Amyloid β Protein-(1–42) Forms Calcium-permeable, Zn2+-sensitive Channel*

Amyloid β protein (AβP) forms senile plaques in the brain of the patients with Alzheimer’s disease. The early-onset AD has been correlated with an increased level of 42-residue AβP (AβP1–42). However, very little is known about the role of AβP1–42 in such pathology. We have examined the activity of AβP1–42 reconstituted in phospholipid vesicles. Vesicles reconstituted with AβP show strong immunofluorescence labeling with an antibody raised against an extracellular domain of AβP suggesting the incorporation of AβP peptide in the vesicular membrane. Vesicles reconstituted with AβP showed a significant level of 45Ca2+ uptake. The 45Ca2+ uptake was inhibited by (i) a monoclonal antibody raised against the N-terminal region of AβP, (ii) Tris, and (iii) Zn2+. However, reducing agents Trolox and dithiothreitol did not inhibit the 45Ca2+uptake, indicating that the oxidation of AβP or its surrounding lipid molecules is not directly involved in the AβP-mediated Ca2+ uptake. An atomic force microscope was used to image the structure and physical properties of these vesicles. Vesicles ranged from 0.5 to 1 μm in diameter. The stiffness of the AβP-containing vesicles was significantly higher in the presence of calcium. The stiffness change was prevented in the presence of zinc, Tris, and anti-AβP antibody but not in the presence of Trolox and dithiothreitol. Thus the stiffness change is consistent with the vesicular uptake of Ca2+. These findings provide biochemical and structural evidence that AβP1–42 forms calcium-permeable channels and thus may induce cellular toxicity by regulating the calcium homeostasis in Alzheimer’s disease.

A pathological hallmark in brain tissue from patients with Alzheimer's disease (AD) 1 is the accumulation of amyloid ␤ protein (A␤P), a 39 -43-amino acid-long polypeptide, as mor-phologically heterogeneous neuritic plaques and cerebrovascular deposits (1,2). A␤P is derived primarily from a proteolytic cleavage of the ␤-amyloid precursor protein (A␤PP), a highly conserved and widely expressed integral membrane protein with a single membrane-spanning polypeptide. The amount and the nature of polypeptides vary considerably among various forms of ADs: A␤P 1-40 and A␤P  are differentially accumulated in sporadic Alzheimer's disease and non-demented brain samples (3) and a mutation in presenilins is linked with an increased ratio of A␤P 1-42 /A␤P  in familial Alzheimer's disease (4 -7). The early-onset familial AD has been correlated with an increased level of A␤P  . However, very little is known about the role of A␤P  in such pathology and about the mechanism(s) of its action.
Accumulating evidence suggests an early and causative role of A␤Ps in the pathogenic cascade (8 -11). Postulated mechanisms of A␤P toxicity include, by its interaction with the tachykinin neuropeptide system, a surface membrane effect (12); by changing cellular ionic concentration via formation of plasma membrane channels (13)(14)(15); and by activating oxidative pathways and making cells more responsive to oxidative stress (for review see Refs. 16 and 17). Reactive oxygen species and the antioxidant defenses work probably by altering the lipid peroxidation and membrane composition. However, A␤P polypeptides associated with the reactive oxygen hypothesis have produced conflicting effects on cytoskeletal organization and cell lysis (18 -23).
The commonly observed change in the cellular ion concentration involves increased calcium level (24 -26) either indirectly via modulating the existing Ca 2ϩ channel or directly via cation-selective channels formed by A␤Ps. Support for the cation-selective A␤P channels are accumulating. Arispe and his collaborators (13)(14)(15)27) have reported cation-selective channels formed by A␤P 1-40 when reconstituted into lipid bilayers and in the membrane patches excised from hypothalamic gonadotropin-releasing hormone neurons. Kagan and his collaborators (28) have also recorded channel-like activity when A␤P [25][26][27][28][29][30][31][32][33][34][35] was reconstituted in lipid bilayers as well as for both A␤P 1-40 and A␤P 1-42 reconstituted in lipid bilayer, 2 though, with less reliability and reproducibility than the A␤P 25-35 current (28). Whether A␤P 1-42 toxicity is also mediated via A␤P 1-42 forming calcium-permeable ion channel is unclear.
The molecular structure of A␤P oligomers, especially as an ion channel, is unknown. Durell et al. (29) have developed theoretical models for the structure of ion channel formed by the membrane-bound A␤P  . However, no direct structural data from EM, NMR, x-ray diffraction, or other microscopic techniques are available to support the presence of the A␤P channel.
We have used an atomic force microscope (AFM) (30) integrated with a light and fluorescence microscope (31) to examine the mechanism(s) of A␤P 1-42 toxicity. A␤P 1-42 was reconstituted in phospholipid vesicles and were imaged in buffer to reveal the A␤P-membrane complex and a channel-like structure. Consistent with the possibility of fresh A␤Ps forming ion-permeable channels (i) fluorescently labeled anti-A␤P antibodies were localized in A␤P-reconstituted vesicles, (ii) vesi-* This work was supported by grants from the State of California, Department of Health Services, Alzheimer's Disease Program, 95-23336, and Yeungnam University in Korea. Portions of this work have been published as an abstract from the Biophysical Society annual meeting, 1998. The costs of publication of this article were defrayed in part by the payment of page charges. This article must therefore be hereby marked "advertisement" in accordance with 18 U.S.C. Section 1734 solely to indicate this fact.

EXPERIMENTAL PROCEDURES
Reconstitution of A␤P  into Liposomes-Human A␤P 1-42 and phospholipids were purchased from Bachem (Torrance, CA) and Avanti Polar (Birmingham, AL), respectively. Liposomes were prepared from both palmitoyloleoylphosphatidylethanolamine and palmitoyloleoylphosphatidylserine. 10 l of phospholipids (10 mg/ml palmitoyloleoylphosphatidylethanolamine: palmitoyloleoylphosphatidylserine::1:1) in chloroform was dried under argon gas, and then 2 ml of buffer (10 mM HEPES, pH 7.4) was added. The mixture was then bath sonicated for 20 min. For the incorporation of A␤P 1-42 into liposomes, the phospholipids were first dried under argon gas, then 1.5 ml of buffer was added, followed by 20 min of bath sonication. The A␤P stock solution (5 mg/ml) was added, and the mixture was sonicated for another 20 min. The final concentration of phospholipids and A␤P is 1 mg/ml.
Immunolabeling with Anti-A␤P Antibody-A mouse monoclonal antibody raised against the NЈ terminus of A␤P (named 3D6 antibody [anti-A␤P 1-5 (DAEFR)]) was obtained from Dr. Russell Rydel at Athena Neurosciences (San Francisco, CA). Goat anti-mouse IgG conjugated with Cy3 (1 mg/ml) was purchased from Chemicon International (Temecula, CA). Liposomes were adsorbed on glass coverslips and fixed with 4% paraformaldehyde for 10 min, then washed with phosphate-buffered saline (PBS), and then blocked with PBS containing 3% BSA and 1% goat serum to minimize nonspecific binding. Primary antibody (diluted 500-fold) was added in the presence of 3% BSA and 1% goat serum and incubated for 1 h at room temperature. After washing with PBS, the sample was incubated with secondary antibody (500-fold dilution) at the same condition as primary incubation. Fluorescent images were obtained using 40ϫ high numerical aperture objective lens with an inverted Olympus inverted fluorescence microscope.
Measurement of 45 Ca 2ϩ Uptake into Liposomes- 45 Ca 2ϩ uptake was measured by the modified method of Nakade et al. (32). 25 l of liposomes reconstituted with/without A␤P 1-42 was incubated with 75 l of HEPES buffer (10 mM, pH 7.4) containing 2 Ci of 45 Ca 2ϩ . Calcium influx was measured separately for each perturbation: anti-A␤P-antibody, Zn 2ϩ , Tris, Trolox, and DTT, respectively. After incubation for 1 min at 30°C, in the presence of a perturbation, the calcium influx reaction was stopped with a blocking solution containing 300 l HEPES buffer, 0.3 mM CaCl 2 , 5 mM ZnCl 2 , 10 mM Tris, and 15 g/ml 3D6 antibody. The 400-l mixture of liposomes and the blocking solution was immediately loaded onto a Chelex 100 column (bead volume: 3 ml) (Bio-Rad) which was pre-equilibrated with a buffer containing 200 mM sucrose, 20 mM Tris-HCl (pH 7.4 at room temperature), and 0.3% BSA. The column was then washed immediately with 5 ml of a buffer containing 200 mM sucrose and 20 mM Tris-HCl (pH 7.4) to take the liposomes. The 45 Ca 2ϩ content of the liposomes was measured with a Beckman liquid scintillation counter.
Imaging Liposomes with Atomic Force Microscopy-AFM images were obtained as described (33), 3 using a prototype of Bioscope AFM and a Multimode AFM (Digital Instruments, Santa Barbara, CA). Contact mode AFM was used for most of the images. Oxide-sharpened silicon nitride tips with a nominal spring constant of about 0.06 newton/m (Digital Instruments) were used for most experiments. All imaging was conducted on wet and hydrated liposomes. For liposomes reconstituted with/without A␤P 1-42 , 20 -50 l of sample was deposited on a clean glass Petri dish and left for 30 min. The surface of the Petri dish was then rinsed with a buffer (10 mM HEPES, pH 7.4) and imaged in the buffer. The imaging force was regularly monitored and kept to a minimum. The imaging force varied from subnanonewton to tens of nanonewtons. All imaging was performed at room temperature (22-24°C).
Measuring Viscoelastic Properties-We measured stiffness of A␤P vesicles by the AFM force-mapping technique as described (34) using the Nanoscope III software (Version 4.23R2; Digital Instruments, Santa Barbara). Force maps were taken by raster scanning the tip over the sample with 64 ϫ 64 measuring points with a pixel resolution of 95 nm. Each force map thus consists of a topographical image of 64 ϫ 64 pixels, and a force curve is stored for every pixel. Force maps were obtained alternately with the regular height and error modes of imaging the surface topography, using the same cantilever.
Vesicle stiffness was calculated from the force-volume imaging data. Force maps were exported to and analyzed with IGOR Pro software (Wavemetrics, Lake Oswego, OR) and macros written by Dan Laney (34). To calculate the elastic modulus of vesicles from the force curves, the Hertz model was used and the vesicle-tip system was equated to a sphere-sphere interaction (35,36). First, the indentation of the tip into the sample was determined by comparing force curves on the vesicles with those on the hard glass substrate. Then, the elastic modulus was extracted from the portion of the force curve where the indentation was still very small (ϳ10 nm).

RESULTS AND DISCUSSION
Immunolocalization of A␤P 1-42 on the Reconstituted Liposomes-Liposomes reconstituted with fresh A␤P 1-42 show strong immunolabeling with an anti-A␤P antibody. Fig. 1D shows a fluorescence labeling image of liposomes reconstituted with A␤P 1-42 . For comparison, no immunofluorescence labeling was observed in the liposomes prepared without A␤P 1-42 (Fig. 1B). Also, there was no immunofluorescence labeling observed in vesicles reconstituted with aged A␤P 1-42 (A␤P stored for 24 h or longer). All immunolabeled liposomes have a well defined vesicular structure as revealed by AFM imaging (Fig.  1, A and C). Vesicles with strong immunofluorescence signals appear to have double-layered (two membranes) disc-shaped structures in AFM images. Vesicle size ranged from 0.5 to 1 m. Vesicles without A␤Ps are on average half in diameter but more spherical compared with the vesicles with A␤Ps. The average height of vesicles without A␤P is 40 nm and 31 nm for vesicles with A␤P. The vesicular flattening and larger size are probably due to the protein-lipid interactions (37,38) as is the case with other membrane proteins such as gap junctions and acetylcholine receptors.
Since A␤PP is a membrane protein, it was believed that the membrane-bound A␤P portion would not be found as a free peptide except in case of membrane injury leading to a proteolytic cleavage of A␤PP and also that the hydrophobic and self-aggregating nature of A␤P would prohibit it from existing as a soluble, circulating peptide in normal biological fluids. Subsequent studies, however, revealed secreted soluble A␤P in 3 Y. J. Zhu, Y. Zhang, and R. Lal, submitted for publication. the conditioned media of a variety of primary or transfected A␤PP-expressing cells under normal metabolic conditions (39 -41). Our result is consistent with the presence of soluble A␤Ps that could interact with lipid membrane and form channels.
A␤P Channel-specific Calcium Uptake-Liposomes reconstituted with A␤P 1-42 show a significantly larger (Ͼ4-fold) increase in the influx of 45 Ca 2ϩ compared with the liposomes without A␤P 1-42 (compare A and B in the top panel of Fig. 2). In the presence of an antibody raised against the amino-terminal domain of A␤P, there was no significant influx of 45 Ca 2ϩ in the liposomes reconstituted with A␤P (compare A, B, and C in the top panel of Fig. 2). Such inhibition of calcium uptake by the anti-A␤P antibody shows the specificity of A␤P 1-42 -induced calcium uptake. The level of inhibition by anti-A␤P antibody would have been greater if all epitopes of the reconstituted A␤Ps were oriented outside of the liposomes.
We examined the mechanisms of A␤P 1-42 -specific calcium uptake. Recent studies suggest that A␤Ps form cation-selective ion channels which can be inhibited by zinc, Tris, and other related compounds. Consistent with such a possibility, calcium uptake was prevented when the liposomes reconstituted with A␤P 1-42 were incubated with Tris or zinc (compare A, B, and C, in the lower panel of Fig. 2). Whether this is also true at the cellular level needs to be examined. A␤P 1-40 -specific cationic channels are reported in AD-free fibroblasts and neuronal patches (24 -27, 47). 3 A␤P is reported to bind specifically and saturably with zinc in a biphase mode: at high affinity (K D ϭ 107 nM) or at low affinity (K D ϭ 5.2 M) (42). The zinc-binding site was mapped to a stretch of contiguous residues between positions 6 and 28 of the A␤P sequence. Zinc, at more than 1 M concentration in the A␤P-solubilized solution, is thought to facilitate the precipitation of A␤P (43,44). This property is important because zinc is abundant in the same neocortical regions where A␤P deposits are most commonly found, and high micromolar concentration of zinc is achieved during glutamatergic neurotransmission (45,46), providing one possible explanation for the propensity of A␤P to deposit close to the neocortical synaptic vicinity.
The influx of 45 Ca 2ϩ into liposomes containing A␤P  was not prevented by anti-oxidants Trolox (100 mM) and dithiothreitol (2 mM) (compare A, D, and E in the lower panel of Fig. 2). Rather, these anti-oxidants often appear to increase the calcium uptake. Anti-oxidants have been proposed to interact with the membrane lipid components and change the membrane permeability (for review see Refs. 16 -18). However, in our study, there was no significant change in calcium uptake in the liposomes prepared without A␤Ps. Our results thus strongly suggest that the anti-oxidants have little to no effect on A␤P 1-42 channel-mediated ionic exchange in vitro. Whether this is also true at the cellular level needs to be determined.
Calcium Uptake Induced Change in Vesicular Elasticity-In cells, calcium uptake would lead to altered metabolic load and an imbalance in the ionic homeostasis. In reconstituted vesicles, calcium uptake should change the physicochemical properties, especially vesicle stiffness and charge-charge interactions. We examined the change in vesicle stiffness using the force-mapping feature of the AFM imaging on vesicles reconstituted with/without A␤P 1-42 (Fig. 3) with a well defined morphology as revealed in AFM images. The shape and size of the vesicles appear to vary, but the apparent height of the unilammellar vesicles was similar.
Change in stiffness was examined for each treatment, e.g. anti-A␤P antibody, Zn 2ϩ , Tris, Trolox, and DTT. Stiffness was measured on the same vesicles before and after the application of a perturbation, thus each vesicle serving as its own control. Stiffness change was normalized with respect to the average stiffness in the calcium-free medium. In parallel with the change in the calcium uptake, stiffness of the vesicles reconstituted with A␤P 1-42 increased significantly compared with that for the vesicles without A␤Ps (compare A, B, and C in the top panel of Fig. 4). This stiffness change was inhibited by the anti-A␤P antibody suggesting the presence of A␤P. The calcium uptake-induced increased stiffness was inhibited by the presence of zinc or Tris (compare A, B, and C in the bottom panel in Fig. 4). On the other hand, anti-oxidants, such as Trolox and DTT, had diverse effects on the calcium uptakeinduced stiffness increase. DTT inhibited the stiffness change whereas Trolox enhanced the stiffness change. The inhibitory effect of DTT, however, is weaker than the inhibitory effect of zinc, Tris, and antibody. Such a complex effect of these antioxidants and the channel blockers may be reflected in the modes of action of these perturbations.
The change in stiffness is expected to result from (i) calcium ion-induced charge-charge repulsion energy inside liposomes, (ii) the binding of calcium to the lipids and proteins, and also (iii) the increased efficacy of lipid-protein interactions. However, the binding between calcium and the lipids or peptides on the external surface of the liposomes is unlikely the main driving force to increase in stiffness of liposomes because the same divalent cation zinc blocked the increase in stiffness of liposomes (Fig. 4, lower panel). It is possible that antioxidants do not prevent calcium binding to lipids on the internal face of the vesicle whereas zinc, Tris, and antibody (perhaps by steric effect) all prevent the binding of calcium to the lipids and proteins on the external face of the vesicles. Further study is required to examine this issue. In summary, our study strongly suggests that A␤P 1-42 forms calcium-permeable, Zn 2ϩ -sensitive channels in vitro and allows calcium transport. Such calcium exchange could destabilize cellular calcium homeostasis and lead to the cell toxicity.