Oxidative Stress and Cell Membranes in the Pathogenesis of Alzheimer's Disease

Paul H. Axelsen, Hiroaki Komatsu, Ian V. J. Murray


Amyloid β proteins and oxidative stress are believed to have central roles in the development of Alzheimer's disease. Lipid membranes are among the most vulnerable cellular components to oxidative stress, and membranes in susceptible regions of the brain are compositionally distinct from those in other tissues. This review considers the evidence that membranes are either a source of neurotoxic lipid oxidation products or the target of pathogenic processes involving amyloid β proteins that cause permeability changes or ion channel formation. Progress toward a comprehensive theory of Alzheimer's disease pathogenesis is discussed in which lipid membranes assume both roles and promote the conversion of monomeric amyloid β proteins into fibrils, the pathognomonic histopathological lesion of the disease.

The literature of Alzheimer's disease (AD) research is vast, complex, and often contradictory. Nevertheless, a broad perspective on this literature, and on what typically does and does not happen in biological systems, suggests either that lipid membrane damage is directly involved in the pathogenesis of AD or that it is an important consequence of AD. Therefore, investigations into the relationship between membrane damage and amyloidogenesis may yield important insights into AD pathogenesis.

The “amyloid hypothesis” has driven much of the research on AD pathogenesis since it was proposed in 1991 (123, 124). In its simplest form, this hypothesis suggests that the accumulation of amyloid beta (Aβ) proteins in brain tissue drives the pathogenesis of AD. Aβ proteins originate from a large transmembrane protein of unknown function known as amyloid precursor protein (APP) by the action of β-secretase and γ-secretase activity (FIGURE 1). The specific cleavage site of γ-secretase is somewhat variable, yielding proteins ranging from 39 to 43 residues in length, with 40 and 42 residue forms predominating (Aβ40 and Aβ42). The non-amyloidogenic pathway for APP processing involves α-secretases, a family of enzymes that cleave APP near the middle of the Aβ protein segment.


Processing of the amyloid precursor protein

Aβ proteins are 39- to 43-residue segments within the 770-residue amyloid precursor protein (APP), beginning at residue 672. The non-amyloidogenic processing pathway is catalyzed by α-secretase, which cleaves in the midst of the Aβ protein segment. The amyloidogenic pathway is catalyzed by β-secretase, which cleaves an extracellular site of APP, and γ-secretase, which cleaves at points within the transmembrane segment.

The amyloid hypothesis was initially based on genetic studies of familial AD and the occurrence of AD-like pathology in Down Syndrome. Among the most cited experimental studies in support of this hypothesis are those in which Aβ proteins impair physiological and cognitive function when injected into rodent brains (72, 136, 184, 282, 320, 336). Nevertheless, the amyloid hypothesis has been criticized for being incomplete and vague. For example, it is unable to explain nonfamilial or sporadic cases of AD (the most prevalent form), how fibril formation leads to neuronal death, the poor clinicopathological correlation between amyloid plaque burden and cognitive impairment (279), the hyperphosphorylation and aggregation of tau proteins as neurofibrillary tangles, and axonopathies that precede amyloid deposition (302). These criticisms notwithstanding, there are no alternative hypotheses with nearly as much support, and the amyloid hypothesis is readily adapted to answer some of these criticisms. For example, the pathologies that precede amyloid deposition may be due to prefibrillar intermediate forms of Aβ protein, and plaques may represent an inert form of the protein (62, 125, 337). Supporting this idea is the observation that a variant form of Aβ that oligomerizes but does not fibrillize can nonetheless cause several pathological manifestations of AD other than fibril formation (318).

It has been suggested that there is a tenuous balance among the rates of Aβ protein production, aggregation, and elimination in brain tissue, such that a small disturbance applied over sufficient time causes the pathological accumulation of Aβ protein as amyloid fibrils in AD. However, the concentrations of Aβ proteins in the brain appear to be far lower than their aqueous in vitro solubility (143, 238, 281). Therefore, Aβ proteins should not form amyloid fibrils in the brain even if concentrations are modestly elevated. Yet, fibrils do form with advancing age in virtually everyone, indicating that factors other than Aβ protein concentration must have a role in causing them to form. If small changes in protein concentration cannot account for fibril formation, then we must consider factors such as macromolecular crowding (216, 228), locally high concentrations of Aβ proteins at synapses (71), chemical modification (30, 161, 231, 287, 360), cross-linking (17, 22, 290, 291), or metal complex formation (6, 31, 92, 105, 153, 272, 284, 306). Understanding the chemical mechanisms that reduce the solubility of Aβ proteins in brain tissue is key to understanding why AD pathology develops in some individuals and not others.

Oxidative Stress

Oxidative stress is caused by a dense, complex and heterogeneous network of oxidizing reactions running counter to the reducing conditions that otherwise prevail in cells and tissues. It has been suggested that the accumulation of Aβ proteins in the brain may be a protective response to oxidative stress (16, 28, 162, 164, 273, 292). The density of plaques containing Aβ protein correlates inversely with markers of oxidative damage (75), and the cortical deposition of Aβ proteins correlates with reduced oxidative damage in Down syndrome (236). Moreover, Aβ proteins prevent lipoprotein oxidation (163) and metal-induced neuronal death in culture (363). However, it is inherently difficult to quantify oxidative stress, and attempts to do this are subject to many methodological pitfalls. For example, a study concluding that the primary mechanism of Aβ toxicity does not involve oxidative pathways measured thiobarbituric acid reactive substances (TBARS) to assess lipid peroxidation (248). However, assays for TBARS measure only a small subset of the diverse oxidation products generated, as well as substances not derived from lipids. Other studies suggesting that Aβ has antioxidant activity were not controlled for the inherent redox activity of peptide groups or conducted in the presence of biologically relevant electron donors and acceptors to drive the reactions (23, 24).

Overall, it seems more likely that Aβ proteins promote oxidative stress (42, 50, 91, 201, 202, 254). The brain in AD appears to sustain more oxidative damage than normal brain (200, 211, 331), exhibits an increased susceptibility to oxidative stress (190, 220, 277), and has relatively low levels of naturally occurring antioxidants such as α-tocopherol (144, 358). Regions of the brain rich in Aβ proteins also have increased levels of protein oxidation (129). The overexpression of Aβ proteins in transgenic mice, in C. elegans, and in cell culture, increases biomarkers of oxidative stress (256, 345). Aβ proteins cause H2O2 and lipid peroxides to accumulate in cells (27). Catalase protects cells from Aβ toxicity, and cell lines selected for resistance to Aβ toxicity also become resistant to the cytotoxic action of H2O2. Aβ proteins promote the oxidation of compounds such as dopamine, phospholipids, and cholesterol (40, 134, 135, 235, 241, 257, 356) as long as an intact methionine side chain is present (5, 21, 45, 47, 49, 232).

Aβ proteins and amyloid plaques bind redox-active transition metals that are likely to be the actual site of redox activity (76, 273). Copper levels are significantly increased in amyloid plaques (183, 191, 304), although comparatively subtle increases in copper transport across cell membranes may cause significant changes in Aβ protein turnover (321). Excess dietary copper increases AD-like pathology in a mouse model (160). Cu2+ and Zn2+ chelators appear to inhibit Aβ deposition in the brains of mouse models (64), and metal depletion promotes the disaggregation of Aβ fibrils (65). Cu2+ potentiates the neurotoxicity of Aβ42 > Aβ40 in embryonic rodent neurons, and its effect is mediated by H2O2 (135, 355). Studies of copper in the pathogenesis of AD have been extensively reviewed (39, 90, 100, 305). In contrast, Zn2+ ions are redox-inert and able to protect/rescue human cells in tissue culture from Aβ and Cu2+ toxicity (75). A recent study, however, observed that the zinc released at synapses with neurotransmitters caused the accumulation of toxic Aβ oligomers to these membranes (86).

It has been suggested that Aβ proteins split into fragments that are both neurotoxic and able to generate additional oxygen radicals (48, 128), although these findings have been strongly refuted (89). Nevertheless, electron paramagnetic resonance spectroscopy has shown that there is a strong correlation between the intensity of radical generation by Aβ and neurotoxicity (219). In these studies, preincubation of Aβ to form fibrils increased its toxicity. In contrast, replacing the redox-active sulfur atom in residue Met35 with methylene (CH2) resulted in a peptide that formed fibrillar structures but had no demonstrable toxicity toward cultured hippocampal neurons. In the same experimental system, vitamin E (presumably acting as an antioxidant) neutralized the neurotoxicity of Aβ but had no effect on its ability to form fibrils (332). Another study concluding that free radicals do not mediate Aβ-induced neurotoxicity used ginkgolides (a purported anti-oxidant component of Ginkgo biloba leaves) and vitamin E to inhibit oxidation (351). Although neither agent protected cells from apoptosis and death, it is unlikely that they completely halted radical-mediated oxidative damage.

Proteomics studies of oxidative stress have focused on proteins damaged by oxidation, nitration, and other reactive substances (41, 44, 50, 51, 57, 58, 60, 80, 107, 243, 285, 293, 317). In most cases, reactions with protein carbonyl groups are taken for evidence of damage by oxidation, and the nature of the protein modification is not explicitly defined. Two studies have examined the ability of Aβ proteins to induce oxidative modifications in other proteins (34), and Aβ proteins may also undergo oxidative damage, particularly His, Tyr, and Met side chains (15, 32, 43, 132, 139). However, it is difficult to imagine how oxidative damage to proteins, especially Aβ proteins, would lower the aqueous solubility of Aβ proteins and promote fibril formation.

Enter Lipid Membranes

Investigations into the role of lipids in AD are experimentally challenging because they are chemically diverse, with thousands of distinct molecular species present in every cell. They are also physically diverse, rarely existing as monomeric species in solution unless protein bound. In vitro, the vast majority form micelles or vesicles, and the latter may be unilamellar or multilamellar. Thus lipid suspensions generally have two or more phases and complex interphase equilibria. This physical heterogeneity complicates most types of physicochemical analysis, even when pure synthetic lipid preparations are used, and it makes some others outright impossible. It also impedes the measurement of fibril formation. Turbidity measurements, for example, are confounded by ambiguity over whether the species causing the turbidity is protein or lipid. Lipids markedly increase the fluorescence of the thioflavin T apart from fibril formation, whereas assays based on Congo Red absorbance ratios entail practical restrictions on protein and lipid concentrations that limit the flexibility of this assay. The presence of membranes can also complicate sample preparation, and they yield artifacts when the halogenated solvents used to disaggregate Aβ proteins are not completely removed (56, 310).

Due in part to these challenges, the evidence that cell membranes are involved in the pathogenesis of AD remains largely circumstantial, even though the amount of evidence pointing to some type of link is overwhelming. Aβ proteins are produced by cleavage of APP at two sites, one site being located approximately at the midpoint of the transmembrane segment (FIGURE 1). As a consequence, the COOH-terminal residues of Aβ proteins that were part of the APP transmembrane segment are uniformly hydrophobic. Following their cleavage from APP, some investigators have observed that Aβ proteins remain associated with detergent-resistant lipid membrane domains in the brain (181), or with membrane-anchored APP (189). It has also been hypothesized that Aβ proteins bind to the transmembrane helices of membrane proteins and cause their dysfunction (199). Ultrastructural studies suggest that amyloid fibril formation tends to occur first in portions of diffuse amyloid deposits that are closest to membranes (234, 319, 346). Aβ40 with the E22Q mutation (responsible for hereditary cerebral hemorrhage with amyloidosis-Dutch type) will fibrillize on the surface membrane of human cerebrovascular smooth muscle cells (330).

Despite a substantial hydrophobic segment, the general conclusion reached by most investigators is that Aβ proteins have little affinity for neutral lipid membranes (205). Techniques such as the hydration of a mixed protein-lipid film must be used to induce the penetration of Aβ proteins into a neutral lipid bilayer (83). One laboratory investigating Aβ40 and another investigating Aβ42 have documented that the proteins situate differently in membranes depending on whether they are embedded in a bilayer membrane or allowed to associate with the surface of a preformed membrane (33, 106, 176). In general, anionic lipids tend to induce Aβ proteins to adopt extended β structure (35, 63, 66, 76, 130, 167169, 204, 213, 214, 313, 314, 344, 352). Possible reasons for the inducement of β structure and fibril formation by protein-lipid interaction have been reviewed, including the ability of such interactions to serve as templates for structural change, to increase the local concentration of protein, and to orient protein monomers relative to each other (2, 113). Aβ proteins adopt α-helical structure in association with lipid membranes at low lipid-to-protein ratios (315), at high cholesterol concentrations (145), or in conjunction with metal ions (21, 76, 77). Some investigators have found that detergent micelles promote α-helical structure (283), whereas others find that it promotes the formation of oligomers with β structure (261). Spontaneous insertion into various membranes has been observed for Aβ segments (82) and for Aβ proteins at relatively low membrane surface pressures (94).

The relative abundance of various lipid classes in membranes is altered in AD (121, 227, 247, 252, 350), and altering membrane composition protects PC12 cells from toxic effects of Aβ proteins (338). Altered physical properties, presumably arising from compositional differences, have been observed in hippocampal membranes of AD brains (364). Plasmalogen deficiency is frequently associated with AD (102, 110112, 119, 120, 229). In addition to having an effect on membrane physical properties, plasmalogens are particularly susceptible to oxidative damage (156158). This susceptibility may confer on plasmalogens the ability to protect other lipid species by diverting and trapping oxidizing agents (36, 117, 118, 170, 195, 196, 222, 230, 242, 264, 362). Cerebral white matter is enriched in plasmalogens for unknown reasons (101).

Human epidemiological studies support a link between AD and the consumption of ω-3 polyunsaturated fatty acyl (PUFA) chains (103, 141, 223, 224, 237, 250, 275, 323), and this association is supported by animal model and cell culture studies (5355, 126, 146, 185, 193, 350). The most prevalent ω-3 PUFA in the brain is docosahexaenoic acid (DHA), and dietary DHA supplementation alleviates both AD-like histopathology and cognitive impairment in animal models (53, 55, 126, 185). The metabolism of DHA in normal brain is remarkable in several respects (159), and it is difficult to induce a measurable deficiency of DHA-containing lipids in the brain tissue of animal models through dietary restriction (84). Animal studies have shown that the brain responds to a dietary deficiency of DHA by elongating arachidonic acid (ARA) chains. Because there are no enzymes capable of desaturating the distal end of these PUFA chains, the result is a marked increase in docosapentaenoic acid (DPA) chains (138).

Aβ proteins appear to have special relationships or interactions with specific lipid species. For example, the ganglioside GM1 is bound together with Aβ proteins in diffuse amyloid plaques (348). GM1-containing membranes promote the formation of α structure (212, 215), β structure (206), or fibrils in vitro (6769, 115, 127, 148151, 165, 207, 218, 239, 240, 334, 335). In several of these reports, GM1 is presented in raft-like membrane subdomains in which cholesterol is a significant component, and rafts have also been discussed as an influence on the processing of APP into Aβ proteins (74). Apart from rafts, cholesterol has been epidemiologically (289, 311) and experimentally (263, 300) linked to the incidence of AD, and it may influence the production (108, 361), location (77, 87, 145), behavior (79, 214, 352), or toxicity (38, 78, 137, 308, 338) of Aβ proteins. In models of Niemann-Pick Type C disease, Aβ proteins accumulate along with cholesterol (347).

Cholesterol depletion reduced the β-cleavage of APP and the production of Aβ proteins in a study of neurons in the rat hippocampus (288). Treatment with lovastatin/mevalonate alone, however, was insufficient to induce a significant effect; treatment with β-cyclodextrin to achieve a 70% reduction of cellular cholesterol content was also required. It has been suggested that the dependence of Aβ protein production on cholesterol is due to the selective activity of β-secretase activity in cholesterol-dependent raft-like subdomains of the plasma membrane, whereas α-secretase activity appears to predominate in non-raft domains (95). Some β-secretase stimulation has also been attributed to neutral glycosphingolipids and anionic phospholipids (152). The efficacy of a β-secretase inhibitor has been increased by linking it to cholesterol and thereby targeting it to membranes (260).

The Membrane as Villain

The lipids in cell membranes are often regarded as being chemically unreactive and merely a physical barrier or a support matrix for proteins. That view is misleading on many levels, of course, but particularly so when the membrane lipids contain PUFA chains and are subjected to oxidative stress. Reviews of the role of membranes in AD pathogenesis frequently overlook the effects of oxidative stress and chemical modification of PUFA chains on membrane properties. Aβ proteins have a prooxidant activity toward polyunsaturated lipids that can be neutralized by lipophilic antioxidants, chelation of metal ions, anaerobic conditions, mutation of His13 or His14 to Ala, or modification of the Met35 side chain (232). Lipid oxidation products and the susceptibility of lipids to oxidative damage are both increased in AD (78, 104, 210, 309). Several reports have outlined the potential for a complex interplay between cholesterol oxidation products and Aβ proteins (38, 116, 235, 241, 301, 329, 356, 360).

The myriad mechanisms of oxidative stress yield diverse chemically reactive products from PUFA chains including hydroxy- and hydroperoxy-lipids that undergo spontaneous decomposition, lipid free radicals that may participate in free-radical chain reactions, as well as fragments such as malondialdehyde, acrolein, and hydroxynonenal with the potential to form adducts with proteins and nucleic acids (61, 270, 274, 293, 295, 296). These materials are direct toxic threats to the tissues in which they are generated. Brain is the most lipid-rich organ in the body, and it contains more lipids bearing PUFA chains than any other organ. Therefore, brain tissue is at particularly high risk for chemical injury due to highly reactive lipid oxidation products.

Evidence that lipid peroxidation may be involved in the pathogenesis of AD has been extensively reviewed (13, 42, 46, 201). Lipid hydroperoxides undergo spontaneous (nonenzymatic) decomposition, and ω-6 PUFA chains yield reactive α,β-unsaturated aldehydes such as 4-oxo-2(E)-nonenal (180) and 4-hydroxy-2(E)-nonenal (HNE) (97, 98), as well as eicosanoids such as isoprostanes (225, 226). The prooxidant activity of Aβ proteins mentioned above can catalyze the formation of HNE from ω-6 PUFAs in the presence of copper ions (232). Isoprostanes are relatively unreactive compounds that have been used as biomarkers of oxidative stress (179, 221, 244, 268). Some reports suggest that isoprostanes are specifically elevated in AD (253, 255, 256), although these findings have not been confirmed by all investigators (221, 357). Analogous products derived from ω-3 PUFA chains have been dubbed “neuroprostanes” and also have been put forward as indexes of oxidative stress in the brain (14, 266, 267, 298).

Compared with isoprostanes, HNE is chemically reactive, with a well known propensity to form adducts with the side chains of various amino acid residues (37, 98, 328). For this reason, HNE-protein adducts have also been used as biomarkers of oxidative stress (325). HNE concentrations in human ventricular fluid are ∼15 μM and are elevated in AD (190, 203, 210). HNE modification is also known to inhibit proteasome function (286) and glutamate transport in synaptosomes (177). Together with the observation that immunoreactivity of antibodies to HNE-modified His residues localizes to amyloid plaques (7, 294), these observations suggest that Aβ proteins not only promote lipid oxidation but that there may also be a mechanistic link between the lipid oxidation products formed during oxidative stress and Aβ misfolding (161).

Aβ proteins increase HNE production from PUFA chains in vitro that, in turn, causes covalent modification of the three His residues in Aβ proteins and fibril formation (231). These modifications promote the aggregation of Aβ proteins into fibrils (FIGURE 2) (165, 231, 232). Moreover, they increase the ability of Aβ42 to seed fibril formation by unmodified Aβ40 (166). The effects of HNE on Aβ fibril formation depend on the size or hydrophobicity of the modification because the corresponding analog from ω-3 PUFA chains, 4-hydroxy-2(E)-hexenal, does not have this effect (187). The addition of an octanoyl group to specific Lys residues of Aβ proteins also appears to have a pro-amyloidogenic effect (258).


Electron micrograph of fibrils induced by HNE modification

Aβ42 was induced to fibrillize with HNE, then treated with mouse anti-His-HNE, and gold-tagged anti-mouse antibody. The focus in this image is on the gold particles, but various fibril morphologies are evident, and the gold particles making anti-HNE-His antibodies are preferentially associated with long, relatively straight fibils.

In vivo, HNE-His epitopes are concentrated in the vicinity of amyloid plaques, but they do not precisely co-localize with the plaques (FIGURE 3). One possible explanation for this distribution is that the presence of HNE-His adducts and Cu-His complexes in the same protein molecule are mutually exclusive. Presumably, the formation of Cu-His complexes precedes the generation of HNE and HNE-His adducts, if indeed they are responsible for generating HNE from ω-6 PUFAs. Consequently, HNE-His epitopes would develop around preexisting amyloid plaques, not within them. A second possibility is that Aβ-specific antibody epitopes are masked by HNE modification, as observed in vitro for 4G8 and 6E10 epitopes (231). A third possibility is that Aβ fibrils induced to form by HNE modification have relatively sparse HNE-His epitopes (as suggested by FIGURE 2), yielding sparse fluorescence among fibrils. The fluorescence of HNE-11S seen in FIGURE 3 may be to HNE produced in the plaque but bound to HNE-His adducts in proteins other than Aβ proteins.


Immunofluorescent staining of the posterior parietal association area in a transgenic mouse model of Alzheimer's disease that expressed amyloid precursor protein with the “Swedish” K670N/M671L mutation, and presenilin 1 with the M146L mutation

A: 2H4 antibodies (Covance) specific for the amino terminus Aβ proteins (green). B: HNE 11S specific for HNE-His adducts (red). C: A and B superimposed. Possible reasons for the close association of HNE-His epitopes and Aβ proteins, but not precise colocalization, are discussed in the text.

Some investigators have focused on the role of acrolein in the pathogenesis of AD. Acrolein is a well known environmental toxin and a recently recognized product of lipid peroxidation (324, 326, 327). It reacts spontaneously with Lys residue side chains, forming a cyclic Nε-(3-formyl-3,4-dehydropiperidino) derivative that has been considered a biomarker for measuring oxidative stress in AD (52). Acrolein is elevated in AD brain, it appears to be more neurotoxic than HNE, it causes elevated intracellular calcium levels (192, 342), and it appears to inactivate flippase, inducing a breakdown of lipid membrane asymmetry and apoptosis (1, 59).

The Membrane as Victim

Although the foregoing discussion suggests that AD may be caused by membrane-derived neurotoxic agents, a contrasting view is that AD is due to physical attacks on the barrier function of membranes (265). For example, Aβ proteins or their fragments have been observed to penetrate the membrane surface (82, 94), disrupt raft-like domains (67), cross the membrane (341), induce isotropic phases (176), increase the permeability of membranes to various materials (154, 344), cause physical thinning (81), decrease the mobility of hydrophobic membrane probes (169), activate phospholipase A2 (182), and promote membrane fusion (81, 249). Membrane binding by Aβ proteins appears to mediate some forms of neurotoxicity (21, 70, 316). Plasma membranes isolated from human brain accelerate Aβ fibrillogenesis (339), whereas fibrillizing Aβ proteins disrupt the structure of membranes formed from both synthetic lipids (168, 169) and whole-brain lipid extracts (353). The action of Aβ proteins on membranes has been compared with the action of antimicrobial peptides, and homogenates of brain with AD have elevated antimicrobial activity (299). Among the effects of Aβ proteins on membranes, however, the formation of ion channels has received the most attention (1012, 93, 147, 186, 251).

Aβ proteins share an ability to induce channel-like activity in membranes with a wide variety of other amyloid-forming proteins (259). In the case of Aβ, the channels exhibit only modest cation selectivity but very long lifetimes. Aβ cation channels can be blocked by Zn2+, suggesting the possibility that they may be therapeutic targets (9), and small molecule inhibitors of the Aβ calcium channel have been described (88). Ring-like structures possibly representing these pores or channels have been observed in membranes by AFM (173, 186, 259) and EM (174, 175). Although such features are not always observed with Aβ proteins (114, 341), there are many reports suggesting that they are formed by other amyloidogenic proteins such as the insulin-associated polypeptide, α-synuclein, and serum amyloid A protein (259). It has been suggested that the channel-forming toxic properties of Aβ proteins on a membrane depends on the extent to which it has aggregated into oligomers and that this extent is concentration dependent (276). Model structures of the purported channels have been patterned after β-barrel pore-forming toxins (8, 93). This comparison is suggested not only by the EM/AFM images but by the reactivity of Aβ and α-hemolysin with so-called “conformation-specific” antibodies (354). Recent studies confirming the formation of zinc-sensitive ion channels have also exposed the potential for artifact due to residual amounts of halogenated solvent in the Aβ protein samples (56).

It may be significant that truncated Aβ peptides that are unable to form fibrils can nevertheless induce ion permeability (142) and that membranes formed from highly compressible lipids are most sensitive to Aβ-induced changes in permeability (297). It has been suggested that channel openings are merely protein-induced defects in the bilayer structure or organization that heal after a time, giving the appearance of a transient channel opening (114). This mechanism would be consistent with oligomer-induced membrane thinning (297) and would help resolve the paradoxical appearance of many simultaneous pore structures in membranes that exhibit only single channel conductances (96).

In living cells, as opposed to synthetic systems, one must consider the possibility that Aβ may profoundly affect the behavior of naturally occurring channels and transporters. For example, Aβ42 increases HNE-modification of a glutamate transporter in synaptosomes (177). Aβ oligomers have been implicated in causing a pathological calcium influx through hippocampal NMDA receptors, followed by the activation of calpain and the degradation of dynamic 1-a GTPase involved in synaptic vesicle recycling (155). Aβ proteins have also been shown to depress the surface concentration of AMPA receptors involved in calcium regulation at the synapse (133, 188). The dysregulation of calcium has long been associated with AD pathogenesis (172) and is thought to be involved with the accumulation of amyloid deposits (85, 208, 209) and the hyperphosphorylation of tau (25). Calcium also promotes the conversion of oligomeric Aβ to fibrils (140). An extensive literature exists on the role of mitochondria in AD pathogenesis, and the accumulation of calcium by mitochondria followed by damage to the mitochondrial membrane structure is a prominent theme in this literature (131, 246, 262).

The Lipids Associated With Apolipoprotein E

The lipids in nascent lipoprotein particles are arranged in a membrane-like bilayer, surrounded by apolipoprotein molecules in a configuration often called the “belt” model (FIGURE 4) (280). Apolipoprotein E (ApoE) is a focus of considerable interest because its isoforms are strongly associated with the risk of developing AD. In plasma, ApoE is found with other proteins in lipoprotein particles as diverse as chylomicrons and high-density lipoproteins. In the brain and cerebrospinal fluid, ApoE is the most abundant apolipoprotein and is found in particles that resemble high-density lipoproteins in both density and size (269). Among the three common isoforms, ApoE3 is the most common (77% of the alleles) and is therefore considered to be the wild type. ApoE2 has an R158C substitution, whereas ApoE4 has a C112R substitution, and both are associated with forms of hyperlipidemia (197). One copy of the ApoE4 gene confers a threefold increased risk of Alzheimer's disease, whereas two copies confer an eightfold increased risk. One copy of the ApoE2 gene, however, reduces the risk by 60% (73).


Nascent lipoprotein E particles consist of α-helical “belts” around the perimeter of a lipid bilayer disk

For this arrangement and the number of protein and lipid molecules in a particle, there is excess protein, which may situate across the face of the disk. When the protein is apoE3, there is one thiol group for every 45 lipids, concentrating a large antioxidant capacity in close proximity to the lipids.

In addition to the epidemiological evidence, there is an abundance of experimental evidence linking ApoE to the pathogenesis of Alzheimer's disease. For example, the ApoE4 allele is associated with increased Aβ deposits in the brain, and a distinct neuropathological phenotype (278), whereas a lack of ApoE reduces Aβ deposition in mice (19). Amyloid plaques bind anti-ApoE antibodies, suggesting that ApoE is present in these plaques (3). Indeed, ApoE copurifies with Aβ from amyloid plaques (26) and may exist as a bound complex with Aβ proteins (245, 349). Some have suggested that ApoE4 accelerates fibril formation by Aβ40 (18, 194, 271, 343), but its behavior in this regard may depend on whether it is monomeric or dimeric. Others have suggested that dimeric forms inhibit fibril formation and that ApoE3 is much more effective at doing this because a large fraction circulates as a disulfide-linked dimer, whereas ApoE4 cannot form disulfide bonded dimers due to its C112R substitution (99, 340). Despite all of these observations, the reason that different isoforms of ApoE have different levels of risk for AD is not known. There are no isoform-dependent differences in the apparent structure of lipoprotein particles (280). A case for isoform-specific direct interactions between ApoE and Aβ proteins has been made (233), but clear conclusions are elusive due to problems with protein purity, aggregation and denaturation of the ApoE protein, and indirect assay methods (4, 171, 198, 245, 303, 307).

Others have suggested that risk of AD with ApoE isoforms may be related to differences in antioxidant activity (178, 217, 312). ApoE4 has no Cys residues and, hence, no free thiol groups; ApoE3 has one Cys, whereas ApoE2 has two. Free thiol groups have significant antioxidant activity via mechanisms that differ from those of glutathione (109, 333, 359). In apoA-I, variants with a free Cys residue side chain exhibit significantly greater antioxidant activity than variants without a free Cys residue (29). Therefore, it is intriguing to speculate that ApoE4 confers increased risk of AD, whereas ApoE2 confers decreased risk, because Cys residue side chains protect lipids in lipoprotein E particles against oxidative stress.

Toward a Comprehensive Theory of AD Pathogenesis

The cause of AD is known in cases of familial disease, but that knowledge has not clarified how sporadic disease develops, nor has it so far yielded an effective therapy for either form of the disease. Clearly, we need a more detailed theory of AD pathogenesis, and it seems likely that lipid membranes will have a prominent role in any such theory. As summarized above, the interactions between lipid membranes and Aβ proteins are bi-directional: lipids damage Aβ proteins, and Aβ proteins damage lipid membranes. These processes can operate in tandem to accelerate fibril formation in human brain lipid extracts (231). Even if fibril formation is only an inert by-product of the disease-causing process, it is nonetheless a pathognomonic finding of AD. Therefore, it is vital to understand how fibrils form.

As illustrated in FIGURE 5, the interaction of lipid membranes may constitute an amplification system for fibril formation. Aβ proteins in vitro only form fibrils at micromolar concentrations and after days of incubation, whereas membranes containing PUFA-chains lower the protein concentration requirement by three orders of magnitude and shorten the time required to minutes (165, 166). The same system may also amplify multiple mechanisms of neurotoxicity that operate independently of fibril formation and link many of the seemingly unrelated observations reviewed above. For example, inflammation due to trauma or other factors will promote oxidative stress and the production of reactive oxygen species that exert direct neurotoxicity or lipid damage and Aβ protein aggregation. Lipid oxidation products may exert direct toxicity. ApoE4 proteins are deficient in thiol-mediated antioxidant activity; this deficiency would allow excess oxidative damage to the lipids in lipoprotein particles that likewise promote Aβ protein aggregation. Aging is associated with reduced tissue antioxidant levels, accumulated oxidative lipid damage, and low levels of adventitial Aβ protein aggregation, with each process promoting further Aβ protein aggregation. Various aggregated forms of Aβ protein may form pores, channels, or other neurotoxic structures in neuronal membranes.


Membrane-mediated amplification of fibrillogenesis, illustrating possible relationships between meachanisms involved in Aβ protein aggregation and neurotoxic mechanisms involved in the pathogenesis of Alzheimer's disease

Lipid oxidation products modify the three His residues of Aβ42, increasing its membrane affinity and accelerating the conversion of Aβ40 into oligomers and fibrils. Oligomeric Aβ proteins bind copper ions while they undergo redox cycling. Highly reactive oxygen species may be generated by electrons from copper or as by-products of inflammation, including inflammation induced by trauma. ApoE4 alleles lack thiol-mediated antioxidant activity and may allow excess oxidative damage to the lipids in lipoprotein particles. Aging by itself is associated with reduced tissue antioxidant levels and accumulated oxidative lipid damage. During any or all of these processes, direct neurotoxins may be produced, and Aβ proteins may form pores, channels, or other disruptive neurotoxic structures in neuronal membranes.

It is not yet clear whether all cases of AD arise through the operation of one primary pathogenic mechanism that diverges only at a late stage into various subtypes or whether disparate mechanisms operate from the beginning. For example, neurofibrillary tangles of tau protein (20, 322) occur in only 70–80% of cases of AD (122), suggesting that tangles reflect a response to some but not all pathogenic paths. In any case, one or more additional factors must be incorporated into the amyloid hypothesis to explain the pathogenesis of AD, and confronting the experimental challenges presented by lipids and membranes may be necessary to identify such factors.


  • No conflicts of interest, financial or otherwise, are declared by the author(s).

  • The authors are supported by grants from the National Institute on Aging, the Alzheimer's Association, and the American Health Assistance Foundation.


  1. 1.
  2. 2.
  3. 3.
  4. 4.
  5. 5.
  6. 6.
  7. 7.
  8. 8.
  9. 9.
  10. 10.
  11. 11.
  12. 12.
  13. 13.
  14. 14.
  15. 15.
  16. 16.
  17. 17.
  18. 18.
  19. 19.
  20. 20.
  21. 21.
  22. 22.
  23. 23.
  24. 24.
  25. 25.
  26. 26.
  27. 27.
  28. 28.
  29. 29.
  30. 30.
  31. 31.
  32. 32.
  33. 33.
  34. 34.
  35. 35.
  36. 36.
  37. 37.
  38. 38.
  39. 39.
  40. 40.
  41. 41.
  42. 42.
  43. 43.
  44. 44.
  45. 45.
  46. 46.
  47. 47.
  48. 48.
  49. 49.
  50. 50.
  51. 51.
  52. 52.
  53. 53.
  54. 54.
  55. 55.
  56. 56.
  57. 57.
  58. 58.
  59. 59.
  60. 60.
  61. 61.
  62. 62.
  63. 63.
  64. 64.
  65. 65.
  66. 66.
  67. 67.
  68. 68.
  69. 69.
  70. 70.
  71. 71.
  72. 72.
  73. 73.
  74. 74.
  75. 75.
  76. 76.
  77. 77.
  78. 78.
  79. 79.
  80. 80.
  81. 81.
  82. 82.
  83. 83.
  84. 84.
  85. 85.
  86. 86.
  87. 87.
  88. 88.
  89. 89.
  90. 90.
  91. 91.
  92. 92.
  93. 93.
  94. 94.
  95. 95.
  96. 96.
  97. 97.
  98. 98.
  99. 99.
  100. 100.
  101. 101.
  102. 102.
  103. 103.
  104. 104.
  105. 105.
  106. 106.
  107. 107.
  108. 108.
  109. 109.
  110. 110.
  111. 111.
  112. 112.
  113. 113.
  114. 114.
  115. 115.
  116. 116.
  117. 117.
  118. 118.
  119. 119.
  120. 120.
  121. 121.
  122. 122.
  123. 123.
  124. 124.
  125. 125.
  126. 126.
  127. 127.
  128. 128.
  129. 129.
  130. 130.
  131. 131.
  132. 132.
  133. 133.
  134. 134.
  135. 135.
  136. 136.
  137. 137.
  138. 138.
  139. 139.
  140. 140.
  141. 141.
  142. 142.
  143. 143.
  144. 144.
  145. 145.
  146. 146.
  147. 147.
  148. 148.
  149. 149.
  150. 150.
  151. 151.
  152. 152.
  153. 153.
  154. 154.
  155. 155.
  156. 156.
  157. 157.
  158. 158.
  159. 159.
  160. 160.
  161. 161.
  162. 162.
  163. 163.
  164. 164.
  165. 165.
  166. 166.
  167. 167.
  168. 168.
  169. 169.
  170. 170.
  171. 171.
  172. 172.
  173. 173.
  174. 174.
  175. 175.
  176. 176.
  177. 177.
  178. 178.
  179. 179.
  180. 180.
  181. 181.
  182. 182.
  183. 183.
  184. 184.
  185. 185.
  186. 186.
  187. 187.
  188. 188.
  189. 189.
  190. 190.
  191. 191.
  192. 192.
  193. 193.
  194. 194.
  195. 195.
  196. 196.
  197. 197.
  198. 198.
  199. 199.
  200. 200.
  201. 201.
  202. 202.
  203. 203.
  204. 204.
  205. 205.
  206. 206.
  207. 207.
  208. 208.
  209. 209.
  210. 210.
  211. 211.
  212. 212.
  213. 213.
  214. 214.
  215. 215.
  216. 216.
  217. 217.
  218. 218.
  219. 219.
  220. 220.
  221. 221.
  222. 222.
  223. 223.
  224. 224.
  225. 225.
  226. 226.
  227. 227.
  228. 228.
  229. 229.
  230. 230.
  231. 231.
  232. 232.
  233. 233.
  234. 234.
  235. 235.
  236. 236.
  237. 237.
  238. 238.
  239. 239.
  240. 240.
  241. 241.
  242. 242.
  243. 243.
  244. 244.
  245. 245.
  246. 246.
  247. 247.
  248. 248.
  249. 249.
  250. 250.
  251. 251.
  252. 252.
  253. 253.
  254. 254.
  255. 255.
  256. 256.
  257. 257.
  258. 258.
  259. 259.
  260. 260.
  261. 261.
  262. 262.
  263. 263.
  264. 264.
  265. 265.
  266. 266.
  267. 267.
  268. 268.
  269. 269.
  270. 270.
  271. 271.
  272. 272.
  273. 273.
  274. 274.
  275. 275.
  276. 276.
  277. 277.
  278. 278.
  279. 279.
  280. 280.
  281. 281.
  282. 282.
  283. 283.
  284. 284.
  285. 285.