In the past several years, a number of genetic epidemiology studies have been undertaken on probands with AD and their families.1–7 Cumulatively, these studies (see review in reference6) strongly argue that the familial aggregation of AD is not due simply to the high frequency of AD in the general population. These studies suggest that the age-dependent risk and the overall lifetime risk for AD in first-degree relatives of AD probands varies from 10 to 50 percent. The most comprehensive recent study suggests an age-dependent risk curve asymptotic to a final risk of 38 percent by age 85 years.7 The latter study, as well as several other earlier epidemiologic studies, make it difficult to assign a pure mendelian mode of transmission in the majority of AD cases. Instead, these studies imply that the majority of cases of familially clustered AD probably reflect a complex mode of transmission such as one or more common independent, but incompletely-penetrant, single autosomal gene defects; a multigenic trait; or a mode of transmission in which genetic and environmental factors interact. Nevertheless, there is a small proportion of AD cases (≈10 percent) which appear to be transmitted as a pure autosomal dominant mendelian trait with age-dependent but high penetrance. Analysis of these pedigrees with molecular genetic tools has led to the discovery of at least four different genetic loci associated with inherited susceptibility to AD (MIM 104300, 104760, 104310, 104311, 600759, 602096).
The Amyloid Precursor Protein
The first gene to be identified in association with inherited susceptibility to AD was the amyloid precursor protein gene (βAPP). The βAPP gene on chromosome 21 encodes an alternatively spliced transcript, which in its longest isoform encodes a single transmembrane spanning polypeptide of 770 amino acids (GenBank Y00264).8–11 Alternative splicing of exon 7 (which encodes a Kunitz serine protease inhibitor protein domain), exon 8 (which encodes a sequence homologous to the ox-2 antigen), and exon 13a result in polypeptides of 695 amino acids (which are expressed predominantly in brain), and 751 and 770 amino acids (are the predominant isoforms in nonneuronal tissues), and several other minor transcripts.12 There have been considerable advances in the understanding of the processing of the βAPP protein; they are reviewed in detail elsewhere.13 The βAPP precursor protein undergoes a series of endoproteolytic cleavages (Fig. 234-4). One of these results from the putative membrane-associated α-secretase, which cleaves βAPP within the Aβ peptide domain between residues Lys687 and Leu688 (codon numbering of βAPP695), that is, between residues 16 and 17 of Aβ, and liberates the extracellular N-terminus of βAPP (which was previously identified at protease nexin II, a protease inhibitor possibly involved in coagulation). This pathway is “nonamyloidogenic” because the cleavage precludes the formation of Aβ peptide.
Topology of the β-amyloid precursor protein and its principal routes of endoproteolytic cleavage.
The other cleavage pathway, which occurs in part in the endosomal-lysosomal compartment, involves β- and γ-secretases, which give rise to a series of peptides that contain the 40 to 42 amino acid Aβ peptide (β-secretase cleaves between Met671 and Asp672; γ-secretase cleaves after Ile712, Thr714, or Val715 to generate Aβ40, Aβ42 or Aβ43). Aβ ending at residue 40 is the predominant isoform produced during normal metabolism of βAPP.14–19 Current evidence suggests that Aβ40 is predominantly produced in endosomal-lysosomal systems.20,21 Aβ peptides ending at residue 42 or 43 (Aβ42/43 or long-tailed Aβ), on the other hand, are thought to be more fibrillogenic and more neurotoxic.19 Substantial evidence suggests that these longer isoforms may be generated at intracellular sites such as the ER and cis-Golgi, which, in neurons at least, are distinct from the sites of Aβ40 synthesis,22–24 although lipid-rich raft domains (caveolae) have also been suggested as sites of Aβ generation.25 These secretases, and especially the specific γ-secretases giving rise to the more fibrillogenic and potentially neurotoxic long-tailed Aβ1–42, appear to play a central role in the pathogenesis of at least some of the genetic forms of AD associated with presenilin and βAPP mutations, and possibly also in nongenetic forms (see below and reviews in references13 and 26).
The function of βAPP is currently unknown. Knockout of the murine βAPP gene has not been illuminating because it leads only to subtle phenotypes including minor weight loss, decreased locomotion, abnormal forelimb motor activity, and minor nonspecific reactive gliosis in the cortex.27 In vitro studies in cultured cells suggest that secreted βAPP (βAPPs) can function as an autocrine factor stimulating cell proliferation, cell adhesion, and supporting NGF induced neurite outgrowth of PC12 cells.28,29 Other studies have implied a role for βAPP in signal transduction by association of βAPP with heterotrimeric GTP-binding proteins.30
Several lines of evidence led to the suspicion that the βAPP gene was the site of mutations associated with AD. First, patients with Down syndrome (Trisomy 21) almost invariably develop the neuropathologic attributes of AD by age 40 years (there is considerable variation in the age of onset of actual dementia, which may in part be modulated by the genotype at APOE; see below).10,31–34 Second, the gene encoding the full-length βAPP protein is located on chromosome 21.8 Third, genetic linkage studies had shown weak but suggestive evidence for linkage of a familial AD locus on chromosome 21 near the markers D21S1/D21S11, which map near the βAPP gene.35,36 Fourth, genetic linkage and mutational studies of the βAPP gene identified a Glu693Gln missense mutation of the βAPP gene (codon numbering of the βAPP770 isoform) in affected and at-risk members of families with hereditary cerebral hemorrhage with amyloidosis of the Dutch type (HCHWA-D).37
Subsequently, direct nucleotide sequencing led to the discovery of several different missense mutations in exons 16 and 17 of the βAPP gene in families with early-onset AD (Table 234-1). Some of these missense mutations are probably not pathogenic mutations either because they have also been detected in normal elderly relatives, or because they are not present in all affected members of these pedigrees. Nevertheless, the missense mutations at codon 670/671 (Swedish mutation),38 at codon 692 (Flemish mutation),39 at codon 716,40 and at codon 71741–45 seem quite clearly to be pathogenic. The mutations at codon 670/671 and at codon 692 are rare, having been seen only in single families. Mutations at codon 717 on the other hand have been seen in approximately 20 unrelated pedigrees from different ethnic origins. The majority of such mutations at codon 717 were seen in Anglo-Saxon, Italian, and Japanese subjects. The reason for this aggregation is unclear because within each ethnic group there is no common haplotype of genetic markers within or surrounding the βAPP gene.46 The absence of a conserved haplotype around the βAPP gene in βAPP717 carriers in the same ethnic group argues against a common founder for the βAPP717 mutation in each ethnic group as the explanation for apparent clustering in these ethnic groups.
Table 234-1: Missense Mutations in the APP Gene |Favorite Table|Download (.pdf) Table 234-1: Missense Mutations in the APP Gene
|Codon ||Mutation ||Phenotype |
|655 ||Gln → Asp ||Late onset AD; no segregation |
|670/671 ||Lys-Met → Asn-Leu ||FAD; increased Aβ production |
|673 ||Ala → Thr ||No disease phenotype |
|692 ||Ala → Gly ||FAD + cerebral haemorrhage; increased Aβ |
|693 ||Gln → Gly ||Late onset AD; no segregation |
| ||Gln → Gln ||HCHWA-D |
|713 ||Ala → Val ||Schizophrenia; no segregation |
| ||Ala → Thr ||AD; no segregation |
|716 ||Gln → Val ||FAD |
|717 ||Val → Ile ||FAD; increased long Aβ isoforms |
| ||Val → Phe ||FAD |
| ||Val → Gly ||FAD |
That all of the clearly pathogenic mutations cluster close to the β-secretase site after Met671 (K670N/M671L), α-secretase site after Lys687 (A692G and E639Q), or γ-secretase site after Thr714 (I716V and V717I) led to the hypothesis that these mutations might cause AD by influencing the processing of βAPP.38 Further investigation confirmed that mutations at codons 716 and 717 led to a selective increase in the production of Aβ peptides ending at residue 42/43.40,47–51 The K670N/M671L mutation, on the other hand, appears to augment the production of both Aβ40 and Aβ42/43.52 The A692G mutation has a more complicated effect on βAPP processing. This mutation causes impaired α-secretase cleavage, increased heterogeneity of secreted Aβ species, and increased hydrophobicity of the Aβ.50 The A692G mutation has clinical features that in some cases are similar to those of HCHWA-D, and that, in other cases, are more similar to AD but with somewhat subtle differences in the size of the amyloid cores.53 The Glu693Gln mutation causes an increased propensity for Aβ to form fibrils.54
The relative or absolute overproduction of Aβ peptide, and in particular of Aβ42, as an effect of βAPP mutations that leads to neurodegeneration is an attractive hypothesis. This hypothesis is supported by the observation that mutations in presenilin 1 and presenilin 2 (see below) also affect βAPP metabolism and lead to overproduction of β42. Moreover, there is substantial evidence that Aβ42 peptides have an enhanced propensity to fibril formation16,18,19 and that such a conformational change is necessary to change the inert (or even marginally neurotrophic) soluble Aβ into toxic Aβ.18,19 Multiple molecular mechanisms have been postulated to explain the neurotoxic effects of Aβ. These include induction of apoptosis by direct effects upon cell membranes, by indirect effects such as potentiation of neurotoxic effects of excitatory amino acids, oxidative stress, and increases in intracellular calcium and free radicals.55–57 However, although the preponderance of available evidence supports the notion that presenilin mutations cause AD neurodegeneration via a mechanism involving overproduction of Aβ peptides, at least one experimental result suggests that βAPP mutations could cause apoptosis by a mechanism that is independent of Aβ peptide production. Specifically, a series of experiments has shown that, for the βAPP717 mutations at least, the mutations are able to induce apoptosis by causing constitutive activation of the heterotrimeric GTP-binding protein, G0.58,59 Furthermore, although some βAPP and PS1/PS2 mutations are associated with overproduction of Aβ peptides, such overproduction of Aβ peptide (as opposed to increased deposition and/or decreased clearance) is not a universal feature of all cases of sporadic AD.60
Genetic linkage studies in pedigrees with predominantly late-onset familially clustered AD provided suggestive evidence (z = +2.5 at θ = 0.00) for the existence of a second AD susceptibility locus near the markers BCL3 and ATP1A3, which map to chromosome 19q12-q13 (GenBank M10065, K00396).61 While this initial localization was both crude and tentative, Strittmatter et al. isolated proteins from the CSF that were capable of binding the Aβ peptide.62 Microsequencing of these peptides revealed that one was apolipoprotein E (APOE).62 The gene for APOE maps on chromosome 19q13, very close to the markers showing evidence for linkage and/or association with late-onset AD. Simultaneously, APOE was found in the senile plaques of AD, inferring that APOE may be a pathologic chaperone.63 Together, these observations suggested that the APOE gene itself was the AD susceptibility locus.
The APOE gene in humans contains three common coding sequence polymorphisms. The most common coding sequence variant, ∈3, reflects the presence of a cysteine at codon 112 and arginine at codon 158, and is present in approximately 75 percent of Caucasians. A second coding sequence variant, ∈4, reflects substitution of arginine for cysteine at codon 112, and is present in approximately 15 percent of Caucasians. The third coding sequence variant, ∈2, contains cysteine at codons 112 and 158, and is present in approximately 10 percent of Caucasians. Analysis of these coding sequence polymorphisms in normal control populations and in patients with AD has consistently shown that (a) there is an increase in the frequency of the ∈4 allele in patients with AD (∈4 allele frequency in AD is approximately 40 percent),64 and (b) there is a smaller reduction in the frequency of the ∈2 allele (to about 2 percent in AD).65 More significantly, there is a dose-dependent relationship between the number of copies of ∈4 and the age-of-onset of AD, such that ∈4/∈4 subjects have an earlier onset (mean age of onset is less than 70 years) than heterozygous ∈4 subjects (mean onset of after 70 years).66 Subjects with an ∈2 allele, on the other hand, have a later onset (mean age of onset for ∈2/∈3 is greater than 90 years).65 The association between ∈4 and AD has been robustly confirmed in numerous studies and in several different ethnic groups (reviewed in reference 67), and currently it seems that APOE ∈4 accounts for ≈ 30 to 50 percent of the genetic risk for AD in population studies. The association is weaker with advanced age of onset, and the putative protective role of the ∈2 allele is less clear at younger ages of onset (where ∈2 may even be associated with a more aggressive course).68,69 Currently, the major exception to the association of APOE ∈4 with AD arises from studies in African-Americans and Hispanics, which studies have generated conflicting results.70–72 It is unclear whether these conflicting results reflect small sample sizes or whether there is a true lack of association between AD and APOE ∈4 in subsets of these populations.
Although the association between APOE ∈4 and AD is robust (with the possible exceptions noted above), it is not entirely specific. Observations of patients with head injury73,74 and spontaneous intracerebral hemorrhage,75 and of patients undergoing elective cardiac bypass surgery,76 suggest a poorer cognitive outcome for patients with the ∈4 allele. There is also evidence for synergistic effects of a history of head injury and APOE ∈4 on risk for AD,73 such that patients with APOE ∈4 and a head injury have a tenfold increase in risk for AD compared to a twofold increase with APOE ∈4 alone, and no increase for head injury alone. Reports of association between the ∈4 allele and Creutzfeldt-Jacob disease77 and multi-infarct dementia78 are conflicting, although the majority of studies appear not to detect an association between ∈4 and these diseases.79,80 However, there is a confirmed association between the ∈4 allele and the Lewy body variant of AD, which has an overlapping but subtly different clinical phenotype from classical AD (e.g., more frequent hallucinations and sensitivity to neuroleptics).81
The mechanism by which the ∈4 allele is associated with an earlier onset of AD and by which the ∈2 allele is associated with a later onset is unclear. As is discussed below, the genetic data (the association of AD exclusively with the ∈4 allele, a protective effect for the ∈2 allele, and dose-dependent relationship between the ∈4 copy number and age of onset of AD) provide a strong argument that the ∈4/∈2 polymorphism is an actual biologic effector in the APOE gene. Currently, however, the possibility that the true pathogenic mutations/polymorphisms in nearby sequences are simply in linkage disequilibrium with ∈4/∈2 cannot be entirely excluded (see below). Nevertheless, a large body of biochemical evidence has been accumulated to support various hypotheses on how these APOE coding sequence polymorphisms might promote/protect against AD. The most obvious hypothesis is that APOE ∈4/∈2 polymorphisms might influence the production, distribution, or clearance of the Aβ peptide. This hypothesis is supported by observations that the genotype at APOE accounts for some of the variation in age-of-onset in subjects carrying the βAPP Val717Ile mutation (but not the APP692 mutation), suggesting a direct biochemical interaction between APOE and βAPP or its metabolic products.82–85 Second, subjects with one or more APOE ∈4 alleles have a higher Aβ peptide plaque burden than do subjects with no ∈4 alleles.86 In vitro studies suggest that delipidated APOE ∈4 binds Aβ more avidly than APOE ∈3.62,87 Third, there is evidence that both APOE and Aβ may be cleared through the low-density lipoprotein-related (LRP) receptor and that APOE ∈4 and the Aβ peptide may compete for clearance through the LRP receptor.88 Finally, transgenic mice that have an intact endogenous APOE gene and that overexpress human βAPP with the Val717Phe mutant under the control of the PDGF β-subunit promoter, develop profuse deposits of extracellular Aβ by 9 months of age. In contrast, when the same transgene is expressed on an APOE−/− background, there is a dramatic reduction in extracellular Aβ deposition, thus supporting the hypothetical role for APOE in sequestering extracellular Aβ.89
While theories attempting to explain a role for APOE ∈4 mediated by alterations in brain Aβ peptide levels have received the greatest attention, there is also some biochemical evidence to suggest a relationship between APOE and neurofibrillary tangles and synaptic density. In vitro APOE isoform-specific binding experiments with Tau and MAP2 suggest that ∈3 binds to both Tau and MAP2 better than the ∈4 isoform.90–92 This has led to suggestions that ∈2 and ∈3 may protect and sequester microtubule associated proteins better than ∈4, thereby reducing the ability of Tau to bind to itself, become hyperphosphorylated, and form paired helical filaments. This is supported by the fact that neurofibrillary degeneration begins earlier in ∈4 carriers than in non-∈4 carriers.93 APOE immunoreactivity has been detected within the neuronal cytoplasm, and particularly at the base of proximal dendrites, suggesting that although APOE is not transcribed in neurons, it can be imported into neurons from extracellular sources.94 This would place APOE in the right location for interactions with microtubule associated proteins.
Finally, there is a good correlation between the degree of clinical dementia and the decrease in synaptic density in AD as measured by both MAP2 and synaptophysin immunoreactivity.95 It has been suggested that APOE may be involved in synaptic plasticity during regeneration and repair, and that the ∈4 allele is less efficient in this role. Thus, APOE knockout mice also show an age-dependent decrease in synaptic density and spontaneous Aβ peptide aggregation within astrocytic processes.96 Several types of neural tissue culture cells demonstrate decreased neurite outgrowth in the presence of APOE ∈4 in the media rather than APOE ∈3.97,98 Perhaps, then, the APOE isoforms might differentially affect synapse formation in response to injury, learning, and aging.
In addition to the ∈2/∈3/∈4 coding sequence polymorphisms, several polymorphisms have been discovered in the 5′ -promoter of the APOE gene. These sequence polymorphisms have attracted interest because of preliminary data from quantitation of APOE levels in the brains, CSF, or serum of patients with AD. Data from some, but not all, studies suggest that subjects with AD have a 1.5- to threefold higher levels of APOE than do subjects with the same genotype but without AD.99–103 Two different studies provide a potential explanation for this observation. Thus, the Th1/E47cs polymorphism104 and homozygosity for the −491 A allele105 are suspected to increase risk for AD independently of APOE ∈4, and to cause this increased risk for AD by altering the transcriptional activity of APOE (although direct proof of the latter in vivo remains to be obtained). A follow-up study by the original group provided supporting evidence for an effect of the Th1/E47cs T allele and a weaker effect for the −491 T allele.106 However, several other independent studies using large sample sizes (>88 AD patients) were unable to replicate these findings, although they did confirm that the −491 A/T polymorphism at least, is in linkage disequilibrium with the APOE ∈2/∈3/∈4 polymorphism.107,108 Another polymorphism in the intron 1 enhancer region also failed to show association with AD beyond what was anticipated through its linkage disequilibrium with the ∈2/∈3/∈4 polymorphism.109
After the discovery that βAPP missense mutations and the APOE polymorphisms did not account for all cases of FAD, several groups undertook a survey of the nonsex-linked chromosomes other than chromosomes 19 and 21. These studies identified a series of polymorphic markers located on chromosome 14q24.3 (D14S43, D14S71, D14S77 and D14S53) that showed robust evidence of linkage to a particularly aggressive early-onset (onset 25 to 65 years; mean 45 years) form of familial Alzheimer disease (z >23.0 at θ = 0.01).110–112 Subsequent genetic mapping studies narrowed the region containing this third Alzheimer susceptibility locus (AD3) to a region of approximately 10 cM between the marker D14S271 at the centromeric end and D14S53 at the telomeric end, a physical distance of approximately 7 Mb. The actual disease gene (presenilin 1) was then isolated using a positional cloning strategys,113 and a homologue (presenilin 2) was subsequently mapped to chromosome 1 (see below).
The chromosome 14 AD3 subtype gene, presenilin 1 (PS1), is highly conserved in evolution, being present in C. elegans 114 and D. melanogaster, 115 and appears to encode a polytopic integral membrane protein (GenBank L42110). Theoretical predictions based upon Kyte-Doolittle hydrophobicity analysis suggest that there are between 5 and 10 membrane-spanning domains, that the N-terminus is acidically charged, and that there is a large hydrophilic, acidically charged loop domain between the putative sixth and seventh transmembrane domains.113,116 Partial direct experimental support for a polytopic structure has been obtained from studies in transfected cells (see below).
The presenilin 1 gene is transcribed at low levels in many different cell types, both within the central nervous system and also in non-neurological tissues.113 In the central nervous system, PS1 transcripts can be detected by in situ hybridization in the neocortex (especially in cortical neurons in layers II and IV), neurons of the CA1 to CA3 fields of the hippocampus, granule cell neurons of the dentate gyrus, subiculum, cerebellar Purkinje and granule cells, and deep nuclei, as well as lesser amounts in the olfactory bulb, striatum, some brainstem nuclei, and thalamus. Despite intense signals on northern blots of the corpus callosum, there is very little in situ hybridization signal detectable in oligodendrocytes in white matter.117 However, following injury such as kanaic acid injection into cerebral white matter there is induction of significant PS1 mRNA expression in white matter.118
The genomic structure of the PS1 gene has been elucidated and some of the transcriptional regulatory elements have been defined.119 Like the βAPP gene, there is evidence for alternate splicing of the PS1 transcript. Thus, there is a variably present 4-amino acid VRQS insert which arises from use of an alternate splice donor site at the 3′ end of exon 4.119–121
Immunoblotting and immunohistochemical studies suggest that the PS1 protein is approximately 50 kDa in size and is predominantly located within intracellular membranes in the endoplasmic reticulum, perinuclear envelope, the Golgi apparatus, and some as yet uncharacterized intracytoplasmic vesicles.122,123 The tissue-specific expression patterns of PS1 protein largely reflect those of the mRNA.124 Studies of the topology of PS1 suggest that the N-terminus and the residues in the TM6–TM7 loop are both located in the cytoplasm (Fig. 234-5).122,125–127
Putative topology of the PS1 protein showing the location of several Alzheimer disease-related mutations.
The orientation of the C-terminus is not yet completely resolved.122,125–127 However, the predominance of opinion suggests that it is oriented to the cytoplasm and that the preceding hydrophobic residues either are membrane associated or represent two additional transmembrane domains (TM7 and TM8). Studies of the PS1 protein in brain tissue, as well as many other peripheral tissues, reveal that only very small amounts of the PS1 holoprotein exist within the cell at any given time.128,129 Instead, the holoprotein is actively catabolized, possibly by at least two different proteolytic mechanisms. One of these mechanisms appears to involve the proteasome.130 Another proteolytic mechanism involves a series of heterogeneous endoproteolytic cleavages near residue 290 within the TM6–TM7 cytoplasmic loop domain.128,129 This endoproteolytic cleavage generates a series of N- and C-terminal heterogeneous fragments of approximately 35 kDa and 18 to 20 kDa, respectively. Remarkably, the stoichiometry of the N- and C-terminal fragments is tightly maintained on a 1:1 ratio, and the absolute abundance of the N- and C-terminal fragments is also tightly regulated such that artificial overexpression of PS1 results in only a modest increase in N- and C-terminal fragments.128 This has led to the suggestion that the process of endoproteolytic cleavage involves a tightly regulated saturable process.131
A third proteolytic mechanism acting upon the presenilin holoproteins involves members of the caspase 3 family of proteases. Activation of apoptosis by a variety of means that culminate in activation of caspase 3 results in cleavage of PS1 near residue aspartate 345 and the equivalent aspartate residue in PS2 (Asp 329).132–134 It is currently unclear whether caspase mediated endoproteolytic cleavage of the presenilins is actively involved in the regulation of apoptotic signal pathways (preliminary data suggest an antiapoptotic effect for the C-terminal derivative of PS1 cleaved by caspase).135 Alternatively, the presenilins might simply represent innocent bystanders that are cleaved by the caspase enzyme once apoptosis is activated. Certainly, caspase-mediated cleavage of the presenilins is not required for the effects of the presenilins on amyloidogenesis and Notch signaling.134
It is currently unclear whether the holoprotein, the endoproteolytic fragments, or both have biologic functions. However, it is clear that the native form of the holoprotein and its endoproteolytic fragments are components of high molecular weight, multimeric protein complexes. Thus, the holoprotein appears to be a component of an ≈180-kDa complex that is predominantly resident within the rough endoplasmic reticulum (rER).131,136,137 Both the N-terminal fragment (NTF) and the C-terminal fragment (CTF) appear to associate with each other as heterodimeric components of a larger (≈250 kDa) multimeric protein complex that appears to be resident in the endoplasmic reticulum, Golgi apparatus, and some additional intracellular membranous domains whose identity has not been entirely clarified.25,131,136,137 The latter may include detergent insoluble glycosphingolipid-rich domains (which have also been called lipid-rich rafts). It has been suggested that inclusion of the presenilin proteins in these high molecular weight membrane-bound complexes plays a critical role in the processing of PS1. Specifically, it has been suggested that the incorporation of the holoprotein and its subsequent endoproteolytic cleavage and incorporation into a larger complex represents a rate-limiting step in the processing pathway.138 However, once incorporated into these high molecular weight complexes, the endoproteolytic fragments remain together with a stable 1:1 stoichiometry and with very long half-lives.131,136,137 Holoprotein monomers that fail to get incorporated in these complexes are rapidly degraded with a half-life of less than 1 h via a proteasome-dependent mechanism.130 It also appears likely that incorporation of the presenilin proteins and their derivatives into these complexes is necessary for the biologic activity of the presenilins.139
The identity of the other components of the presenilin complexes is currently under investigation. However, a combination of yeast-two-hybrid, coimmunoprecipitation, immunohistochemistry, and biochemical fractionation studies reveals that, in peripheral tissues and in brain, the presenilins associate with β-catenin, a member of the armadillo protein superfamily.137,140 In brain, the presenilins also associate with a novel armadillo protein termed neuronal plakophillin-related armadillo protein (NPRAP) or δ-catenin.137,140 The functional significance of the presenilin:armadillo interactions is not entirely clear because the armadillo proteins have diverse functions, ranging from a structural role in stabilization of intercellular junctions (including synapses) to intracellular transduction of receptor-mediated signals (e.g., Wnt and certain growth factors) to participation in apoptotic cell death pathways. There is evidence from some laboratories that presenilins may also directly interact with βAPP 141,142 and filamen.143 However, not all laboratories confirm the latter results.131 More recently, a novel, Type 1 transmembrane glycoprotein, nicastrin, has been identified as another functional component of the presenilin complexes.332 Nicastrin is necessary for some aspects of Notch/glp-1 signalling during embryonic development in C. elegans, and absence of nicastrin results in embryonic lethality. Mutations of a conserved hydrophilic domain in the N-terminus of nicastrin can significantly modulate γ-secretase cleavage of βAPP and the production of Aβ in vitro. Nicastrin can bind to the C-terminal stubs of βAPP, which are derived from α-secretase cleavage (C83-βAPP) or β-secretase cleavage (C99-βAPP), the immediate precursors of Aβ. Similarly, nicastrin can bind the furin- and kuzbanian-cleaved, membrane-tethered fragments of Notch, which are the substrates for the signal-dependent S3-cleavage that liberates the Notch intracellular domain during signal transduction. Currently, it seems likely that there is a bi-directional interaction between the presenilins and nicastrin which serves to regulate the subsequent γ-secretase or S3-cleavage events.
The normative function of presenilin 1 has not yet been defined. Functional analogies have been made to the weakly homologous SPE4 protein of C. elegans, which is involved in maintenance of a Golgi-derived membranous organelle thought important in the partitioning of protein and cell membrane products in the maturing spermatocyte of C. elegans.144 This has led to speculation that the PS1 protein might serve a similar role in protein and membrane trafficking.113 This hypothesis is supported by two different lines of direct experimental evidence. First, ablation of functional PS1 expression by homozygous targeted disruption of the murine PS1 gene causes aberrant processing of βAPP with the failure of γ-secretase cleavage of the C-terminal stubs of βAPP derived from α-secretase or β-secretase cleavage.145 Failure of γ-secretase cleavage results in the accumulation of uncleaved α-secretase or β-secretase stubs in a variety of intracellular loci including the endoplasmic reticulum, Golgi, and lysosomes.145,146 A similar effect has been achieved by (artificially) mutating either of two conserved aspartate residues (D258 and D387) in putative TM domains of PS1 and PS2. Second, as described earlier, the presenilins form multimeric protein complexes with the armadillo protein β-catenin,137,140 and missense mutations in PS1 and PS2 cause mistrafficking of this presenilin ligand.147 These results suggest that PS1 might be directly involved in trafficking of βAPP, βAPP derivatives or components of the putative γ-secretase enzyme. Alternatively, PS1 might serve as an adaptor for, or an activator of γ-secretase. Finally, the loss-of-function effect caused by the artificial D258E or D387E mutations suggests that these residues may be part of the active site of a novel aspartyl protease and that PS1 is itself γ-secretase. This result, however, must be tempered with the knowledge that mutation of these aspartate residues also has a structural effect on the presenilin complexes.333 Nevertheless, functional assays have shown that γ-secretase proteolytic activity co-purifies with presenilin 1 in a high molecular weight complex. More impressively, several transition-site analogue inhibitors of aspartyl proteases, which block γ-secretase activity, can be chemically cross-linked to the presenilins, indicating that if the presenilins are not themselves the catalytic component of γ-secretase, they are likely to be physically very close to the catalytic site within a multimeric protein complex.
Other putative roles for presenilin 1 have included a role in the regulation of intercellular signal transduction during development, in apoptosis, and possibly in intracellular calcium ion homeostasis. The former suggestion arose because null mutations in a second presenilin orthologue in C. elegans (sel12) exert a suppressor effect on abnormalities in vulva progenitor cell fate decisions induced by activated Notch mutants.114 Notch is involved in intercellular signaling during development. SEL12 protein shows stronger amino acid sequence identity to the human presenilin proteins than does SPE4. A role for mammalian presenilins in Notch-mediated signal transduction is further supported by the fact that homozygous targeted knockout of the murine PS1 gene (using homologous recombination) causes embryonic lethality around day E13 and is associated with: (a) severe developmental defects in somite formation and axial skeleton formation; (b) the occurrence of cerebral hemorrhage; and (c) reduced Notch and delta transcription in selected cell types.148,149 Similar phenotypes have been observed in mice with targeted knockouts of the murine Notch1 and DLL1 genes, supporting the hypothesis that PS1 has either a direct or an indirect role in intercellular signal transduction.150 A role in the suppression of apoptosis has been suggested from studies in transfected cells. Overexpression of full-length wild-type PS1 or wild-type PS2 can cause apoptosis in transfected cells, and mutations further sensitize these cells to apoptosis, possibly through a mechanism involving heterotrimeric G-coupled proteins sensitive to pertussis toxin.151 It is of note that mutations in βAPP are also thought to cause constitutive activation of programmed cell death pathways involving heterotrimeric G-coupled proteins.58,59
To date, more than 40 different mutations have been discovered in the PS1 gene (Table 234-2). The majority of these mutations are missense mutations giving rise to the substitution of a single amino acid. These mutations are predominantly located in highly conserved transmembrane domains, at or near putative membrane interfaces, or in the N-terminal hydrophobic or C-terminal hydrophobic residues of the putative TM6-TM7 loop domain. Two splicing-defect mutations have been identified. One involves a point mutation in the splice acceptor site at the 5′ end of exon 10 (in some exon numbering systems, exon 10 is labeled exon 9).152–154 Because exon 9 and exon 11 are in-frame, this mutation allows exon 9 to be fused in-frame with exon 11, thereby removing a series of charged residues at the apex of the hydrophilic acidically-charged TM6–TM7 loop domain. Interestingly, this mutation removes residues near the endoproteolytic cleavage site at residue 290 and results in the production of increased quantities of uncleaved holoprotein, which appears to be incorporated normally in the high molecular weight ER and Golgi-resident complexes.128,136 The second splice-site mutation arises from deletion of a G nucleotide from the splice donor site at the 3′ end of exon 5 (reference155 and Rogaeva, unpublished). An alternate GT splice donor site four base pairs further downstream in intron 5 is used, which causes the in-frame insertion of an extra tyrosine residue within the luminal TM1–TM2 loop domain (Rogaeva et al., unpublished).
Table 234-2: Missense Mutations in the Presenilin Genes |Favorite Table|Download (.pdf) Table 234-2: Missense Mutations in the Presenilin Genes
|Condon ||Location ||Mutation ||Phenotype |
| || || ||PRESENILIN I |
|79 ||N-term loop ||Ala → ? ||FAD, onset 64 years |
|82 ||TM1 ||Val → Leu ||FAD, onset 55 years |
|95 ||TM1 ||Val → Phe ||FAD |
|115 ||TM16TM2 loop ||Tyr → His ||FAD, onset 37 years |
|117 ||TM16TM2 loop ||Pro → Leu ||FAD, onset 28 years |
|120 ||TM16TM2 loop ||Glu → Asp ||FAD, onset 48 years |
|139 ||TM2 ||Met → Thr ||FAD, onset 49 years |
|139 ||TM2 ||Met → Val ||FAD, onset 40 years |
|143 ||TM2 ||lle → Thr ||FAD, onset 35 years |
|146 ||TM2 ||Met → Leu ||FAD, onset 45 years |
|146 ||TM2 ||Met → Val ||FAD, onset 38 years |
|146 ||TM2 ||Met → lle ||FAD, onset 40 years |
|163 ||TM3 interface ||His → Arg ||FAD, onset 50 years |
|163 ||TM3 interface ||His → Tyr ||FAD, onset 47 years |
|171 ||TM3 ||Leu → Pro ||FAD, onset 40 years |
|209 ||TM4 interface ||Gly → Val ||FAD |
|213 ||TM4 interface ||lle → Thr ||FAD |
|231 ||TM5 ||Ala → Thr ||FAD, onset 52 years |
|233 ||TM5 ||Met → Thr ||FAD, onset 35 years |
|235 ||TM5 ||Leu → Pro ||FAD, onset 32 years |
|246 ||TM6 ||Ala → Glu ||FAD, onset 55 years |
|260 ||TM6 ||Ala → Val ||FAD, onset 40 years |
|263 ||TM66TM7 loop ||Cys → Arg ||FAD, onset 47 years |
|264 ||TM66TM7 loop ||Pro → Leu ||FAD, onset 45 years |
|267 ||TM66TM7 loop ||Pro → Ser ||FAD, onset 35 years |
|280 ||TM66TM7 loop ||Glu → Ala ||FAD, onset 47 years |
|280 ||TM66TM7 loop ||Glu → Gly ||FAD, onset 42 years |
|285 ||TM66TM7 loop ||Ala → Val ||FAD, onset 50 years |
|286 ||TM66TM7 loop ||Leu → Val ||FAD, onset 50 years |
|del291-319 ||TM66TM7 loop ||Short loop ||FAD, |
|384 ||TM66TM7 loop ||Gly → Ala ||FAD, onset 35 years |
|392 ||TM66TM7 loop ||Leu → Val ||FAD, onset 25-40 years |
|410 ||TM7 ||Cys → Tyr ||FAD, onset 48 years |
| || || ||PRESENILIN II |
|141 ||1 TM2 ||Asn → lle ||FAD, onset 50-65 years |
|239 ||4 TM5 ||Met → Val ||FAD, onset variable 45-84 yrs |
|Adapted with permission from St George-Hyslop P: Genetics of Alzheimer disease, in Markesbery W, (Ed): Neuropathology of Dementing Diseases. London, Edward Arnold, 1997. |
The wide scattering of missense mutations led to speculation that the effect of most of the FAD-related mutations is a gain-of-function effect.156 This is partially borne out by two observations in PS1 knockout animals (PS1−/−). First, these animals have a phenotype of early perinatal mortality without evidence of Alzheimer disease.148,149 This loss-of-function phenotype in PS1−/− animals can be completely rescued by both wild-type and mutant PS1 transgenes.157,158 Second, PS1−/− mice have a defect in βAPP processing manifest by the failure of γ-secretase cleavage and the accumulation of the C-terminal stubs of βAPP following α- and β-secretase cleavage (α- and β-stubs).145 This defect in βAPP processing is completely reversed by both wild-type and mutant PS1 transgenes. A gain of function is imparted by the mutant transgenes because they also induce an increase in Aβ 42, which, as described below, is a consistent biochemical effect of PS1 mutations.157,158 However, a very different perspective is provided by studies using human PS1 cDNAs in complementation assays of mutant sel12 in C. elegans. These studies suggest that the wild-type human PS1, but not mutant human PS1 cDNAs are able to complement the loss-of-function sel12 mutants.159,160 The latter result argues that the human PS1 mutants may not be fully functional (but do not fully preclude a gain-of-function effect as well).
Regardless of whether PS1 (and PS2) mutations cause a gain of function or a loss of function, it seems likely that one effect is to alter the processing of βAPP by preferentially favoring the production of potentially toxic long-tailed Aβ peptides ending at residue 42 or 43.60,161–164 Thus, fibroblasts from heterozygous carriers of PS1 mutations, various cell lines transfected with βAPP and PS1 cDNAs, as well as the brain from transgenic mice overexpressing mutant PS1 transgenes, contain or secrete increased quantities of long Aβ peptide isoforms with only a variable but minor increase in short-tailed Aβ peptides.60,161–164 Direct measurements of Aβ peptide isoforms in the postmortem brain tissue of patients dying with PS1-linked FAD also show marked increases in the amount of long-tailed Aβ isoforms compared to control brain tissue and to brain tissue from subjects with sporadic Alzheimer disease.165
As with βAPP mutations, while there is a substantial body of work to suggest that defective processing of βAPP and overproduction of Aβ 42 are intimately associated with PS1 mutations, evidence has also emerged that suggests that PS1 and PS2 mutations modulate cellular sensitivity to apoptosis induced by a variety of factors including staurosporine, Aβ peptide, serum withdrawal and NGF withdrawal.118,151,166,167 Currently, these data are still evolving and the apparent paradox of a putative “apoptosis promoting effect” for the presenilins and the existence of viable transgenic mice overexpressing mutant or wild-type presenilin cDNAs, but lacking widespread apoptosis, remains to be explained.
A third way in which PS1 mutations might lead to AD is by altering the metabolism of the presenilin proteins in a manner analogous to the effects of βAPP mutations on the metabolism of βAPP holoprotein. Evidence to support this is somewhat conflicting. It is clear that the exon 10 splicing mutation does inhibit the formation of N- and C-terminal fragments.128 This observation clearly indicates that the formation of N- and C-terminal fragments is not necessary for the genesis of AD. Interestingly, however, the mutant PS1ΔExon10 holoprotein is incorporated into the higher molecular weight complex which contains the N- and C-terminal fragments.136 Preliminary studies of cultured cells overexpressing missense mutant presenilin cDNAs had suggested that missense mutations might also inhibit the physiological endoproteolytic cleavage events.168 Further studies, however, have yielded conflicting results. Some studies suggest that only a subset of mutations (e.g., C410Y) inhibit endoproteolytic cleavage while others (e.g., G384A) did not.169 Other studies of the levels of mutant presenilin proteins in the brains of transgenic mice seemed to suggest the opposite result: namely PS1 missense mutations cause an increased accumulation of physiological endoproteolytic cleavage fragments,170 while PS2 mutations have no effect on the physiological endoproteolytic process.171 Preliminary studies also suggest that PS2 mutations might increase the sensitivity of PS2 to caspase 3-mediated cleavage.132 However, this was not borne out in subsequent studies in cultured cells or in homogenates of brain tissue from subjects with PS1 mutations (the levels of various PS1 derivatives were indistinguishable from controls).172
During the cloning of the presenilin 1 gene on chromosome 14 a very similar sequence was identified in the public nucleotide sequence databases (GenBank L44577).173,174 Further analysis revealed that this similar nucleotide sequence was derived from a gene on chromosome 1q42.1, and encodes a polypeptide whose open reading frame contained 448 amino acids. The sequence of this peptide showed substantial amino acid sequence identity with that of the presenilin 1 protein (overall identity approximately 60 percent), and would be predicted to have a structural organization very similar to that of PS1 protein. Within the TM domains, the amino acid sequence identity between this new gene and presenilin 1 was even higher (approximately 90 percent). However, the pattern of transcription of this novel gene was slightly different from that of presenilin 1, being expressed less homogeneously in the brain and in peripheral tissues. In fact, PS2 was maximally expressed in cardiac muscle, skeletal muscle, and pancreas. Nevertheless, when the genomic organization of this novel gene was worked out it was apparent that many of the intron-exon boundaries (especially those relating to the highly conserved transmembrane domains) were identical between this gene and presenilin 1.119,175,176 Cumulatively, these observations suggested that this novel gene, which became known as presenilin 2,173 was a homologue of the presenilin 1 gene on chromosome 14. Not surprisingly, the predicted topology of PS2 is similar to that of PS1, and it also forms similar but independent multimeric protein complexes which contain β-catenin.136,137
Mutational analyses uncovered two different missense mutations in the presenilin 2 gene in families segregating early onset forms of Alzheimer disease. The first mutation (Asn141Ile) was detected in a proportion of families of Volga German ancestry, in which the FAD locus had been independently mapped by genetic linkage studies to chromosome 1.173,174 The second mutation (Met239Val) was discovered in an Italian pedigree.173 However, in contrast to the frequency of presenilin 1 mutations, screening of large data sets reveal, that presenilin 2 mutations are likely to be rare.176
Another profound difference between the presenilin 2 mutations and those in the βAPP and PS1 gene is that the phenotype associated with PS2 mutations is much more variable.176,177 Thus, in the vast majority of carriers, the illness begins between the ages of 35 and 65 years for PS1 mutations, and between 40 and 65 years for βAPP mutations. In contrast, the range of age-of-onset in heterozygous carriers of PS2 mutations is between 40 and 85 years of age, and there is at least one instance of apparent nonpenetrance in an asymptomatic octogenarian transmitting the disease to affected offspring.176,178,179 A similar, but less profound variation in age-of-onset within families segregating the βAPP V717I mutation has been ascribed to a modifying effect by the APOE gene.82–84 Thus, carriers of the βAPP V717I mutation who have one or more ∈4 alleles at APOE have an earlier onset than do heterozygous carriers of the V717I mutation who have the ∈2 allele and no ∈4 alleles of APOE. However, because the effect of APOE ∈4 on the age-at-onset in PS2 mutations is either absent or less profound, modifier loci other than APOE probably account for much of this variation.
The relationship of the normal function of presenilin 2 to that of presenilin 1 remains unknown. Although both proteins reside within the perinuclear envelope, endoplasmic reticulum, Golgi, and some as yet uncharacterized intracytoplasmic vesicles, they appear to form distinct high molecular weight complexes,131,137 and their tissue-specific patterns of expression are different.173,175 Nevertheless, given the strong similarities in structure and in amino acid sequence of the respective proteins, it would seem likely that PS1 and PS2 have similar or overlapping activities. This is supported by the fact that the effect of PS2 mutations is similar to those of PS1 mutations and by the fact that the residues mutated in PS2 are conserved in PS1. Thus, PS2 mutations, like PS1 mutations, increase the secretion of long-tailed Aβ peptides.60,164 PS2 mutations may also cause increased sensitivity to apoptosis, but it is unclear whether this is an effect independent of their ability to cause increased Aβ peptide secretion.118,166 Finally, although PS2−/− mice have essentially no detectable Notch-deficient embryonic phenotype, and have apparently normal βAPP processing (in contrast to PS1−/− mice), double knock-out mice (PS1−/− and PS2−/−) have very profound and more severe Notch-deficient embryonic phenotypes and essentially complete cessation of γ-secretase activity.334–336 This result suggests that the activities of PS1 and PS2 are likely to be overlapping and slightly redundant.
The observation that presenilin 1, presenilin 2, βAPP, and APOE mutations/polymorphisms accounted for approximately half of the genetic variance of AD prompted a number of groups to undertake genome surveys to identify other susceptibility loci. Using two independent data sets of late-onset AD pedigrees (n = 16; n = 38), a novel AD locus was identified in the pericentromeric region of chromosome 12 between the markers D12S1042 and D12S390 (z = 3.5).180 Follow-up studies in a dataset of 53 independent pedigrees confirmed the presence of an AD susceptibility gene within a larger (approximately 60-cM) region of chromosome 12 in a subset of pedigrees.181 This latter study also indicated that pedigrees showing evidence for linkage to chromosome 12 generally had an onset after the age of 60, an unambiguous pattern of familial clustering with at least three affected family members that was independent of their APOE status. The subset of pedigrees that did not show linkage to chromosome 12 presumably reflects the presence of one or more genetic susceptibility factors elsewhere in the genome.
The identity of the chromosome 12 locus at the moment is unclear. Several candidate genes, including α2-macroglobulin (A2M), low-density lipoprotein receptor-related protein (LRP1), ARF2 and so on, map within this interval. Some of these genes (e.g., ARF3, Wnt1, plakophillin 2, ITR2) were excluded by the failure to find nucleotide sequence changes in their open reading frames that are either enriched in, or unique to, patients with the chromosome 12 form of AD. Some biochemical studies have suggested a role for A2M protein in AD (through its ability to bind Aβ and through competition with both Aβ and APOE for clearance through the LRP1 receptor). This work led to speculation that A2M might be the site of mutations associated with AD (see reference182 for a summary). Considerable support for this concept (which links Aβ, APOE, A2M, and LRP1 in one biochemical cascade) was provided when a preliminary family-based association study detected an association between AD and an intronic insertion/deletion polymorphism at the 5′-splice site of exon 18 of A2M in a dataset of 104 pedigrees collected by the National Institute of Mental Health.182 However, follow-up studies in a larger dataset of NIMH pedigrees (n = 142), in several datasets of independent pedigrees (>60 pedigrees each), and in several sporadic AD case-control datasets (>100 subjects each), failed to confirm an association between AD and this insertion/deletion polymorphism in A2M.183 Furthermore, no biologic effect could be discerned for the A2M polymorphism.183 Together, these results suggested that the original observation may have been an artifact. Another potential candidate gene on chromosome 12, LRP1, has also generated ambiguous results when studied using conventional allelic association (case-control) methods. Thus, the 87-bp allele of tetranucleotide repeat polymorphism in the 5′-promoter of LRP was reported to be significantly increased in late onset sporadic AD cases.184 However, in an independent collection of French AD cases, the 87-bp allele was significantly decreased in frequency,185 and a study of more than 130 patients with sporadic AD showed no significant association with either allele.186 Similarly, several studies have suggested a weak, borderline statistically significant association (e.g., p <0.01 not corrected for multiple testing) for homozygosity of the C,C allele of the C766T polymorphism in exon 3 of LRP.187–189 However it was not significantly different from controls in several other comparable studies.190–192
Although five different genetic loci associated with inherited susceptibility to AD are now known, several large surveys of subjects with familial Alzheimer disease have indicated that they still do not account for the disease in all pedigrees. Pedigrees lacking mutations in any of the four known AD genes and failing to cosegregate with chromosome 12 markers do not display a singular phenotype, but, instead, comprise a mix of early-onset autosomal dominant pedigrees and late-onset multiplex pedigrees. Consequently, it is likely that there are several FAD genes yet to be found. Some of these FAD loci will probably be associated with rather rare, but highly penetrant defects similar to those seen with mutations in PS1 and βAPP. Other genes may result in incompletely penetrant autosomal dominant traits like that associated with PS2. However, it is likely that a significant proportion of the remaining genes will have effects similar to that of APOE, in which the presence or absence of other genetic and environmental risk factors influences the ultimate phenotype.
Attempts to identify novel AD susceptibility genes have followed two strategies. One strategy, a continuation of the conventional positional cloning strategy, has attempted to show cosegregation of marker alleles with the disease phenotype (identity by descent) in pedigrees multiply affected by AD. These studies have assessed cosegregation using both conventional parametric lod score methods as well as newer nonparametric methods.180,181,193 However, because these methods work best for pedigrees multiply affected with AD, increasing emphasis has been placed on the use of simple case-control studies (such as those so effectively used to discover the association between AD and APOE). More recently, the case-control method for discovering allelic associations was supplemented by novel statistical methods (family-based association methods) that enable examination of allele sharing between affected sibs compared to unaffected sibs, such as the sibship disequilibrium transmission test (SDT) and the sib-transmission disequilibrium test (S-TDT).182,194 These family-based association methods examine the parental alleles transmitted to unaffected siblings as a source of control chromosome information that, theoretically, is better matched for ethnicity and genetic background.
The conventional case-control allelic association tests yielded positive results on a significant number of genes, many of which are plausible biochemical candidate genes. However, most of these studies have not received the same robust replication as the association between AD and the ∈4 allele of APOE. As a result, it is difficult to discern whether the reported associations are true but perhaps limited to particular subsets of AD, or whether they represent statistically significant but biologically incorrect results. The possibility of biologically false positive results in allelic association studies is a well-recognized problem in human genetics, and can arise when the test and control populations are not drawn from identical genetic backgrounds (i.e., due to population stratification). Recently, in an attempt to rectify the high false positive rate for simple allelic association studies, some studies have begun to use the newer family-based association methods.182
A partial list of candidate genes provisionally identified as putative AD susceptibility loci includes: homozygosity for the A allele of an intronic polymorphism in α1-chymotrypsin;195 5 repeat allele of an intronic insertion-deletion polymorphism in the very-low-density lipoprotein receptor;196 neutral coding sequence and intronic polymorphisms in low-density lipoprotein receptor-related protein;187 homozygosity for the common “1” allele of an intronic polymorphism in presenilin 1;197 K-variant of butyrylcholinesterase;198 and homozygosity for the Val/Val variant of the V443I polymorphism in bleomycin hydroxylase.199 However, most of these candidate genes have not received widespread confirmation when tested in independent but comparable datasets.107,108,200–204
The Mitochondrial Genome and Alzheimer Disease
The role of mitochondria in some stages of apoptosis and the slight preponderance of females with AD suggest that mutations in the mitochondrial genome might play a role in the genesis of AD. Analysis of mitochondrial DNA has led to conflicting results. Thus, heteroplasmic mutations at nucleotide 5460 in codon 331 of subunit 2 of NADH dehydrogenase-ubiquinone oxidoreductase were found in mtDNA from 10 of 19 brain samples affected by AD, but were absent in all 11 normal brain samples.205 Follow-up studies, however, either failed to find these mutations in 14 of 14 sporadic AD patients or observed them at a lower frequency in patients with AD (4.4 percent) as compared to normal controls (8.6 percent).206,207 A variant at position 4336 within the tRNA(Gln) gene was found in ≈ 5 percent of AD patients, but only 0.7 percent of controls.207 A variant at position 3397 that converted methionine to valine within the ND1 gene was reported in 2.7 percent of AD subjects but 0 percent of controls.207 Again, follow-up studies yielded inconsistent results, some appearing to support the existence of mutations at position 4336 as a risk for AD,208 others failing to find differences in the frequency of mutations at position 4336 between controls and AD affected patients.209,210 Finally, missense mutations within the mitochondrial cytochrome oxidase genes MTCO1 and MTCO2 were proposed to be associated with late-onset AD.211 However, follow-up studies revealed that these sequence differences actually arose from contamination of the mtDNA with nuclear DNA, and that the contaminating nuclear DNA contained these sequence differences within nuclear pseudogenes.212
Role of βAPP, APOE, and the Presenilins in the Pathogenesis of “Sporadic” AD
The appearance of derivatives of βAPP, APOE,63,86,213 and PS1 proteins124,214 in pathologic structures in brain tissue of subjects with “sporadic” AD argues that these proteins are involved in the disease, including “sporadic” AD. The most pressing evidence for a role of these genes in sporadic AD obviously derives from the population-based studies showing an association between the ∈4 allele of APOE as described above. There is some evidence to suggest that homozygosity for the more common 1 allele of an intronic polymorphism within the PS1 gene might be associated with nearly doubling of risk for late-onset “sporadic” AD.197,215,216 However, multiple follow-up studies in several different populations have generally not confirmed this association.204,217–220 Nevertheless, a small number of missense mutations have been observed in the PS1 gene (but not in the PS2 gene) of patients with apparently “sporadic” AD (e.g. PS1 H163R and PS2 M139K).221–223
A prevailing theory of the pathogenesis of AD is that the various causes of this disease affect a biochemical pathway that leads to AD and to mismetabolism of βAPP with the subsequent pathologic deposition of Aβ peptide. Such a single biochemical pathway hypothesis would predict the existence of interactions between the known AD susceptibility genes, a prediction for which there is now increasing evidence. Thus, the APOE genotype influences age of onset in carriers of the βAPPV717I mutation,82,84 but not the βAPP692 missense mutation.85 This interaction between APOE and βAPP is also evident in transgenic mice (see above).89 There is a small degree of variation of age at onset in families with PS1 mutations and a greater degree of variation in age at onset in families segregating PS2 mutations that appears to arise from genetic modifiers, but the role of APOE is somewhat unclear.176,177 However, clear evidence for additive effects of mutations in PS1 and βAPP is provided by the analysis of Aβ peptide production in transfected cells and cerebral Aβ deposition in transgenic mice expressing both mutants.164,224–226 Thus, Aβ42 production is dramatically increased in cell lines and in the brains of transgenic mice expressing both mutant PS1 and mutant βAPP cDNAs, indicating that these mutations are likely to act in different places within the same biochemical cascade.
Transgenic mice have been created in which mutant human βAPP or mutant human PS1 transgenes have been overexpressed in mouse brain. The first such transgene used a spliceable human βAPP transgene that carried the Val717Phe missense mutation expressed under the platelet-derived growth factor receptor β subunit promoter.227 This transgene produced abundant quantities of mutant βAPP and resulted in the accumulation of diffuse deposits of Aβ peptide in the brain and abnormal synaptic morphology at about 9 months of age, but was not associated with obvious cognitive deficits or the appearance of neurofibrillary tangles.96,228 A second transgenic mouse overexpressing the human βAPP695 transcript containing the 670/671 missense mutation also resulted in the widespread accumulation of Aβ peptide and the appearance of dysmorphic synapses and developed a very subtle behavioral deficit by about 9 months of age, although they did not show neurofibrillary pathology.228 A third transgenic mouse overexpressing mutant human βAPP was recently described, in which there is some evidence for neuronal loss and staining for abnormally phosphorylated Tau, as well as prominent extracellular Aβ deposition possibly due to the higher levels of βAPP expression in this line.229 However, as in the other two lines, no neurofibrillary tangles were observed. Finally, overexpressing mutant forms of PS1 or PS2 has created several transgenic animals.162,164 While at the time of writing, these animals are too young to develop any neuropathology, biochemical measurements show that transgenic mice with mutant presenilins have increased levels of long-tailed isoforms of Aβ peptide in the brain. Interestingly, several groups have found that cross-breeding mice that overexpress mutant PS1 with mice that overexpress mutant βAPP results in accelerated Aβ deposition,164,224–226 arguing that these genes act at different points within the same biochemical pathway leading to AD. One of the principal uses of transgenic murine models of Alzheimer disease is that they provide a means to rapidly test and develop new therapies for the disorder. The growing evidence that the accumulation of Aβ-peptide, and especially Aβ42, assembled into oligomeric aggregates (or protofibrils), is an essential event in the pathogenesis of Alzheimer disease, has led to attempts to inhibit the production, or accelerate the removal of Aβ. One somewhat unexpected strategy to achieve the latter goal has been to immunize transgenic mice with Aβ42.337 Several groups have now shown that this results in a dramatic reduction in the number of amyloid plaques and an improvement in the cognitive deficits of murine models of Alzheimer disease.337–339 Significantly, the antibodies which are most effective in this regard are those directed toward fibrillar forms of Aβ.338,339 Furthermore, these beneficial effects on cognition and on neuropathology can be achieved with only modest changes in Aβ-peptide levels in brain (presumably because protofibrillar forms of Aβ are a low abundance, labile species).339