Introduction
Alzheimer’s disease (AD) is a neurodegenerative disorder characterised clinically by a progressive loss of memory and cognition, accompanied by functional impairments of orientation and praxis. Pathologically, the major changes involve a deposition of amyloid β protein (Aβ) in brain parenchyma (as amyloid plaques) and hyperphosphorylated tau within neurones (as neurofibrillary tangles). Additionally, most cases display deposits of Aβ within blood vessel walls—a change known as cerebral amyloid angiopathy (CAA). While more than 90% cases of AD are without obvious genetic cause, and termed ‘sporadic’, the remainder is associated with mutational events involving either the Amyloid Precursor Protein (APP) or presenilin (PSEN) genes.
With respect to the transmembrane protein
APP, missense mutations changing the amino acid sequence at either the amino- or carboxy-terminal points of the Aβ sequence (e.g.
APP670/671,
APP717) result in increased catabolic breakdown of APP by β- and/or γ-secretase into Aβ, and confer a pathological picture similar to that seen in sporadic AD. On the other hand, mutations lying in the juxtamembrane region, such as
APP692 or
APP693, are more associated with CAA than plaques, and often manifest clinically as acute stroke. There are also rare French [
6,
14,
37], Dutch [
41], Finnish [
38], Japanese [
19], Swedish [
48] and British [
30] families where AD is linked to duplications at the
APP locus, resulting in APP overproduction. In most of these families, the duplication has been validated only in living patients and confirmed cases with brain donation are scarce. An
APP duplication has also been reported in a Spanish patient with apparently sporadic AD and severe CAA [
21], but other studies of sporadic AD with CAA have not identified such duplications [
3,
11]. It has long been known that most individuals with Down syndrome (DS), who live into middle age and beyond, show a pathological picture indistinguishable from that of AD [
24,
25]. In most DS individuals, there is a complete triplication of chromosome 21, including the
APP locus. In both
APPdup and DS individuals, it is presumed that the early deposition of Aβ plaques and CAA stems from an overexpression of
APP and the consequent degradation of an excessive production of APP. In addition, recent work suggests that a mutation in the 3′untranslated region of
APP also result in APP overexpression and might act as a genetic determinant in some cases of CAA [
33].
Although all cases of AD are defined pathologically by the presence of numerous plaques and tangles, and usually CAA, throughout the cerebral cortex and hippocampus, the morphological appearance of these changes, especially plaques and CAA, can vary according to the underlying genetic background. For example, a much more severe CAA is seen in patients with presenilin-1 (
PSEN-
1) mutations located beyond codon 200 compared to those where the mutation lies before codon 200 [
28]. Also, certain
PSEN-
1 mutations are associated with an unusual morphological form of amyloid plaques, known as ‘cotton wool’ plaques, and often present clinically with a spastic paresis [
29].
The cardinal clinical presentation in
APP duplications (
APPdup) is that of progressive dementia frequently accompanied by seizures and intracerebral haemorrhage (ICH) [
6,
14,
19,
37,
38,
41], and neuropathological studies have revealed severe CAA in association with abundant Aβ plaques and neurofibrillary tangles, and occasionally Lewy bodies [
6,
14]. CAA is also often prominent in individuals with DS [
25] where there is also an additional
APP copy number. Nevertheless, CAA can present in sporadic and familial AD in various phenotypic histological forms, involving differing combinations of leptomeningeal and parenchymal arterial pathology, sometimes extending into the capillary bed [
2,
46].
We have recently reviewed epidemiological data on stroke (including haemorrhagic stroke) related to CAA to make comparisons between DS and
APPdup [
5]. Although ICH, the main clinical consequence of vascular amyloidosis, is a common clinical occurrence in
APPdup, this is a more poorly defined feature of individuals with DS, suggesting the presence of a mechanism(s) that acts protectively [
5]. This might seem somewhat paradoxical given that DS only differs from
APPdup in the ~ 270 other genes located on chromosome 21 that are also triplicated.
However, a direct comparison of vascular amyloidosis at pathological level between DS, APPdup, and other APP mutations has never been undertaken as far as we are aware, and consequently the degree and nature of tissue differences in CAA, ICH, and Aβ deposition between these disorders remain unclear. Therefore, in this analysis, we aim to compare the severity of amyloid plaque formation and CAA, and the subtype pattern of CAA pathology itself, and the relationship between CAA severity and CAA phenotype, in APP genetic causes of AD (APPdup, APP mutations), DS and sporadic (early and late onset) AD (sEOAD and sLOAD, respectively). The observations might help to elucidate important differences between the patient groups, and thus provide mechanistic insights related to clinical and neuropathological phenotypes. Since lipid and cholesterol metabolism is implicated in AD as well as vascular disease, we additionally aimed to explore the role of APOE genotype in CAA severity and subtypes.
Discussion
In the present study, we have compared the extent and pattern of Aβ deposition, both as plaques and as CAA, in patients with duplications in
APP, others with missense mutations in
APP, patients with sEOAD and sLOAD, older individuals with DS and elderly controls. The main findings to emerge were:
1.
The degree of plaque formation was greater in both DS and missense APP mutations than in sEOAD and sLOAD cases, while the degree of plaque formation was not significantly different between sEOAD and sLOAD.
2.
Conversely, the severity of CAA was significantly greater in both APPdup and missense APP mutations, and DS, compared to sEOAD and sLOAD. Again, all five pathological (AD) groups had a significantly greater degree of CAA than controls, and the degree of CAA was also significantly greater in sEOAD than sLOAD, and in APPdup compared to DS.
3.
When stratified by CAA subtype, no significant differences in overall plaque scores were seen between each CAA subtype for any of the six groups. However, in both APP mutations and sEOAD there was a significantly greater level of CAA as types 2 and 3 CAA compared to type 1. Conversely, in DS, sLOAD and controls there was a significantly greater level of CAA in type 1 CAA than types 2 and 3. In APPdup type 3 was the predominant CAA phenotype.
4.
CAA severity scores progressively increased across CAA types 1–4 for all cases combined and for each pathological group individually.
5.
APOE ε4 allele frequency was overall significantly higher in sEOAD than in DS, missense APP and controls, as was ε4 allele frequency in sLOAD compared to DS, missense APP mutations and controls. There were no significant differences in APOE ε4 allele frequency between DS, missense APP and controls. None of the APPdup patients bore APOE ε4 allele.
6.
All blood vessels stained for CAA by 4G8 appeared to be detected by BA27 but fewer were detected with BC05. Conversely, all plaques detected by 4G8 appeared to be detected by BC05, but only a subset was detected by BA27.
7.
APOE ε4 allele frequency varied numerically between each of the CAA phenotypes in APPdup, missense APP mutations, sEOAD, sLOAD, DS and controls, but not significantly so. Nonetheless, APOE ε4 homozygosity was more commonly associated with type 3 CAA than types 1 and 2 CAA, and was associated with a greater severe of CAA overall in sLOAD but not in sEOAD or DS.
In a recent study, Head and colleagues [
15] compared the overall extent of CAA, atherosclerosis and arteriolosclerosis in 32 individuals with DS, ranging in age from 43 to 70 years, and in 80 individuals mostly with late onset sporadic AD (sLOAD) and 37 controls. Younger patients with sEOAD and
APP mutations were not specifically investigated in this latter study. Nonetheless, like ourselves, these authors found that CAA occurred at significantly higher frequencies in the brains of individuals with DS compared to sLOAD cases and controls, with the DS cohort being 1.2 times more likely to have CAA relative to sLOAD cases, and 4.6 times more likely to have CAA compared to control cases. On the other hand, atherosclerosis and arteriolosclerosis were rare in cases with DS. Such observations of significantly more frequent CAA, and a greater severity of CAA when present, in people with DS relative to sLOAD and control cases are consistent with the hypothesis that such changes are driven, at least partially, by an overexpression of
APP. The lack of AD neuropathology and CAA, even at greater than 70 years of age, in rare cases of DS with partial trisomy 21 where
APP is not overexpressed [
12,
35] would be consistent with this argument.
Nonetheless, how an overexpression of
APP might be translated into enhanced CAA remains unclear, but could involve deficiencies in clearance mechanisms when faced with such an overload of Aβ. In this latter context, it has been postulated that the strong association between age, CAA and AD pathology in the general population is driven, at least partially, by an impaired efficiency of cerebral vessels in later life in expelling extracellular fluid containing soluble forms of Aβ as a consequence of atherosclerosis/arteriosclerosis within such vessels [
49]. Nonetheless, individuals with DS appear to be protected against hypertension [
1] and atherosclerosis, and show less cerebrovascular pathology typically associated with cardiovascular risk factors, including atherosclerotic lesions and arteriolosclerosis [
15,
22]. Paradoxically, this ought to result in a better preservation of perivascular drainage channels in DS, and consequently a less severe, rather than more severe, CAA compared to sLOAD. Potential inefficiencies in perivascular drainage might not appertain to individuals with
APP mutations or DS since these would only be anticipated to occur beyond an age at which most would generally survive to. However, Aβ can also be cleared from the brain through several other routes, involving endocytosis by microglia and astrocytes, or enzymatic degradation. It is, therefore, possible that failures in these latter pathways, in conjunction with the overexpression of APP, might foster an inability to expel Aβ and result in severe CAA.
Despite the commonality of sharing duplication at the
APP locus, and an overexpression of APP protein, there are clear differences in clinical presentation between
APPdup and DS individuals. Stroke and ICH are the main clinical consequences of vascular amyloidosis in
APPdup, occurring in at least one-third of all cases, but both are uncommon in individuals with DS, with only a handful of case reports of this in the literature (see [
5]). Indeed, it has been estimated that haemorrhagic stroke occurs in only about 3–4% of older people with DS [
43], some ten times less frequent than in patients with
APPdup. Why these clinical differences should occur is unclear, given that DS differs from
APPdup only in the number of other genes located on chromosome 21 that are also triplicated. In most instances of DS, a full triplication of chromosome 21 is present, whereas in
APPdup triplication of
APP locus is variable, generally ranging from 0.5 to 6.5 Mb [
48], with the region of triplication in some instances being limited to
APP gene alone [
41], while in others it may extend to include up to 12 other genes [
37]. This raises the question as to whether the possession of other triplicated genes in some way confer some degree of protection against the likelihood of stroke or ICH in DS. These genes may be involved in the production of Aβ, of which there are two main pathways—the secretory pathway or the endo-lysosomal pathway. In the latter, Aβ cleavage occurs in the endosomal compartment where pH is optimal for β-secretase activity. Enlargement of the early endosomal compartment is considered one of the earliest morphological alterations detectable in postmortem tissues in sporadic AD, in most
APP mutations and in DS [
8].
APP overexpression is implicated in the formation of enlarged endosomes, but the mechanism is uncertain. Interestingly, it was recently shown that increase of β-CTF, the C-terminal fragments of APP generated after β-secretase cleavage, can produce enlarged endosomes in fibroblasts from DS individuals [
20] while in lymphoblastoid cell lines from individuals with
APPdup endosomes appeared to be of normal size [
9].
However, with regards to clinical phenotype, it is notable that in the Dutch family [
41] where triplication of
APP locus was restricted to
APP gene alone, and in the Swedish family the region of triplication extended to cover only those two genes either side of
APP, no instances of ICH were reported [
48]. Conversely, ICH commonly occurred in a Finnish family with 0.55 Mb duplication covering
APP and four other neighbouring genes [
38]. By contrast, in the French families where there is a greater and more variable extension of triplication of
APP locus [
6,
37], ICH occurred in about 30% of patients though, notably, dementia but not ICH defined the clinical course of the four
APPdup patients studied here in whom the region of triplication was 0.78 Mb [
6]. Hence, factors other than size of
APP triplication per se may account for the low prevalence of ICH in DS compared to
APPdup.
Alternatively, it may be differences in the actual CAA phenotype present that are responsible. Type 1 CAA tended to be more common in DS and sLOAD, and type 2 CAA was more common in missense
APP mutations and sEOAD than in the other groups. Type 3 CAA was more common in individuals with
APPdup than those with DS as well as others with sEOAD or missense
APP mutations. Type 4 was only seen in patients with
APP692 mutation, though 2 of the
APPdup patients (patients #1 and 3), although designated as CAA type 3, showed a particularly severe phenotype that was reminiscent of type 4 CAA. Consequently, patients with
APPdup in the present study had a more severe CAA phenotype than most individuals with DS, despite the similar ages, with only 30% DS individuals showing this type 3 phenotype, and then not to the same degree of severity as in patients with
APPdup. It is possible, therefore, that it is the lesser extent of CAA in DS (compared to
APPdup) that lowers the risk of CAA-related stroke and haemorrhage in such individuals. Correlations between CAA severity scores and CAA phenotype suggest that the different CAA phenotypes exist on a continuum with type 1 being the least ‘aggressive’ form, and types 2–4 following progressively as Aβ becomes deposited further along the arterial tree reaching into parenchymal arteries and arterioles (type 2) and finally into capillaries (types 3 and 4). Certainly, this would accord with the proposal put forward by Weller et al. [
49] based on a progressive slowing of perivascular drainage leading to build up of deposits in vessel walls. However, if so, this would not explain the relative absence of amyloid plaques in types 3 and 4 CAA, and the highest levels of plaques in type 1 CAA.
APP is processed by proteolytic enzymes known as secretases. Cleavage within the APP domain containing the Aβ sequence by α-secretase precludes its formation, whereas sequential cleavage at the amino- and carboxyl- termini of the Aβ sequence by β- and γ-secretases, respectively, releases Aβ into the extracellular fluid of the brain. Most of this occurs as a more slowly aggregating form, Aβ
40, compared to the longer, more rapidly aggregating form, Aβ
42(3). It has been shown that in both AD and DS the Aβ in CAA is composed mostly of Aβ
40, whereas in plaques it is Aβ
42(3) that predominates [
17,
18,
40,
44]. The differential pattern of composition of Aβ within brain parenchyma and blood vessel walls can be explained by the relative aggregation properties of Aβ
40 and Aβ
42(3), with the less aggregation prone Aβ
40 travelling further along perivascular drainage channels and ultimately reaching blood vessel walls, compared to the less abundant though more rapidly aggregating Aβ
42(3) which coalesces into plaques within the brain parenchyma. In familial AD due to certain missense mutations in
APP (for example those at or around codon 717) proteolytic processing of APP elevates levels of Aβ
42(3) relative to Aβ
40 [
39] with excessive numbers of Aβ
42(3) containing plaques being formed as a consequence [
26]. Other missense mutations (for example, those around codons 692 and 693) enhance the aggregation properties of both Aβ
40 and Aβ
42(3) without affecting levels of production. As mentioned earlier triplication at
APP locus will increase production of both Aβ
40 and Aβ
42(3).
Hence, the different CAA phenotypes seen here in the different forms of AD might in some way reflect the relative proportions of Aβ
40 and Aβ
42(3) being generated. In
APPdup and DS, where excess amounts of Aβ
40 (and Aβ
42(3)) are produced, this could lead to failure to expel this from the extracellular fluid leading to a massive build up in smaller arteries and capillaries evidenced as the more extensive type 3 CAA. In missense
APP mutations such as those at codon 692, the mutated, more highly aggregation prone, form of Aβ
40 would promote its excessive deposition in vessel walls, and again result in the severe type 4 CAA (see [
26] for
APP693 mutations), whereas no excess of Aβ
40 is generated in missense mutations in
APP occurring around codon 717, and in these cases the less extensive type 1 and 2 CAA predominate. In sEOAD and sLOAD, where normal levels of both Aβ
40 and Aβ
42(3) are produced, the different CAA phenotypes present might reflect the relative efficiencies in which Aβ
40 is cleared through the perivascular drainage channels. Present observations that all blood vessels stained for CAA by 4G8, irrespective of genetic or pathological group, or CAA phenotype, also appeared to be strongly immunoreactive for Aβ
40 but less strongly for Aβ
42(3). On the other hand, all plaques detected by 4G8 were strongly immunoreactive for Aβ
42(3) but only a subset (of cored plaques) appeared to contain Aβ
40. Such findings are consistent with our previous studies in familial AD and DS [
17,
18]. Consequently, in addition to potential differences in Aβ production, differences in CAA phenotypes between DS and
APPdup might also involve amyloid clearance, but alternative mechanisms could involve a unique oxidative stress profile or immune response in DS [
52].
Interestingly, as others have shown [
34,
40,
46], possession of two copies of
APOE ε4 allele was associated with a greater severity of CAA in sLOAD patients alone. However,
APOE genotype per se did not greatly influence the actual CAA phenotype in any pathological group. Although there was a numerical increase in ε4 allele frequency from type 1 to type 3 CAA in sEOAD and sLOAD, these differences were not substantiated statistically. Nonetheless, it is known that an increase in ε4 allele copy from 0 to 2 is associated with higher levels of Aβ
40 deposition (as plaques) in sLOAD [
13,
27], possession of ε4 allele/ApoE E4 isoform decreases brain clearance of Aβ [
7], and ApoE E4 isoform promotes fibrillogenesis [
23]. In these respects, possession of
APOE ε4 allele/E4 isoform could potentiate the development of CAA, as type 3 CAA, in those patients with sLOAD bearing
APOE ε4ε4 genotype, a suggestion in keeping with previous studies [
46]. The absence of type 3 CAA in elderly controls and missense
APP mutations involving codon 717, would also accord with the relative infrequency of
APOE ε4 allele and ε4ε4 homozygosity in such individuals. However, having said that, overall severity of CAA and prevalence of type 3 CAA were equivalent in DS individuals as in patients with sEOAD and sLOAD, despite there being in DS a relative lack of
APOE ε4 alleles, and a complete absence of ε4ε4 homozygotes.
While there were no significant differences in overall plaque scores between sEOAD and sLOAD cases, overall CAA scores were lower in sLOAD than sEOAD. Furthermore, age at death did not vary significantly between CAA phenotypes for these two groups though the proportion of cases with the less severe type 1 CAA was greater in sLOAD than sEOAD, suggesting that advancing age per se may, if anything, lessen the severity of CAA and phenotype present, at least in AD. This observation was not due simply to a shorter duration of disease in sLOAD compared to sEOAD which might potentially have terminated disease progression at an early stage in the later onset cases.
Hence, the factors that determine CAA phenotype are complex and remain unclear, possibly involving differential levels of production or clearance of Aβ
40 or shorter sized peptides, or factors which promote its aggregation such as ApoE E4 isoform, or some combination of all of these. Moreover, why there should be a distinction in CAA phenotype profiles between DS and
APPdup is puzzling and it is curious why sEOAD should be more strongly associated with type 2 CAA compared to sLOAD (and vice versa for type 1 CAA), but differential possession of
APOE ε4 allele does not appear to determine this. Interestingly, previous studies [
28] have shown type 2 CAA to be particularly common in early onset familial AD associated with
PSEN-
1 mutations, especially in those where the mutation is located after codon 200, in the absence of any
APOE ε4 allele modifying effect. Possibly sEOAD shares some genetic or mechanistic risk affinity with such
PSEN-
1 mutations, which ultimately translate into a similar CAA phenotype. The neuropathological differences between the different forms of AD highlighted in this study require further study to elucidate the underlying mechanisms. The scientific value in knowing what CAA phenotype is present will help to reduce variability of findings, and provide greater consistency of results, when factors relating to promotion of CAA are being investigated.