How Cholesterol Constrains Glycolipid Conformation for Optimal Recognition of Alzheimer's β Amyloid Peptide (Aβ1-40)

Membrane lipids play a pivotal role in the pathogenesis of Alzheimer's disease, which is associated with conformational changes, oligomerization and/or aggregation of Alzheimer's β-amyloid (Aβ) peptides. Yet conflicting data have been reported on the respective effect of cholesterol and glycosphingolipids (GSLs) on the supramolecular assembly of Aβ peptides. The aim of the present study was to unravel the molecular mechanisms by which cholesterol modulates the interaction between Aβ1–40 and chemically defined GSLs (GalCer, LacCer, GM1, GM3). Using the Langmuir monolayer technique, we show that Aβ1–40 selectively binds to GSLs containing a 2-OH group in the acyl chain of the ceramide backbone (HFA-GSLs). In contrast, Aβ1–40 did not interact with GSLs containing a nonhydroxylated fatty acid (NFA-GSLs). Cholesterol inhibited the interaction of Aβ1–40 with HFA-GSLs, through dilution of the GSL in the monolayer, but rendered the initially inactive NFA-GSLs competent for Aβ1–40 binding. Both crystallographic data and molecular dynamics simulations suggested that the active conformation of HFA-GSL involves a H-bond network that restricts the orientation of the sugar group of GSLs in a parallel orientation with respect to the membrane. This particular conformation is stabilized by the 2-OH group of the GSL. Correspondingly, the interaction of Aβ1–40 with HFA-GSLs is strongly inhibited by NaF, an efficient competitor of H-bond formation. For NFA-GSLs, this is the OH group of cholesterol that constrains the glycolipid to adopt the active L-shape conformation compatible with sugar-aromatic CH-π stacking interactions involving residue Y10 of Aβ1–40. We conclude that cholesterol can either inhibit or facilitate membrane-Aβ interactions through fine tuning of glycosphingolipid conformation. These data shed some light on the complex molecular interplay between cell surface GSLs, cholesterol and Aβ peptides, and on the influence of this molecular ballet on Aβ-membrane interactions.


Introduction
Alzheimer's disease is a neurodegenerative pathology of the central nervous system currently affecting more than 25 millions of individuals worldwide. It is characterized by the presence of neuritic plaques and neurofibrillary tangles which contribute to neuronal and synaptic loss. Although the molecular and cellular mechanisms responsible for Alzheimer's disease are not fully understood, the formation of insoluble deposit of the b-amyloid peptide (Ab) fragments seems to play a major role in the pathogenesis of the disease [1,2]. What we call Ab is in fact a family of peptides derived by the proteolytic cleavage of the amyloid precursor protein (APP), a transmembrane protein expressed in neuronal but also in non-neuronal tissue [3]. The most abundant forms of Ab are 40-and 42-amino acid peptides respectively referred to as Ab  and Ab  . Both peptides are found in amyloid plaques that, according to the amyloid cascade hypothesis [2,4], eventually lead to the neurodegeneration. The aggregation process of amyloid peptides is driven by the supramolecular assembly of beta sheet structures. Non-fibrillar oligomers of Ab are also toxic [5], and it is difficult to assess which molecular species among dimers, oligomers and fibrils are the most pathogenic.
Given that these amyloid peptides are released in the immediate vicinity of the plasma membrane in which APP is anchored, it is not surprising that several membrane lipids can interact with Ab and affect the oligomerization process [6]. Lipid rafts, which are membrane microdomains enriched in cholesterol and sphingolipids [7], have been proposed to act as surface catalysts able to accelerate the aggregation of Ab [8]. Ab peptides interact with several glycosphingolipids (GSLs), including both neutral GSLs such as asialo-GM1 [9] or galactosylceramide (GalCer) [10,11] and gangliosides such as GM1 [12]. Conflicting data have suggested that binding to GM1 induces either a conformational transition from random coil to a protective a-helical structure [13] or to a fibrillationprone b-structure [14,15]. As a matter of fact, the structure of Ab bound to GM1 depends on several physicochemical parameters including pH, membrane fluidity, GM1-carrier lipid ratios, Ab concentration, and the presence of cholesterol, which could explain the discrepancies reported in the literature.
Although cholesterol has been identified as a major risk factor for Alzheimer's disease [16], its effects on Ab fibrillogenesis and toxicity are not well understood and the results reported so far are controversial [17]. Cholesterol stimulates the insertion of APP into phospholipid monolayers [18] and it binds to Ab 1-42 protofibrils [19]. However, whether cholesterol accelerates [20] or decreases [21,22] Ab polymerization is still uncertain. Moreover, the generation of Ab peptides through APP proteolysis occurs within lipid rafts and is sensitive to inhibitors of cholesterol biosynthesis [16], so that the involvement of cholesterol homeostasis in Alzheimer's disease cannot be simply ascribed to the regulation of Ab fibrillogenesis.
There is one important issue that has not been addressed in these investigations. GSLs exhibit a high biochemical diversity in both their glycone and ceramide moieties [23]. Perhaps the most widely neglected biochemical characteristic of GSLs in amyloid studies is the existence of two distinct types of ceramide backbones which are defined by the absence or presence of an OH group bound to the C2 of the fatty acid chain [24]. Thus, major brain GSLs such as GalCer actually exist as 2-hydroxy (2-OH) fatty acid (HFA) or nonhydroxy fatty acid (NFA) species, in roughly equivalent amounts [25]. The hydroxylation status of the ceramide has a major impact on the conformation of the GSL, as elegantly demonstrated by I. Pascher and co-workers for GalCer [26,27]. An intramolecular Hbond network in GalCer-HFA stabilizes the parallel orientation of the galactose ring with respect to the lipid membrane, giving the molecule a typical L-shape [27]. In contrast, the lack of this OH group in GalCer-NFA allows the galactose head group to freely adopt an extended conformation in a layer-perpendicular position. Proteins that recognize the L-shape conformation of GalCer-HFA can totally ignore GalCer-NFA, as perfectly demonstrated for the HIV-1 surface envelope glycoprotein gp120 [28].
Here we studied the interaction of Ab 1-40 with various HFA and NFA GSLs and the effects of cholesterol on these interactions. We show that Ab 1-40 has a marked preference for HFA vs. NFA GSLs. This effect was reproducibly observed with both natural and synthetic GSLs with defined fatty acid content. Cholesterol was found to inhibit the interaction of Ab 1-40 with the HFA species but to strongly stimulate the interaction with NFA GSLs. Using a combination of physicochemical and molecular modeling approaches, we show that the 2-OH group of the fatty acid and the OH group of cholesterol have a similar conformational effect on various GSLs.

Surface Pressure Measurements
Surface pressure measurements revealing peptide-lipid interactions were studied by the Langmuir film balance technique with a fully automated microtensiometer (mTROUGH SX, Kibron Inc. Helsinki, Finland). Indeed, the interaction of a peptide (or a protein) with a glycolipid monolayer is an interfacial phenomenon which can be studied by surface pressure (p) measurements [6]. Namely, the insertion of the peptide in the lipid monolayer can be detected, at constant area, by an increase in the surface pressure (Dp). This increase in the surface pressure is caused by the insertion of the peptide between the polar heads of vicinal glycolipids in the monolayer, which is not counterbalanced by an increase of the area of the monolayer. This effect can be followed kinetically by real-time surface pressure measurements after injecting the peptide into the aqueous subphase underneath the lipid monolayer as described previously [29]. Such [Dp = f(t)] curves allow to assess the actual interaction of a peptide (or a soluble protein) with a given glycolipid. The initial velocity (v i ) of the insertion process is expressed as mN.m 21 .min 21 . The difference between the maximal (p max ) and the initial (p i ) surface pressure values allows to calculate the maximal surface pressure increase (Dp max ) induced by the peptide (expressed in mN.m 21 ). Mixed monolayers were prepared from stock solutions of lipid mixtures (1:1, mol:mol). Unless otherwise indicated, the subphase underneath the lipid monolayer is ultrapure water (pH 6.9 for all experiments). In any case, the subphase was prepared extemporaneously from disposable sterile units. All experiments were conducted at 20uC in triplicate. Results were expressed as means 6 standard deviation (S.D.).

Molecular Modeling Simulations
The structure of Ab 1-40 has been retrieved from the PDB entry # 1BA4 [30]. The structure of GSLs were derived from those published by Pascher & Sundell [26]. The whole structure of Ab 1-40 was merged with GalCer-HFA or GalCer-NFA + cholesterol. Molecular dynamics studies of GSLs and Ab-GSL interactions were performed with the Hyperchem program, using with the MM+ force field as described previously [29,31]. The Polak-Ribiere algorithm was used to calculate the minimal energy of the individual molecules (GSLs and peptides) in each system. Molecular dynamics simulations were then performed for 1 ps. The introduction of water molecules in the periodic box did not influence the establishment of sugar-aromatic stacking interactions [29]. Indeed, similar data were obtained when the modeling of this interaction was performed in vacuo and in water. To improve clarity, water molecules were not represented in the molecular models.

Chemical Structures
All chemical structures were drawn with ChemDraw.

Effect of Cholesterol on the Interaction between GalCer and Ab 1-40
In a first set of experiments we investigated the interaction of Ab 1-40 with GalCer-HFA ( Figure 1a) and GalCer-NFA ( Figure 1b) purified from bovine brain (see Table 1 for a description of the fatty acid content of the GSLs used in the present study, and Figure 1 for GalCer structures). We used the Langmuir monolayer technique which has been widely validated for lipid-protein studies and has proven particularly useful for reconstituting sphingolipid-cholesterol systems (for review, see [6]). Monolayers of these GSLs were prepared at the air-water interface on a subphase of ultrapure water (the use of water subphases for such surface pressure studies has been extensively characterized and validated in previous studies [28,29,31]). The peptide was injected in the subphase and the surface pressure was continuously measured with a microtensiometer. As shown in Figure 1a, the surface pressure started to increase immediately after the injection of the peptide underneath the GalCer-HFA monolayer, with an initial velocity (v i ) of 0.26660.023 mN.m 21 .min 21 (n = 3, Table 2). A maximal value of 12.260.9 mN.m 21 (n = 3) was reached after 65 min of interaction (Dp max ). In contrast, the peptide interacted very weakly with a monolayer of GalCer-NFA ( Figure 1b). Following an initial slow increase (v i = 0.0336 0.004 mN.m 21 .min 21 (n = 3, Table 2), the surface pressure peaked after 18 min of interaction (Dp max = 1.560.1 mN.m 21 , n = 3, Table 2). This indicates that Ab 1-40 has a high affinity for GalCer-HFA and a very low affinity for GalCer-NFA. The presence of cholesterol in these monolayers (i.e. mixed GalCer/ cholesterol monolayers) had opposite effects for both GalCer species. In the case of GalCer-HFA (Figure 1a), cholesterol markedly inhibited the interaction of Ab 1-40 with the monolayer (v i = 0.03160.004 mN.m 21 .min 21 ; Dp max = 3.260.2 mN.m 21 , n = 3, Table 2). Compared with pure monolayers of GalCer-HFA, this corresponds to a 8.1-fold decrease in v i and a 3.8-fold decrease in Dp max . In contrast, cholesterol markedly stimulated the interaction of Ab 1-40 with GalCer-NFA ( Figure 1b). During the first 15 min following the injection of the peptide underneath a mixed GalCer-NFA/cholesterol monolayer, the interaction occurred with a v i = 0.10760.008 mN.m 21 .min 21 (n = 3, Table 2). Then the velocity of the interaction further increased to reach 0.30860.027 mN.m 21 .min 21 (n = 3). A Dp max of 10.4 mN.m 21 6 0.6 (n = 3, Table 2) was reached after 47 min of incubation, corresponding to a 6.9-fold increase. That the interaction proceeded as a two-step phenomenon with an initial slow velocity followed by a more rapid one suggests that a conformational adjustment is required for an optimal interaction.
The data obtained with natural GalCer-NFA purified from bovine brain were fully confirmed with synthetic GalCer-NFA, i.e. GalCer with a C12:0 nonhydroxylated fatty acid ( Figure 1c). Indeed, in absence of cholesterol, the surface pressure moderately increased after the injection of Ab 1-40 (v i = 0.09160.008 mN.m 21 .min 21 , n=3, Table 2), then reached a plateau at 2.160.1 mN.m 21 (n = 3, Table 2) after 20 min and finally decreased to null values after 60 min of incubation. This decrease could be due to the secondary release of the small amounts of peptide initially adsorbed onto the monolayer, consistent with the low affinity of Ab for GalCer-C12. As for natural GalCer-NFA, cholesterol stimulated the interaction between Ab and synthetic GalCer-C12 (v i = 0.11760.009 mN.m 21 . min 21 ; Dp max = 12.261.2 mN.m 21 , n=3, Table 2). This corresponded to a strong stimulation of Dp max (65.8 times) with only a minor effect on v i (61.28 times), in perfect agreement with the data obtained with natural GalCer-NFA. Indeed, as for GalCer-NFA, the interaction followed a two-step kinetic with an initial slow velocity ( Table 2) followed by a more rapid one (0.20360.014 mN.m 21 .min 21 , n = 3, Figure 1c).
Several control experiments were conducted with various compositions of the lipid monolayer ( Table 2). Firstly, we demonstrated that cholesterol by itself, at the surface pressure used in the experiments described in Figure 1 Table 2).

Effect of Cholesterol on the Interaction between Gangliosides and Ab 1-40
Since gangliosides are known to bind to Alzheimer's amyloid peptides and to affect their aggregation [12,13,32,33], we studied the effect of cholesterol on the interaction between Ab 1-40 and the monosialylated gangliosides GM3 and GM1 ( Figure 3). These gangliosides, which were purified from bovine brain (GM1) or buttermilk (GM3), contained only NFA species (see Table 1 for fatty acid content and Figure 3 for chemical structures). In absence of cholesterol, both gangliosides were recognized by Ab 1-40 as shown by the kinetics of surface pressure increase after Table 1. Typical fatty acid composition of the natural glycosphingolipids used in this study*.

Fatty Acids
GalCer-NFA GalCer-HFA GM3 GM1  injection of Ab 1-40 underneath a monolayer of GM3 (Figure 3a) or GM1 (Figure 3b). However, the values of Dp max were of low magnitude ( Table 2), which indicated a moderate affinity of the peptide for these gangliosides compared with GalCer-HFA. As for GalCer-NFA and GalCer-C12, cholesterol had a stimulatory effect on Ab 1-40 insertion within GM3 (Figure 3a) and, although to a lesser extent, GM1 monolayers ( Figure 3b). Namely, cholesterol induced a 2.5-fold increase for GM3 and a 1.55-fold increase for GM1, as assessed by comparing the Dp max values. A detailed analysis of the kinetic parameters of these interactions is presented in Table 2.
Overall, these data indicated that cholesterol reproducibly facilitates the interaction of Ab 1-40 with all the GSLs that have a non-hydroxylated fatty acid. This effect was reproducibly observed with various NFA-GSLs (GalCer-NFA, GalCer-C12, LacCer-C8, GM3, and GM1) containing biochemically distinct fatty acids (Table 1). On the opposite, cholesterol decreases the interaction of Ab 1-40 with GalCer-HFA. Thus the OH group linked to the C2 of the fatty acid seems to play a critical role for Ab 1-40 recognition by GSLs, and cholesterol probably acts at this level. To unravel the role of cholesterol on GSL conformation in the context of fatty acid hydroxylation, we performed a series of molecular modeling simulations.

Molecular Modeling Simulations of Cholesterol-GSL Interactions
X-ray diffraction studies have revealed that the galactose ring of GalCer-NFA protrudes at 180u with respect to the main axis of the ceramide backbone [27]. Molecular dynamics studies of GalCer-NFA remarkably converged to the same type of conformation (Figure 4a). A remarkable fit between GalCer-NFA and cholesterol could be found (Figure 4b), the complex being stabilized by both van der Waals forces and H-bonds. In particular, a H-bond network involving i) the OH group of cholesterol, ii) the NH of sphingosine, and iii) the oxygen atom of the glycosidic bond, could be predicted (Figure 4g-h). To better illustrate the effect of cholesterol on the conformation of GalCer-NFA, the GSL alone (in green) has been superimposed on the structure of the GalCer-NFA/cholesterol complex (Figure 4d-e).
In GalCer-HFA, the 2-OH group restricts the conformation of the galactose ring so that the molecule adopts a typical L-shape structure (Figure 4c). This is in full agreement with crystallographic data which demonstrated that the orientation of the galactose ring in GalCer-HFA is constrained by a network of Hbonds involving i) the 2-OH group of the acyl chain, ii) the NH of sphingosine, and iii) the oxygen atom of the glycosidic bond [26,27]. This H-bond network is shown in Figure 4f which shows the original structure of GalCer-HFA deduced from X-ray diffraction studies [26]. Overall, these data strongly suggest that the 2-OH group of the acyl chain and cholesterol have a comparable conformational effect on GalCer. The molecular mechanism is similar but a detailed analysis shows that in the case of GalCer-HFA, the 2-OH group is a H-bond acceptor (Figure 4f) whereas the OH group of cholesterol is a H-bond donor (Figure 4g-h). Yet in both cases, the OH group reinforces the strength of the H-bond between the NH group of sphingosine and the oxygen atom of the glycosidic bond, thereby forcing the GSL to adopt an L-shape structure.

Molecular Modeling Simulations of Ab 1-40 -GSL Interactions
Our physicochemical studies (Figures 1-3) suggested that only the OH-constrained L-shape structure of GalCer is recognized by Ab 1-40 . So we performed a series of molecular modeling simulations to investigate the possibilities of interaction between the amyloid peptide and each type of GalCer ( Figure 5). In these experiments, we worked with the structure of Ab 1-40 which has been obtained by NMR studies of micellar systems [30]. We did not find any significant fit for Ab 1-40 and a single molecule of GalCer-NFA. We observed that individual GalCer-NFA molecules could readily form a patched structure, each galactose ring stacked onto its neighbours (Figure 5a-b). Even in this case, there was no obvious fit between these GalCer-NFA molecules and the peptide. This suggests that the peptide and the GSL do not have any geometric nor chemically compatible domains. Indeed, the only possible mode of interaction between Ab and a cluster of GalCer-NFA was a single H-bond which left the aromatic side chain of residue Y10 fully exposed to the solvent (Figure 5b).
In contrast, the planar apolar surface of galactose in GalCer-HFA [31] allowed the establishment of a coordinated CH-p stacking interaction with the aromatic side chain of the Y10 residue (Figure 5c-d). As a consequence, the whole structure of Ab 1-40 spread onto the surface of the reconstituted GalCer-HFA membrane. In this orientation, residues H14 and F20 are facing the membrane, whereas residue F19 is rejected on the opposite side. A similar interaction could also be predicted between the conformer of GalCer-NFA constrained by cholesterol and the

Physicochemical Experiments in Support of Molecular Modeling Data
Several experiments were performed in order to confirm the interpretations based on the molecular modeling studies. Firstly, we used NaF (from 0.1 to 1M) as a breaker of H bonds, this effect being due to the high electronegativity of the fluor atom [34]. In presence of NaF, there was a marked inhibition of both v i and Dp max in the monolayer assay, showing that destabilizing the hydrogen bond network decreased the affinity of Ab for GalCer-HFA ( Figure 6 and Table 2). Secondly, we studied the effect of high concentrations of NaCl (from 0.1 to 1M) on the interaction between Ab 1-40 and GalCer-HFA. By increasing the ionic strength, this salt destabilizes electrostatic interactions whereas it reinforces hydrophobic interactions. As shown in Figure 6 and Table 2, the presence of NaCl in the subphase resulted in a marked increase in Ab 1-40 -GalCer-HFA association, as assessed by both the values of v i and Dp max . These data suggested that hydrophobic forces, rather than electrostatic interactions, are involved in the stabilization of the peptide-GSL complex. This is consistent with the hydrophobic stacking interaction between the galactose ring of GalCer and the aromatic side chain of Y10 in Ab ( Figure 5). However, the stimulatory effect of NaCl could also be interpreted as a salting out effect of the peptide, i.e. an increased concentration of the peptide at the interface which could change the surface pressure independently of the association between Ab and GalCer-HFA. To investigate this possibility, we studied the effect of various NaCl concentrations on the spontaneous tensioactivity of Ab  . In this experiment Ab 1-40 (5 mM, i.e. 5 times the concentration used to study Ab-GSL interactions) was injected in the water phase containing increasing concentrations of NaCl (ranging from 100 to 1,000 mM). No lipid was spread at the interface so that any change in the surface pressure reflected exclusively the interfacial recruitment of Ab 1-40 . The surface pressure increase measured after 65 min of incubation was plotted against the NaCl concentration (Figure 7). A maximal value of 8.1 mN.m 21 was obtained between 200 and 1,000 mM NaCl. At 100 mM NaCl, Ab 1-40 increased the surface pressure by only 2.1 mN.m 21 (to be compared with the data of Figure 6). Thus, although the above-mentioned salting out effect did exist, it could not, by itself, account for the stimulatory effect of NaCl on the association between Ab 1-40 and GalCer-HFA. Overall these physicochemical data were in good agreement with the predictions of our molecular modeling studies, which underscored the role of hydrophobic stacking aromatic interactions (reinforced by NaCl), and H-bonds (destabilized by NaF) in Ab-GalCer interactions.

Discussion
Although it is widely admitted that cholesterol and GSLs play a pivotal role in Ab release, conformation, oligomerization, and fibrillogenesis [16], the molecular mechanisms controlling the complex interplay between Ab peptides and these lipids are not clear. In particular, the biochemical diversity of GSLs in their ceramide moiety, especially the hydroxylation status of the C2 in the acyl chain, has not been appreciated in Ab-GSLs binding studies. We believe that this issue is critical for at least three reasons: i) quantitatively, NFA and HFA species of the same GSL can be expressed in roughly similar amounts in brain tissues, as it is the case for GalCer in myelin [35]; ii) the 2-OH group has a major impact of the conformation of the GSL, due to intramolecular Hbonding possibilities [36]; iii) gender-specific expression of NFA vs. HFA ceramides have been observed in a mouse model of Alzheimer's disease, and this biochemical feature could be related to the increased propensity of women to develop Alzheimer's disease [37].
Using the Langmuir monolayer technique [6], we studied the interaction between Ab 1-40 and various HFA-and NFA-GSLs with known fatty acid composition. We showed that Ab 1-40 interacted more efficiently with a pure monolayer of GalCer-HFA than with GalCer-NFA, as assessed by both the initial velocity of the insertion reaction and the increase in the surface pressure induced by the peptide (Figure 1). Similar results were previously obtained with HIV-1 gp120, a protein sharing with Ab peptides a common sphingolipid-binding domain (SBD) involved in GalCer recognition [10]. On the basis of these data, Hebbar et al. [11] have designed a fluorescent SBD probe targeting lipid raft domains on live cells. This probe, derived from the part of Ab containing the SBD we have defined [10,38], recognized a mixture of bovine brain GalCer containing both GalCer-HFA and GalCer-NFA, but not synthetic GalCer-NFA. Therefore, our data are in perfect agreement with those of Hebbar et al. [11]. No obvious fit could be found, the only predicted interaction being the H-bond between the phenolic OH group of Y10 and the CH 2 OH group of one of the galactose rings. Note that the aromatic side chain of Y10 cannot stack onto any galactose ring, none being accessible. The stacking of the galactose headgroups of vicinal GalCer-NFA molecules (Gal-Gal stacking) is indicated by an arrow. c-CH-p stacking interaction between the galactose ring of GalCer-HFA and the aromatic side chain of the Y10 residue in Ab  . Note that the peculiar geometry of the complex leaves residues H14 and F20 accesible for complementary interactions with membrane lipids, whereas residue F19 is rejected on the opposite side. d-Detailed view of the complex between GalCer-HFA and Ab (to improve clarity only residues 8-22 of Ab 1-40 are shown). The perfect geometry of the CH-p stacking interaction between the galactose ring and the aromatic side chain of Y10 is illustrated in the inset. e-In presence of cholesterol (Chol), the galactose headgroup of GalCer-NFAadopts a specific conformation which renders the galactose ring accessible for a CH-p stacking interaction with the Y10 residue of Ab  . f-Detailed view of the complex between cholesterol, GalCer-NFA, and Ab (residues 8-22 of Ab    In the present study, the interaction of Ab 1-40 with lipid monolayers has been performed at a surface pressure (13-15 mN.m 21 ) that is lower than the value of 30 mN.m 21 currently considered as representative of natural membrane bilayers [39]. The rationale for working at 13-15 mN.m 21 is that at these pressures, Ab 1-40 does not penetrate into pure cholesterol monolayers ( Table 2). This allows a clear-cut assessment of the effect of cholesterol on the interaction of Ab 1-40 with various natural and synthetic GSL species. Previous data indicated that Ab 1-40 readily interacts with a monolayer of GalCer-HFA with an initial surface pressure of 30 mN.m 21 [10]. In contrast, under the same conditions, the peptide showed very little interaction with GalCer-NFA and GalCer-C12 monolayers (data not shown). Therefore, one of the most important findings of the present study (strong interaction of Ab 1-40 with GalCer-HFA monolayers, weak interaction with GalCer-NFA and GalCer-C12) was confirmed at high surface pressure values that are representative for natural biomembranes.
The other major finding of this study is that cholesterol could transform the inactive GalCer-NFA into a fully active GSLcholesterol complex recognized by Ab 1-40 . This stimulatory effect of cholesterol was observed with both natural and synthetic GalCer-NFA ( Table 2). Control experiments with cholesterol alone and with PS:cholesterol mixed monolayers confirmed that the peptide specifically recognized GalCer-NFA complexed with cholesterol ( Table 2). The previously established 3D structures of GalCer-NFA and GalCer-HFA [26] immediately suggests a molecular mechanism accounting for these physicochemical data. In GalCer-HFA, the galactose ring is parallel to the membrane, consistent with the establishment of a typical CH-p stacking interaction [40] between the apolar face of galactose and an aromatic side chain of the peptide [29]. This type of interaction has been evidenced for various sugar-protein complexes, including sugar-lectin and GSL-peptide systems [41]. Structure similarity searches combined with various physicochemical approaches suggested that the interaction between GalCer-HFA and Ab involved the aromatic side chain of Y10 [10]. Interestingly Hebbar et al. [11] who showed that a double mutant Ab peptide (R5A/ Y10A) did not bind to GalCer. Altogether, these data are in line with the molecular modeling study of Figure 5, in which we assigned a major role for Y10 in GalCer recognition. We conclude that the binding of Ab to GalCer requires that the galactose ring of the GSL is in a parallel orientation with respect to the membrane, the GSL adopting a typical L-shape. This conformation can be stabilized by an intramolecular H-bond network involving the 2-OH group of GalCer-HFA. In absence of this H-bond network, GalCer-NFA cannot adopt this active conformation (Figures 4-5). The galactose rings are not accessible to the Y10 residue, and there is no possibility for a stable interaction with Ab. Yet cholesterol can transform this inactive conformation by forcing the galactose ring to adopt a parallel orientation compatible with the stacking of Y10. As expected, this hydrophobic CH-p stacking interaction was reinforced in presence of high concentrations of NaCl in the subphase underneath the monolayer ( Figure 6 and Table 2). Finally, destabilizing the intramolecular H-bond network with NaF strongly impaired the insertion of Ab, which strongly supported the conclusions drawn on the basis of our modeling studies. In total agreement with our data, Ikeda & Matsuzaki [42] also identified H-bonding and hydrophobic interactions as the driving forces responsible for Ab 1-40 -glycolipid interactions in lipid bilayer systems.
The specific effect of cholesterol on GSL conformation is not necessarily related to the condensing activity of the sterol on GSL monolayers. Indeed, comparative studies of mixed monolayers suggested that at 22uC, cholesterol displays a high condensing effect on dihexosylceramides, but on GalCer monolayers [43]. At 37uC, cholesterol was found to specifically condense GalCer-NFA, but not GalCer-HFA [24]. Thus our data cannot be interpreted in terms of cholesterol-induced condensation of GSL species, but definitely in terms of GSL conformation, which can be finely tuned by vicinal cholesterol molecules. That the incorporation of cholesterol in a GalCer-NFA monolayer could modulate the average orientation of the sphingolipid is in line with the subsequent studies of GalCer-cholesterol interactions conducted by Smaby et al. [44].
Finally, the regulatory activity of cholesterol on the conformation of GSL and the resulting impact on Ab recognition was also observed for more complex GSLs such as LacCer, GM3 and GM1. This suggests that cholesterol exerts a wide regulatory activity on Ab-GSL interactions, whose direction depends on the hydroxylation status of the fatty acid chain of the GSL: stimulation of Ab binding for NFA-GSLs, inhibition of Ab binding for HFA-GSLs. That ''variations in GSL fatty acid composition may mediate aglycone regulation of GSL membrane receptor function by a differential interaction with cholesterol and other membrane components'' has been recently discussed by Lingwood et al. [45]. As stated above, fatty acid hydroxylation of GSLs is generally associated with improved ligand binding [11,28,29,[46][47][48], and this effect has been correlated with the conformation of the sugar head group of GSLs based on the crystallographic studies published by Pascher and co-workers [26,27]. In line with all these studies, our data demonstrate that the ceramide part of GSLs is critical for Ab binding, and that cholesterol can, for those GSLs which are normally not recognized by Ab (i.e. NFA-GSLs), forces them to adopt a conformation compatible with Ab. This fine tuning of GSL conformation may be highly relevant to the neuropathology of Alzheimer's disease.