Yang et al., 1997) and support the hypothesis about the role of apoE in the pathogenesis of AD. The apoE2 and E3 isoforms may be “chaperone” proteins regarding Aβ aggregation and neurotoxicity. These isoforms might also facilitate the clearance of Aβ from the extracellular space through the interaction of apoE with its cellular receptors. Conversely, apoE4 is considered as a major risk factor in AD development due to the lack of interaction of apoE4 with Aβ.
The aim of this study was to identify the apoE peptide domain involved in Aβ binding. Here, we demonstrate that the C-terminal lipid-binding domain of apoE (e.g., residues 200-299) specifically interacts with Aβ and inhibits its fusogenic properties, whereas the N-terminal receptor-binding domain of apoE displays no Aβ-binding activity.
MATERIALS AND METHODS
l-α-Phosphatidylethanolamine (PE) from egg yolk, l-α-phosphatidyl-l-serine (PS) from bovine brain, l-α-phosphatidylcholine (PC) from egg yolk, dimyristoylphosphatidylcholine (DMPC), cholesterol, and bovine serum albumin were purchased from Sigma. Bovine sphingomyelin (SM) was from Matreya (U.S.A.). 1-Palmitoyl-2-pyrene(14)-phosphatidylcholine (Pyr-PC) was a kind gift from Somerharju (Helsinki, Finland). All reagents for peptide synthesis and sequencing were purchased from Applied Biosystems. The 2,2,2-trifluoroethanol (TFE) and hexafluoro-2-propanol (HFP) used for peptide solubilization were of the highest grade from Sigma.
Synthesis and purification of peptides
Peptides were synthesized as previously described (Pillot et al., 1996a) by the standard F-moc solid-phase method on an Applied Biosystems model 431A peptide synthesizer. The purity and correct sequences of all peptides were verified by electron spray ionization mass spectrometry using a Fisons/VG Platform (Manchester, U.K.) mass spectrometer.
Circular dichroism (CD) measurements
To overcome problems of amyloid peptide solubility at high concentration, fresh peptide stock solutions (5 mg/ml) were prepared in HFP. CD spectra were obtained at 23°C in a Jasco 710 spectropolarimeter calibrated with 0.1% (wt/vol) D-10-camphorsulfonic acid solution (Pillot et al., 1996b). Portions of peptide stock solutions were removed and diluted in a 10 mM sodium phosphate buffer (pH 7.4). Nine spectra were recorded 20 min after the peptides were dissolved in the aqueous buffer solution at a final concentration of 0.1 mg/ml. The percentages of secondary structure were estimated by curve fitting on the entire elliptical curve between 184 and 260 nm according to the variable selection, as previously described (Pillot et al., 1996a).
Small unilamellar vesicles (SUVs) were prepared from a mixture of PC/PE/PS/SM/cholesterol (10 : 5 : 7.5 : 7.5 : 16, by wt), when necessary in the presence of 2.5 mol% Pyr-PC, as previously described (Pillot et al., 1996a). All lipids were dissolved in chloroform at 10 mg/ml and dried under a stream of nitrogen. Pyr-PC was dissolved in chloroform/ethanol at a concentration of 2 μM. Dried mixed lipids were hydrated in a 10 mM Tris-HCl buffer (pH 8.0) containing 150 mM NaCl, 0.1 g/L Na-EDTA, and 1 mM NaN3. The lipid suspension was sonicated at 23°C, using a Branson sonifier, under nitrogen, at 32 W for 4 × 15 min. After sonication, the labeled and unlabeled vesicles were applied to a Sepharose CL 4B column. Phospholipid concentration was determined by an enzymatic assay (BioMérieux, France). Fusion of pyrene-labeled SUVs together with unlabeled vesicles, at a 1 : 4 (wt/wt) ratio, was measured using a fluorescence probe dilution assay (Pillot et al., 1996a, 1997b). The pyrene excimer/monomer intensity ratio (E/M) was measured as a function of time after addition of increasing quantities of the peptides dissolved at a concentration of 1 mg/ml in either 50 or 20% (vol/vol) TFE. Fusion resulted in a decrease of the excimer intensity and a slight increase of the monomer fluorescence. Emission spectra were obtained on an Aminco SPF 500 spectrofluorimeter at 25°C. The E/M was calculated from the excimer and monomer fluorescence intensity at 475 and 398 nm, respectively, with an excitation wavelength of 346 nm. The excitation and emission bandwidths were set at 2 and 5 nm, respectively. Under these conditions, unlabeled vesicles, in the presence of the peptides, gave no significant signal due to light scattering. All experiments were performed in a 10 mM Tris-HCl buffer (pH 8.0) containing 150 mM NaCl and 0.1 g/L Na-EDTA in a final volume of 500 μl. The final TFE concentration in the reaction mixture was <2.5%, and a correction was made for the effect of the solvent on the E/M.
Monitoring of lipid-apoE association
The kinetics of lipid-apoE peptide association were followed by monitoring the rate of clearance of the turbidity of DMPC liposomes as a function of temperature (De Pauw et al., 1995 ; Pillot et al., 1996b). DMPC liposomes were mixed with apoE peptide at a 4 : 1 DMPC/apoE (wt/wt) ratio, and the optical density was measured at 325 nm as a function of temperature in a Uvikon 941 spectrophotometer. For measurements in the presence of Aβ peptides, stock solutions of Aβ and of apoE peptides were previously mixed in buffer to give increasing molar ratios. The reaction was initiated by addition of DMPC liposomes into the reaction mixture.
Tryptophan fluorescence measurements
apoE(200-299) contains four tryptophan residues, whereas Aβ contains no tryptophan. The blue shift of the maximal fluorescence emission wavelength of the apoE tryptophan residues was used to demonstrate direct binding between the apoE peptide and the Aβ fusion peptides, as previously described for the different apoE isoforms (Pillot et al., 1997a). Aβ peptides were added to apoE peptides at increasing Aβ/apoE peptide molar ratios and incubated at 25°C for 1 h. Tryptophan fluorescence emission measurements, in the presence or absence of Aβ, were performed at 25°C on an Aminco SPF 500 spectrofluorimeter using an excitation wavelength of 295 nm. Denaturation experiments were followed by adding increasing amounts of guanidine HCl (GdnHCl) to apoE(200-299) in the absence or in the presence of Aβ, at an apoE/Aβ molar ratio of 1 : 100, and incubated for 16 h at 4°C. Tryptophan fluorescence emission spectra were recorded as described above. Five different spectra were recorded and averaged.
Sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE) was performed under nonreducing conditions, i.e., without addition of β-mercaptoethanol, using standard methodology (Laemmli, 1970). Protein bands were visualized after Coomassie Blue staining, and gels were scanned using a laser densitometer (Pharmacia, Uppsala, Sweden).
To determine which domain of apoE2 and E3 interacts with Aβ, we synthesized different apoE peptides (Fig. 1) and first tested their effects on the fusogenic properties of Aβ. Membrane-destabilizing properties have been described for the fusion peptides of viral proteins and are related to the abilities of such peptides to penetrate the lipid bilayer with a tilted orientation (for review, see Brasseur et al., 1997). Insertion of peptides in the lipid phase induces a perturbation of the regular packaging of the phospholipid acyl chains and results in the fusion of membrane. We showed that the C-terminal domain of Aβ (e.g., residues 29-40 and 29-42) is responsible for Aβ fusogenic properties due to a direct interaction of the nonaggregated peptide with model membranes (Pillot et al., 1996a). Moreover, we demonstrated that apoE2 and E3 isoforms are able to interact with Aβ to inhibit both its fusogenic properties and aggregation in vitro, whereas apoE4 had no effect (Pillot et al., 1997a).
C-terminal domain of apoE inhibits Aβ-induced liposome fusion
The extent of lipid mixing induced by Aβ peptides, as a measure of their fusogenic activity, was followed by a fluorescent probe dilution assay, as described in Materials and Methods. When SUVs were mixed with Aβ, the intensity of the E/M decreased to 36 and 42% of its original value for the Aβ(29-40) and Aβ(29-42) peptides, respectively, after 10-min reaction (Fig. 2A). This represents a measure of liposome-induced fusion. Addition of apoE(200-299) (140 nM) inhibited vesicle fusion induced by Aβ, as the percentage of E/M decrease amounted to only 14 and 26% for the Aβ(29-40) and Aβ(29-42) peptides, respectively (Fig. 2A). Addition of apoE(200-299) alone to the SUVs did not affect the E/M intensity ratio (data not shown). The decrease of the E/M intensity ratio upon the concentration of apoE(200-299) was investigated by adding increasing amounts of apoE peptide to a fixed amount of the SUV/Aβ mixture (Fig. 2B). Maximal inhibition of the vesicular fusion induced by Aβ was reached by concentrations of apoE(200-299) of 140 nM, corresponding to an apoE(200-299)/Aβ molar ratio of 1 : 100. We next investigated the effect of apoE(200-299) on the fusogenic activity of the Aβ(22-42) and Aβ(12-42) peptides (Table 1). Addition of increasing amounts of apoE(200-299) to the Aβ peptide/SUV mixture also reduced the fusogenic activity of these peptides.
Table 1. Effects of apoE(200-299) on fusogenic properties of Aβ(22-42) and Aβ(12-42) peptidesLipid mixing induced by the Aβ peptides (14 μM) was monitored as previously described (Pillot et al., 1996a), and the % E/M decrease was measured in the presence of increasing amounts of apoE(200-299) after 10-min reaction time.
We moreover tested the effects of an N-terminal apoE peptide (e.g., residues 129-169, Arg158), which includes the receptor-binding domain of apoE3 or E4 isoforms (Fig. 1), on the fusogenic activity of Aβ. Addition of apoE(129-169) to the reaction mixture had no effect on the extent of vesicular fusion induced by the four amyloid peptides listed before (data not shown).
Aβ inhibits lipid-binding properties of apoE(200-299)
The formation of lipid-apoE peptide complexes was monitored as previously described (Pillot et al., 1996b, 1997a). The effect of the addition of Aβ on the lipid-binding properties of apoE was estimated by plotting the total turbidity decrease of DMPC liposomes upon the addition of apoE(200-299) as a function of the Aβ/apoE(200-299) molar ratio (Fig. 3). Aβ peptides inhibited the association of apoE(200-299) with lipids in a concentration-dependent manner, and significant inhibition was observed at an Aβ/apoE(200-299) molar ratio of >15. By contrast, Aβ peptides did not modify the association between apoE(129-169) and lipids even at the highest Aβ peptide concentrations used, confirming the lack of interaction between the N-terminal apoE peptide and Aβ (data not shown).
We synthesized peptides encompassing the two pairs of α-helices and single α-helices of the C-terminal domain of apoE [e.g., apoE(204-246), (248-286), and apoE(263-286), (269-286), respectively (Fig. 1)]. CD measurements of the isolated apoE peptides suggest that they are mainly α-helical (data not shown). The apoE(204-246), (248-286), and (200-299) peptides, dissolved at 0.1 mg/ml in a 10 mM phosphate buffer (pH 7.4), displayed typical negative CD bands at 208 and 222 nm and a positive CD band at 192 nm. On the contrary, the apoE(263-286) and (269-286) peptides were random coil when dissolved under the same conditions. We thus determined the effects of these apoE peptides on the fusion properties of Aβ. Fusion induced by Aβ decreased significantly upon addition of increasing amounts of apoE(248-286) (Table 2). Maximal inhibition of vesicular fusion induced by Aβ peptides is reached at an apoE peptide concentration above 280 nM, corresponding to an apoE/Aβ peptide molar ratio of 1 : 50. Similar results were obtained with the apoE(204-246) peptide (data not shown), whereas the apoE(263-286) and (269-286) had no effects. As for apoE(200-299), we followed the effect of Aβ on the lipid-binding properties of apoE(248-286). Addition of increasing amounts of Aβ peptides to a mixture of apoE(248-286) and DMPC liposomes significantly decreased the formation of complexes between apoE(248-286) and lipids (not shown). As previously described, the apoE(204-246) peptide did not form stable complexes with lipids (Sparrow et al., 1992), as it might be too hydrophilic to generate peptide/DMPC interactions.
Table 2. Effects of apoE(248-286) on fusogenic properties of Aβ peptidesLipid mixing induced by the amyloid peptides (14 μM) was monitored as previously described (Pillot et al., 1996a), and the % E/M decrease after 10 min was measured in the presence of increasing amounts of apoE(248-286).
The denaturation of apoE(200-299) by increasing concentrations of GdnHCl was followed, as previously described (Pillot et al., 1997a), by measuring the maximum Trp emission fluorescence of apoE(200-299). In the absence of Aβ, the maximum Trp emission fluorescence of apoE(200-299) is at 348-350 nm, in good agreement with previous works (Wetterau et al., 1988 ; De Pauw et al., 1995). Following incubation with Aβ, the maximal Trp emission fluorescence of apoE(200-299) is blue shifted to 336-338 nm (Fig. 4). In the absence of Aβ, the midpoint of the denaturation curve of apoE(200-299) was ~1.0-1.2 M GdnHCl. In the presence of Aβ, the midpoint of the denaturation curve was shifted to 2.8-3.0 M GdnHCl (Fig. 4). These data suggest that apoE(200-299) interacts directly with Aβ peptides.
This interaction was visualized by SDS-PAGE. apoE(200-299) migrated as a single polypeptide band of 9.8 kDa on 15% SDS-PAGE under nonreducing conditions (Fig. 5, lane 1). Under these experimental conditions, the amyloid peptides used gave no detectable band within the molecular mass range detected on the gel (data not shown). After 1-h incubation at 25°C of apoE(200-299) with Aβ(29-40) or Aβ(29-42), a new band appeared with an apparent molecular mass of 17-18 kDa corresponding to stable apoE(200-299)/Aβ complexes containing four to five Aβ peptide chains (Fig. 5, lanes 3 and 4). The incubation of apoE(200-299) with Aβ(22-42) and Aβ(12-42) led to the formation of stable complexes with an apparent molecular mass of 21-22 kDa (Fig. 5, lanes 5 and 6). The complexes detected here should have a higher molecular mass, but we cannot rule out that under the SDS-PAGE conditions used, a partial dissociation of the complexes occurred during electrophoresis. Aβ(13-28) did not form a complex with the C-terminal domain of apoE (Fig. 5, lane 2), confirming the lack of interaction of the 13-28 residues of Aβ peptide with apoE(200-299). We previously described mutant peptides of Aβ(29-42), designed to lose their fusogenic properties (Pillot et al., 1996a). These peptides, named Aβ(29-42, 0° or 85°), have the same amino acid composition as the wild-type Aβ(29-42) peptide, but their sequence was modified to abolish their fusion activity. Interestingly, no complex was formed between apoE(200-299) and the nonfusogenic Aβ(29-42, 0° and 85°) mutant peptides or with the simian immunodeficiency virus (SIV) and prion fusion peptides (Pillot et al., 1997b) (Fig. 5, lanes 7-10), stressing the specificity of the interaction described above.
Based on epidemiological and immunohistological studies, apoE4 has been strongly associated with both sporadic and familial AD (Corder et al., 1993 ; Strittmatter et al., 1993a). Although apoE has been identified as one of the numerous amyloid-associated proteins (Namba et al., 1991), the exact role of apoE in the AD etiology remains unclear. One hypothesis is that apoE may function in AD as a pathological chaperone for Aβ, controlling partially the metabolism and clearance of Aβ in the human brain (Wisniewski and Frangione, 1992). We and others have demonstrated that the in vitro interaction between apoE and Aβ is apoE isoform specific (LaDu et al., 1997 ; Pillot et al., 1997a). The apoE2 and E3 isoforms formed stable complexes in vitro with Aβ, whereas apoE4 isoform did not interact with Aβ. These complexes have been recently detected in vivo. The role of apoE in the etiology of AD might be related to a protective role of apoE2 and E3 isoforms by complexing Aβ and inducing the clearance of Aβ from the extracellular space, limiting in that way Aβ peptide aggregation and neurotoxicity. apoE4 is considered a major risk factor in the development of AD due to its incapacity to play the protective “chaperone” role displayed by apoE2 and E3 (Weisgraber and Mahley, 1996).
In this article, we clearly demonstrate that the C-terminal lipid-binding domain of apoE interacts specifically with the Aβ peptides to form stable complexes in vitro. This association impairs the interaction of Aβ with model membranes and decreases its fusogenic properties. On the contrary, an N-terminal peptide of apoE, containing the apoE receptor-binding site, fails to interact with Aβ. Taken together, our results are in good agreement with a preliminary study by Strittmatter et al. (1993b), suggesting that the 244-274 residues of apoE might be involved in Aβ binding. Moreover, recent studies demonstrated the presence of carboxyl fragments of apoE as major species in amyloid deposits (Aizawa et al., 1997). Other groups also confirmed this observation (Naslund et al., 1995), and interestingly, such a carboxyl fragment of apoE can form amyloid-like fibrils in vitro (Wisniewski et al., 1995).
The three-dimensional structure of the C-terminal domain of apoE has not been determined yet, but secondary structure prediction and biochemical studies strongly suggest that this apoE domain consists of four to five amphipathic α-helices (Fig. 1 ; Nolte and Atkinson, 1992 ; Brasseur et al., 1993). In the present study, we showed for the first time that pairs of amphipathic α-helices of the C-terminal domain of apoE directly interact with Aβ and thereby mimic the interaction of the apoE(200-299) domain with Aβ. As previously described, the Aβ peptides mainly exhibited a β-sheet structure in aqueous buffer (Pillot et al., 1996a). We showed that the α-helical structure of the apoE peptides is critical for their interaction with Aβ in a soluble, nonaggregated β-sheet conformation, as demonstrated by CD measurements. According to this, it has been previously reported that apoE preferentially binds to Aβ peptides in a β-sheet conformation (Golabek et al., 1996). We thus propose that the 204-246 and 248-286 regions of apoE contain preferential sites of interaction with the amyloid peptide. Moreover, neither apoE(263-286) nor apoE(269-286) peptides (Fig. 1) had any effect on the fusogenic properties of Aβ, suggesting that the interactions between apoE and Aβ require definite structure of the C-terminal domain of apoE.
We have investigated the effect of several apolipoprotein A-I (apoA-I) peptides, exhibiting α-helical amphipathic repeats present in the exchangeable apolipoproteins (Nolte and Atkinson, 1992 ; Brasseur et al., 1993) on the fusogenic properties of Aβ. Under similar experimental conditions, neither the apoA-I (87-124), (99-124), (107-124), (142-182), nor (225-243) peptides displayed any effect on the lipid-mixing activity of Aβ (data not shown). These results confirm our previous findings indicating specific interactions between native apoE2 and E3 with Aβ (Pillot et al., 1997a). Moreover, we showed that both fusion peptides from the SIV or the prion protein (Pillot et al., 1997b), which exhibit properties similar to those of Aβ, are unable to interact with apoE peptides as demonstrated by SDS-PAGE or liposome fusion assay (data not shown). Reciprocally, we showed that Aβ inhibits the capacity of apoE C-terminal peptides to form discoidal complexes with DMPC, at the same apoE/Aβ molar ratios used for fusion experiments. Thus, the association described here might involve mainly hydrophobic interactions between apoE α-helices and Aβ, as suggested by SDS-PAGE and GdnHCl denaturation experiments.
The C-terminal domain of apoE is able to form stable complexes with Aβ peptides of different lengths (e.g., 12-42 to 29-42). These interactions occur with the C-terminal fusogenic domain of Aβ (e.g., residues 29-40 and 29-42), as the 13-28 amyloid peptide failed to bind the apoE peptides. Interestingly, previous work has demonstrated that residues 29-42 of Aβ are critical for amyloid fibril formation and stabilization (Lansbury et al., 1995). Biochemical studies have also shown that a major component of preamyloid lesions, which are the Aβ deposits appearing first in Down syndrome patients, is Aβ(17-42) (Lalowski et al., 1996). Thus, the interaction described in our study is likely to take place in vivo, according to co-purification of apoE C-terminal fragments with Aβ from neuritic plaques (Wisniewski et al., 1995) or from other systemic amyloidoses (Castano et al., 1995a, b).
We and others have demonstrated that apoE4 is unable to efficiently form complex with Aβ (Zhou et al., 1996 ; LaDu et al., 1997 ; Pillot et al., 1997a). Thus, the mechanism by which apoE contributes to the pathogenesis of AD may be by means of an isoform-specific interaction with Aβ. This hypothesis proposes a protective role for the apoE2 and E3 isoforms in AD by promoting the clearance of Aβ, a function that apoE4 lacks. The results presented here do not give a structural explanation for the preferential binding of Aβ to the apoE2 and E3 isoforms. However, the C-terminal domain of apoE, for which we provided evidence in this study of its Aβ-binding properties, is common to the three apoE isoforms. It might thus be possible that domain interactions specific to apoE4 result in a deficient Aβ-binding activity of apoE4. This question has to be addressed by the crystal structure resolution of the three full-length apoE isoforms. To date, attempts to crystallize the entire apoE protein remain unsuccessful. However, several observations could help us to understand why the C-terminal domain of apoE4 did not interact with Aβ. First, it has been demonstrated that the substitution Cys/Arg at position 112 in the apoE4 isoform influenced the lipid-binding properties of apoE4 and triggered the preferential association of apoE4 with very low density lipoprotein, modifying the physiological properties of this apoE isoform as compared with the apoE2 and E3 isoforms (Weisgraber, 1990). These authors have suggested that the substitution at position 112 in apoE might modify the conformation of the full-length protein due to specific interactions between the N- and the C-terminal domain of apoE4. This hypothesis has been supported by determining the crystal structure of the 22-kDa N-terminal domain of apoE4 and comparing its three-dimensional structure with those of the apoE2 and E3 proteins (Dong et al., 1994). In the apoE3 isoform, with a Cys at position 112, the Arg61 forms a salt bridge with the Glu109. In the case of apoE4, with an Arg at position 112, the only difference is the establishment of a new salt bridge between Arg112 and Glu109. This substitution in the apoE4 isoform did not modify extensively the three-dimensional structure of the N-terminal domain of apoE. However, due to this substitution in apoE4, the Arg61 residue was freed to form another salt bridge with the Glu255, localized in the C-terminal domain of apoE (Dong and Weisgraber, 1996). This interaction does not exist in the apoE2 and E3 isoforms. The authors concluded that the Cys/Arg substitution in apoE4 leads to the subsequent formation of a salt bridge between Arg61 and Glu255, which is an intrinsic property apoE4 and might modify the overall three-dimensional structure of the apoE4, as compared with apoE2 and E3. The Cys/Arg substitution at position 112 in apoE4 might induce specific interactions between the N- and C-terminal domains of apoE4, preventing the binding of the C-terminal domain of apoE4 with Aβ. Moreover, we have clearly established that the preferential apoE motive involved in Aβ binding is a pair of amphipathic α-helices, especially residues 248-286. This apoE peptide contains the Glu255 implicated in a salt bridge in apoE4 but not in the apoE2 and E3 isoforms. Thus, it is possible that the protective role ascribed to apoE2 and E3 is related to the specific three-dimensional organization of the apoE isoforms, modulating the interaction of the C-terminal domain of apoE with Aβ.