Amyloid-bearing plaques, neurofibrillary tangles and neuronal dystrophy and loss characterize the pathology of Alzheimer's disease

It is becoming increasingly clear that AD represents a syndrome with well-defined clinical and neuropathological hallmarks but with an array of specific molecular defects which can initiate the pathology. Despite this etiological heterogeneity, growing evidence suggests that there is a common and rather stereotyped pathogenetic cascade which can result from distinct gene defects and/or as yet unknown environmental factors. External examination of the brains of AD patients usually reveals considerable cortical atrophy, particularly in the limbic and association cortices, together with enlargement of the lateral ventricles. However, the hallmarks of the disorder that confirm the diagnosis are observed on microscopic examination, usually with the aid of a silver stain (Fig. 46-1). These include loss of neurons, particularly medium- and large-sized pyramidal cells, and the presence of intraneuronal NFTs and extracellular deposits of amyloid filaments surrounded by altered neuritic processes and glia. These are termed senile plaques and are not specific for AD. Senile plaques can be seen in many functionally normal individuals beginning after age 60 years in small numbers and in limited topographic distribution, particularly in the hippocampus, amygdala and other limbic structures.

Figure 46-1

The classical histopathological lesions of Alzheimer's disease demonstrated by the modified Bielschowsky silver stain. A 6-μm paraffin section of the amygdala from a 69-year-old man with a 6-year history of progressive dementia. Darkly staining (more...)

Senile plaques are structurally complex lesions, the temporal development of which is only partially understood. Many, if not all, senile plaques probably begin as amorphous, largely nonfilamentous aggregates of a 40- to 42-residue protein, the amyloid β protein (Aβ). After a period of time, the length of which is not known, some of these “diffuse” Aβ deposits become increasingly fibrillar and gradually acquire the classical features of amyloid plaques, namely, relatively compacted bundles of ~8-nm filaments which bind certain histochemical dyes, such as Congo red and thioflavin, and have principally a β-pleated sheet protein conformation. Compacted or mature amyloid plaques are frequently associated with numerous dystrophic axons and dendritic processes that lie within or immediately around the fibrous amyloid deposit. These neurites are often thickened and intensely silver-positive. Such mature plaques also display activated microglia intimately associated with the central amyloid deposit and fibrous astrocytes rimming the plaque. The finding of large numbers of such senile or neuritic plaques in limbic and association cortices is probably the single most reliable neuropathological marker of the diagnostic entity of AD.

In the large majority of AD brains, senile plaques are accompanied by argyrophilic bundles of intraneuronal cytoplasmic fibers, the NFTs. Electron microscopy of such neurons demonstrates that the tangles are generally composed of masses of paired, helically wound, ~10-nm filaments (PHFs), often intermixed with ~15-nm straight filaments. Anatomical studies have shown that NFTs are frequently present in the cell bodies of neurons whose axons project to the sites of neuritic plaques, for example, the entorhinal → hippocampal perforant pathway and the basal forebrain → hippocampal/neocortical pathway. NFTs are a less specific histological marker of AD than are neuritic plaques. They can occur in the absence of neuritic plaques in a number of etiologically diverse neurological disorders, such as subacute sclerosing panencephalitis, Kufs' disease and Hallervorden-Spatz disease. Moreover, a minority of AD brains, perhaps 10 to 15%, show abundant amyloid-bearing neuritic plaques but few or no NFTs in the neocortex. Thus, there can be a clear dissociation of plaques and tangles under some circumstances. The wide variety of neuropathological disorders in which NFTs occur suggests that PHF formation is a relatively nonspecific marker of certain kinds of neuronal injury.

In addition to a variety of morphological types of Aβ protein deposits present in the brain parenchyma, the cortical and meningeal arteries, arterioles, capillaries and, to a lesser extent, venules may contain multifocal deposits of amyloid filaments composed of Aβ. The amyloid deposits appear to be preferentially localized to the abluminal basement membrane of these microvessels. The number of amyloid-bearing cortical vessels can vary dramatically among AD cases having relatively similar densities of amyloid plaques. The association of microvascular amyloidosis with parenchymal amyloidosis in AD has aroused interest in view of the discovery of missense mutations in the β-amyloid precursor protein (βAPP) in both families with AD and families with severe amyloid angiopathy causing cerebral hemorrhages (see below).

Neurons in the limbic and association cortices and the subcortical nuclei that project to them are particularly vulnerable to neurofibrillary tangle formation

In most cases of AD, the innumerable NFTs found in the limbic and association cortices are accompanied by NFTs in neurons of subcortical nuclei that project to these regions. These nuclei include the cholinergic basal forebrain complex, the locus ceruleus and the median raphe nuclei. NFTs are very rarely found in regions of brain that are only minimally involved, both pathologically and clinically, in AD, for example, the cerebellum. In such largely unaffected areas, diffuse or “preamyloid” forms of Aβ deposits (see below) may occur but there is little surrounding neuritic dystrophy and usually no NFT formation.

In addition to containing NFTs, neurons in the limbic and association cortices and in the subcortical nuclei that project to them often undergo perikaryal shrinkage and actual cell death. It is likely that some or many neurons which shrink and die in AD do not pass through a stage of actual NFT formation.

Most cases of AD that have extensive NFTs also show widespread dystrophic neurites, sometimes called neuropil threads or curly fibers, that are scattered in the cortical neuropil and not specifically localized to amyloid plaques. An abundance of dystrophic neurites in the cerebral cortex has been correlated to some extent with both the presence of NFTs and the occurrence of clinical dementia. Such dystrophic neurites are not specific for AD and have been found in other neurological disorders in which NFTs occur in the absence of amyloid plaques.

Multiple neurotransmitter systems are affected in a pattern that correlates with the cellular pathology

The topographically widespread and cytologically heterogeneous populations of neurons affected in the AD brain are associated with a complex array of neurotransmitter deficits. The first transmitter alteration to be defined was a marked decline in the activities of choline acetyltransferase and acetylcholinesterase, indicating dysfunction and loss of basal forebrain cholinergic neurons and their cortical projections. Although the decline in these cholinergic markers has been correlated with both the degree of dementia and the number of neuritic plaques, cholinergic loss should not be considered the preeminent neurotransmitter alteration in AD because many neurons releasing monoamine or neuropeptide transmitters also become morphologically abnormal and undergo attrition. These neurons include, for example, noradrenergic and serotonergic cells in the brainstem, cells producing somatostatin or corticotropin-releasing factor in the neocortex and neurons which release glutamate, GABA, substance P and/or neuropeptide Y. The degrees of decline in the concentrations of these transmitters and their biosynthetic and catabolic enzymes vary markedly among AD brains. This heterogeneity of neurotransmitter alterations together with possible involvement of their second-messenger systems helps to explain why attempts at replacement therapy aimed at just one of these neurotransmitters, most commonly the use of cholinergic agents, have met with very limited success in terms of measurable and sustained improvement in objective cognitive tests. It is clear that AD does not follow the patterns of certain other neurodegenerative disorders, such as Parkinson's disease (Chap. 45), which are more specific in their neurotransmitter profile.

The search for etiologies has resulted in a focus on genetic factors

Ever since the original description of the disorder by Alzheimer, there has been a lively debate as to what factor or factors could initiate this complex, multicellular degeneration. Although AD shows limited parallels with the infectious/inherited spongiform encephalopathies, such as Creutzfeldt-Jakob disease and Gerstmann-Straussler-Scheinker syndrome (see below, “Prion Diseases”), no compelling evidence that AD is caused by an infectious pathogen has been presented. The possibility that an environmental toxin could precipitate the disorder has revolved largely around the role of metal ions, particularly aluminum. Aluminum initially became a candidate simply because it was found to induce silver-positive bundles of neurofilaments in neurons upon injection into rabbit CNS. However, these filamentous lesions are now known to be distinct from Alzheimer-type NFTs, both structurally and biochemically. Some investigators have reported elevated concentrations of aluminum in cortical regions affected by Alzheimer pathology, particularly within the NFTs themselves. However, equal or greater aluminum accumulation has been found in the NFTs of Guam Parkinson dementia complex and certain variants of Hallervorden-Spatz disease, disorders in which few or no amyloid plaques or amyloid angiopathy of the Alzheimer type appear. This observation suggests that aluminum deposition is not unique to AD but, rather, that aluminum can associate secondarily with NFTs, regardless of the specific disease that causes the tangle formation.

Investigators who have demonstrated aluminum within NFTs have sometimes reported abnormal concentrations of other metals, particularly magnesium, calcium and iron, in AD neurons. The presence of aluminum in amyloid plaque cores has been reported, but this association has been less clearly confirmed than that with NFT. It should also be noted that some investigators have reported no substantial elevation of aluminum in the AD brains which they examined. The pathological role of neuronal aluminum deposition in AD remains unclear, and there is little compelling evidence that it serves as a primary toxin which can initiate the disease. In this regard, it is worth noting that aluminum is ubiquitous in our environment, including in the drinking water of many communities. There are currently no rigorous data indicating an unusual or specific source of aluminum, such as antacids, aluminum containers or deodorants, as a risk factor for AD in the general population.

It has been known for decades that some cases of AD occur in an autosomal dominant Mendelian pattern, and this has turned out to be a fruitful clue for the etiological study of the disorder. The percentage of all AD cases that clearly shows such a pattern has often been reported as 10 to 15%, but a much higher percentage of patients has a clinical history of one or more first-degree relatives with a highly similar dementing syndrome. Because of the late onset of most AD cases, it has been difficult to ascertain whether members of previous generations were actually afflicted with the disease. Mounting evidence suggests that a high percentage of subjects, although by no means all, have inherited some type of genetic predisposition to the disease. In a growing number of families, specific DNA mutations have now been identified.

Support for the hypothesis that AD could be genetically based came initially from the observation that virtually all subjects with trisomy 21, also termed Down's syndrome, develop typical histopathological lesions of AD if they survive into their 40s and beyond. This single clinicopathological clue has been perhaps the most significant factor in unraveling the mechanism of AD. Numerous important discoveries about the causation and biochemical mechanism have derived from the link between trisomy 21 and AD lesions. Of particular interest in this regard has been the realization, since the mid-1980s, that subjects with Down's syndrome dying in their teens or 20s may show low to moderate densities of diffuse Aβ deposits in the limbic and association cortices in the absence of detectable surrounding neuronal or glial changes, neuritic dystrophy or NFTs. This important observation has strengthened the concept that deposition of Aβ is a very early event in the disease, preceding other histological changes.

Clues to the mechanisms of familial Alzheimer's disease have arisen from biochemical analyses of the neuropathological phenotype

Although early attempts to purify and chemically analyze pathological structures in AD brain tissue focused on the study of NFTs (see next section), these have not yet led to clues about the molecular etiology of this disease. Instead, the purification of amyloid deposits from meningeal blood vessels by Glenner and Wong in 1984 [2] provided the seminal biochemical information that led both to the identification of the first specific molecular cause of AD and to a plausible model for the pathogenetic cascade. Prior to the study of the cerebral amyloid, widespread doubts had been expressed as to whether biochemical analysis of the histopathological lesions would lead to useful insights about early events in the disease. However, knowledge of other human amyloidoses suggested that extracellular deposits of amyloid-forming proteins could cause local cytopathology and organ dysfunction. Particularly in cases where there were genetic defects in the amyloidogenic protein, such as transthyretin in familial amyloidotic polyneuropathy, these deposits could explain the etiology of the disease in which the amyloid appeared. This scenario has now proven to be the case in AD. As the study of the disease continues, it is becoming apparent that it follows in considerable part the rules of systemic amyloidoses, while in other respects the β-amyloid process is distinct from these disorders.

Neurofibrillary tangles and dystrophic cortical neurites contain post-translationally modified forms of tau proteins

Biochemical analysis of the PHFs which accumulate as perikaryal NFTs and within dystrophic cortical neurites, both in neuritic plaques and outside of plaques in the cortical neuropil, began well before the characterization of the amyloid. A large number of immunochemical and biochemical experiments have led to the conclusion that the principal protein subunit of PHFs is an altered form of the microtubule-associated protein tau (Chap. 8). Tau normally copurifies with tubulin during repetitive cycles of assembly and disassembly of microtubules, and it is known to bind to tubulin and promote the assembly and stability of microtubules. The identification of tau as the major antigenic constituent of PHF was made using antibodies both to purified PHF and to tau (see, for example [3]). Subsequently, harsh methods were used to extract and sequence fragments of tau from purified PHF, many of which are highly insoluble in strong detergents and solvents. These studies demonstrated that tau, particularly portions from the carboxyl third of the molecule containing its microtubule-binding domains, were incorporated into PHF. A separate line of investigation came to a similar conclusion, that the tau protein is the major or sole component of PHF. Wolozin and Davies [4] raised monoclonal antibodies to particulate fractions of AD brain and identified a particular antibody, designated Alz 50, that bound to NFTs, to a large number of dystrophic cortical neurites and to some abnormal neuronal cell bodies that did not contain NFTs. In extracts of AD cortex, Alz 50 recognized a group of proteins having electrophoretic mobility slightly slower than normal tau; these were designated A68 because their relative migration clustered around 68 kDa [4]. The subsequent use of Alz 50 to probe protein fractions from normal brain demonstrated that the antibody recognized normal tau proteins, both in their phosphorylated state (see Chap. 24) and following their in vitro dephosphorylation [5]. These results suggested that A68 proteins represent an altered phosphorylation state of tau that causes its slower electrophoretic mobility on gels. This conclusion was supported by the development of a method to purify a subset of PHFs that were soluble in ionic detergents and the demonstration that such PHFs were solely composed of the A68 forms of tau [6]. Analyses of purified A68, also designated PHF-tau, from AD cerebral cortex have demonstrated that the altered tau molecules contain additional phosphate groups beyond those normally present on tau. The mechanism by which the hyperphosphorylated tau arises may include the increased activities of several kinases which can putatively phosphorylate tau as well as the decreased activities of certain cellular phosphatases (see Chap. 24).

There has been considerable speculation about polypeptides besides tau that may contribute to the PHF structure. Presently, the most clear and reproducible data have suggested that tau is the principal, if not the sole, structural constituent of the filaments. However, like a variety of neuronal inclusions in diseases other than AD, such as Lewy bodies in dopaminergic neurons in Parkinson's disease, NFTs contain ubiquitin. In addition, glycosaminoglycans (GAGs) have been detected in NFTs and in adjacent neurons not bearing tangles per se, and these have been shown to be capable of enhancing the polymerization of tau into PHF structures in vitro [7]. On this basis, it has been postulated that GAGs or proteoglycans bearing them may serve as a critical nidus for the initiation of PHF formation in neurons.

Amyloid in Alzheimer's disease plaques is composed of a 40- to 42-amino-acid portion of an integral membrane glycoprotein, the β-amyloid precursor protein

The initial sequence of the Aβ protein, obtained from the meningovascular amyloid deposits by Glenner and Wong, [2] extended to 24 residues and was not homologous with previously described proteins. Shortly after this observation, the partial purification of the amyloid plaque cores and their solubilization in high concentrations of formic acid or guanidine thiocyanate demonstrated that their subunit protein had an amino acid composition indistinguishable from meningovascular Aβ, although there appeared to be considerable heterogeneity of the amino terminus, including blocked species [8,9]. The sequence of the vascular amyloid was later extended to 40 residues, whereas that of the plaque amyloid was found to be heterogeneous but principally 40 or 42 residues in length [10].

Because of the difficulty of purifying the highly self-aggregating amyloid protein from postmortem cerebral tissue, only a limited number of biochemical analyses have been published. It is likely that even more heterogeneity than is currently recognized, that is, the presence of various truncated Aβ fragments, occurs in many amyloid deposits. Nonetheless, the major species identified to date appear to be Aβ_1–40 and Aβ_1–42. There are reproducible biochemical differences between meningovascular and plaque core amyloid; for example, the former, but not the latter, is soluble in 6 M guanidine hydrochloride, and the former is composed in very large part of Aβ peptides ending at residue 40, whereas the latter contains both Aβ₄₀ and Aβ₄₂ species.

The establishment of the amino-terminal sequence of the Aβ protein led to the cloning of its precursor polypeptide by four laboratories independently in 1987 [10,11]. The first full-length cDNA which was isolated encoded βAPP, a 695-residue protein that contained a single domain with a hydrophobic, putative transmembrane sequence near its carboxy terminus (Fig. 46-2). Subsequent studies of βAPP itself in human tissues and cultured cells demonstrated that it comprised a heterogeneous group of polypeptides ranging from ~105 to 140 kDa [12] and that the protein underwent N- and O-glycosylation as well as tyrosine sulfation during its post-translational maturation in the secretory pathway [13].

Figure 46-2

Schematic diagram of the primary structure of the β-amyloid precursor protein (βAPP). The molecule depicted here is the largest of the known alternate transcripts, comprising 770 amino acids. Several regions of interest are indicated at (more...)

The cloning of βAPP led immediately to the localization of its gene to the long arm of chromosome 21. This finding offered an explanation for the long-standing neuropathological observation that patients with trisomy 21 develop amyloid-bearing plaques and other lesions of AD. Subsequent cloning of βAPP cDNAs from other mammals demonstrated a high degree of evolutionary conservation of this gene. Indeed, the 695-amino-acid isoform, βAPP₆₉₅, is 100% conserved between the cynomologus monkey and human, while rat and mouse show more than 95% conservation. Gene products with considerable homology to βAPP occur in Drosophila and Caenorhabditis elegans; these molecules do not, however, contain the Aβ sequence.

cDNA and genomic cloning demonstrated that βAPP polypeptides arise by alternative exon splicing. The initially cloned isoform, βAPP₆₉₅, is almost exclusively expressed in brain, primarily in neurons. Alternate transcripts of 751 and 770 amino acids that have an inserted exon encoding a Kunitz-type serine protease inhibitor (KPI) motif are the major expressed isoforms in virtually all peripheral cells and are expressed in brain cells. Additional alternative transcripts of low abundance have been identified. Examination of the exon/intron structure of the βAPP gene reveals that the 40- to 42-amino-acid Aβ fragments contain portions of two adjacent exons and, thus, must arise by proteolytic processing rather than by alternative splicing.

Deposition of amyloid β protein precedes the lesions of Alzheimer's disease and arises from alternative proteolytic processing of β-amyloid precursor protein

The fact that some subjects with Down's syndrome dying in their teens show amorphous deposits of Aβ, termed diffuse plaques, in the absence of any other cytological lesions of AD has supported the concept that Aβ deposition can occur prior to AD-type neuronal or glial alteration. It is not possible to establish a precise temporal sequence of pathological changes directly in AD because the brain can be examined only at the end of the disease. However, the presence of large numbers of diffuse plaques, outnumbering compacted, neurite-containing plaques, in AD brains and the fact that the amyloid lesions of Down's syndrome are indistinguishable from those of AD support the notion that diffuse plaques represent the earliest discernible morphological change also in AD. As will be discussed shortly, strong support for this hypothesis has come from the identification of point mutations in the βAPP gene, which segregate with the AD phenotype in certain autosomal dominant pedigrees.

Normal cellular processing of βAPP has been shown to include a pathway that involves maturation of the protein in the Golgi apparatus (Chap. 2), trafficking to the plasma membrane and cleavage at residue 16 within the Aβ domain (residue 687 of βAPP₇₇₀), liberating the large amino-terminal hydrophilic portion of the precursor into the medium [13] and retaining an ~10-kDa membrane-associated carboxy-terminal fragment [12] inside the cell (Fig. 46-2). This so-called α-secretory processing, which appears to occur primarily at or near the cell surface, precludes the formation of intact Aβ. There is evidence that α-secretory processing also occurs in intracellular organelles [14]. βAPP contains an Asn-Pro-Thr-Tyr motif in its cytoplasmic tail that resembles a consensus sequence for the internalization of cell-surface receptors via clathrin-coated vesicles (see Chap. 9). Based on this knowledge, experiments have demonstrated an alternative processing route which involves the reinternalization of holo-βAPP from the cell surface and its trafficking to endosomes and lysosomes. When late endosomes/lysosomes are purified from cultured cells, an array of carboxy-terminal fragments of βAPP can be detected. Indeed, such Aβ-containing fragments have been directly identified in human tissues.

The high degree of insolubility of Aβ isolated from senile plaque and meningovascular deposits and the fact that Aβ is derived from an integral membrane sequence led to the widely held assumption that Aβ must arise from aberrant proteolysis of βAPP following membrane injury. It was therefore surprising to discover that small amounts of Aβ are released continuously from a variety of cultured cells under normal metabolic conditions and that this Aβ in the media is entirely soluble [15–17] (Fig. 46-2). Moreover, soluble Aβ has also been detected in normal and AD cerebrospinal fluid and plasma [16–18]. These data indicate that the Aβ peptide is a normal metabolic product of βAPP throughout life. Furthermore, the findings suggest that cultured human cells can be used as simple in vitro screening systems to identify compounds which decrease Aβ production. Drugs that inhibit the still unidentified protease(s) that creates the C terminus of Aβ, termed γ-secretase, are now being prepared for therapeutic trials in patients. Another outcome of the discovery of soluble Aβ in spinal fluid and plasma has been the development of sensitive ELISA assays, which reveal that many, but not all, AD patients have a lower CSF concentration of the soluble Aβ₄₂ peptide than do normal elderly subjects. To what extent this finding and analogous assays in plasma will be useful in establishing risk for the disease in elderly humans and monitoring the efficacy of anti-amyloid drug therapy in affected patients will soon become clear.

The normal generation of Aβ by cultured cells has enabled studies of the mechanisms of Aβ production. Considerable evidence has emerged that early endosomes, which internalize and recycle βAPP to the cell surface as if it were a receptor, are a principal site for the formation of the 40-residue form of Aβ (Table 46-1). It is currently not clear whether and to what extent the more amyloidogenic 42-residue form is made during endocytic recycling. However, there is mounting evidence that Aβ₄₂ can be generated early during the secretory trafficking of βAPP, namely, in the endoplasmic reticulum and Golgi [19]. Therefore, the proteases referred to as β- and γ-secretases are apparently distributed to several subcellular compartments. The 99-residue C-terminal fragment of βAPP, which is the product of β-secretase action, may be cleaved by distinct γ-secretases at either residue 40 or residue 42 of the Aβ region. Ongoing work should reveal more details about precisely where in the cell the Aβ₄₀ and Aβ₄₂ peptides are made and the identity of the responsible enzymes.

Table 46-1

Data Supporting Early Endosomes as the Principal Site of Aβ Generation.

A rare form of autosomal dominant Alzheimer's disease is caused by point mutations in the gene that encodes β-amyloid precursor protein

The genetic linkage to AD of DNA markers on the long arm of chromosome 21 in some autosomal dominant pedigrees raised the likelihood that βAPP itself could be a disease-causing gene. In 1991, two families were identified in which affected members had a point mutation at codon 717 of βAPP₇₇₀ (Fig. 46-3) [20]. This observation provided the first specific molecular cause of AD and suggested that alterations in βAPP could initiate β-amyloidosis in the absence of any pre-existing pathological events. Even prior to this discovery, a mutation at codon 693 of βAPP₇₇₀ had been found in individuals with the Aβ deposition disease hereditary cerebral hemorrhage with amyloidosis of the Dutch type (HCHWA-D) [21]. This rare disorder is marked by severe Aβ deposition in meningeal and cerebral vessels plus large numbers of diffuse plaques. Very few or no mature neuritic plaques or NFTs are observed, and no Alzheimer-like dementia has been described in these patients. It appears that HCHWA-D is closely related to AD, both genotypically and phenotypically, although the clinical outcome is multiple cerebral hemorrhages rather than dementia.

Figure 46-3

The sequence of β-amyloid precursor protein (βAPP) containing the amyloid β protein (Aβ) and transmembrane region is expanded and shown by the single-letter amino acid code. The underlined residues represent the Aβ (more...)

Since these discoveries, a few other point mutations in the βAPP gene have been found in affected members of families with autosomal dominant, or familial, AD (see Fig. 46-3). In addition to the original familial AD mutation at codon APP₇₁₇, Val → Ile, two other missense mutations at the same codon have been discovered: Val → Gly and Val → Phe. Also, a missense mutation changing codon 692 from Ala to Gly has been discovered in a Flemish family having both congophilic angiopathy with hemorrhage and progressive dementia as the clinical phenotypes. A double missense mutation immediately preceding the Aβ region has been found in an extended Swedish kindred with familial AD: APP₆₇₀ Lys → Asn plus APP₆₇₁ Met → Leu. Importantly, all of the βAPP mutations discovered to date in AD pedigrees cluster in the Aβ region of the precursor. Indeed, the fact that they are at or near the α-, β- or γ-secretase-cleavage sites strongly supports the hypothesis that they alter the proteolytic processing of βAPP in ways that enhance the production of Aβ, especially its highly amyloidogenic 42-residue form. Such enhancement has been demonstrated in both cultured cells and transgenic mouse models [22,23] (Table 46-2).

Table 46-2

β-Amyloid Precursor Protein Mutations: Genotype-to-Phenotype Conversions.

Mutations in the presenilin 1 and 2 genes represent the most common cause of early-onset, autosomal dominant Alzheimer's disease

Shortly after the discovery of the βAPP mutations, a sizable portion of early-onset AD families was linked to an unknown gene on chromosome 14. Further linkage analyses and positional cloning led to the identification of the responsible gene product as a polytopic membrane protein of 467 amino acids currently designated presenilin 1 (PS1) [24]. A large number of missense mutations (close to 45 at this writing) and one exon deletion in PS1 have been found in families of diverse ethnic origins. Immediately after the cloning of PS1, a highly homologous gene called presenilin 2 (PS2) was identified on chromosome 1, and just two missense mutations have been discovered to date in these familial AD kindreds [25]. The presenilins are in turn homologous to the sel-12 gene in C. elegans, a facilitator of the function of the lin-12/Notch proteins that are involved in cell-cell recognition and specification of cell fate during embryogenesis. These discoveries have led to a rapidly growing number of studies of the functional activities of the presenilins in worms and mammals [26], as well as attempts to understand whether these functions are affected by the mutations in PS1 and PS2 which cause early-onset AD.

It is clear that PS1 expression is required for proper mammalian embryogenesis and survival, but it appears that mutant human PS1 can convey these vital developmental functions as well. Rather than conferring a fundamental loss of function of presenilin, the PS1 and PS2 mutations linked to familial AD seem to lead to a gain of toxic function: the disregulation of γ-secretase(s) in a way that selectively enhances the proteolysis of βAPP at Aβ₄₂. A roughly twofold increase in relative Aβ₄₂ production has been documented in transfected cells and transgenic mouse brains harboring PS1 or PS2 mutations [27,28], and this elevation has been directly demonstrated in the plasma, skin fibroblast media and brains of humans with such mutations [18,29]. How presenilin mutations increase Aβ₄₂ generation is not yet known, but the effect may involve the formation of complexes between presenilin and βAPP, and perhaps additional polypeptides, and may occur in the endoplasmic reticulum and/or Golgi apparatus.

The ϵ4 allele of apolipoprotein E is a major genetic risk factor for late-onset Alzheimer's disease

Biochemical studies searching for CSF proteins capable of binding to Aβ identified apolipoprotein E as one such protein [30]. Subsequent genetic analyses showed that the naturally occurring ϵ4 polymorphism of the ApoE gene was substantially over-represented in sporadic AD subjects compared to age-matched controls and, thus, appeared to represent a major risk factor for the development of the disease [30]. It has since been confirmed that inheritance of one or two ApoE ϵ4 alleles significantly increases the likelihood of developing late-onset sporadic AD and decreases its age of onset. Conversely, inheritance of the infrequent ApoE ϵ2 allele appears to confer a decreased risk of developing the disorder compared to that seen in humans harboring the common ϵ3 allele.

The ApoE4 protein, which lacks cysteines and therefore cannot undergo intramolecular or intermolecular disulfide crosslinking, increases the likelihood of AD whereas the ApoE3 and ApoE2 proteins, which do contain cysteines, do not; however, the molecular mechanism remains unclear. A major clue to the mechanism has come from the observation, now confirmed in numerous laboratories, that AD subjects with two ϵ4 alleles have a significantly higher number and density of Aβ deposits in the brain than subjects with no ϵ4 alleles, while subjects with one ϵ4 allele generally fall in between [31]. In vitro biochemical studies have suggested that the ApoE4 protein may be less effective in retarding the aggregation of Aβ into amyloid fibrils than ApoE2 or ApoE3. Alternative hypotheses for the effect of ApoE4 in AD have been proposed. These are based on evidence that (i) ApoE4 does not support neurite outgrowth in vitro and is less salutary for normal neuronal structure and function than ApoE3 and (ii) ApoE4 may permit tau to become dissociated from microtubules and participate in enhanced PHF formation. However, the latter hypothesis is inconsistent with the observation that amyloid plaque density, not NFT density, correlates with ApoE4 gene dosage in AD patients [31].

Genotype-to-phenotype relationships implicate β-amyloidosis as an early and necessary factor in all known forms of familial Alzheimer's disease

To summarize at this juncture, four genes that are unequivocally associated with the development of AD have been identified to date, and linkage analyses of other families make it clear that additional genes can be responsible (Table 46-3). Three of the known genes, APP, PS1 and PS2, can be said to be causative of AD in the respective families in which mutations in these genes occur. In each of these three cases, there is now compelling evidence that the pathogenic mechanism involves altered APP catabolism to generate increased amounts of Aβ peptides, particularly the highly aggregation-prone, 42-residue form (Table 46-3). In the case of the ApoE gene on chromosome 19, its ϵ4 allele is a major genetic risk factor for the development of AD, perhaps contributing to the development of the disorder in some 30 to 40% of all patients. However, ApoE4 is not causative per se because some patients with one or two ApoE4 alleles show no signs of the clinical disease even late in life and, conversely, more than half of all AD patients do not bear an ϵ4 allele.

Table 46-3

Genetic Factors Predisposing to Alzheimer's Disease: Relationships to the β-Amyloid (Aβ) Phenotype.

Our discussion thus far has emphasized the possible role of APP metabolism and the gradual accumulation of insoluble Aβ deposits in the pathogenesis of the disease. However, many other biochemical and structural abnormalities have also been observed in the brains of AD patients. Although it is impossible to determine a precise temporal sequence of progression, the outlines of a pathogenic cascade are emerging (Fig. 46-4). Insights into the temporal course of the disorder in its preclinical phase derive primarily from three sources: (i) the study of the accrual of AD-type brain changes in patients with trisomy 21 who have died of other causes at various ages from early childhood to late adulthood; (ii) similar analyses of the development of AD-type lesions during the normal aging process in humans, other primates, dogs and cats; and (iii) studies of transgenic mice which overexpress mutant forms of human APP that cause early-onset AD in humans [22,23].

Figure 46-4

A hypothetical sequence of the molecular pathogenesis of familial forms of Alzheimer's disease.

Analyses of Down's syndrome (DS) brains have provided perhaps the most relevant information about how AD may progress. Numerous investigators have reported that the earliest AD-like morphological change found in very young DS brains, 12 to 15 years old, is the accrual of amorphous, largely nonfibrillar forms of Aβ deposits, the diffuse plaques. Some or many such deposits are found in limbic and association cortices and often in striatum, cerebellum and elsewhere in trisomic individuals dying after age 12 years or so. Importantly, such diffuse plaques also accumulate in the brains of nontrisomic people with normal cognition who have died of other causes after the age of 60 years. Diffuse plaques are also abundantly present in typical AD brains at the end of the disease.

Light and electron microscopic studies of diffuse plaques in AD and in DS demonstrate very little or no structural alteration of axons, dendrites, astrocytes and microglia within and immediately surrounding these amorphous Aβ deposits. This lack of cytopathology appears to correlate with a relative dearth of fibrillar amyloid in the diffuse deposits. When the brains of DS patients of increasing age are examined, fibrillar plaques with surrounding neuritic and glial dystrophy are detected increasingly after approximately age 30. At about the same time, NFTs also begin to appear. Although such temporal correlation is imprecise in the relatively limited number of DS patients reported to date, a consensus has emerged that diffuse Aβ plaques clearly precede the other AD-type changes that occur in DS. The early accumulation of diffuse plaques is assumed to be caused by the elevated APP gene dosage and the consequent increase in APP expression and Aβ concentrations documented in these patients.

APP transgenic mice experience high brain expression of APP from birth and are thus analogous in some respects to patients with DS. However, the mice reported to date have the additional influence of familial AD-linked missense mutation flanking the Aβ region of APP [22,23]. Although such animals have high neuronal expression of the APP transgene as well as high amounts of soluble Aβ within their brains from birth, they develop diffuse and compacted Aβ plaques resembling those of AD beginning at around 5 to 8 months (mice normally live about 2 to 2.5 years). During the next several months, the transgenic mice show increasing numbers of Aβ deposits, many of which are now Congo red-positive, suggesting that they contain fibrillar amyloid; and electron microscopy clearly reveals filamentous amyloid cores. Moreover, after Aβ plaques develop, the mice show morphologically and immunocytochemically abnormal neurites intimately associated with the amyloid plaques [22,23]. Cytoskeletal proteins such as the microtubule-associated protein 2 (MAP2), the neurofilament protein and even the tau protein can show abnormal immunoreactivity in these dystrophic neurites and in some nearby neuronal cell bodies [22,23], although the formation of human-type NFT has not been reported to date. A brisk reactive astrocytosis occurs within and around the Aβ plaques, and activated microglial cells occur near the centers of many of the plaques. Confocal microscopy of the mouse plaques and immunostaining for synaptic proteins indicate that degeneration and loss of synapses is occurring, particularly in the vicinity of the plaques [22]. The extent of behavioral impairment is not yet clear [22,23].

Although the rather rapid acquisition of AD-like lesions in transgenic mice, resulting from high expression of APP from birth, cannot be considered an ideal model of human AD, these mice clearly provide a highly useful and manipulatable representation of the AD process.

Assuming that studies of disease progression in DS and in transgenic mice are relevant to the mechanism of AD, one may postulate that the gradual accrual of Aβ peptides in the form of first diffuse and then fibrillar plaques may result in local cellular effects that include activation of microglial cells, reactive astrocytosis and alterations of nearby axons and dendrites. A hypothetical sequence of the pathogenic cascade leading to clinical cascade is shown in Figure 46-4. The extent to which these cytotoxic events derive from properties of the aggregated Aβ protein itself or from the numerous β-amyloid-associated proteins that have been detected in plaques is yet unclear.

These associated polypeptides, some of which have been referred to as “pathological chaperones” because of their putative role in enhancing the aggregation, deposition and toxicity of Aβ, include the normally secreted proteins α1-antichymotrypsin, ApoE, serum amyloid P component, heparan sulfate proteoglycans and various components of the classical complement pathway. Activated microglia, which become associated with maturing plaques, are capable of releasing a number of well-characterized cytokines that can, in turn, stimulate local astrocytes to release yet other proteins, including α1-antichymotrypsin and ApoE. The serum amyloid P protein, which is associated with all forms of central and peripheral amyloid deposits, is not expressed in the brain and thus must come to the plaque via passage across the blood-brain barrier. To what extent other circulating molecules, including Aβ itself, breach the blood-brain barrier to contribute to the pathological changes is unclear.

It can be concluded that many proteins potentially capable of exerting biological activity on surrounding neurons and glia accumulate within the amyloid plaque. We thus face an embarrassment of riches in terms of potential effectors of AD cytopathology. At exactly which point axons and dendrites in the vicinity, as well as their cell bodies of origin, undergo an activation of kinases, deactivation of phosphatases or both, resulting in the hyperphosphorylation of tau proteins underlying tangle formation, is difficult to say [32]. In all probability, the multiple molecular and cellular alterations found in the AD cortex develop at varying rates but in reasonable proximity to each other.

Biochemical and morphological changes can presumably also occur in cortical and subcortical neurons and their processes, which are not intimately associated with amyloid deposits. Subcortical neurons in regions such as the cholinergic nucleus basalis of Meynert, the noradrenergic locus ceruleus and the serotonergic median raphe nuclei, whose axons all project into plaque-rich cortical areas, often show shrinkage, NFT formation and cell loss. The complex array of plaque-associated and non-plaque-associated cytopathology observed by the end stage of AD may ultimately be very difficult to order into a precise sequence of temporal evolution.

The multiple neurotransmitter alterations in AD brain that were uncovered beginning in the late 1970s are now known to include several monoaminergic and neuropeptide deficiencies beyond the loss of cholinergic function that was first described. In the context of the complex, multicellular pathogenic cascade discussed above, it comes as no surprise that AD does not affect a single neurotransmitter system. Indeed, morphological studies have demonstrated that any one amyloid-bearing neuritic plaque may contain altered neurites derived from neurons of multiple neurotransmitter specificities. These considerations provide one explanation for the general lack of robust symptomatic improvement in patients given cholinergic replacement therapy, such as acetylcholinesterase inhibitors.

The fact that NFTs arise in a variety of etiologically unrelated diseases in the absence of Aβ deposits suggests that they represent a response of neurons to a range of insults and are not specific for the amyloidotic process. The same may also be true of the tau-positive neuropil threads in the cortex, which can also occur in certain other degenerative diseases bearing tangles but lacking amyloid. However, the neuritic plaque, with its severe microglial and astrocytic cytopathology, is more specific for AD and DS. The abundant Aβ deposits that can be found in some cognitively normal elderly subjects are predominantly of the diffuse type, lacking neuritic and glial alteration. This distinction may explain why they are not associated with clinical dysfunction.

Therapeutic strategies arise from understanding the molecular basis of Alzheimer's disease

The development of therapies expected to slow or arrest the progression of a disease requires as true and detailed an understanding of the molecular pathogenesis as possible. The tools for screening a wide array of compounds for possible efficacy in AD include cell culture systems and the transgenic mouse models. These are used to test an array of potential therapeutic targets, which are discussed in the following section.

The inhibition of Aβ secretion from neuronal and non-neuronal cells has been actively pursued. One way this can be accomplished is by designing specific inhibitors of the β- and γ-secretases after these proteases are identified and cloned. However, there may be other ways of lowering Aβ production that do not involve direct inhibition of these enzymes. In any event, the testing of potentially therapeutic chemicals on cells that continuously secrete Aβ has already led to the identification of compounds which lower the production of the peptides. One question regarding this general approach relates to whether chronic treatment with cholinergic agonists will result in increased processing of APP molecules by the α-secretase pathway and, thus, in diminished production of Aβ. Such a possibility can initially be examined in transgenic mice, in which both cerebral and CSF Aβ concentrations can be measured, but it must then be confirmed in treated patients by measuring CSF Aβ concentrations. A number of other first messengers that affect intracellular regulation (see Chaps. 20–22) could turn out to be at least as effective as cholinergic agonists in shifting APP processing from the β- to the α-secretase pathway.

A therapeutic approach that seems particularly attractive is to attempt to slow the aggregation of the secreted Aβ peptide into its fibrillar, putatively cytotoxic form. In vitro studies indicate that certain small molecules, including the amyloid-binding dye Congo red, can retard the aggregation of synthetic Aβ peptides into high-molecular-weight aggregates. Compounds which interfere with Aβ assembly into amyloid fibrils could act in the extracellular space of the brain and, thus, avoid interference with the metabolism of APP and other molecules inside cells. Full-length APP and its various metabolites, including APP_s and perhaps even Aβ, have normal functions, whereas the aggregated forms of Aβ that plaques are composed of are believed to represent solely pathological moieties. Thus, interfering with Aβ aggregation, if it can be done in a selective fashion, could avoid effects on the metabolism of APP and other molecules. The transgenic mouse models, which have fibrillar amyloid plaques, should be a reasonable system in which to evaluate the efficacy and safety of such compounds.

Yet another therapeutic approach based on the growing understanding of presymptomatic events in AD is the use of anti-inflammatory drugs that could interfere in part with the microglial activation, cytokine release and acute phase responses that occur in maturing amyloid plaques. Epidemiological evidence suggests that individuals who have been on nonsteroidal anti-inflammatory drugs may have a lower likelihood of developing the pathological and clinical features of AD. One may assume that the inflammatory process which appears around amyloid plaques is sufficiently distinct from peripheral forms of inflammation that it will require specialized anti-inflammatory compounds, which could again be identified and characterized in transgenic mice.

The devastating impairment of higher cortical functions which characterizes AD must ultimately be attributed to profound neuronal dysfunction and degeneration. Therefore, a variety of neuroprotective strategies can be envisioned in this disorder. One relatively specific approach would be to attempt to design compounds that interfere with any altered signal-transduction pathways that are proven to mediate the effects of extracellular amyloid filaments and their closely associated molecules on neuronal homeostasis. However, no single cell-surface receptor for Aβ in its monomeric or aggregated form has yet been found to fully mediate the toxicity. Agents that could hypothetically interfere with Aβ toxicity would include compounds that coat the amyloid aggregates in a way that makes them “invisible” to the cell and molecules which inhibit a downstream effector pathway inside the neuron.

In addition to such approaches directed at the specific neurotoxic cascade putatively induced by amyloid, general therapies could be applied that might be equally applicable to AD and to other neurodegenerative disorders. Such strategies might include the use of inhibitors of excitotoxicity, agents that block calcium entry into cells, free radical scavengers and antioxidant treatments (see Chap. 34). Evidence is mounting from in vitro studies that aggregated Aβ induces multiple features of oxidative injury in cultured neurons.

Another general approach to retarding neurodegeneration that might also be applicable to AD would be neurotrophic therapy. However, technical hurdles must be overcome to chronically deliver neurotrophic peptides to the appropriate sites in the brains of elderly subjects. In addition, more must be learned about all of the effects of trophic factors on APP expression and on the complex effects of upregulation of trophic influences.

The combined power of genetics, molecular biology and biochemistry has produced remarkable advances in our understanding of the causes and mechanisms of AD, and all of these sciences are required for therapeutic approaches.