The minimal amyloid-forming fragment of the islet amyloid polypeptide is a glycolipid-binding domain



This article is corrected by:

  1. Errata: Corrigendum Volume 274, Issue 7, 1878, Article first published online: 13 March 2007

J. Fantini, Université Paul Cézanne, Laboratoire de Biochimie et Physicochimie des Membranes Biologiques, Faculté des Sciences de St-Jérôme, 13397 Marseille Cedex 20, France
Fax: +33 491 288 236
Tel: +33 491 28 27 54


Several proteins that interact with cell surface glycolipids share a common fold with a solvent-exposed aromatic residue that stacks onto a sugar ring of the glycolipid (CH–π stacking interaction). Stacking interactions between aromatic residues (π–π stacking) also play a pivotal role in the assembly process, including many cases of amyloid fibril formation. We found a structural similarity between a typical glycolipid-binding domain (the V3 loop of HIV-1 gp120) and the minimal amyloid-forming fragment of the human islet amyloid polypeptide, i.e. the octapeptide core module NFGAILSS. In a monolayer assay at the air–water interface, the NFGAILSS peptide specifically interacted with the glycolipid lactosylceramide. The interaction appears to require an aromatic residue, as NLGAILSS was poorly recognized by lactosylceramide, whereas NYGAILSS behaved like NFGAILSS. In addition, we observed that the full-length human islet amyloid polypeptide (1–37) did interact with a monolayer of lactosylceramide, and that the glycolipid film significantly affected the aggregation process of the peptide. As glycolipid–V3 interactions are efficiently inhibited by suramin, a polyaromatic compound, we investigated the effects of suramin on amyloid formation by human islet amyloid polypeptide. We found that suramin inhibited amyloid fibril formation at low concentrations, but dramatically stimulated the process at high concentrations. Taken togther, our results indicate that the minimal amyloid-forming fragment of human islet amyloid polypeptide is a glycolipid-binding domain, and provide further experimental support for the role of aromatic π–π and CH–π stacking interactions in the molecular control of the amyloidogenesis process.


maximal variation of surface pressure






human islet amyloid polypeptide




initial surface pressure


prion protein


transmission electron microscopy


thioflavin T

Stacking interactions, which consist of overlapping arrangements of parallel planar molecules at van der Waals distances of separation, play a major role in biology. On the one hand, base stacking is an enthalpy-driven process that significantly contributes to the stability of the DNA double-helical conformation in water [1]. On the other hand, the three-dimensional structure of globular proteins is often stabilized by entropy-driven π–π stacking interactions between aromatic residues located in their hydrophobic core [2]. Stacking interactions can also involve surface-exposed aromatic amino acid residues. Computational studies have recently shown that solvent-accessible aromatic residues generally form part of a binding site for an external ligand [3]. This is probably due to the weak interactions between aromatic rings and water molecules, which can be advantageously replaced by more stable stacking interactions. In this respect, π–π stacking interactions between surface-exposed aromatic residues were suggested to play a major role in the aggregation process of amyloidogenic peptides [4]. The first experimental demonstration of this hypothesis was provided by substitution of the Phe residue with Ala in the context of a minimal amyloid-forming octapeptide (NFGAILSS) of the islet amyloid polypeptide. This relatively minor substitution between two hydrophobic residues induced a dramatic decrease in its amyloidogenic potential to form fibrils [5]. In contrast, replacing each of the other residue with alanine had no effect. Later studies that used high-resolution NMR and X-ray tools provided further direct experimental support for the key role of aromatic interactions in the assembly of various amyloidogenic polypeptides [6–8]. The role of aromatic interactions in the hierarchical assembly of the amyloidal structures is also supported by molecular dynamics simulations performed by various groups [9–11].

Surface-exposed aromatic residues have also been shown to be involved in the binding of proteins to the sugar headgroup of glycolipids [12]. In this case, the six-carbon sugar ring belonging to the polar head of the glycolipid receptor provides a complementary surface for the aromatic side chain. The stacking interaction is driven by the proximity of the aliphatic protons of the sugar ring, which carry a net positive partial charge, and the π-electron cloud of the aromatic ring. It can thus be considered as a limiting case of hydrogen bonding, in which the acceptor group is the electron cloud of the aromatic ring and the donor group is a C–H of the sugar ring. Thus, it is referred to as a CH–π interaction [13]. This particular type of stacking interaction appears to play an important role in the recognition of glycolipids by several cellular and pathogen proteins, including the HIV-1 surface envelope glycoprotein gp120, the prion protein, and the β-amyloid peptide of Alzheimer's disease [14,15]. The last of these findings suggests that there may be structural and functional homology between glycolipid-binding domains and amyloid peptides. Indeed, it is striking that the hallmark of these biochemical structures is a surface-exposed aromatic residue.

This may have profound biological significance. By providing a complementary surface for amyloid and/or pathogenic proteins, cellular glycolipids may constitutively stabilize their ‘inoffensive’ conformation and inhibit their aggregation [12]. Replacing these protective heterologous CH–π interactions with homologous π–π interactions would then trigger the pathologic process [16]. A theoretical model accounting for the chaperone/protective activity of membrane glycolipids for amyloid peptides and prions has been recently proposed [12]. The identification of a short amyloid core peptide derived from human islet amyloid peptide (hIAPP) gave us an opportunity to establish a correlation between the potential to form amyloid fibrils (a typical π–π stacking process) and glycolipid recognition (CH–π interaction). In the present article, we show that the minimal NFGAILSS amyloid core peptide derived from hIAPP is also the shortest glycolipid-binding domain so far characterized. We also show that suramin, a polyaromatic compound known to bind to the glycolipid-binding domain of HIV-1 gp120, has a bimodal effect on amyloid formation by hIAPP. Taken together, our data support the view that the surface-exposed aromatic residues of amyloid peptides may be involved in both π–π and CH–π stacking interactions.


Structural similarity between hIAPP and the glycolipid-binding domain (HIV-1 gp120 V3 loop)

The three-dimensional structures of both the hIAPP20−29 peptide and the V3 loop of HIV-1 gp120 have been studied in solution by NMR spectroscopy [17,18], and the atomic coordinates were downloaded from the Protein Databank (entries 1CE4 and 1KUW, respectively, for V3 and hIAPP20−29). As for hIAPP20−29, three distinct families of conformers could be identified (Fig. 1A,B). In all cases, the peptide adopted a distorted turn structure, leaving the aromatic side chain of the Phe23 residue available for a potential intermolecular interaction. Therefore, the three-dimensional structure of the glycolipid-binding domain of the HIV-1 gp120 V3 loop could be superposed on each type of hIAPP conformer. An example of such superposition is shown for the hIAPP conformer C3 (Fig. 1C). According to this structural study, a common motif consisting of a turn with a solvent-accessible aromatic residue (Phe20 for V3, Phe23 for hIAPP) was observed in the three-dimensional structure of these protein domains, which do not share any obvious sequence homology. As a matter of fact, the turn is induced by the presence of Gly and/or Pro residues in the motif. Molecular mechanics simulations were then conducted in order to generate a model of interaction between the glycolipid galactosylceramide (GalCer) and either V3 or hIAPP (Fig. 1D). One should note that, in both cases, the sugar ring of galactose in the polar head of the glycolipid provides a complementary surface for the aromatic side chain of Phe, allowing the establishment of a typical CH–π interaction. This suggested that hIAPP possesses a structural glycolipid-binding domain containing the Phe23 residue. As the hIAPP22−29 (NFGAILSS) octapeptide is the minimal amyloid-forming fragment of hIAPP, these modeling studies raised the intriguing possibility that this motif could also be a glycolipid-binding domain.

Figure 1.

 Structural similarity between the glycolipid-binding domain (HIV-1 gp120 V3 loop) and hIAPP20−29. (A) Superposition of the backbone of the 40 lowest-energy conformers of the hIAPP peptide20−29 as determined by two-dimensional solution NMR spectroscopy according to Mascioni et al. [17]. (B) Structure of three hIAPP20−29 conformers, C1, C2 and C3, each representative of a distinct family of conformers. In all cases, the peptide adopts a turn structure with the aromatic side chain of Phe23 oriented towards the solvent. (C) Comparison of HIV-1 gp120 V3 loop (in orange) and hIAPP conformer C3 (in blue) structures and sequences. (D) Molecular modeling of V3 and hIAPP complexed with GalCer (molecular mechanics simulations). The Protein Data Bank entries used were 1CE4 (V3) and 1KUW (hIAPP fragment).

Specific interaction between hIAPP22−29 and glycolipids

The ability of hIAPP to interact with glycolipids was assessed by incubating the synthetic NFGAILSS peptide with a monolayer of lactosylceramide (LacCer) spread at the air–water interface. In this experiment, the monomolecular film of LacCer had a stable surface pressure of 18 mN·m−1. This value was taken as the initial surface pressure (πi) of the monolayer. As shown in Fig. 2, addition of NFGAILSS in the aqueous phase underneath the LacCer monolayer induced a dramatic increase in the surface pressure. The maximal surface pressure increase (Δπmax) was 6.25 mN·m−1. This interaction was specific, as the peptide did not alter the pressure of sphingomyelin or phosphatidylcholine monolayers prepared under the same conditions. Conversely, the LacCer monolayer was recognized by a synthetic peptide derived from the human prion protein (PrP) (segment 185–208 of human PrP). This peptide, which shares a structural similarity with the V3 loop of HIV-1 gp120 and the Alzheimer β-amyloid peptide [14], also forms amyloid aggregates [19]. Thus the ability of hIAPP22−29 to interact with monomolecular films of glycolipids is not unique among amyloid peptides. As molecular mechanics simulations suggested a critical role for aromatic residues in glycolipid recognition, we tested the interaction of a peptide analog in which Phe23 was replaced with a Leu residue. This peptide did not interact with a monolayer of LacCer prepared at a πi of 18 mN·m−1 (Fig. 2). These experiments were then performed with LacCer prepared at various values of πi(Fig. 3). For the wild-type NFGAILSS peptide, the interaction with LacCer was still detected at high πi values, and the critical pressure of insertion (extrapolated for Δπmax = 0) was 31 mN·m−1. For comparison, the mean surface pressure of the plasma membrane has been estimated to be 30 mN·m−1[20]. In contrast, the interaction of the peptide analog NLGAILSS with LacCer rapidly decreased as πi increased, with a critical pressure of insertion of < 20 mN·m−1. Thus the aromatic Phe23 residue of NFGAILSS appeared to be critical for LacCer recognition. To further evaluate the importance of the aromatic side chain, we studied the interaction of a peptide analog with a Tyr residue in place of Phe23. As shown in Fig. 3, this peptide behaved exactly like the wild-type NFGAILSS, suggesting that it is the aromatic nature of this residue rather than its exact structure that is needed for glycolipid recognition. This is consistent with the molecular model presented in Fig. 1D, as sugar–aromatic CH–π interactions may occur with both Phe and Tyr side chains.

Figure 2.

 Physicochemical studies of the interaction between amyloid peptides and membrane lipids. (A) Monolayers of LacCer, sphingomyelin or dipalmitoylphosphatidylcholine were prepared at an initial pressure of 18 mN·m−1. After stabilization of the film, the indicated peptide was added to the aqueous subphase at a final concentration of 8 µm. Surface pressure changes were continuously recorded with a microtensiometer. Wild-type hIAPP core peptide NFGAILSS under LacCer (•), sphingomyelin (bsl00084) or phosphatidylcholine (bsl00066) monolayers; NLGAILSS analog under a LacCer monolayer (▪); amyloid prion peptide P1 under a LacCer monolayer (○).

Figure 3.

 Variations of maximal surface pressure increase (Δπmax) as a function of the initial surface pressure (πi) of LacCer monolayers incubated with NFGAILSS (○), NYGAILSS (•), or NLGAILSS (▪).

LacCer interacts with full-length hIAPP1−37 and inhibits its aggregation

In order to check whether the glycolipid-binding domain of hIAPP was accessible and functional in the full-length peptide, we probed a monolayer of LacCer with hIAPP1−37(Fig. 4A). After a lag phase of 20 min, the surface pressure of the film started to increase, and the maximal effect was observed after 2 h. This indicated that the full-length hIAPP recognized the monomolecular film of glycolipid. As the microtensiometer trough is compatible with real-time microscopic obervations, the aggregation process of hIAPP1−37 could be followed under the inverted microscope during surface pressure recordings. As shown in Fig. 4B, the presence of typical aggregates of hIAPP was obvious after 90 min of incubation of the peptide in aqueous solution. In contrast, when the same amount of peptide was injected in the aqueous phase underneath a monolayer of LacCer, the aggregation process of hIAPP was significantly affected. Taken together, these data indicated that the full-length peptide did interact with LacCer, and that the glycolipid could interfere with the aggregation process of hIAPP.

Figure 4.

 Interaction of full-length hIAPP1−37 with LacCer. (A) Kinetics of interaction of full-length hIAPP1−37 (8 µm) with a LacCer monolayer. Phase contrast micrographs of hIAPP1−37 aggregates as observed after 90 min of incubation of the peptide in the microtensiometer trough. (B) Without LacCer. (C) Underneath a LacCer monolayer. Magnification: × 20.

Suramin, a polyaromatic compound that binds to glycolipid-binding domains, may interact with hIAPP22−29

The key role of aromatic residues in CH–π stacking interactions provides a molecular interpretation for the inhibitory activity of water-soluble aromatic compounds on glycolipid–protein interactions [16]. An interesting example of such aromatic compounds is suramin (Fig. 5), which binds to the V3 loop of HIV-1 gp120 and interferes with a wide range of HIV-1–glycolipid interactions [21]. Molecular modeling studies using a molecular mechanics approach allowed us to predict the regions of suramin that could potentially interact with the V3 loop, as well as the physical nature of the interactions involved in the formation of a V3–suramin complex. As shown in Fig. 6A,B, there are three main regions of interaction between V3 and suramin: (a) an electrostatic interaction between Arg18 and one sulfate group of suramin (zone 1); (b) a hydrophobic interaction between the methyl group linked to ring B of suramin and the side chain of Ala19 (zone 2); and (c) a T-shaped π–π stacking interaction involving Phe20 and suramin ring A (zone 3). An hIAPP–suramin complex was also modeled (Fig. 6C,D). Interestingly, in the model obtained, the region of suramin interacting with hIAPP was the same as that involved in V3 binding. Indeed, three zones of interaction could also be predicted in this case: (a) a CH–π interaction between the methylene group of the Asn21 side chain and ring B of suramin (zone 4); and (b) two T-shaped π–π stacking interactions involving Phe23 and suramin rings A and B (zones 5 and 6). Altogether, these data suggested that suramin could interact with hIAPP and thus affect the amyloidogenic properties of this peptide. Indeed, hIAPP fibril formation was efficiently inhibited by phenol red, a small polyphenol molecule [22]. These data prompted us to evaluate the ability of phenol red to interfere with the formation of a complex between suramin and a glycolipid-binding domain. This was analyzed by using a solid-phase radioassay in which the V3 peptide was adsorbed onto a poly(vinyl chloride) surface and probed with [3H]suramin [23]. To ensure that suramin had free access to the V3-binding motif (GPGRAF), we used a multimeric synthetic V3 loop peptide with eight GPGRAF motifs radially branched on an inert matrix [24]. As shown in Fig. 7, phenol red induced dose-dependent inhibition of the binding of [3H]suramin to the glycolipid-binding domain of the HIV-1 gp120 V3 loop. Taken together, these data strongly support the view that glycolipid-binding domains are recognized by both suramin and phenol red through heteroaromatic π–π stacking interactions.

Figure 5.

 Chemical structure of suramin. Figure created using chemdraw (Cambridge Soft, Gainesville, FL, USA).

Figure 6.

 Molecular modeling of suramin complexed with V3 (A,B) or hIAPP fragment (C, D). Numbers 1–6 indicate the main zones of interactions.

Figure 7.

 Phenol red-induced inhibition of [3H]suramin binding to immobilized multivalent V3 loop peptide. The eight-branched multimeric V3 GPGRAF motif was coated onto a poly(vinyl chloride) 96-well plate and then incubated with the indicated concentrations of phenol red. The wells were then incubated with [3H]suramin. The radioactivity associated with the wells at the end of the incubation was counted, and the data were expressed as the mean ± SD of three independent determinations.

Effect of suramin on hIAPP fibril formation

Given the prominent role of π–π stacking interactions in amyloid formation [4,5] and the suppressive effect of various aromatic compounds, including phenol red [22,25], it was logical to investigate whether suramin could inhibit hIAPP amyloid formation. The spontaneous aggregation of hIAPP1−37 alone was followed with a thioflavin T (ThT) assay. Under these conditions, amyloid formation was evidenced by a fluorescence increase occurring after a lag phase of approximately 20 h (Fig. 8A). Addition of suramin to hIAPP1−37 resulted in opposite effects on fibril formation, according to the concentration of suramin in the assay (Fig. 8). Strong inhibition was observed for suramin concentrations lower than 5 µm (Fig. 8B). However, at concentrations higher than 5 µm, suramin induced a marked increase of fluorescence, indicating significant stimulation of the aggregation process. Morphologic transmission electron microscopy (TEM) studies confirmed the opposite effects of suramin on amyloid formation: stimulation in presence of 50 µm suramin (Fig. 8C,D), and inhibition in the presence of 2 µm and 3 µm suramin (Fig. 8E–G). The bimodal effect of suramin on amyloid formation was particularly evident when the percentage of fluorescence was plotted as a function of suramin concentration (Fig. 9). Indeed, it is clearly apparent in Fig. 9 that suramin induced dose-dependent inhibition of hIAPP1−37 fibril formation in the 2–4 µm range, and then induced dose-dependent activation of amyloid formation at concentrations higher than 5 µm.

Figure 8.

 Aggregation of the hIAPP peptide1−37 in the presence of various concentrations of suramin. Human IAPP was dissolved in HFIP and diluted to a final concentration of 5 µm with a gradual increase in concentration of suramin. (A, B) ThT fluorescence kinetic values of hIAPP1−37 in the absence or in the presence of the indicated concentration of suramin. (C–G) Ultrastructural morphology determined using TEM of 5 µm hIAPP incubated alone for 24 h (C) or with 50 µm suramin for 24 h (D). TEM of 5 µm hIAPP was also performed after 48 h when it was incubated alone (E) or in the presence of 3 µm suramin (F) and 2 µm suramin (G). Suramin solutions of different concentrations were used as references in the assay. Samples were negatively stained with 2% uranyl acetate. Scale bar, 200 nm.

Figure 9.

 Biphasic effects of suramin on the aggregation of hIAPP1−37. The relative endpoint fluorescence values after incubation of hIAPP1−37 for 5 days in the presence of the indicated suramin concentrations (the 100% value was measured in the absence of suramin) are shown.


In the present study, we show that the minimal NFGAILSS octapeptide core of hIAPP is a glycolipid-binding domain. The interaction of the synthetic peptide with LacCer, a prototype glycolipid, has been studied at the air–water interface by using Langmuir film balance technology. This method has been used previously to formally demonstrate the interaction of the V3 domain of HIV-1 gp120 with various glycolipids [26]. The increase in surface pressure following the introduction of the peptide in the aqueous phase underneath the LacCer monolayer is indicative of the interaction. The specificity of the interaction was assessed by: (a) the lack of surface pressure increase when the hIAPP peptide was incubated with a monolayer of sphingomyelin, i.e. a sphingolipid with phosphorylcholine as polar head instead of carbohydrate, or with a monolayer of the glycerophospholipid phosphatidylcholine; (b) the gradual decrease of measured Δπmax as the initial pressure of the LacCer film increased; and (c) the decreased interaction of the synthetic peptide when Phe was replaced with Leu, indicating that the aromatic residue played a major role in the interaction. Taken together, these data strongly suggest that the minimal amyloid NFGAILSS peptide interacts specifically with LacCer. Thus, it is clearly a glycolipid-binding peptide. As a CH–π interaction appeared to be involved in the recognition of LacCer, it is not surprising that the replacement of Phe by a Tyr residue did not affect the binding of the amyloid peptide to LacCer. Indeed, it has been shown previously that Tyr has a higher propensity to interact with sugar rings than does Phe [27]. However, the NYGAILSS peptide has a reduced ability to form amyloid fibrils in comparison to the wild-type NFGAILSS motif [28]. Thus, similar aromatic-containing peptide fragments derived from amyloid proteins may bind to the same glycolipid but differ with respect to their amyloidogenic properties. Understanding the biochemical specificities of the π–π and CH–π stacking interactions will be a major issue in future research on conformational diseases. However, one of the major outcomes of the present study is the structural similarity between the hIAPP core amyloid peptide and the domain of HIV-1 gp120 (i.e. the crown of the V3 loop) involved in glycolipid recognition (Fig. 1). In both cases, a Phe residue oriented towards the solvent provides a complementary planar surface that is optimally presented to establish a CH–π stacking interaction with the ring of a sugar belonging to the polar head of the glycolipid. As aromatic residues are usually found in the functional core of many amyloid peptides [5], our data suggest that glycolipid recognition could be an intrinsic property of amyloid proteins. We are aware that further studies will be necessary to confirm this potential correlation. Nevertheless, the recent finding that the Alzheimer β-amyloid peptide interacted with GalCer in a monolayer assay [14] is in line with this theory.

In the second part of this study, we focused on the potential link between polyaromatic compounds, amyloid peptides and glycolipid-binding domains. In a first set of experiments, we showed that phenol red, which is known to inhibit amyloid fibril formation [22], recognized a glycolipid-binding domain (HIV-1 gp120 V3 motif) in an indirect solid-phase assay (Fig. 7). We used tritiated suramin as a ligand for immobilized synthetic V3 motifs [23], and we demonstrated that phenol red induced dose-dependent inhibition of the binding of suramin to V3. This allowed us to establish that both suramin and phenol red recognized the glycolipid-binding region of the V3 loop. As phenol red has previously been shown to interact with hIAPP and to inhibit its amyloidogenesis [22], these data demonstrate that hIAPP and the glycolipid-binding domain were recognized by the same polyaromatic compound. It should be emphasized that the use of a microtensiometer in glycolipid-binding studies did not allow the direct evaluation of polyaromatic compounds as potential inhibitors of peptide–glycolipid interactions. Indeed, we observed that phenol red alone dramatically increased the pressure of a LacCer monolayer (data not shown). Thus, although phenol red actually binds to hIAPP, it was not possible to detect any inhibitory effect of this compound on the hIAPP–LacCer reaction by means of surface pressure measurements. We encountered the same problem with suramin. Nevertheless, the results obtained with the solid-phase radioassay (Fig. 7) strongly support the notion that phenol red and suramin could interact with various glycolipid-binding peptides including hIAPP. Our data are also consistent with two recent studies showing that suramin can interfere with the amyloidogenesis process of PrP and β2-microglobulin [29,30].

Molecular modeling studies using molecular mechanics simulations suggested a potential mechanism of recognition between suramin and hIAPP, through heteroaromatic π–π stacking interactions (Fig. 6) [22,25]. Thus, it was logical to investigate the potential effects of suramin on amyloid formation by hIAPP. The data obtained in this study were somewhat surprising. A bimodal effect of suramin on amyloid formation was observed, depending on the concentration of suramin used in the assay. At concentrations ≤ 3 µm, suramin induced dose-dependent inhibition of amyloid fibril formation. However, at concentrations > 3 µm, suramin appeared to enhance fibril formation by hIAPP (Figs 8 and 9). Phenol red did not show such a biphasic effect, as it gradually interfered with the process of amyloid formation over the concentration range 0.5–40 µm[22]. The chemical structures of phenol red and suramin may shed some light on this strikingly distinct behavior. Both molecules are polyaromatic, but suramin is a symmetric compound (Fig. 5). In addition, it is a polysulfonated compound, whereas phenol red has just one –SO3 group. Our data suggest that at high concentrations, suramin could cross-link several hIAPP molecules, probably through nonspecific electrostatic interactions involving its numerous –SO3 groups. This could lead to the establishment of a complex network including both suramin and hIAPP fibrils. Consistent with this hypothesis, hIAPP1−37 is a basic peptide (theoretical pI of 8.9) with two basic residues (Lys1 and Arg11) and neither Asp nor Glu residues. In any case, our data showed that under specific conditions, multivalent aromatic compounds did not inhibit but instead promoted amyloid formation. This should be kept in mind before considering using polyaromatic molecules as therapeutic π–π breaker agents.

Our data may have several biological implications. First, we have demonstrated that the minimal NFGAILSS amyloid core peptide is also the shortest monomeric peptide able to interact with a glycolipid in our monolayer binding assay. Most importantly, this minimal peptide has no charged residues (either acid or basic), so its ability to interact with glycolipids could be essentially ascribed to its unique aromatic Phe residue. This means that the stacking CH–π interaction is by itself sufficient to initiate and stabilize a glycolipid–peptide interaction, provided that the aromatic residue is accessible enough to the sugar headgroup of the glycolipid. This conclusion is in line with our modeling study presented in Fig. 1, where the CH–π stacking interaction represents the main force for stabilization of the glycolipid–peptide complex. We recently obtained similar data with a synthetic glycolipid-binding peptide derived from a bacterial adhesin [31].

Second, the finding of a glycolipid-binding domain in hIAPP may be relevant to the etiology of type 2 diabetes. Indeed, hIAPP is synthesized in the β-cells of the pancreas and cosecreted with insulin [32]. It is the major protein of the islet amyloid deposits frequently seen in the pancreas of patients affected by type 2 diabetes [33]. Recently, it has been shown that the glycolipid sulfatide mediates the conversion of insulin hexamers to the biologically active monomers [34]. These data indicate that sulfatide has an important chaperone activity for the insulin hormone. Consistently, sulfatide treatment could prevent the development of type 1 diabetes in a mouse model [35]. On the basis of our data, one can tentatively propose that glycolipids expressed by the β-cells of the pancreas (either sulfatide or other glycolipid species) could also interact with hIAPP through CH–π interactions, thereby impairing the establishment of the π–π stacking interactions leading to amyloid formation. The preliminary observation that LacCer seemed to interfere with the aggregation process of hIAPP (Fig. 4) is consistent with this view. As amyloid deposits are assumed to play a key role in the pathogenesis of type 2 diabetes, such CH–π interactions between hIAPP and glycolipids may prevent (or delay) the development of this disease. Altogether, these data support the concept that glycolipids of the β-cells of the pancreas act as chaperones that are able to lock insulin and hIAPP in their functional and/or inoffensive conformations through a common molecular mechanism involving CH–π interactions. A therapeutic option for type 1 and type 2 diabetes would be to restore these protective interactions with nontoxic polyaromatic compounds [22] or soluble synthetic analogs of glycolipids [23].

Experimental procedures

Synthetic peptides and preparation of stock solutions

The synthetic peptides derived from hIAPP (NFGAILSS and analogs NYGAILSS and NLGAILSS) were purchased from Peptron Inc. (Taejeon, Korea). Human IAPP1−37 was purchased from Calbiochem (San Diego, CA, USA). The synthetic peptide P1 (KQHTVTTTTKGENFTETDVKMMER), derived from human PrP, [14] was purchased from Euro Sequence Gene Service (Evry, France). The peptides were purified by HPLC (purity > 95%) and characterized by ESI MS. Stock solutions of NFGAILSS and peptide analogs were prepared by dissolving the lyophilized form of the peptides in dimethylsulfoxide at a concentration of 100 mm. The stock solution of hIAPP was prepared by dissolving the lyophilized form of the peptide in 1,1,1,3,3,3-hexafluoro-2-propanol (HFIP) at a concentration of 500 µm. To avoid any preaggregation, all stock solutions were sonicated for 2 min before each experiment, as previously reported [24]. The multivalent eight-branched V3 peptide (GPGRAF)8-(K)4-(K)2-K-βA [20] was generously provided by J-M Sabatier (IFR Jean Roche, Marseille, France). LacCer, sphingomyelin and dipalmytoyl phosphatidylcholine of the highest available purity were purchased from Sigma (St. Louis, MO, USA).

Molecular modeling

Molecular structures were visualized using the swiss-pdb viewer program [36]. Protein Data Bank identification numbers were 1CE4 (HIV-1 gp120 V3 loop peptide) and 1KUW (hIAPP fragment). Molecular mechanics simulations (with the MM+ field force) of V3 and hIAPP peptides complexed with GalCer and suramin were performed with hyperchem 7 (Cambridge Soft), and the models were visualized with the swiss-pdb viewer as previously reported [37].

Surface pressure measurements

The surface pressure was measured with a fully automated microtensiometer (µTROUGH SX; Kibron Inc., Helsinki, Finland). All experiments were carried out in a controlled atmosphere at 20 °C ± 1 °C. Monomolecular films of LacCer and sphingomyelin were spread on pure water subphases (volume of 800 µL) from hexane/chloroform/ethanol (11 : 5 : 4, v/v/v) as described previously [38]. After spreading of the film, 5 min was allowed for solvent evaporation. To measure the interaction of synthetic peptides with lipid monolayers, various concentrations of the ligand (dissolved in pure water containing 1% dimethylsulfoxide) were injected into the subphase with a 10 µL Hamilton syringe, and the pressure increases produced were recorded until equilibrium was reached. The data were analyzed with the filmware 2.5 program (Kibron Inc.). The accuracy of the system under our experimental conditions was ± 0.25 mN·m−1 for surface pressure.

[3H]Suramin-binding assay

A 50 µL aliquot of the eight-branched V3 peptide (100 ng) was incubated in poly(vinyl chloride) 96-well plates overnight at 4 °C, as described previously [23]. The wells were washed three times with 200 µL of NaCl/Pi, and subsequently treated with NaCl/Pi containing 1% gelatin for 90 min at 37 °C to reduce nonspecific binding. The plates were rinsed in NaCl/Pi and incubated with various concentrations of phenol red for 1 h at 37 °C. After washing, the plates were incubated with 100 µL of [3H]suramin (1 µCi·mL−1). After 1 h at 37 °C, the plates were washed five times with 200 µL of NaCl/Pi, each well was individualized, and the radioactivity was determined in a β-scintillation counter (Packard United Technologies, Downers Grove, IL, USA).

ThT fluorescence assay

After sonication, dissolved hIAPP was diluted with 10 mm sodium acetate buffer (pH 6.5) or in suramin solutions, also in 10 mm sodium acetate buffer (pH 6.5), to a final concentration of 5 µm, with a final HFIP concentration of 1% (v/v). After dilution, the protein was centrifuged for 20 min at 20 817 g and 4 °C (Eppendorf centrifuge 5417R, rotor F-45-30-11). After centrifugation, the supernatant fraction was taken for incubation at 25 °C for the inhibition experiment. ThT was added to the incubated samples, giving an hIAPP final concentration of 0.5 µm and a ThT final concentration of 0.4 µm. Fluorescence was measured using an Yvon Horiba Fluoromax3 fluorimeter (Edison, NJ, USA) (excitation at 450 nm, 2.5 nm slit; emission at 480 nm, 5 nm slit).


Ten-microliter samples were placed on 400-mesh copper grids covered by carbon-stabilized formvar film (SPI Supplies, West Chester, PA, USA). After 2 min, excess fluid was removed, and the grids were negatively stained with 2% uranyl acetate in water for a further 2 min. Samples were viewed in a JEOL 1200EX electron microscope (Jeol, Croissy-sur-Seine, France) operating at 80 kV.