© 2000 Oxford University Press
Human Molecular Genetics, 2000, Vol. 9, No. 16 Review 2383–2394
Transcriptional regulation of Alzheimer’s disease genes: implications for susceptibility Jessie Theuns and Christine Van Broeckhoven+ Flanders Interuniversity Institute for Biotechnology (VIB), Born-Bunge Foundation (BBS), University of Antwerp (UIA), Department of Biochemistry, Universiteitsplein 1, B-2610 Antwerpen, Belgium Received 19 June 2000; Accepted 6 July 2000
In recent years, important progress has been made in uncovering genes implicated in Alzheimer’s disease (AD). Three causal genes have been identified in which mutations cause familial presenile AD: the amyloid precursor protein gene and the presenilin 1 and 2 genes. Additionally, the ε4 allele of the apolipoprotein E gene was shown to be a major risk factor for AD. Despite the genetic heterogeneity, all of these genes work through a common mechanism, i.e. increasing the amount and deposition of the amyloid β peptide (Aβ) in brain triggering AD-related neuronal degeneration. Therefore, the levels of Aβ and of the factors involved in its production and deposition are important in the neuropathogenesis of AD. Regulation of transcription of AD genes might therefore be an important player in the neurodegenerative process. In this review, we describe the major features of transcriptional regulation of the known AD genes and the implications of variable expression levels on susceptibility to AD.
INTRODUCTION Alzheimer’s disease (AD) is a neurodegenerative disorder of the brain characterized by neuronal loss, extensive deposition of amyloid β peptide (Aβ) in the brain parenchyma and in vessel walls, and the appearance of neuronal inclusions of abnormally phosphorylated tau. The regions that are most affected are the hippocampus and cerebral cortex. Clinically, AD patients show a gradually progressive decline in memory and cognitive functions that is diagnosed on neurological examination, neuropsychological testing and neuroimaging. Among the dementias, AD is the most frequent, with 70% of cases affected. The prevalence of AD increases with age, with 40% of the population older than 85 years affected (1). Apart from aging per se, two other well defined risk factors are gender and family history of AD (2). The genetic etiology of AD is complex, with both genetic and environmental factors influencing the expression of the phenotype. Nevertheless, in a small percentage (60 PSEN1 and 4 PSEN2 mutations have been identified in EOAD patients, scattered over the entire coding region of the genes (http://molgen-www.uia.ac.be/ADMutations/ ). Both in vitro and in vivo, PSEN mutations lead to increased Aβ42 production, suggesting a role for PSEN in the processing of APP (31–45). Together, these findings led to the hypothesis of a gain of function for mutant PSEN (20) and provided further evidence for the central role of Aβ42 in AD pathogenesis. PSENs
whom correspondence should be addressed. Tel: +32 3 8202601; Fax: +32 3 8202541; Email:
[email protected]
2384 Human Molecular Genetics, 2000, Vol. 9, No. 16 Review
Table 1. AD genes Gene
APP
Gene
cDNA
Length of 5′-UTR (bp)
Coding sequence (nucleotides)
Length of 3′-UTR (bp)
Length (kb)
GenBank accession no.
Length (bp)
GenBank accession no.
287
D87675
3579
NM000484
147 (exon 1)
148–2460 (exons 1–18)
1120 (exon 18)
PSEN1
83
AF109907
2756
NM000021
248 (exons 1A–3); 541 (exons 1B–3)
249–1652 (exons 3–12)
1104 (exon 12)
PSEN2
25
U50871
2236
NM000447
367 (exons 1–3)
368–1714 (exons 3–12)
522 (exon 12)
M10065
1156
NM000041
61–1014 (exons 2–4)
143 (exon 4)
APOE
3.6
have also been shown to be death substrates undergoing caspase cleavage during apoptosis (46–48). Although the role of apoptosis in neuronal cell death in AD remains to be proven, AD-linked PSEN mutations, as well as decreased expression of PSEN1 and overexpression of PSEN2 (46), result in apoptosis (49). However, the homology with SEL-12 of Caenorhabditis elegans suggested a possible role for PSEN in NOTCH signaling (50). Further evidence was provided by the study of embryonically lethal PSEN1-deficient mice, which show abnormal somite segmentation, a phenotype shared with NOTCH1-deficient mice (51,52). Moreover, mutations in other genes involved in NOTCH signaling also lead to central nervous system (CNS) disorders with onset ages in adulthood (53). In addition to the three causative genes, the ε4 allele of the apolipoprotein E gene (APOE) on chromosome 19 was identified as a genetic risk factor for both EOAD (54,55) and late-onset AD (LOAD) (56–58). The risk associated with the ε4 allele is dose dependent, which is reflected in the increased risk and decreased onset age with the number of ε4 alleles (59). Several observations indicated that APOE ε4 is also involved in increased Aβ deposition. It was shown that APOE promotes Aβ fibril formation in vitro (60). Also, APOE ε4 has a higher binding affinity for Aβ than has APOE ε3 (56,61), possibly making Aβ insoluble and therefore more prone to deposition. Although these in vitro data are controversial, AD patients homozygous for APOE ε4 have indeed been shown to have more Aβ deposits than APOE ε3 homozygotes (62). Also, it was shown that APOE ε4 influences the onset age of EOAD patients carrying an AD-related APP mutation but not that of PSEN1 mutation carriers (63). Since mutations in the causal EOAD genes explain 80% of promoter activity, indicating the importance of protein–protein interactions between TFs binding to upstream and downstream cis-elements. Preliminary data on murine PSEN1 promoter activity and in situ hybridization suggest that PSEN1 is expressed and transcribed preferentially in neurons. The structure and expression of PSEN1 are highly conserved between mice and human. Multiple mRNA transcripts, originating from TSSs of alternatively transcribed first exons, have been reported for both species. Reporter gene analysis showed that both TSSs are most probably controlled by one single promoter spanning the +1 position of exon 1A (130). In human, only one exon 1B transcript has been reported so far; therefore, it is difficult to conceive how exon 1B transcription can be driven from the same promoter as the abundant exon 1A transcripts, as described for the mouse transcripts. Hence, it is possible that transcription from the human exon 1B is controlled by its own promoter or at least by exon 1B-specific cis-elements, possibly located downstream of +178. Until now, no data on functional cis-elements controlling transcription from human PSEN1 exon 1B have been reported.
Human Molecular Genetics, 2000, Vol. 9, No. 16 Review 2387
In the regulatory region, maximal similarity is found in the –39/+117 sequence of the mouse PSEN1 promoter region. The +20 GC box (SP1) and the –12 Ets motif are strictly conserved, and reporter gene analysis indicated that this region contributes to the neuron-preferred promoter activity (130). The mouse sequence upstream of position –39 differs significantly from the human sequences. Absence of long stretches of sequence homology is one of the main problems in promoter recognition and it is conceivable that more functional cis-elements, although not located at the corresponding positions, are shared between the mouse and human PSEN1 promoter. Furthermore, sequences from –22 to +178, which confer >80% of human PSEN1 promoter activity, coincide with the region of high homology with the mouse promoter. Also, the major human +1 TSS is located only eight nucleotides downstream of the mouse TSS (130,131), emphasizing the importance of this region for the function of the promoter in both species. Despite the striking similarities between PSEN2 and PSEN1 in genomic structure, alternative transcripts and use of multiple TSSs (116,135,136), there is little homology in the 5′-flanking region, and the first two exons of PSEN2, although alternatively transcribed, are not mutually exclusive (116). There are two distinct cis-elements regulating transcription from the predicted TSS. PSEN2 basal promoter activity resides between –403 and +13, a TATA-less, GC-rich region containing numerous putative AP-2 and SP1 sites (137). In this region, a functional nerve growth factor (NGF)-responsive element that mediates PSEN2 promoter activation following NGF treatment also resides. The observed 2-fold up-regulation by NGF confirmed the predicted involvement of PSENs in neuronal differentiation (138). Whether the AP-2 or SP1 sites clustered in this region or another as yet unidentified cis-element is responsible for the NGF responsiveness is not yet clear. Further deletion analysis provided evidence for the existence of a second PSEN2 promoter located in intron 1, directing transcription from the start sites in exon 2, though no functional elements have been identified so far. APOE is expressed in a variety of tissues and cell types and its expression is highly regulated by nutritional, hormonal, tissue- and cell-specific factors, and intracellular cholesterol levels (139,140). In contrast to the causal AD genes, the 5′-flanking sequence of APOE harbors a functional TATA box (141). Multiple general and specific cis-elements have been mapped to the APOE promoter (Table 2, Fig. 1C) (140). The proximal GC box binds SP1 and is required for maximum transcriptional activity (142). Three non-specific enhancer elements, active in both neuronal and non-neuronal cells (143), were identified. Two of these, upstream regulatory elements 1 (URE1) and 2 (URE2), reside in the proximal promoter region and one resides in the first intron, the intron regulatory element 1 (IRE1). The dominant regulatory sequence in URE1 is located from –161 to –141 and is termed the positive element for transcription (PET). This fragment binds at least two TFs, one of which is SP1. Although SP1 is the only protein required for enhancer activity of PET, a second as yet unknown protein competes with SP1, in this way possibly negatively regulating APOE expression (142). A fourth regulatory domain in the APOE promoter, URE3, binds the 300 kDa URE3-binding protein, but needs further characterization (144). In astrocytic cells, the activity of the proximal promoter of APOE is up-regulated synergistically by cAMP and retinoic
Table 3. Promoter variations in AD genes Gene
Region screened
Variations
APP
–802/+268
–209C→G
PSEN1
3.5 kb upstream of exon 1B –48C→T
References 161 207
–1789G→A –2154G→A –2319(T) n –2823D/I –280C →G –2818A→G APOE
–1017/+406
+113C →G
71–73,164,175
–219G →T –427C →T –491A →T –456C →T –557C →T Potential AD-related mutations are shown in bold.
acid (RA), which is mediated by two AP-2 sites located in the proximal promoter (145). cAMP mimics the changes occurring in reactive gliosis (146,147) and RA is a potent morphogenetic agent on the developing nervous system and a known regulator of AP-2 expression (148). It is therefore reasonable to presume that the observed synergistic effect of cAMP and RA on the APOE promoter is probably due to an RA-promoted increase in AP-2, followed by a post-transcriptional activation of AP-2 mediated by cAMP (148,149). More tissue-specific cis-elements identified at present seem to reside downstream of APOE, with a potential brain-specific transcriptional activator in the APOE–APOCI intergenic region (150–154). The 5′ upstream region of rodent APOE is homologous to human APOE up to 200 bp upstream of the TSS (155,156), with the TATA box, the proximal GC box and URE3 at corresponding positions (Table 2) (141,156,157). PROMOTER VARIATIONS AND TRANSCRIPTIONAL ACTIVITY Screening for mutations in the –802/+268 APP promoter fragment in sporadic, and familial EOAD and LOAD cases did not reveal any AD-specific mutation (158–161). However, a C→G transversion was detected at position –209 in both affected and unaffected subjects (Table 3) (161). Although this variation is not unique for AD, it may have an effect on APP transcriptional activity associated with AD. Since the number of patients and controls was small and the individual analyses covered only parts of the APP promoter, one cannot yet exclude the existence of AD-related variations in the APP promoter altering transcriptional activity. Recent sib-pair analyses suggested that genetic variability at the APP locus may contribute to risk for LOAD (66,67). It is obvious that a systematic screening of the APP regulatory sequences in extended AD populations is necessary.
2388 Human Molecular Genetics, 2000, Vol. 9, No. 16 Review
Besides the amount of full-length APP available, the amount and activity of APP processing factors also influence the production of the amyloidogenic Aβ42. Since PSENs were assigned a pivotal role in APP processing, altered PSEN expression due to variations in regulatory regions is considered a risk factor for AD. Genetic association studies in a population-based EOAD case–control sample showed association of the single nucleotide polymorphism (SNP) –48C→T with EOAD (Table 3) (68). Systematic screening of 3 kb of the PSEN1 upstream region in the same population revealed, in addition to –48C/T (22), four novel polymorphisms (–1789G/A, –2154G/A, –2319Tn and –2823I/D), two of which (–2154G/A and –2823I/D) were also shown to be associated with increased risk for EOAD (Table 3). Linkage disequilibrium allowed for the identification of an EOAD risk haplotype (–48C/–2154G/–2823D). Additionally, two potentially ADrelated mutations (–280C→G and –2818A→G) were identified (Table 3) (69). The effect of –280C→G and –48C→T on the transcriptional activity of PSEN1 was studied in a transient transfection system. Luciferase reporter gene analysis demonstrated a neuron-specific 30% decrease in promoter activity for the –280G mutant and a neuron-specific 50% decrease in promoter activity for the –48C risk allele, which in homozygous individuals can lead to a critical decrease in PSEN1 expression. Notably, the genetic association of –48C with EOAD was explained by an over-representation of the CC genotype in EOAD patients. These studies provide evidence that increased risk for EOAD associated with PSEN1 may result from genetic variations in the regulatory region leading to altered expression levels of PSEN1 in neuronal cells due to differential binding of nuclear proteins (69,162,163). These data suggest that the increased risk for EOAD associated with PSEN1 may result from decreased expression levels of the PSEN1 protein. Although PSEN2 has also been shown to be involved in APP processing, and changes in its expression levels might be important in AD pathology, no association was found with AD and no PSEN2 promoter variations have been reported to date. It was shown that the relative APOE ε4 mRNA level is increased in AD compared with controls, and it was suggested that genetic variability in the neuronal expression of APOE contributes to disease risk (70). Multiple studies suggested that genetic variability in the regulatory region of APOE may modulate the risk associated with the APOE ε4 isoform. The first variation detected in the APOE regulatory region was a C→G transversion at position +113 in the intron 1 enhancer element (IE1), although no statistically significant association independent of APOE ε4 could be detected (Table 3) (164). In a population-based study, three SNPs (–491A→T, –427T→C and –219T→G alias Th1E47cs) and two heterozygous mutations (–557C→T and –456C→T) were identified (Table 3) (71,73). Reporter gene analysis and electrophoretic mobility shift assays (EMSAs) demonstrated that the three SNPs alter transcriptional activity of the APOE promoter due to differential binding of TFs (71,72). For all three SNPs, genetic association with AD, independent of APOE ε4, was reported (71–73). Although several studies attempted to confirm this association, most reported either absence of association or association due to linkage disequilibrium with APOE ε4 (114,165–172). However, population-based differences of APOE ε4 frequencies, giving rise to differences in relative risk for AD, have been documented
previously (173,174). It is therefore conceivable that there is a wide variation in relative risk for AD associated with APOE promoter polymorphisms. In vivo studies demonstrated that the deleterious effect on disease risk of both the –219T and –491A risk alleles correlated with an increased expression of the ε4 allele in brain (175). Later it was shown that the –491AA risk genotype is associated with increased levels of APOE in plasma, independently of APOE ε4 or AD status, though more pronounced in AD patients (176). These data provide evidence that, in addition to the qualitative effect of the APOE ε2/ε3/ε4 isoforms on risk for AD, the quantitative variation of expression of these isoforms due to functional APOE promoter variations is a key determinant in AD development. IMPLICATIONS OF VARIABLE GENE EXPRESSION ON AD PATHOGENESIS The most favored hypothesis suggests a pivotal role for increased Aβ42 secretion in AD pathology. Since the amount of APP and of factors involved in its processing are crucial for this elevation, it is conceivable that the transcriptional regulation of their genes plays an important role in AD pathology. A number of studies indicate that the amount of APP mRNA is indeed increased in AD brains (177–181). Moreover, trisomy 21 in Down syndrome (DS) patients leads to a 4- to 5-fold overexpression of APP, resulting in a 50 year decrease in onset age of AD in DS patients compared with the normal population (182). These results imply that a fundamental component of the molecular etiology of AD may lie in the expression of APP, its biogenesis and turnover, since the induction of the pathway leading to Aβ production will depend on the amount of APP present. Since overlapping cis-elements are known to be important for differential gene expression (183–185), the presence of overlapping SP1- and USF-binding sites in the APP promoter suggests that two different and independent regulatory pathways for APP expression might exist, one mediated by SP1 and the other by USF (104). SP1 has been shown to be ubiquitously expressed, however, with a substantial variation in different cell types and during development (186). The low levels of SP1 detected in different brain regions suggest that USF might also contribute substantially to the high expression of APP in neuronal cells, which was confirmed by EMSA with nuclear extracts from rat brain showing binding to USF but not to SP1. A number of factors have been reported that are able to influence SP1 activity in certain cell types, in this way leading to preferred usage of one pathway (187,188). Therefore, a deregulated overexpression of APP in brain might simply be caused by a local increase in SP1 activity. Additionally, AD brains exhibit numerous features which indicate that neurons affected by AD exist under conditions of stress. Since APP expression is regulated by stress factors, one can speculate that APP may be one of the genes coordinately modulated in brain in response to situations that require a defensive reaction. Stress-induced overexpression of APP can then lead to increased Aβ production. It is therefore conceivable that an imbalance between different regulatory pathways for APP expression, caused by a variation in a functional ciselement or by altered expression levels of TFs in specific brain regions, might be a risk factor for AD.
Human Molecular Genetics, 2000, Vol. 9, No. 16 Review 2389
During aging, the expression of PSENs decreases (120), and an even more significant decrease in PSEN has been reported in neurons from brain areas adversely affected by AD (189,190). In contrast, astrocytes reacting to neurodegeneration express elevated levels of PSEN (189,191). Several lines of evidence indicate that PSEN1 is closely linked to the γ-secretase processing of APP and that decreased expression (
