Mechanism of cholesterol-assisted oligomeric channel formation by a short Alzheimer β-amyloid peptide



Alzheimer β-amyloid (Aβ) peptides can self-organize into oligomeric ion channels with high neurotoxicity potential. Cholesterol is believed to play a key role in this process, but the molecular mechanisms linking cholesterol and amyloid channel formation have so far remained elusive. Here, we show that the short Aβ22-35 peptide, which encompasses the cholesterol-binding domain of Aβ, induces a specific increase of Ca2+ levels in neural cells. This effect is neither observed in calcium-free medium nor in cholesterol-depleted cells, and is inhibited by zinc, a blocker of amyloid channel activity. Double mutations V24G/K28G and N27R/K28R in Aβ22-35 modify cholesterol binding and abrogate channel formation. Molecular dynamic simulations suggest that cholesterol induces a tilted α-helical topology of Aβ22-35. This facilitates the establishment of an inter-peptide hydrogen bond network involving Asn-27 and Lys-28, a key step in the octamerization of Aβ22-35 which proceeds gradually until the formation of a perfect annular channel in a phosphatidylcholine membrane. Overall, these data give mechanistic insights into the role of cholesterol in amyloid channel formation, opening up new therapeutic options for Alzheimer's disease.


Aβ22-35 peptide, which encompasses the cholesterol binding domain of Aβ, induces a specific increase of Ca2+ level in neural cells. Double mutations V24G/K28G and N27R/K28R modify cholesterol binding and abrogate channels formation. Molecular dynamic simulations suggest that cholesterol induces a tilted α-helical peptide topology facilitating the formation of annular octameric channels, as schematically shown in the graphic (with a hydrogen bond shown in green for two vicinal peptides). Overall, the data give insights into the role of cholesterol in amyloid channel formation and open up new therapeutic options for Alzheimer's disease.

Abbreviations used

amyloid protein precursor


In the past few years, the paradigm that associated directly amyloid fibril formation and the neurotoxicity of Alzheimer's β (Aβ) peptides has been seriously questioned (Lal et al. 2007; Fantini and Yahi 2010). The demonstration that various oligomers of Aβ peptides can form ion channels and perturb Ca2+ homeostasis in cultured neural cells has led to a re-examination of the dogma (Arispe et al. 1993, 2010; Jang et al. 2010; Fändrich 2012). Moreover, oligomeric Aβ species have been found in the brain of Alzheimer's disease patients (Shankar et al. 2008). Interestingly, Esparza et al. (2012) have recently reported that cognitively normal elderly patients can have comparable amounts of Aβ plaques than patients with clinical dementia of the Alzheimer type. In contrast, the concentration of Aβ oligomers in cortex samples was higher in Alzheimer's disease patients than in cognitively normal patients. These striking observations definitely suggest that Aβ oligomers, rather than Aβ fibrils or plaques, are the main neurotoxic species of Alzheimer's disease.

Even if a correlation between plasma cholesterol levels and the cholesterol content of neural membranes has not been formally established, it is important to assess the role of this lipid in the pathogenesis of Alzheimer's disease. Several studies suggest that plasma membrane cholesterol control the cytotoxicity of Aβ peptides (Arispe and Doh 2002; Abramov et al. 2003, 2011). Cholesterol is known to bind to the amyloid protein precursor (APP) (Beel et al. 2010), to regulate its membrane insertion (Lahdo and De La Fournière-Bessueille 2004) and to facilitate its processing into Aβ peptides in lipid raft microdomains (Ehehalt et al. 2003). Cholesterol also binds to monomeric Aβ peptides (Barrett et al. 2012), oligomers (Ashley et al. 2006), aggregates (Avdulov et al. 1997) and fibrils (Harris 2008). Recently, we identified the 20–35 fragment of Aβ1-42 as a linear cholesterol-binding domain and we showed that several peptides encompassing this domain bind cholesterol with high affinity (Di Scala et al. 2013). One of these peptides, i.e. Aβ22-35, has particularly retained our attention since it exerts neurotoxic effect on hippocampal neurons (Takadera et al. 1993), and it induces dementia in mice when injected intra-cerebroventricularly (Zhao et al. 2011). Although the Aβ22-35 peptide has been shown to self-aggregate into amyloid fibrils (Fraser et al. 1991), we wondered whether this peptide could also form oligomeric ion channels in the plasma membrane of neural cells and perturb cellular Ca2+ homeostasis. Using a combination of complementary techniques including intracellular Ca2+ levels measurements, lipid-protein interaction studies, and molecular dynamics simulations, we showed that cholesterol stimulates Aβ22-35 insertion and ion channel formation in the plasma membrane of neural cells. This effect was dependent upon the presence of membrane cholesterol. We used molecular dynamics simulation to examine the binding of cholesterol to Aβ22-35 in a realistic membrane environment. On this basis, we could construct a perfect annular channel resulting from the oligomerization of eight peptide/cholesterol subunits. This model was validated by the experimental data obtained with a series of synthetic Aβ22-35 peptides (wild-type, single, and double mutants). Overall, we propose a molecular mechanism accounting for the role of cholesterol as a key cofactor of amyloid ion channel formation.

Materials and methods


Aβ22-35 and Aβ1-42 peptides were purchased from rPeptide (Bogart, GA, USA). All other synthetic peptides with a purity > 95% were obtained from Schafer-N (Copenhagen, Denmark). Ultrapure apyrogenic water was from Biorad (Marnes La Coquette, France). Cholesterol was purchased from Matreya (Pleasant Gap, PA, USA).

Cell culture

SH-SY5Y cells were obtained from the ATCC. They were maintained in Dulbecco's Modified Eagle Medium: Nutrient Mixture F-12 (Gibco, Saint-Aubin, France) supplemented with 10% fetal calf serum, glutamine (2 mM) and penicillin (50 U/mL)/streptomycin (50 μg/mL) and cultured at 37°C with 5% CO2. Cells were passaged twice a week and not used beyond passage 20. For the cholesterol-dependent peptide insertion study, cells were treated with methyl-β-cyclodextrin 1 mM during 24 h.

Calcium measurements

Cells were plated at a density of 45 000 cells/dish in 35 mm culture dishes and grown during 72 h at 37°C. They were loaded with 5 μM Fluo-4AM (Invitrogen, Saint-Aubin, France) for 30 min in the dark, washed three times with HBSS (Gibco) and incubated for 20 min at 37°C. The calcium fluxes were estimated by measuring the variation of cells fluorescence intensity after amyloid peptide injection (220 nM) into the recording chamber directly above an upright microscope objective (BX51W Olympus, Tokyo, Japan) equipped with an illuminator system MT20 module. Fluorescence emission at 525 nm was imaged by a digital camera CDD (Hamanatsu ORCA-ER, Hamamatsu, Japan) after fluorescence excitation at 490 nm. Times-lapse images (1 frame/10 s) were collected using the CellR Software (Olympus). Fluorescence intensity was measured from the region of interest centered on individual cells. Signals were expressed as fluorescence after treatment (Ft) divided by the fluorescence before treatment (F0) multiplied by 100. Then, the results are averaged and the percentage of fluorescence of control is subtracted for each value. All experiments were performed at 30°C during 60 min.

In silico studies of cholesterol/Aβ22-35 interactions and channel formation

Molecular modeling studies were performed with the Hyperchem 8 program (ChemCAD, Obernay, France). The starting conditions for generating a cholesterol/Aβ22-35 complex were those used to model the cholesterol/Aβ1-40 complex (Di Scala et al. 2013). Geometry optimization of the cholesterol/Aβ22-35 complex was achieved using the unconstrained optimization rendered by the Polak–Ribière conjugate gradient algorithm (Fantini et al. 2011). Molecular dynamics simulations were then performed for iterative periods of 1 ns in vacuo with the Bio+ (CHARMM) force field (Singh et al. 2009). The resulting complex was then immersed in a phosphatidylcholine matrix consisting of 350 molecules of palmitoyl-oleoyl-phosphatidylcholine and 3500 molecules of water. The formation of an oligomeric channel was obtained in this matrix after 100 ns of simulations. The energy of interaction was determined with the Molegro Molecular Viewer (Thomsen and Christensen 2006). The helical wheel diagram was obtained from the Emboss Pepwheel website (

Lipid monolayer assay

Peptide-cholesterol interactions were studied with the Langmuir film balance technique using a Kibron microtensiometer as previously described (Fantini and Yahi 2011).

Statistical analysis

The results are expressed as mean ± SD. anova and the Fisher multiple-comparison post hoc tests were conducted. Difference with < 0.05 was considered significant.


Molecular modeling of Aβ22-35/cholesterol channels

Molecular dynamics simulations have been successfully used to determine how cholesterol interacts with Aβ1-40, and the models have been sufficiently robust to predict which amino acid mutations could affect cholesterol binding (Di Scala et al. 2013). We used the same strategy to construct a realistic annular channel, starting from a minimized model of cholesterol-associated Aβ22-35 (Fig. 1a and b). The use of a 14-mer peptide as a surrogate for longer channel-forming Aβ peptides (i.e. 40/42-mer for full-length Aβ peptides, 32-mer for Aβ11-42, and 26-mer for Aβ17-42) considerably simplified the molecular modeling simulations of an oligomeric channel. We could thus merge the peptide with cholesterol in a realistic membrane-like matrix, and perform the modeling for 100 ns of running time. Under these conditions, a perfect annular channel consisting of eight Aβ22-35/cholesterol subunits surrounded by phosphatidylcholine lipids could be constructed (Fig. 1). These channels have an outer diameter of ca. 5.8 nm and a pore diameter of ca. 1.8 nm. The molecular details of the Aβ22-35/cholesterol complex are given in the (Figure S1). These values are close to the one measured by Jang et al. (2010) in their molecular models of Aβ9-42 and Aβ17-42 channels. Atomic force microscopy data were also consistent with these geometric features (Quist et al. 2005; Jang et al. 2010).

Figure 1.

Molecular dynamics simulations of Aβ22-35 oligomeric channels in presence of cholesterol and phosphatidylcholine. (a) The association of eight Aβ22-35/cholesterol subunits forms an oligomeric channel with a pore size diameter of 1.46 nm and an external diameter of 4.4 nm (including cholesterol). A calcium ion has been placed in the center of the channel. Cholesterol is colored in yellow. The atoms of the peptide are colored in red (oxygen), green (carbon), light gray (hydrogen), and blue (nitrogen). (b) Topology of the octameric Aβ22-35/cholesterol channel dipped in a phosphatidylcholine matrix. The surface representation of the channel includes both the peptides and cholesterol. Acidic regions are in red, basic regions in blue, and apolar domains in light gray. A calcium ion has been placed in the center of the channel. (c) In each Aβ22-35/cholesterol complex, the peptide has a tilted orientation with respect to the lipid (mean angle 40°). This tilted orientation allows a series of stabilizing peptide–peptide contacts resulting in the formation of a circular octameric channel. (d) The principal stabilizing force of the oligomeric channel is a hydrogen bond formed between the ε-amino group of Lys-28 of one peptide and the amide group of Asn-27 of its neighbor. The peptidic bond between Ser-26 and Asn-27 also forms a hydrogen bond with Lys-28.

Interestingly, the Aβ22-35 peptide did not interact with cholesterol as the full length Aβ1-40 did. Among important structural changes we can notice the loss of interaction with Glu-22 and Gly-25, the important decrease of the energy of interaction of Val-24 and of Lys-28 and the establishment of a new contact involving Asn-27 (Table 1). Therefore, the interaction between cholesterol and Aβ22-35 led to a specific geometry of the complex resulting from a fine conformational tuning of the peptide. This conformational adaptation was possible because Aβ22-35 is shorter than Aβ1-40. As a result, the Aβ22-35 peptide had a tilted orientation with respect to the longitudinal axis of cholesterol (Fig. 1c). It formed an angle of 40° with cholesterol, which is typical of the angle of insertion of both amyloid and viral fusion peptides within the plasma membrane of host cells (Crowet et al. 2007; Fantini et al. 2011). This unique tilted topology has a dramatic consequence on the mode of insertion of Aβ22-35 within a cholesterol-containing membrane. Indeed, it is consistent with an oligomerization process of eight Aβ22-35/cholesterol subunits that self-organize into an annular channel (Fig. 1a). In this particular membrane orientation, each Aβ22-35 peptide has its acidic N-ter part tipped in the same direction, which, as a result of the oligomerization process, delineates the acidic mouth of the channel (Fig. 1b). In these channels, the Glu-22 residues of each peptide monomer form a cluster of negatively charged acidic residues which could make up the halldoor of Ca2+ in the cell. The apolar part of the Aβ22-35 peptide is tipped in the opposite direction, allowing the establishment of numerous van der Waals interactions with surrounding phosphatidylcholine lipids forming a membrane-like matrix. These multiple interactions of the Aβ22-35/cholesterol channel with phosphatidylcholine molecules result in a perfect insertion of the channel within the lipid matrix, as shown in Fig. 1b. Finally, our modeling study suggested that the oligomerization process is stimulated by the formation of a hydrogen bond network involving the ε-amino group of Lys-28 of one Aβ22-35 peptide and the amide group of Asn-27 of its neighbor (Fig. 1d). It also involves the peptidic bond between Ser-26 and Asn-27 (see Figure S2 for a full description of the hydrogen bond network and Table S1 for an energetic analysis of the oligomerization process). Incidentally, the oligomerization process reinforced the cholesterol/Aβ22-35 interaction in each subunit, as demonstrated by the increase in the overall energy of interaction from −46 to −57 kJ/mol and in particular a substantial increase of energy of interaction between Asn-27 and cholesterol (Table 1). The better stability of cholesterol-Aβ22-35 subunits in the channels would be the result of the cholesterol-dependent orientation of Lys-28 side chain kept inside the membrane. The Lys-28 side chain could have been expected to snorkel, as described in the APP transmembrane domain(C99)-cholesterol complex, to locate its ε-amino group near the polar lipids head groups (Strandberg and Killian 2003; Barrett et al. 2012). As mentioned in a recent study of Lys residues belonging to transmembrane domains (Gleason et al. 2013), the particular location of ε NH3+ group of Lys-28 in the membrane lipid phase induced a dramatic decrease in its pKa value. This could in turn reinforce its hydrogen bonding potential with Asn-27, thereby facilitating implied the oligomerization process of Aβ22-35.

Table 1. Energetics of interaction of cholesterol with Aβ22-35 monomers and oligomers
Amino acid residueAβ1-40/cholesterolaAβ22-35/cholesterol monomersbAβ22-35/cholesterol oligomersc
  1. a

    Retrieved from the molecular dynamics simulations of the cholesterol/Aβ1-40 interaction (Di Scala et al. 2013).

  2. b

    Calculated from the molecular dynamics simulations of cholesterol-Aβ22-35 monomers.

  3. c

    Calculated from the molecular dynamics simulations of cholesterol-Aβ22-35 octamers at the end of the oligomerization process. The data are expressed in kJ/mol. (−) indicates that the concerned residue does not interact with cholesterol.


Aβ22-35 alters cellular Ca2+ levels

To study the pore-forming properties of Aβ peptides, we have followed Ca2+ exchanges in SH-SY5Y cells loaded with the fluorescent-sensitive dye Fluo-4 and subsequently treated with the peptides. When added to cells bathed in Ca2+-containing buffer, Aβ1-42 induced a progressive, time-dependent increase of intracellular Ca2+ levels. Under similar conditions, this effect was not observed when the cells were treated with Aβ1-16, a peptide previously shown to bind to selected glycosphingolipids. In contrast, Aβ22-35 induced a rapid increase of cellular Ca2+ content (Fig. 2a and b). Both the rate and amplitude of Ca2+ increase were higher with Aβ22-35 compared with Aβ1-42. Overall these data suggest that the pore-forming capability of Aβ1-42 is totally included in the 22–35 fragment.

Figure 2.

Aβ22-35 and Aβ1-42 elevate intracellular calcium. (a) Traces show Ca2+ dependent fluorescence signals averaged from SH-SY5Y cells in response to Aβ22-35 (black, n = 104), Aβ1-42 (red, n = 96) and Aβ1-16 (blue, n = 120) injections (220 nM). Results are expressed as the mean ± SD. (b) The images are pseudo color representations of cells. They were acquired before (0 min) or after (60 min) Aβ treatments. Warmer colors correspond to higher fluorescence. Scale bar 100 μm.

Effect of Ca2+-free medium and Zn2+ on Aβ22-35 channels

The increase of Ca2+ levels induced by Aβ22-35 was no longer observed when SH-SY5Y cells were switched into a Ca2+-free medium (Figure S3). However, when the cells were switched back to normal Ca2+-containing medium, the increase in Ca2+ was immediately recovered and even amplified. These data suggested that Ca2+ comes from external sources through the forming of Aβ22-35 ion channels in the plasma membrane of neural cells. Then we tested the effect of Zn2+, a classical amyloid channel inhibitor (Arispe et al. 1996). As shown in Fig. 3a, Zn2+ significantly inhibited the elevation of intracellular Ca2+ induced by Aβ22-35. The inhibitory effect of Zn2+ was also observed with Aβ1-42 (Fig. 3b), although in this case the inhibition was more efficient than for Aβ22-35. In any case, these data are consistent with the notion that both Aβ22-35 and Aβ1-42 form calcium channels in the plasma membrane of SH-SY5Y cells.

Figure 3.

Zn2+ reduces the intracellular calcium elevation. During loading with Ca2+ indicator Fluo 4AM, cells were incubated with Zn2+ (50 μM). (a) After Aβ22-35 injection, time course of Ca2+ dependent fluorescence was recorded in presence (red, n = 104) or in absence of Zn2+ (black, n = 104). (b) Aβ1-42 Ca2+ signals (black, n = 70) were abolished when cells are treated with Zn2+ (red, n = 65). The results are averaged and expressed as the mean ± SD.

The formation of Aβ22-35 channels is cholesterol-dependent

Since the Aβ22-35 peptide binds cholesterol with high affinity (Di Scala et al. 2013), it was logical to search for a potential effect of cholesterol on the formation of Aβ22-35 channels. In a first series of experiments, we treated SH-SY5Y cells with a non-toxic concentration of the cholesterol depleting agent methyl-β-cyclodextrin (Figure S5). After 24 h of incubation, the cells were washed, loaded with Fluo-4, and treated with the Aβ22-35 peptide (Fig. 4a). The methyl-β-cyclodextrin treatment abolished the elevation of Ca2+ entry induced by the Aβ22-35 peptide. In a second series of experiments, we analyzed the effect of a panel of mutant Aβ22-35 peptides on Ca2+ fluxes. The mutated amino acid residues were those appearing critical for cholesterol binding, that is, Val-24, Lys-28, and Asn-27 (Table 1). As shown in Fig. 4b, the single Aβ22-35 mutants V24G and K28G induced a slight inhibition of the Ca2+ response whereas the N27R mutant behave like the wild-type peptide. In contrast, Ca2+ influx was totally suppressed with Aβ22-35/V24G/K28G double mutant and greatly inhibited with Aβ22-35/N27R/K28R double mutant. To understand why the double mutants had lost the channel-forming capability of Aβ22-35, we analyzed and compared the interaction of all mutant peptides with cholesterol.

Figure 4.

The formation of Aβ22-35 channels is cholesterol-dependent. To evaluate the potential role of cholesterol on channel formation, SH-SY5Y cells were treated with methyl-β-cyclodextrin. (a) In response to addition of Aβ22-35 (black; n = 74), Aβ22-35/V24G/K28G (red; n = 70), and Aβ22-35/methyl-β-cyclodextrin (blue; n = 102), Ca2+ dependent fluorescence were recorded. (b) The histogram bars show averages of percentage of fluorescence at 60 min in response to Aβ22-35 (n = 74), Aβ22-35/V24G (n = 90), Aβ22-35/K28G (n = 85), Aβ22-35/K28R (n = 80), Aβ22-35/V24G/K28G (n = 70), Aβ22-35/N27R (n = 60), and Aβ22-35/N27R/K28R (n = 176). Data shown are mean ± SD. *** Significant difference; < 0.001.

Physico-chemical measurements of cholesterol binding to Aβ22-35 peptides

In these experiments, cholesterol was prepared as a lipid monolayer at the air-water interface and the peptides were added in the aqueous subphase underneath the monolayer. The interaction between the peptides and cholesterol was quantified by measuring the increase in the surface pressure of the lipid film with a microtensiometer (Thakur et al. 2009). The experiment was first performed at a fixed value of the initial surface pressure of the cholesterol monolayer, that is, 20 mN/m. When the wild-type Aβ22-35 peptide was added underneath the cholesterol film, an important increase of the surface pressure was measured, reaching a plateau value of 18 mN/m (Fig. 5a). When the same experiment was performed with the mutant Aβ22-35 peptides, the increase in the surface pressure was just above 5 mN/m for the single mutants V24G and K28G/R and around 10 mN/m for the Aβ22-35/V24G/K28G double mutant, indicating that all these mutations affected the interaction with cholesterol. To investigate further this point, we prepared several cholesterol monolayers at various initial surface pressure values, ranging from 10 to 35 mN/m (Fig. 5b). When the initial pressure increases, there are more cholesterol molecules in the monolayer. Thus, it is increasingly difficult for the peptide to penetrate into the cholesterol monolayer. The critical pressure of insertion of the wild-type Aβ22-35 peptide, that is, the extrapolated value of the initial surface pressure at which the insertion process does no longer occur, is 42 mN/m. For the single point Aβ22-35 mutants, we showed in a previous work (Di Scala et al. 2013) that with V24G and K28G/R mutants the surface pressure increase was similar at all values of the initial surface pressure, for example, ca. 5 mN/m. The same behavior can be observed with N27R mutant (Fig. 5b). This indicates that these peptides have no particular affinity for cholesterol and does not make any difference between cholesterol-rich and cholesterol-poor domains. For this reason, such patterns of surface pressure measurements are generally interpreted as a loss of insertion power of the cholesterol monolayer. The behavior of the double mutant peptides Aβ22-35/V24G/K28G and Aβ22-35/N27R/K28R were markedly different with regard to single mutants (Fig. 5b). Indeed, Aβ22-35/V24G/K28G still interacted with cholesterol, and a critical pressure of insertion similar to Aβ22-35 wild-type peptide could be measured. However, at all values of the initial surface pressure, the pressure increase induced by the double mutant was significantly lower than the one induced by the wild-type Aβ22-35 peptide. In the case of Aβ22-35/N27R/K28R, a critical pressure of insertion less important than the wild-type peptide could also be observed (37 mN/m). These data indicated that the double mutants could still interact with cholesterol, but through distinct molecular mechanisms compared with the wild-type peptide.

Figure 5.

Interaction of wild-type and mutant Aβ22-35 peptides with cholesterol. The interaction between wild-type and mutant Aβ22-35 peptides and cholesterol has been quantified by the increase in the surface pressure induced by the addition of each peptide underneath a cholesterol monolayer. (a) Mean surface pressure increase induced by the addition of the indicated peptide (10 μM) under a cholesterol monolayer prepared at an initial surface pressure of 20 mN/m (± SD, n = 3). (b) Cholesterol monolayers were prepared at various values (range 10–35 mN/m). After equilibration of the monolayer, wild-type Aβ22-35 (full squares), Aβ22-35/N27R (full circles), Aβ22-35/V24G/K28G (open squares), and Aβ22-35/N27R/K28R (open circles) mutant peptides were added in the subphase at a concentration of 10 μM. The maximal surface pressure was determined after reaching the equilibrium. ** Significant difference; < 0.001.


As recently pointed out by Lukiw (2013), Alzheimer's disease is primarily ‘a disorder of the plasma membrane’. The proteolytic cleavage of the amyloid protein precursor APP into neurotoxic Aβ peptides occurs in cholesterol-enriched lipid raft nanodomains (Ehehalt et al. 2003). Correspondingly, cholesterol depletion inhibits the generation of Aβ peptides (Simons et al. 1998). Cholesterol also stimulates Aβ deposition in neural tissues by modulating the activities of Aβ-producing enzymes (Wang et al. 2012). Both physico-chemical and microscopy techniques have shown that cholesterol physically interacts with APP (Beel et al. 2010), with Aβ peptides (Beel et al. 2010) and with fibrils (Harris 2008). Interestingly, APP and Aβ peptides have distinct, although partially overlapping cholesterol-binding domains, that is, Aβ17-40 for APP (Barrett et al. 2012) and Aβ20-35 for Aβ1-42 (Di Scala et al. 2013). Moreover, Gly residues play a key role in the interaction of APP with cholesterol (Barrett et al. 2012; Fantini and Barrantes 2013), whereas Val-24 and Lys-28 are critical for cholesterol binding to Aβ1-42 (Di Scala et al. 2013). Recently, we showed that the synthetic Aβ22-35 peptide encompassing most of the linear cholesterol-binding domain of Aβ1-42, binds cholesterol with high affinity (Di Scala et al. 2013). Because it displays both neurotoxicity properties (Takadera et al. 1993; Zhao et al. 2011) and cholesterol-binding potency, Aβ22-35 is a unique tool for studying the mechanistic relationship between cholesterol and Aβ toxicity (Hartmann 2001). Shorter neurotoxic peptides such as Aβ25-35 also have amyloid potential (Simmons and Schneider 1993), but do not contain key amino acid residues such as Val-24 that are critical for the binding of cholesterol to Aβ1-42. Thus, even if Aβ25-35 has retained some of the cholesterol-binding properties of full-length Aβ peptides, it could not be used as a surrogate for Aβ1-42 in mechanistic studies focused on cholesterol. Accordingly, only Aβ22-35 simultaneously offers the neurotoxicity potential, amyloidogenic properties and cholesterol-binding capability of full length Aβ peptides.

Our in silico data suggest that the driving force of the membrane insertion/oligomerization of Aβ22-35 is a combination of three parameters: (i) a cluster of apolar residues which limits the energetic cost of the membrane penetration process, (ii) the deprotonation of Lys-28 which finely tunes the helix tilt and facilitates the oligomerization process by its interaction with Asn27 through inter-peptide driven hydrogen-bond and (iii) cholesterol, which boosts the whole process, through direct binding to the peptide through three main anchoring sites including Val-24, Asn-27, and Lys-28, inducing a favorable tilted topology and the reorganization of acidic residues in the N-ter part of the peptide. An α-helical wheel representation of the Aβ22-35 peptide complexed to cholesterol (Figure S4) showed residues Glu-22, Asp-23, and Ser-26 in a relatively short arc. These polar residues are grouped in the hydrophilic 90° sector and constitute a pathway for Ca2+ permeation. The remaining 270° sector, which faces membrane lipids, is chiefly hydrophobic, except for the presence of Asn-27 and Lys-28 whose polar head groups are neutralized by the hydrogen bond network described above. For the sake of comparison, a similar hydrogen bond network retrieved from the structure of a ligand-bound dimerized Fibroblast growth factor receptor (Plotnikov et al. 1999) is shown in the Supplementary information (Figure S2). Finally, it is noteworthy that the helical topology of Aβ22-35 in the channel is strikingly similar to the α-helical transmembrane segments of a synthetic tetrameric calcium channel (Zhorov and Ananthanarayanan 1996).

Given the propensity of various truncated Aβ peptides to form oligomeric structures (Jang et al. 2010), we considered the possibility that the neurotoxic properties of Aβ22-35 could be due to the formation of oligomeric channels in the plasma membrane of neural cells. To evaluate the reliability of our docking study, we have analyzed calcium cellular fluxes induced by a collection of synthetic Aβ peptide. We showed that Aβ1-42 and Aβ22-35 rapidly induced a rapid, progressive, and time-dependent increase of intracellular Ca2+ levels. We believed that this effect is due to the cholesterol-dependent formation of an oligomeric Ca2+ channel in the plasma membrane of SH-SY5Y cells for the following reasons: the Ca2+ source was extracellular, as demonstrated by the lack of calcium fluxes changes in cells incubated in Ca2+-free medium; Aβ1-42 and Aβ22-35-induced Ca2+ flux were inhibited by zinc, a blocker of amyloid ion channel activity which also decreases the uptake of Ca2+ induced by Aβ11-42, Aβ17-42 (Jang et al. 2010); Aβ1-42 and Aβ22-35 had no effect in cholesterol-depleted cells treated with methyl-β-cyclodextrin prior to peptide addition; Aβ1-16, a truncated peptide which interacts with GM1 (Fantini and Yahi 2011) but not with cholesterol (Di Scala et al. 2013), had no effect on Ca2+ fluxes; mutations that affect cholesterol recognition (V24G/K28G and N27R/K28R) impaired channel formation. Overall, the Aβ22-35 channels were functionally similar to the zinc-sensitive Ca2+ channels formed by longer Aβ peptides (e.g. Aβ11-42, Aβ17-42) in the plasma membrane of neural cells. Thus, we have further delineated the minimal fragment of Aβ peptides able to form oligomeric ion channel to the 22–35 fragment, which also corresponds to the cholesterol-binding domain of Aβ1-42. Moreover, we demonstrated that Aβ22-35 channels are cholesterol-dependent, thereby suggesting a mechanistic link between cholesterol and amyloid ion channel formation.

Our molecular modeling data concerning cholesterol dependence of Aβ22-35 channel formation are strongly supported by the experiments performed with the mutant peptides Aβ22-35/V24G/K28G and Aβ22-35/N27R/K28R. These double mutants have lost the ability to form Ca2+ channels in neural cells (Fig. 4b) through both altered interaction with cholesterol and impaired oligomerization. Indeed, Val-24, Asn-27, and Lys-28 are involved in cholesterol recognition and Lys-28 plays a central role in the oligomerization process of Aβ22-35 (Fig. 1b). Interestingly, Aβ22-35/V24G/K28G and Aβ22-35/N27R/K28R have retained some capability of interaction with cholesterol (Fig. 5b) but have lost its aggregating properties. This might reflect a distinctive type of association with cholesterol, most likely a partial insertion, preventing channel formation. Conversely, the single mutants (V24G, N27R, K28G, and K28R), which have lost the cholesterol monolayer insertion potency in vitro, can still form Ca2+ channels in live cells. All together these data suggested that Aβ22-35 channel formation requires both a tight association with cholesterol and an oligomerization capacity in the membrane environment. This further strengthens the key role of cholesterol as a lipid cofactor for Aβ peptide membrane binding/insertion (Ji et al. 2002; Qiu et al. 2009; Yu and Zheng 2012) and as a universal regulator of ion channel function (Levitan et al. 2010).

In conclusion, our study provides a potential molecular mechanism accounting for the cholesterol-assisted amyloid ion channel formation. Apart from Aβ peptides, this mechanism could also account for the cholesterol-dependent membrane insertion/oligomerization of α-synuclein (Fantini and Yahi 2011) and amylin (Trikha and Jeremic 2011). Finally, the identification of a linear cholesterol-binding domain in Aβ suggests a new therapeutic strategy to reduce Aβ neurotoxicity. Indeed, any non-toxic molecule that specifically targets the 22–35 domain of Aβ has potential neuroprotective effects that should be evaluated in cellular and animal models of Alzheimer's disease. Molecular docking studies are currently conducted in our laboratory to identify such possible inhibitors of cholesterol/Aβ interactions that could prevent Aβ membrane insertion and amyloid channel formation.


C. Di Scala is recipient of a doctoral fellowship from the Ministère de l'Enseignement Supérieur et de la Recherche (Paris, France). The authors have no conflict of interest to declare.