Alzheimer's Disease and the Amyloid β Protein

Alzheimer's disease (AD) is characterized by the deposition of amyloid in the extracellular compartment of the brain in the form of congophilic amyloid angiopathy (CAA) and amyloid plaques (APs). Intracellular neurofibrillary tangles (NFTs) (88) formed from the abnormally phosphorylated cytoskeletal protein tau are also seen (52). The identification of the amyloid β protein (Aβ) in CAA and APs (28 ; 58) led to the cloning of the amyloid protein precursor (APP) (44). The discovery of familial AD (FAD) mutations in the APP gene (10 ; 29 ; 63 ; 64 ; 86) has supported the view that a defect in APP metabolism or function is directly involved in AD pathogenesis. The demonstration that mutations in the tau gene can lead to non-Alzheimer dementias with neurofibrillary pathology, lacking Aβ plaques (reviewed by 81), has reinforced the view that the NFTs are a secondary phenomenon in the pathogenesis of AD. It has long been argued that the deposition of amyloid is an early step in AD pathogenesis (58 ; Hardy and Higgins, 1992 ; 57). The term amyloid refers to insoluble proteinaceous deposits that are congophilic and exhibit red-green birefringence in the presence of plane polarized light (45). Implicit in much of the research on the role of APP and Aβ has been the assumption that deposits of amyloid are toxic to the brain (42) and that these deposits are the underlying cause of AD. The observation that Aβ peptides when “aged” (incubated to form amyloid fibrils) become toxic to neurons in culture (94 ; 22 ; 46 ; 66 ; 38) has further supported this view. The amyloid cascade hypothesis of AD, as formalized by Hardy and Higgins (35), states that Aβ“precipitates to form amyloid and, in turn, causes neurofibrillary tangles and cell death.” However, this hypothesis has been challenged (see, e.g., 16 ; 34). It has been argued that the deposition of amyloid does not correlate with dementia (89 ; 2 ; 71 ; 72 ; 5), although the failure to observe a correlation may be related to the method by which AP load is measured (14). Whether amyloid deposits have a pathogenic role remains a controversial issue. Some of the neuropathological changes occurring in the AD brain were first described by Alzheimer (1). Extracellular deposits of amyloid in the form of APs and CAA, as well as intracellular NFTs, are major features of AD pathology (68). The major protein constituent of CAA and APs is a 4-kDa polypeptide termed the amyloid protein or Aβ (28 ; 58). A partial amino acid sequence of Aβ was used to clone a cDNA encoding a protein now referred to as the APP, which has features of an integral type I transmembrane glycoprotein (44). The APP gene contains 18 exons spanning >170 kb (95). The region encoding the Aβ sequence comprises part of exons 16 and 17 and contains between 40 and 43 amino acid residues that extend from the ectodomain into the transmembrane domain of the protein (Fig. 1). Primary structure of APP. Upper panel : APP is expressed initially as a type I transmembrane glycoprotein containing a large N-terminal ectodomain, a short transmembrane (TM) domain of hydrophobic amino acid residues, and a relatively short C-terminal cytoplasmic domain. The Aβ sequence makes up part of the ectodomain and extends partly into the TM domain. The APP gene encodes a protein containing a signal peptide sequence (SP), a cysteine-rich domain, a region rich in acidic residues, a domain with homology to Kunitz-type protease inhibitors (KPI), and a region sharing homology to the OX-2 protein. mRNA splicing can produce forms that lack the KPI or OX-2 domain. Lower panel : Amino acid sequence of the Aβ region of APP. Large arrows show the major sites of cleavage for α-, β, and γ-secretases. Small arrows show known FAD mutations. FIG. 1. It is now known that Aβ is a normal product of APP processing (32 ; 30 ; 21). The major route of APP processing is by α-secretase, an enzyme that cleaves within the Aβ sequence (20). Cleavage by β- and γ-secretases at the N- and C-terminal ends of the Aβ sequence liberates the Aβ polypeptide, which can subsequently be secreted from cells (21 ; 30 ; 31). The major form of Aβ that is secreted contains 40 amino acids (Aβ1-40). However, minor species containing 42 or 43 amino acid residues (Aβ1-42/43) are also produced. These extended forms of Aβ aggregate more readily and may seed amyloid fibril polymerization during the early stages of plaque formation (42). The strongest evidence for a pathogenic role for APP or Aβ comes from genetic studies of early-onset FAD (33). Several FAD mutations have been found in the APP gene. All of these mutations have been found to cluster close to the amyloid sequence in APP. Mutations at codon 716 (Florida), 717 (London), and 723 (Australian) cause an increased proportion of γ-secretase cleavage at position 42 or 43 in the amyloid sequence (Fig. 1 and Table 1). A mutation found at codons 670 and 671 in a Swedish kindred results in increased β-secretase cleavage (8 ; 43), whereas a point mutation at codon 612 (Flemish) inhibits α-secretase cleavage (32). The consequence of the Swedish and Flemish mutations is to increase processing of APP via the β-secretase pathway. All of the FAD mutations in the APP gene result in increased production of Aβ1-42/43 (Table 1). More than 40 FAD mutations in the presenilin 1 and presenilin 2 genes have also been reported (17 ; 75). The common feature of all these mutations is that they also cause an increase in the production of Aβ1-42/43 (73). Thus, presenilins are involved in regulating the proteolytic breakdown of APP by γ-secretase (18). However, the mechanism by which this occurs is unknown. TABLE 1. Studies on transgenic mice that overexpress FAD mutant forms of human APP also strongly argue for a central role of APP or Aβ in disease pathogenesis. Games et al. (26) demonstrated Alzheimer-like pathology in a transgenic mouse overexpressing a mutant (V717F) form of APP. Subsequently, Hsiao et al., (40) demonstrated similar pathology in a mouse expressing another mutant form of APP (Swedish mutation). It is interesting that the Hsiao mouse had behavioral changes indicative of a cognitive defect occurring before the appearance of robust plaque pathology. More recently, AD-like pathology has been reported in two other APP transgenic mice carrying FAD mutations (84). Although these mice do not show NFTs, there are behavioral abnormalities, abnormal neuritic processes, neuron and synapse loss, and biochemical abnormalities reminiscent of AD (26 ; 40 ; 56 ; 26 ; 9 ; 23 ; 41 ; 65 ; 79). The validity of these transgenic mice as models of AD pathology is reinforced by the observation that the AD-like phenotype is accelerated in APP transgenic mice that also contain an FAD mutant presenilin 1 transgene (37). Studies on the genetics of AD and on APP transgenic mice provide compelling evidence that a disturbance in APP metabolism or function is the underlying cause of AD. However, these studies do not prove that Aβ is the causative agent. The strongest evidence implicating Aβ in the pathogenesis of AD comes from the observation that Aβ peptides are toxic to neurons in culture (94 ; 22 ; 46 ; 66 ; 38). This toxicity is enhanced if the peptides are “aged” (incubated from hours to days), a procedure that increases amyloid fibril formation (66). Although the process of aging increases the number of amyloid fibrils formed from Aβ, this is not proof per se that fibrils are the major toxic form of Aβ. It is likely that the levels of soluble oligomeric species of Aβ are also increased by the process of aging (see next section). The mechanism of neurotoxicity is unclear. Some studies suggest that Aβ can disrupt calcium homeostasis (Mattson et al., 59, 61,60), perhaps by interfering with L-type voltage-dependent calcium channels (15 ; 90), Aβ may reduce Na+,K+-ATPase activity (55), thereby influencing membrane depolarization. Furukawa and Mattson (25) have reported that cytochalasin D, a compound that inhibits actin polymerization and calcium entry, can reduce Aβ neurotoxicity. Other studies suggest a role for reactive oxygen species in Aβ toxicity (4 ; 6 ; 36). Disturbances in redox potential may lead to disruption of calcium homeostasis, as reactive oxygen species can impair ATPase activities (55). Aβ may cause lipid peroxidation and affect superoxide dismutase, which may contribute to its neurotoxicity in culture (7). The receptor that transduces the effects of Aβ is unknown, although the receptor for advanced glycation end products (RAGE) (91) has been implicated. Studies by Yan et al. (92) suggest that Aβ may also bind an intracellular hydrosteroid dehydrogenase known as ERAB. The neurotoxicity may be mediated by an indirect action of Aβ on a nonneuronal cell. For example, microglial cells are often found in association with neuritic plaques (69), and Aβ has been shown to activate microglia in culture (11). Therefore, the possibility that Aβ stimulates release of an unidentified neurotoxic agent from a nonneuronal compartment must also be considered. Recent work by Geula et al. (27) has shown that when aged Aβ is injected into the brains of old rhesus monkeys, it is neurotoxic. However, injection of the same material into young monkeys has little toxic effect. This suggests that although Aβ may be pathogenic, there must be other age-related susceptibility factors that are also important to generate a toxic reaction in vivo. There is now considerable evidence that the type 4 allele of the apoE gene is a major susceptibility factor for late-onset AD (82). ApoE is a 299-amino acid glycoprotein that is principally involved in lipid transport and related functions (reviewed by 54). The N-terminal domain (residues 1-191) contains a receptor-binding region (residues 136-150) that exists as a four-helix bundle. The C terminus is hydrophobic and contains the lipoprotein-binding determinants. Genetic heterogeneity leads to three common isoforms, designated as apoE2, apoE3, and apoE4, encoded by three alleles called ε2, ε3, and ε4, respectively. The isoforms differ from each other by cysteine-arginine interchanges at positions 112 and 158 (54). The ε4 allele frequency is significantly increased in late-onset AD patients (82). Furthermore, those individuals with one or two copies of the ε4 allele have higher amounts of Aβ immunoreactivity in their brains than those individuals without ε4 (74), suggesting that apoE4 promotes fibrillation of Aβ to form amyloid. There is evidence from several studies to suggest that apoE4 could be involved in the polymerization of Aβ to form amyloid in vivo. For example, apoE4 has been found to promote Aβ fibrillogenesis in vitro more readily than apoE3 (Strittmatter et al., 82,83 ; 53). Also studies in which apoE knockout mice have been crossed with human APP transgenic mice show that the expression of apoE is necessary for amyloid deposition in vivo (3). However, the mechanism by which apoE4 influences the risk of AD is still unknown. Although apoE can bind Aβ (Strittmatter et al., 82,83), most studies showing that apoE4 stimulates Aβ aggregation have used apoprotein, i.e., delipidated, forms of apoE, which do not possess a native conformation. Indeed, studies using native forms of apoE isoforms suggest that apoE4 binds less well to Aβ than the other isoforms (49 ; 96 ; 93). On the basis of this finding it has been proposed that apoE may be involved in clearance of Aβ (92), although this hypothesis would not explain why apoE knockout inhibits Aβ deposition (3). Molecular genetic studies indicate a central role for Aβ in AD pathogenesis. However, these studies do not indicate the form or site of action of Aβ neurotoxicity. Until the mechanism of Aβ neurotoxicity is understood, it will be difficult to explain the topography of neurodegeneration (77). It has been presumed that deposits of amyloid constitute the entire Aβ load, but recent studies indicate that some of the Aβ in the brain exists in a soluble form. Soluble Aβ is unlikely to be detected by routine fixation and immunostaining. Aβ is probably secreted as a monomer and subsequently aggregates into soluble oligomers or fibrils (67). There is good evidence for the existence of low-molecular-weight Aβ oligomers in the brain. Studies by Kuo et al. (47) have isolated watersoluble Aβ oligomers from normal and AD brains. Perhaps of greatest interest in this study was the finding that not only was the level of soluble Aβ greater in the AD brain compared with controls, but as with some familial mutations, the proportion of soluble Aβ1-42/43 was significantly increased over soluble Aβ1-40 species in AD patients. Similar results have been obtained by Funato et al. (24). In the study by Kuo et al. (47), the watersoluble Aβ species ranged in size from monomers of 100 kDa. Studies by Roher et al. (70) suggest that the watersoluble dimeric species are neurotoxic, whereas in a recent study Lambert et al. (50) found that small, low-molecular-weight oligomers of Aβ1-42 are several orders of magnitude more potent neurotoxins than high-molecular-weight fibrillar species of Aβ1-40. These studies have implications for our view of amyloid toxicity. If oligomeric soluble forms of Aβ have a pathogenic role, then is it possible that APs are not the major toxic form of Aβ in the brain (Fig. 2) ? Model describing pathways of APP processing. APP can be cleaved by β- and γ-secretases to yield Aβ1-42/43 (γ42/43) or Aβ1-40 (γ40), which can be actively secreted. Alternatively, APP can be cleaved by α-secretase (α) to yield sAPPα, which several studies (reviewed by 78) have shown may have neuroprotective or trophic functions. Aβ1-42/43 can aggregate to form soluble oligomeric species or may seed the polymerization of Aβ1-40 to form insoluble amyloid fibrils, which are deposited in the form of APs. FIG. 2. Recently, Crook et al. (13) described an unusual variant of AD involving a deletion of exon 9 of the presenilin 1 gene from the mRNA. Both NFTs and Aβ-immunopositive plaques were present, but the plaques were of the diffuse nonneuritic (nonfibrillar) type. Like the other FAD mutations in the presenilin 1 gene, the exon 9 deletion also increases Aβ1-42/43 production (62). There are at least two possible explanations for the existence of an AD variant with only diffuse plaques. As previously considered, 42- or 43-residue-long forms of Aβ may be secreted to exert a neurotoxic action (Fig. 2). However, a second possibility is that intraneuronal Aβ plays a role in pathogenesis. Although the molecular genetic studies strongly argue for a direct role of Aβ1-42/43 in AD pathogenesis, they do not provide any indication of whether the Aβ is extracellular or intracellular. Skovronsky et al. (76) have shown that Aβ1-42/43 accumulates preferentially in an insoluble intracellular fraction where it is more abundant than Aβ1-40. Lee et al. (51) have localized intracellular Aβ1-42/43 to the endoplasmic reticulum. Clearly, if Aβ accumulates within intracellular organelles, it could have profound effects on normal cellular protein trafficking and metabolism. If Aβ1-42/43 is the real culprit in AD, then an inverse correlation between the amount of Aβ1-42/43 production and the age of onset of the disease might be predicted. Mutations causing high levels of Aβ1-42/43 in cell culture should cause an early age of onset of clinical symptoms. However, studies with fibroblasts taken from FAD patients (73) show that this is not the case (17). There are several possible explanations for this. It is possible that the level of Aβ1-42/43 production in cell culture does not reflect the level in brain. Another possibility is that FAD mutations influence other cellular events that affect the age of onset. A third possibility is that Aβ1-42/43 is not the only (or even the major) pathogenic form of Aβ. Although Aβ peptides terminating in positions 39-43 are the major forms produced, it is possible that very low levels of previously undetected longer forms of Aβ may also exist (12). If they do exist, they could also be neurotoxic. This would be consistent with observations that C-terminal fragments of APP containing the Aβ sequence are more toxic than Aβ (80). There is very good evidence that Aβ accumulation is the underlying cause of FAD, and there is strong circumstantial evidence to suggest that a similar process underlies the pathogenesis of sporadic (late-onset) AD. Although amyloid deposits (APs and CAA) are markers of the disease, insoluble fibrillar Aβ may not be the main neurotoxic form. Low-molecular-weight diffusible forms of Aβ1-42/43 may also be important. The fact that aged Aβ in vitro contains more amyloid fibrils does not necessarily prove that amyloid is neurotoxic. It is likely that aging also produces increased amounts of soluble oligomeric Aβ species. Therefore, more work is needed to define the precise nature of the toxic form of Aβ and to delineate the mechanism of this toxicity. D.H.S. is supported by grants from the National Health and Medical Research Council of Australia and the Rebecca L. Cooper Foundation.