Introduction to Alzheimer’s Disease 1 1 Introduction to Alzheimer’s Disease David Allsop 1. Introduction In 1907, Alois Alzheimer published an account (1)of a 51-year-old female patient, Auguste D., who suffered from strong feelings of jealousy towards her husband, increased memory impairment, disorientation, hallucinations, and often loud and aggressive behavior. After four and a half years of rapidly dete- riorating mental illness, Auguste D died in a completely demented state. Post- mortem histological analysis of her brain using the Bielschowsky silver technique revealed dense bundles of unusual fibrils within nerve cells (neu- rofibrillary tangles or NFTs) and numerous focal lesions within the cerebral cortex, subsequently named “senile plaques” by Simchowicz (2) (Fig. 1). This combination of progressive presenile dementia with senile plaques and neu- rofibrillary tangles came to be known as Alzheimer’s disease (AD), a term that was later broadened to include senile forms of dementia with similar neuro- pathological findings. It was Divry (3)who first demonstrated the presence of amyloid at the center of the senile plaque, by means of Congo red staining. All amyloid deposits were originally thought to be starch-like in nature (hence the name), but it is now apparent that they are formed from a variety of different peptides and proteins (the latest count being 18). All amyloid share the prop- erty of a characteristic birefringence under polarized light after staining with Congo red dye, which is due to the presence of well-ordered 10 nm fibrils. The underlying protein component of these fibrils invariably adopts predominantly an antiparallel β-pleated sheet configuration. Ultrastructural observations have confirmed that the core of the senile plaque consists of large numbers of closely-packed, radiating fibrils, similar in appearance to those seen in other forms of amyloidosis (4,5), and have also revealed the presence of paired heli- cal filaments (PHFs) within the NFTs (6). However, it took more than 50 yr From:Methods in Molecular Medicine,Vol. 32:Alzheimer’s Disease: Methods and Protocols Edited by: N. M. Hooper © Humana Press Inc., Totowa, NJ 1 2 Allsop Fig. 1. (A)Neurofibrillary tangle (Palmgren silver technique). from Divry’s original observation to determine the precise chemical nature of the senile plaque amyloid. Many neuropathologists have regarded this amyloid as a “tombstone” (an inert bystander) of AD. However, the advent of molecular genetics has finally and firmly established the central role of amyloid in the pathogenesis of the disease, although this is still disputed by some workers in the field. This introductory chapter is written in support of what has become known as the “amyloid cascade” hypothesis. 2. Chemical Nature of Cerebral Amyloid and PHFs The first attempts to determine the chemical nature of senile plaque amyloid were based on immunohistochemical methods, which, not surprisingly, gave unequivocal results. A method for the isolation of senile plaque amyloid “cores” from frozen post-mortem brain was first reported in 1983 (7), and around the same time methods were also developed for the isolation of PHFs (8). The unusual amino acid composition of the senile plaque core protein clearly excluded forms of amyloid known at the time (e.g., AA, AL types) as major components of the plaque core (7). In 1984, a 4-kDa protein, termed “β-protein,” now commonly referred to as Aβ, was isolated from amyloid-laden meningeal Introduction to Alzheimer’s Disease 3 Fig. 1. (B) Senile plaque (Anti-Αβ immunohistochemistry, monoclonal antibody 1G10/2/3,ref. 11). Magnification for both x1100. blood vessels (a frequent concomitant of AD), and its N-terminal amino acid sequence was determined to be unique (9). Antibodies raised to synthetic pep- tides corresponding to various fragments of Aβ were found to react with both senile plaque (Fig. 1B) and cerebrovascular amyloid in brains from patients with AD (10,11), and immunogold labeling studies showed that the amyloid fibrils were decorated with gold particles (12). It was soon recognized that synthetic Aβ peptides will assemble spontaneously into fibrils closely resem- bling those seen in AD (13). These observations clearly demonstrated that Aβ is an essential and integral component of the Alzheimer amyloid fibril. The chemical nature of PHFs remained in dispute for some time after the discovery of Aβ, until evidence for the microtubule-associated protein tau as the principal constituent of PHFs became overwhelming (14–17). The demon- stration that structures closely resembling PHFs could be assembled in vitro from tau established beyond reasonable doubt that tau is an integral component of the PHF (18). There are six major isoforms of human tau (seeFig. 2) derived by alternative mRNA splicing from a single gene on human chromosome 17. Alternative splicing of exon 10 gives rise to 3-repeat and 4-repeat forms, which 4 Allsop Fig. 2. Diagrammatic representation of the major isoforms of human tau. refers to the number of microtubule-binding units. All six of these tau isoforms are expressed in the adult brain, but only the shortest isoform (tau-352) is expressed in the fetal brain. Tau can be phosphorylated at multiple sites, and tau from the fetal brain is more heavily phosphorylated than tau from the adult brain. Tau protein extracted from PHFs (PHF-tau) was found to contain all of the six major isoforms (19). NFTs in AD are composed predominantly of tau in the form of PHFs, but a minority of pathological tau can also exist in the form of so-called “straight” filaments. Intraneuronal filamentous inclusions in other neurodegenerative diseases (e.g., progressive supranuclear palsy) can be com- posed almost entirely of straight filaments. The studies of Goedert and cowork- ers (18) on the in vitro assembly of filamentous structures from different tau isoforms suggest that PHFs and straight filaments are formed from 3-repeat and 4-repeat forms of tau, respectively. Numerous studies (reviewed in ref. 20) using antibodies specific for par- ticular phosphorylation-dependent epitopes demonstrated that PHF-tau appears to be abnormally hyperphosphorylated (i.e., more heavily phosphorylated than fetal tau, and at additional unique sites in the molecule). It later became apparent that the abnormal hyperphosphorylation of tau in AD may have been overemphasized in these studies. Some of the supposed AD-specific phospho- rylation sites on tau have now be seen in living neurons. In particular, analysis of human biopsy tissue has suggested that tau protein is more highly phospho- rylated than previously thought in living brain, due to a rapid (1–2 h) postmor- tem dephosphorylation (21). This has led to the conclusion that there may be a Introduction to Alzheimer’s Disease 5 Fig. 3. Structure of APP, showing some of the major functional domains. deficiency (or inhibition) of phosphatase activity in brains from patients with AD (21). However, on balance, it is clear that abnormal aggregates of tau in a highly phosphorylated state are a hallmark of AD pathology, and it remains likely that tau phosphorylation plays a role in NFT formation. Levels of phos- phorylated tau are significantly higher in fresh lumbar puncture samples of cerebrospinal fluid taken from AD patients than in similar samples from age- matched controls (22). Furthermore, a number of studies have now shown that fibrillized forms of Aβ can induce tau phosphorylation in vitro and in vivo. This reinforces the possibility of a direct link between amyloid deposition and tau phosphorylation (considered further below). 3. The Amyloid Precursor Protein (APP) The amino acid sequence of the Aβ peptide was used by Kang et al. (23) to identify from a fetal brain cDNA library a full-length clone that encoded Aβas part of a much larger 695 amino acid precursor (APP ). This precursor was 695 predicted to contain a single membrane-spanning domain towards its carboxyl- terminal end, with the sequence of the Aβpeptide commencing at amino acid residue 597 and terminating part way through the membrane-spanning region (see Fig. 3). Subsequently, a number of slightly longer cDNA clones were isolated by other workers. The 751 amino acid APP sequence (APP ) 751 described by Ponte et al. (24) contained an additional 56 amino acid insert encoding a Kunitz-type serine proteinase inhibitor (KPI). Kitaguchi et al. (25) identified another precursor (APP ) with both the KPI sequence and an 770 additional 19 amino acid insert. These isoforms of APP arise as a result of alternative splicing of exons 7 and 8 during transcription of the APP gene. Additional isoforms generated by alternative splicing of exon 15 have also been described (26). It is not clear if all of these various isoforms of APP can give 6 Allsop Fig. 4. Aβ region of APP, showing the pathogenic APP mutations and the α-, β-, andγ-secretase cleavage sites. rise to amyloid in the brain. DeSauvage and Octave (27) have also found a smaller APP mRNA variant (APP-593) lacking the Aβ coding region. 4. Proteolytic Processing of APP Following discovery of the full-length APP cDNA clone, numerous studies were undertaken to detect the APP protein in cells and tissues. Full-length, membrane-bound forms of APP were readily detected by Western blotting, and it soon became apparent that a large, soluble, N-terminal fragment of APP (sAPPα) is released by the action of a putative “α-secretase” into conditioned tissue culture medium, cerebrospinal fluid, serum, and tissues such as brain (seeFig. 4). Esch et al. (28) and Anderson et al. (29)showed that this was due to cleavage of APP at the Lys16-Leu17 bond in the middle of the Aβ sequence, which would preclude formation of the intact Aβpeptide. This led to specula- tion that the production of Aβ from APP must be a purely pathological event (30). However, it soon became apparent that C-terminally truncated forms of secreted APP completely lacking Aβ immunoreactivity could also be detected (31,32), along with C-terminal membrane-associated fragments of APP appar- ently containing the entire Aβ sequence (33). Seubert et al. (32) demonstrated the existence of a form of secreted APP (sAPPβ) that terminates at the Met596 residue immediately prior to the N terminus of the Aβsequence. This was dem- onstrated by means of a specific monoclonal antibody (termed “92”) to resi- dues 591–596 of APP , the reaction of which depended on the presence of the 695 free carboxyl-terminal Met596. These observations suggested the presence of an alternative “β-secretase” activity that cleaves APP to release the N terminus of the Aβpeptide. The detection of Aβitself in culture medium from cells, and in body fluids (cerebrospinal fluid, blood, urine) from normal individuals (34–37), showed that this peptide is, in fact, a product of the normal metabolism of APP. These findings also inferred the action of a third “γ-secretase” activity that acts within the membrane-spanning domain of APP to produce the C-terminus of Aβ. The detection of “short” (predominantly Aβ40) and “long” (predominantly Introduction to Alzheimer’s Disease 7 Aβ42) forms of Aβ (see, e.g., ref. 38) was also important, given later data on the effects of familial AD mutations on APP processing. The Aβ peptide may be physiologically active in brain, as in its soluble form it has weak neurotrophic properties (see below). The identity of the α-, β-, and γ-secretases is unknown, although it is likely that α-secretase is a zinc metalloproteinase (39). There are numerous reports claiming identification of β-secretase and fewer reports claiming the identifi- cation of γ-secretase, but in no case for the various candidates in the litreature is there strong evidence that they are actually β- or γ-secretase. As far as β-secretase is concerned, the multicatalytic proteinase or “proteasome” has been implicated (40), as have several chymotrypsin-like serine protein- ases (41–43). The metallopeptidase thimet has been proposed (44), but has always been an unlikely candidate, as it seems not to tolerate large substrates such as APP, and can now be discounted (45). Cathepsin D (an aspartyl pro- teinase) has received considerable attention as a potential β-secretase due to its ability to cleave peptide substrates containing the APP Swedish mutant se- quence at a much faster rate than the normal sequence (46). However, the fact that cathepsin D knockout mice still produce Αβ (47) indicates that this en- zyme cannot be β-secretase. Anumber of small peptide aldehydes of the type known to inhibit both cys- teine and serine proteinases have been shown to inhibit Αβ formation from cultured cells, probably through inhibition of the γ-secretase pathway (48–51). The activity of these compounds as inhibitors of γ-secretase cleavage has been shown to correlate with their potency as inhibitors of the chymotrypsin-like activity of the proteasome, suggesting that the latter may be involved, either directly or indirectly, in the γ-secretase cleavage event (52). Further candidates forγ-secretase include prolyl endopeptidase (53), and cathepsinD (54). In the case of γ-secretase, there is the additional complication that there may be separate enzymes responsible for the generation of Αβ40 and Αβ42(50,51). APP is synthesized in the rough endoplasmic reticulum, and follows the con- ventional secretory pathway through the Golgi apparatus where it is tyrosyl sulfated and sialylated (55), and then to secretory vesicles and the cell surface. Studies on the subcellular compartments where the α-, β-, and γ-secretase cleavages take place are complicated by the fact that the sites of processing may well be different in neuronal and nonneuronal cells, and also the fact that many published data were obtained using APP-transfected cells where the overexpressed APP could be forced into a nonphysiological compartment. Cur- rent evidence suggests that in differentiated neuronal cells the formation of Αβ40 occurs in the trans-Golgi network, whereas Αβ42 is synthesized at an earlier point en route to the cell surface within the endoplasmic reticulum (56). This finding that Αβ40 and Αβ42appear to be formed in different subcellular 8 Allsop compartments has strengthened the possibility that they may be derived by differentγ-secretases. However, an alternative possibility is that the intracellu- lar membranes at the sites of production of Αβ40 and Αβ42 by the same γ-secretase are slightly different thicknesses (56). 5. Aggregated Forms of Aβ Show Neurotoxic Properties Whitson et al. (57,58)first reported that Αβhas mild neurotrophic effects in vitro, and Yankner et al. (59) showed that Αβcan also have neurotoxic proper- ties. Initial difficulties in reproducing these findings in other laboratories were largely resolved when it was realized that the physiological properties of Αβare critically dependent on its state of aggregation. Freshly dissolved, soluble pep- tide appeared to promote neuronal survival, whereas peptide that had been “aged” for >24 h (and was therefore in an aggregated, fibrillar form) showed neurotoxic properties (60). The precise mechanism by which aggregated Αβ causes neu- ronal degeneration in vitro is unclear, but the effect is likely to be due to disrup- tion of Ca2+ homeostasis and induction of oxidative free radical damage. Also, Αβcan induce apoptosis or necrosis, depending on the concentration of Αβand the cell type under investigation. There is still no clear evidence that this toxic- ity is mediated via an initial binding between Αβand a membrane-bound recep- tor, although the “RAGE” (receptor for advanced glycation end products) has been suggested to be involved (61). The identity of the precise molecular form of Αβ responsible for its cytotoxic effects is unclear, with both mature fibrils (62) and dimers (63) being implicated. The identification of a protofibrillar intermediate in β-amyloid fibril formation may shed light on this matter (62,64). There is also considerable debate concerning the relevance of these observa- tions to the actual process of neurodegeneration in the brains of patients with AD. Yankner has recently provided compelling evidence that Αβ also shows neurotoxic properties in vivo when injected into the brains of aged primates (65). This effect was not found with younger animals, suggesting that the aged brain may be particularly vulnerable to Αβ-mediated neurotoxicity. This very important finding also casts doubt on the relevance of many of the in vitro Αβ-induced models of toxicity. It has also become increasingly apparent that the in vivo aggregation of Αβ probably precipitates a chronic and destructive inflammatory process in thebrain(66). Activation of both microglia and astrocytes occurs in the imme- diate vicinity of senile plaques in the brains of AD patients. These two cell types are the primary mediators of inflammation in the CNS, through the pro- duction of a wide range of proinflammatory molecules such as complement, cytokines, and acute-phase proteins. Because APP synthesis is upregulated by interleukins such as IL-1, this is likely to lead to a vicious cycle whereby amyloid deposits stimulate microglial activation and cytokine production, leading to Introduction to Alzheimer’s Disease 9 even higher expression of APP (66), with the whole process culminating in the degeneration of neuronal cells, possibly via the production of free radicals by activated microglia, or by complement lysis of neuronal membranes. The initi- ating event in this process may be the Αβ-mediated activation of complement (67,68), or the binding of Αβ peptide to microglia via scavenger (69,70) or RAGE receptors (61). 6. The Normal Functions of APP Many potential functions have been ascribed to either full-length or secreted APP, including protease inhibition, membrane receptor (possibly G coupled), 0 cell adhesion molecule, regulation of neurite outgrowth, promotion of cell sur- vival, protection against a variety of neurotoxic insults, stimulation of synaptogenesis, and modulation of synaptic plasticity (see ref. 71 for a recent review). Kang et al. (23) originally pointed out similarities between full-length APP and cell-surface receptors. This idea has received some support from the find- ing that the cytoplasmic domain of APP can catalyze guanosine triphosphate (GTP) exchange with G suggesting that APP might function as a G -coupled O 0 receptor(72). However, this finding remains to be confirmed by others. If this finding is true, the activating ligand is unknown, but APP is clearly not a con- ventional 7-transmembrane G protein-coupled receptor. The secreted form of APP containing the (KPI) insert was found some time ago to be identical to protease nexin II, a growth regulatory molecule produced by fibroblasts (73). Protease nexin II is an inhibitor of serine proteinases, including factor XIa of the blood clotting cascade (74). APP has also been found to inhibit the matrix metalloproteinase gelatinase A (75), possibly through a small homologous motif between residues 407–417 of APP-695 and Cys3–Cys13 of tissue inhibitor of matrix metalloproteinases (TIMP) (76). Several studies have suggested that APP functions as an adhesion molecule, promoting cell–cell or cell–extracellular matrix interactions (71). APP has at least one high-affinity heparin-binding site (77), a collagen-binding site (78), and an integrin-binding motif (amino acid sequence RHDS at residues 5–8 of Aβ(79) and has been shown to bind to laminin, collagen, and heparan sulfate proteoglycans(80). Agrowth-promoting effect of soluble APP has been shown for fibroblasts and cultured neurons, and this activity has been claimed to reside in the amino acid sequence RERMS at residues 328–332 of APP (81,82). Synthetic 695 RERMS peptide and a 17-mer peptide containing this sequence were reported to retain the neurotrophic properties of soluble APP. In addition, the bioactivity of these peptides was reversed by the antagonist peptide RMSQ, which over- laps the active RERMS pentapeptide at the C-terminal end. Specific and satu-