Introduction

Alzheimer’s disease (AD) is a neurodegenerative disorder that leads to cognitive decline [5, 85]. Definite AD diagnosis depends on the postmortem neuropathological analysis of brain sections and the detection of extracellular plaques that consist of amyloid β-peptide (Aβ) fibrils as well as intracellular neurofibrillary tangles (NFTs) with abnormally phosphorylated τ-protein [5, 47, 58, 81]. The clinical diagnosis of symptomatic AD in living patients is currently based upon signs of dementia and positive AD biomarkers, such as amyloid positron emission tomography (PET) and pathological Aβ or τ-levels in the cerebrospinal fluid [25, 85]. This symptomatic phase of the disease is preceded by preclinical AD (preAD), which can be discriminated from non-AD cases by the presence of the above-mentioned biomarkers [25, 125, 145]. Pathologically defined preclinical AD (p-preAD) encounters all cases that exhibit AD pathology at autopsy, i.e., amyloid plaques and NFTs, but did not exhibit any signs of cognitive decline during life [102, 133].

Aβ is a 36–43 amino acid peptide that results from β- and γ-secretase-mediated cleavage of the amyloid precursor protein (APP) [48, 65, 99, 106]. The most abundant forms of Aβ are the 40- and 42-residue peptide variants Aβ1–40 and Aβ1–42. In addition, several post-translational modifications of Aβ have been identified, including N-terminal truncations and pyroglutamate modifications at residues 3 or 11 (AβN3pE and AβN11pE) as well as phosphorylation at Serine residue 8 and 26 (pSer8Aβ and pSer26Aβ) [69, 89, 111, 112]. Although Aβ occurs mainly in the extracellular space, it can also be found within neurons [36, 42, 153]. Intraneuronal Aβ aggregates may colocalize with tau and contain AβN3pE or pSer8Aβ. They could also impair neuronal function and correlate with neurodegeneration [41, 72, 104, 128, 152].

Aβ peptides have intrinsic tendency to self-assemble into a range of different aggregates that are termed oligomers, protofibrils or mature amyloid fibrils based on their appearance by electron or atomic force microscopy [38, 52, 75]. Another way of defining different forms of Aβ extracted from human brains is based on the separation of soluble and insoluble fractions by differential ultracentrifugation and centrifugation steps with different solvents [121]. The species found in the Tris-buffered supernatant after 175,000×g ultracentrifugation will be referred to as ‘soluble Aβ’ and those in the pellet as ‘insoluble Aβ’. The ‘dispersible Aβ’ fraction belongs to the insoluble Aβ fraction but remains in the supernatant of Tris-buffered brain homogenates after centrifugation at 14,000×g [101]. ‘Membrane-associated Aβ’ is the other part of the ‘insoluble Aβ fraction’ that remains in the pellet after 14,000×g centrifugation and requires extraction with sodium dodecyl sulfate (SDS) or Triton X. Plaque-associated Aβ can only be recovered from the 14,000×g pellet by formic acid treatment [101, 121]. Aβ plaques as well as all types of soluble and insoluble non-plaque-associated, dispersible and membrane-associated Aβ oligomers, protofibrils and fibrils are found in symptomatic AD as well as in p-preAD cases [7, 96, 102, 137, 139]. A detailed protocol for the centrifugation steps required to separate the soluble, dispersible, membrane-associated and plaque-associated fractions can be found in the literature, e.g., Rijal Upadhaya et al. [102].

Here, we will review the neuropathological, molecular structural and biochemical aspects of Aβ aggregation and discuss its relation to AD neuropathology and the clinical stages of the disease.

Neuropathology of amyloid plaque deposition and cerebral amyloid angiopathy (CAA)

Amyloid plaques and CAA are morphologically detectable correlatives of Aβ aggregation in AD (Fig. 1), whereas soluble Aβ oligomers as well as dispersed Aβ oligomers, protofibrils and fibrils are usually not detected by immunohistochemistry with conventional anti-Aβ antibodies [38, 81, 139]. Figure 1a–e shows the most prominent plaque types occurring in the human brain with all relating to AD [91, 138]. The morphological appearance of distinct plaque types is related to their anatomical distribution. For example, fleecy amyloid is restricted to the layers pri-α, pri-β, pri-γ of the entorhinal cortex, and to the CA1-subiculum region, while lake-like amyloid occurs exclusively in the presubicular region [138]. CAA can occur in all types of vessels including capillaries (Fig. 1f, g). However, amyloid plaques and CAA are not specific for symptomatic AD and can be seen in non-demented individuals [7, 96, 137, 139]. For the neuropathological diagnosis of AD, as published by the National Institute of Aging and the Alzheimer Association, all cases in which Aβ plaques are found in the brain are to be diagnosed with AD pathology regardless of their clinical status [91]. However, we classify non-demented cases with pathologically detectable AD pathology as p-preAD cases in contrast to symptomatic AD cases [102, 133, 139]. This definition does not imply that all of these cases would have necessarily converted into symptomatic AD given the chance to live longer but rather describes that non-demented patients can have AD pathology.

Fig. 1
figure 1

Different types of Aβ plaques in the human brain [138]. All are today considered to represent AD pathology [91]. a Diffuse and cored plaques. b Presubicular lake-like amyloid. c Subpial band-like amyloid (arrows). d Fleecy amyloid (arrows). e Neuritic plaques (arrowheads), diffuse non-neuritic plaque (arrow). f Artery with Aβ deposition in the vessel wall, i.e., CAA. g Capillary Aβ deposition (arrows). Stainings: a–d, f, g anti-Aβ17–24 (4G8; Covance, Dedham, USA); e anti-Aβ8–17 (6F/3D, Novocastra, Newcastle upon Tyne, UK) counterstained with Gallyas silver staining, calibration bar in g valid for: a 30 µm, b 160 µm, c 55 µm, d 145 µm, e, g 25 µm, f 45 µm. CA1 Ammon’s horn sector CA1, ER entorhinal regions, OCC occipital isocortex, TI temporal isocortex

Full size image

Neuritic plaques (Fig. 1e) comprise a distinct subgroup of amyloid plaques that is characterized by the combined occurrence of Aβ deposits and dystrophic neurites [22, 90]. These plaques have been considered to have specific pathological value for AD and are, therefore, included in the neuropathological criteria for the postmortem diagnosis of AD [58, 90]. Two types of dystrophic neurites can be associated with neuritic plaques: (1) APP-containing dystrophic neurites and (2) paired helical filaments containing dystrophic neurites consisting of abnormal τ protein [22, 148]. Neuritic plaques can contain both or exclusively either of these types of dystrophic neurites [22, 148], whereby τ-positive neuritic plaques appear more frequently than τ-negative neuritic plaques [28]. Since neuritic plaques occur later in the development of AD-related pathology than diffuse, non-neuritic Aβ plaques [45, 132, 135, 138], it is in our opinion more likely that the occurrence of dystrophic neurites within neuritic plaques represents a reactive lesion due to axonal damage. Similarly, APP-positive dystrophic neurites are observed after head trauma, brain infarction or artificial brain tissue damage by laser irradiation [43, 63, 84, 119]. A further argument that Aβ aggregates could trigger APP-positive dystrophic neurites is the finding in APP transgenic mice that CAA lesions attract sprouting of APP-positive dystrophic neurites [95]. Interactions between Aβ and APP with abnormal τ are likely to ensue because APP accumulation and τ aggregation can occur in the same dystrophic neurite [28, 136] and because dystrophic neurites develop in the close vicinity of extracellular Aβ [22, 23, 154]. At synapses a colocalization of intracellular τ and Aβ has been described [15, 128] probably pointing to the synapse as a critical anatomical correlative for the AD-related neurodegeneration process with synapse loss as a well-known neuropathological feature of AD [21, 130].

Amyloid plaque pathology as well as CAA pathology usually starts in neocortical brain regions before they expand first into allocortical areas and then into the rest of the brain [134, 137]. Neocortical amyloid plaques define the first phase of amyloid plaque pathology, while phase 2 shows an additional involvement of allocortical areas, such as the entorhinal cortex, hippocampus, and cingulate gyrus. In phase 3, further amyloid plaques become detectable in the striatum, hypothalamus, thalamus, and the basal forebrain, while phase 4 shows additional plaque pathology in the midbrain and the medulla oblongata. Finally, in phase 5, plaques are also seen in the cerebellum and the pons [137] (Fig. 2). Similarly, stage 1 of CAA begins in the cortical and leptomeningeal vessels of neocortical areas and then expands into allocortical regions and the cerebellum (CAA stage 2). Finally, vascular Aβ deposits spread into vessels of the basal ganglia, diencephalon, brain stem and/or the white matter (CAA stage 3) [134] (Fig. 2).

Fig. 2
figure 2

Schematic representation of the expansion of Aβ plaques (Aβ phases) [137], vascular Aβ deposition in CAA in the human brain (CAA stages) [134] and biochemical maturation stages of Aβ aggregates as seen in the soluble, dispersible, membrane-associated, and plaque-associated fractions of brain homogenates (B-Aβ stages) from p-preAD and symptomatic AD cases [102]. For Aβ plaque deposition newly involved brain regions are marked in red, already affected areas are painted in black. For CAA newly involved brain regions are marked in blue, already affected areas are black. Except for the cerebellum the regional expansion patterns of Aβ plaques and CAA are quite similar. Parts of this figure are reproduced with permission [102, 137]

Full size image

The progression from p-preAD to the symptomatic phase of AD is associated with the spread of Aβ pathology and changes in the composition of Aβ aggregates. Biochemically, Aβ42 is the first Aβ species (Fig. 3) to accumulate in the human brain [59, 76]. Aβ40 is detected subsequently, followed by N-terminal truncated and pyroglutamate-modified AβN3pE and/or AβN11pE (Fig. 3; Table 1). These modified forms of Aβ are frequently detected in plaques of p-preAD cases and in all AD cases [59, 60, 76, 102]. In some studies, AβN3pE and Aβ42 occurred together within plaques in every case [60]. However, in our sample (n = 74) we found few p-preAD cases (n = 3) with Aβ plaques that did not exhibit AβN3pE (Table 1). pSer8Aβ in plaques was less frequently observed in p-preAD cases than AβN3pE and AβN11pE (Figs. 2, 3). It was mainly restricted to symptomatic AD cases [102]. The α-secretase-cleaved P3 fragment (Fig. 3) was also found mainly in plaques of symptomatic AD patients [60] (Table 1). Biochemical extraction from human neocortex homogenates also revealed a similar differential distribution of Aβ and its modified forms in soluble, dispersible and membrane-associated Aβ aggregates [102]. Notably, the occurrence of modified forms of Aβ in biochemical isolates of soluble and insoluble Aβ aggregates correlated well with its neuropathological detection in plaques following a specific sequence: (1) non-modified Aβ, (2) AβN3pE and (3) pSer8Aβ. This sequence of Aβ aggregate maturation was considered to represent sequential changes in the biochemical composition of Aβ aggregates allowing the identification of three stages of the biochemical composition of Aβ aggregates (B-Aβ stages) [102] (Fig. 2).

Fig. 3
figure 3

Cleavage of APP and production of Aβ, P3 by β-, γ-, and α-secretase cleavage and its modified forms. The distinct Aβ species are ranked by their sequence of occurrence in amyloid plaques. The species found in preclinical (p-preAD) and clinical AD (AD) cases are indicated

Full size image
Table 1 Aβ, AβN3pE, AβN11pE, pSer8Aβ and P3 in plaque maturation
Full size table

Mechanism of Aβ aggregation and structure of Aβ fibrils and intermediates

Mature amyloid fibrils are the major compound of amyloid plaques or Aβ-derived deposits in CAA [38, 81]. These aggregates represent the terminal states of the Aβ fibrillation process, at least in vitro [18, 92]. Fibrils have a width of ~10–20 nm and a length of usually more than 1 μm [110]. Cryo transmission electron microscopy (TEM)-based reconstructions of the three-dimensional electron densities of in vitro-formed Aβ fibrils revealed one or several protofilaments that construct the full-scale fibril. These protofilaments share a relatively conserved cross-sectional architecture in Aβ1–40 and Aβ1–42 fibrils [108, 118], which suggests the presence of similar conformations among the different fibrils. The fibrils are constructed from intermolecular β-sheets in which peptides are arranged with β-strands. These strands are oriented perpendicular to the main fibril axis with the backbone hydrogen bonds located parallel. This type of assembly is termed a cross-β structure. The Aβ residues forming this structure are located within the peptide center and at the C-terminus (approximately residues 16–20 and 31–36) [31]. The N-terminus is conformationally flexible in many, but probably not all types of fibrils [31]. These data have come from the analysis of Aβ filaments that were formed in vitro, while the detailed structures of fibrils from AD brain tissue have remained largely elusive. Fibril polymorphism is another phenomenon, which is also mainly known from in vitro-formed Aβ fibrils. Aβ has been shown to adopt multiple fibril structures that can even be observed within the same sample tube [39, 86]. Polymorphic fibrils can differ in the number of their protofilaments, the relative protofilament–protofilament orientation or in their detailed peptide conformations [30]. Changing the conditions of fibril formation may affect this fibril spectrum and create broader or narrower outcomes or even induce alternate fibril structures [67]. An interesting implication of the observation of different fibril morphologies is that it raises the possibility of a strain-like behavior with Aβ aggregates. The term strain as used in the prion field refers to different phenotypic traits of prion protein aggregates in transmissible spongiform encephalopathies that are transmissible from donor to recipient upon prion infection [1]. Prion strains may have their molecular basis in different prion protein conformations and/or aggregate structures. Observation of Aβ fibril morphologies suggests that different biological effects could arise from differently structured fibrils and that these may be potentially transmissible, at least under laboratory conditions. Indeed, there is evidence for such a strain-type behavior of Aβ aggregates in APP transgenic mouse models for Aβ pathology [55, 127]. One study compared Aβ fibril structures in brain homogenates of two human AD cases and reported case-specific Aβ fibril structures distinguishing between the two individual cases [79], indicating support for strain-type behavior of Aβ in animal models.

Amyloid fibrils arise from a complex self-assembly reaction. Monomeric Aβ peptide is the theoretical precursor of all aggregates and has a random coil-like conformation in aqueous solutions at low peptide concentrations as well as in the absence of salts, as demonstrated by circular dichroism spectroscopy [131]. Molecular dynamic simulations found such Aβ monomers to adopt a collapsed micelle-like conformation where hydrophobic side chains are located mainly within the interior of the otherwise rather flexible peptide conformation [143]. In vitro, monomeric Aβ may be stable for sometime in strong denaturants, such as high molar concentrations of guanidine hydrochloride or neat trifluoroacetic acid. However, it is not very stable in aqueous solutions which mimic physiological conditions by their salt, lipid or sugar composition, and are prone to aggregate into β-sheet conformations. This process is exaggerated if the peptide concentration is relatively high [56]. These effects should be kept in mind when Aβ is claimed to be present as monomers. Apolar environments, such as detergent micelles or lipid bilayers, induce α-helical conformations [131] which reflect the natural origin of the Aβ sequence within the α-helical transmembrane domain of APP.

Monitoring the kinetics of Aβ fibril formation in vitro typically shows three phases: an initial lag phase of little fibril formation, a subsequent growth phase of rapid fibril assembly and a final stationary phase where fibril formation and fibril dissociation are at equilibrium (Fig. 4). Mathematical modeling of experimental kinetic data suggested that three main steps account for these kinetic effects: primary nucleation, fibril elongation and secondary nucleation [68]. Fibril nucleation describes the slow initial assembly of Aβ peptide into nuclei that allow the fast subsequent outgrowth of elongated fibrils by monomer addition. Secondary nucleation describes the ability of already formed fibrils to potentiate the formation of new fibrils. This amplification occurs because preformed fibrils disintegrate and expand the number of active sites that enable fibril elongation or alternatively, the surfaces of the preformed fibrils act as scaffolds to accelerate the generation of new fibril nuclei which then elongate into additional filaments [68]. Experimental evidence which supports the effectiveness of a nucleation–polymerization mechanism is provided by studies in which the addition of preformed fibrils was found to strongly accelerate fibril formation in vitro and to reduce or to eliminate the observable lag phase [51] (Fig. 4).

Fig. 4
figure 4

Schematic representation of the kinetics of fibril formation. Blue spontaneous fibril formation from monomeric precursors; red fibril formation, seeded by preformed fibrils added to the solution of monomers

Full size image

Interestingly, several of these general features of fibril formation reactions in vitro are remodeled by cell culture systems of amyloid plaque formation suggesting the effectiveness of similar mechanisms. Measurement of the formation of single Aβ amyloid plaques in a cell model revealed a growth kinetics consisting of lag, growth and a stationary phase [32]. In addition, plaque formation can be accelerated by the addition of preformed amyloid fibrils to the cell model of Aβ plaque formation [32] or by injection of such material into the brain or peritoneum of Aβ-producing mice [27, 64, 87, 149]. So far, only very little is known about the molecular structure of these fibrillation seeds but it is likely that they capture structural features seen in fibrils.

While monomeric peptides and fibrils mark the starting point and the end products of fibrillation, ‘intermediates’ represent the structural states in between these two. Intermediates either occur in the course of fibril assembly or they are structurally in between a monomer and a fibril. Investigating the time course of Aβ peptide assembly into fibrils with TEM or atomic force microscopy, in vitro, revealed a range of different intermediates from protofibrils to oligomers with low and high molecular weights [29]. In addition, there are several protocols reported to prepare specific intermediates inside the test tube or to extract them from AD brains [49]. Detailed nuclear magnetic resonance-based insights into the structure of such intermediates are so far only available for a few oligomer preparations [2, 20, 53] and for protofibrils [114]. Furthermore, there are data indicating a β-barrel assembly for certain oligomeric states [74].

There are several specific oligomers, e.g., Aβ dimers, Aβ trimers, Aβ*56, and Aβ globulomers, that received special interest and were considered to have specific toxic properties that other intermediates may not have [8, 35, 57, 77, 78, 83, 121]. Some oligomeric Aβ intermediates enter the fibril-forming pathway, e.g., Aβ dimers, whereas others will accumulate off-pathway as stable non-fibrillar oligomers, e.g., Aβ trimers or globulomers [35, 40, 82, 94]. Aβ intermediates, are particularly interesting as they possess a higher specific in vitro toxicity than Aβ fibrils [37]. One argument against the critical role of Aβ plaques or fibrils in the development of AD is that NFTs provide a better neuropathological correlate to the progression of clinical deficits than Aβ plaques [6]. In the light of (1) end-stage Aβ plaque pathology in all symptomatic AD cases [137], (2) the significant increase in soluble oligomers and dispersible Aβ aggregates from p-preAD to symptomatic AD cases [102, 139] and (3) the toxic properties of dispersible Aβ aggregates in APP transgenic mouse models [101] soluble or dispersible Aβ oligomers or protofibrils rather than Aβ plaques seem to be responsible for Aβ toxicity and to contribute to disease progression. Moreover, some, though not all, oligomers appear to be capable of accelerating the lag phase for Aβ fibril formation [10]. Nevertheless, it is likely that fibrils and plaques also contribute to AD as they are able to release toxic Aβ intermediates [113]. In favor of this hypothesis, it has been considered that plaque-associated oligomers or protofibrils induce neuritic changes near amyloid plaques in APP transgenic mouse models [11, 126, 140]. As such, the current data appear to support a critical role for Aβ intermediates such as oligomers and protofibrils in AD pathogenesis probably in “collaboration” with NFTs. Whether Aβ intermediates and τ pathology have impact on one another, whether tau or Aβ is the main driver of the disease is still matter of discussion [6, 13, 50, 124]. Anyway, considering the reports on Aβ and τ effects on neurotoxicity there is strong evidence that both pathologies have impact on the disease [6, 13, 50, 71, 75, 120, 124].

However, it is not yet clear whether Aβ intermediate toxicity is restricted to distinct oligomers, e.g., Aβ dimers or Aβ*56 [78, 83] or whether all kinds of high molecular weight intermediates seen by blue-native-polyacryl gel electrophoresis in symptomatic AD cases [103] contribute to neurodegeneration. Additionally, fibrillar Aβ in CAA-affected vessels appears to damage the vessel wall [17, 144] and to alter perivascular drainage [151] as a consequence of the displacement and consecutive destruction of the vessel wall just by its physical presence. This is consistent with Aβ’s mechanical properties [109].

Modification of the Aβ peptide and its impact on aggregation and toxicity

The conformation, self-assembly and aggregation of Aβ is determined by its amino acid sequence. This is well documented for the two abundant species Aβ40 and Aβ42 [49, 146]. The longer Aβ42 variant aggregates much faster than Aβ40 and is the major species initially detected in AD-associated plaques [59, 105], even though the proteolytic processing of APP generates mainly Aβ40. Several mutations in APP have been associated with familial early onset AD. These mutations are located within or close to the Aβ domain of APP and favor the production and/or aggregation of Aβ [66, 129] (Fig. 5). Interestingly, another mutation localized close to the N-terminus of the Aβ domain (A673T) appears to be protective against AD and decreases the proteolytic generation of Aβ [62]. Thus, genetic evidence strongly supports a critical role of Aβ in the pathogenesis of AD. However, mutations in APP that cause familial AD are very rare and only account for a minority of all AD cases. Here, we focus on variants of Aβ that derive from the most abundant ‘wild-type’ form of APP by post-translational modifications and their effects on aggregate formation and their potential roles in AD pathogenesis (Fig. 6).

Fig. 5
figure 5

Localization of post-translational modifications and disease-causing mutations within and close to the Aβ domain. The amino acid sequence of Aβ42 (boxed) and neighboring N- and C-terminal sequences are given in single letter code. Positions of pyroglutamate (pE) and phosphoserine residues (P) are indicated by yellow triangles and red circles, respectively. Mutations associated with familial forms of early onset AD (black letters) or with cerebral hemorrhages or stroke and severe CAA pathology (red letters) are indicated below the amino acid sequence in single letter code. Notably, all identified disease-causing mutations in APP are localized within or close to the Aβ domain. These mutations commonly affect the proteolytic generation and/or the aggregation of Aβ. The table below lists the position of different mutations and their effects on production and aggregation of Aβ (?, effect not reported or unclear). The effects of the mutations on aggregation are predominantly deduced from in vitro experiments. Thus, it would be interesting to test how the mutations affect the solubility of the different aggregates in vivo. Mutation A673T is associated with a decreased risk of AD and considered to be protective (marked in blue). In vitro, this mutation decreased production of Aβ (Table is adapted from http://www.alzforum.org/mutations). The mutations outside the Aβ domain do not directly affect the aggregation propensity of the monomeric Aβ peptide itself, but by increasing the total production of Aβ or altering the ratio of Aβ42/40 promote oligomerization and fibrillization

Full size image
Fig. 6
figure 6

Potential functional links of Aβ maturation, propagation and increase in concentration. Propagation of Aβ plaque pathology as given by the phases of Aβ deposition is associated with an increase of the Aβ content in the brain (a) as well as with maturation of Aβ aggregates in a given regions (b, c) as depicted for the hypothetical propagation of Aβ pathology from the neocortex (b) to the entorhinal cortex (c) between Aβ phases 1 and 2 (blue box in a). Mature Aβ aggregates that are predominantly found in symptomatic AD cases consist of modified forms of Aβ that increase the stability of the aggregates and, in so doing, impair its clearance and finally cause further Aβ aggregation and increased toxicity of the Aβ aggregates

Full size image

A large variety of post-translational modifications has been shown for Aβ including non-covalent interactions with metal ions, lipids and other proteins or peptides [26, 54, 147]. The binding of Aβ to these ligands could modify the aggregation behavior of Aβ and hence, contribute to the pathogenesis of AD. The Aβ peptide itself can also be modified covalently by oxidation, racemization, isomerization, and as mentioned above by pyroglutamate formation and phosphorylation [73, 93, 111, 112, 122, 123].

We and others have shown that Aβ species with pyroglutamate and phosphoserine are abundantly present in both human AD and APP transgenic mice brains. Pyroglutamate-modified Aβ was initially identified in human AD brains [93, 111, 112]. The formation of pyroglutamate requires precedent trimming of the N-terminus of Aβ by exopeptidases or alternative cleavage of APP by endoproteases to expose the glutamate residue at positions 3 or 11 (Fig. 3). The enzyme glutaminyl cyclase catalyzes formation of a lactam ring resulting in Aβ with pyroglutamate at their newly formed N-termini. When compared to unmodified Aβ, AβN3pE has an increased propensity to form oligomeric and fibrillar assemblies [117]. The use of specific antibodies for AβN3pE revealed the common presence of these species in human AD and p-preAD brains [60, 102, 111, 112]. Studies with transgenic mice and cultured neurons also showed that AβN3pE exerts increased toxicity and deposits early in extracellular plaques and intraneuronal aggregates [4, 9]. In vitro, the lag phase for AβN3pE is much shorter in comparison to that for non-modified Aβ [115].

Aβ species phosphorylated at Ser8 have also been well documented in brains of transgenic mice and human AD cases [69, 72, 102, 104]. Ser8 is efficiently phosphorylated by purified cAMP-dependent protein kinase A (PKA) in vitro. Notably, an Aβ phosphorylating PKA-like activity has also been detected on the surface of primary cortical neurons and in human cerebrospinal fluids [69]. These data suggest that secreted Aβ could undergo phosphorylation by extracellular forms of PKA. In vitro studies using Thioflavin T or Congo red, showed that phosphorylation at Ser8 promotes the formation of Aβ fibrils. Kinetically, Ser8 phosphorylation decreases the lag phase of aggregation, indicating an increased formation of oligomeric nuclei. It is also interesting to note that oligomers of pSer8Aβ not only promote fibrillization of phosphorylated Aβ, but can also seed for the assembly of non-phosphorylated Aβ [69]. Studies with phosphorylation state-specific antibodies have shown co-deposition of phosphorylated and unphosphorylated Aβ species in both extracellular plaques and intraneuronal aggregates. However, pSer8Aβ appears to be selectively enriched in the core of individual plaques suggesting that it might also seed aggregation of other Aβ species in vivo.

In summary, phosphorylation of Ser8 has very similar effects on the aggregation behavior as the pyroglutamate modification strongly favoring the formation of oligomeric and fibrillar Aβ assemblies. Aβ can also be phosphorylated in vitro at Ser26 [89]. Notably, synthetic pSer26Aβ peptides assemble rapidly into “off-pathway” oligomers without proceeding to fibrillar aggregates [100]. It remains unknown if pSer26Aβ is also found in human brain and whether it plays an important role during AD pathogenesis.

The impact of Aβ modification on its aggregate stability and resistance against degradation

Aggregation of modified pyroglutamate and phosphorylated Aβ species results in structures that show, compared with monomeric Aβ, an increased resistance against proteolytic degradation or other clearance mechanisms of the brain. The N-terminal modification of Aβ by pyroglutamate formation could also decrease its proteolytic degradation by amino- or endopeptidases [116]. We have also shown that phosphorylation at Ser8 strongly decreased the cleavage of monomeric Aβ by the insulin-degrading enzyme (IDE) and angiotensin-converting enzyme [70]. Thus, these post-translational modifications increase the biostability of Aβ and eventually their concentration in the brain, which could further promote their aggregation. Whether pyroglutamate modification or phosphorylation could also affect other clearance mechanisms including: phagocytosis and subsequent intracellular degradation or the drainage via the blood brain barrier or the perivascular space, remains to be investigated.

Post-translational modifications of Aβ might also affect the physicochemical characteristics of oligomeric and fibrillar assemblies and thus, their dissociation during biochemical extraction from brains. We showed that pSer8Aβ extracted from brains of APP transgenic mice and separated by SDS-PAGE has increased stability under denaturing conditions as compared to unmodified Aβ [69]. The different characteristics of unmodified and modified Aβ species during extraction from brain tissue and the detection by Western blotting, ELISA or mass spectrometry have to be considered for quantitative analyses. Different reactivities of generic Aβ antibodies for modified and unmodified Aβ species in the monomeric or aggregated state could further complicate the interpretation of experimental results for quantification of different Aβ species.

Mechanisms for the progression of Aβ pathology in the pathogenesis of AD

As described above, Aβ pathology (1) expands from the neocortex into further brain regions following neuronal connections [134, 137], (2) follows a maturation sequence during which the concentration of soluble Aβ increases, post-translational modifications take place and insoluble aggregates accumulate [60, 76, 102], and (3) is associated with an increase in total Aβ amount [19, 102, 139, 141, 142]. However, the critical events that trigger or cause these processes are still under debate.

In the event that expansion of Aβ pathology from one brain region into another is critical for disease progression, the question still remains whether specific forms of aggregates are crucial for propagation of Aβ pathology. Expansion of Aβ pathology means that there is a specific sequence in which Aβ aggregates emerge in different brain regions as extracellular plaques in the brain parenchyma and in blood vessels. In vitro Aβ fibrillization as well as plaque formation in cell models is characterized by a lag phase in which there is only very little amyloid formation [32, 51]. Interestingly, addition of seeds composed of preformed fibrils can shorten this lag phase significantly (Fig. 4). Hence, such a seeding mechanism is a probable explanation for the observed expansion of Aβ plaque pathology from one brain region into the next. An argument favoring such a seeding hypothesis is the observation of a strain-type behavior in mouse models injected with distinct amyloid structures [55, 127] that remained detectable after expansion from one brain region into another [149]. Amyloid seeds might distribute in the brain by passive diffusion and/or by active transport in nerve or glial cells. Diffusion might be particularly relevant for soluble extracellular Aβ aggregates. It has been demonstrated that neuron-derived Aβ accumulates in CAA-affected leptomeningeal vessels without direct contact with the CNS neurons [14] and in cortex grafts from non-transgenic animals transplanted into the cortex of APP transgenic mice [88]. The finding that different brain regions connected with one another by axons become sequentially affected by Aβ plaque pathology [137] does not argue against diffusion as a major mechanism for Aβ propagation because axons may facilitate directed diffusion of Aβ aggregates. An alternative hypothesis is that cells carry and transport Aβ. Intracellular Aβ has not only been detected in neurons but also in astrocytes and microglial cells [3, 33, 42, 156]. Therefore, migrating glial cells could also contribute to the propagation of Aβ. Microglial cells and astrocytes can take up Aβ [80, 155] and are both capable of migrating into other brain regions [12, 61]. For astrocytes, migration along myelinated fiber tracts has been reported [12] which would be compatible with an expansion along axonal tracts. Thus, these cells may be able to distribute Aβ or its assemblies. For astrocytes, it has been shown that they are able to clear Aβ into the perivascular space [107]. Here, Aβ can interact with the vessel wall and ultimately result in the development of CAA [16, 151]. Cell–cell propagation of different Aβ species has been demonstrated in neuroblastoma cell culture [24] and may depend on the activity of exosomes that may contain Aβ or Aβ aggregates [98]. Whether microglial cells indeed contribute to Aβ propagation is not clear, as their deletion had no effect on plaque pathology in APP transgenic mice [44]. However, microglial cells can play an important role in AD pathogenesis as they can phagocytose Aβ and modulate neuroinflammatory processes in the brain [34, 46, 97, 150]. Thus, it is very likely that multiple mechanisms, such as passive diffusion and active transport by neurons and glial cells, contribute to the propagation of Aβ pathology throughout the brain. Therefore, it will be necessary to further dissect the relative contribution of these mechanisms to Aβ propagation and AD pathogenesis in the future.

The maturation of Aβ aggregates from deposits containing mainly non-modified Aβ into aggregates encompassing significant amounts of AβN3pE and pSer8Aβ is associated with the propagation of Aβ pathology from the neocortex to further brain regions as well as with the conversion from preclinical to symptomatic AD. Although current findings already suggest that the maturation stage of Aβ aggregates could be crucial for the toxic properties of Aβ, it will be important to further confirm whether the maturation stage of Aβ aggregates indeed determines their toxicity and if it critically affects AD pathogenesis. The presence of AβN3pE and pSer8Aβ may boost aggregation of Aβ [69, 117]. Our finding of an Aβ phase 3-like distribution pattern indicating expansion of Aβ plaques from the neocortex into other brain regions in a case with the earliest stage of biochemical Aβ aggregate maturation (B-Aβ stage 1), i.e., Aβ aggregates without detectable modified Aβ species may argue against an essential contribution of advanced Aβ aggregate maturation to the spreading and seeding of Aβ pathology. Another case exhibited mature Aβ aggregates (B-Aβ stage 3) containing Aβ, AβN3pE and pAβ only in the neocortex (Table 2). These examples imply that modified Aβ forms are no prerequisite for the propagation of Aβ aggregates. Whether distinct “off-pathway” Aβ intermediates with high toxicity are produced at distinct steps in Aβ maturation is not clear. However, the overall increase of high molecular weight Aβ intermediates in AD cases, as seen by BN-PAGE analysis of brain lysates [103], coincides with the increase of toxic intermediates. It will be interesting to further assess whether AD-related neurodegeneration is driven by several specific Aβ intermediates or whether many kinds of fibrillar and non-fibrillar Aβ aggregates contribute to the development of the disease.

Table 2 Distribution of Aβ phases at given biochemical maturation stage (B-Aβ stage; a) and of B-Aβ stages at given Aβ-MTL phases (n = 40; b) as previously published [102]
Full size table

In addition to the maturation of soluble and insoluble Aβ aggregates, their concentration in the neocortex increased from p-preAD cases to AD cases [102, 139]. AβN3pE and pSer8Aβ have an increased stability against proteolytic degradation [69, 117]. Accordingly, aggregates containing these variants might also have a lower tendency to dissociate. Therefore, these post-translational modifications may decrease the clearance of Aβ and contribute to increased Aβ levels during AD pathogenesis. Increased concentration of Aβ aggregates may promote spreading of Aβ pathology and further accumulation of posttranslationally modified Aβ peptides. The increase in Aβ concentration, its maturation and expansion into further brain regions could be functionally linked to the pathogenesis of AD. Once the aggregation of Aβ is initiated, further seeding and maturation of Aβ aggregates may increase the total Aβ levels in the brain and thereby accelerate the propagation of Aβ into unaffected brain regions and promote disease progression.

Conclusions

In this review, we have demonstrated that the progression from the p-preAD to the symptomatic AD phase is accompanied by an increase in concentration, maturation and expansion (spreading/propagation) of Aβ aggregate pathology. Based upon biochemical and biophysical properties of Aβ together with neuropathological findings, we hypothesize that modified Aβ species such as AβN3pE and pSer8Aβ play critical roles for Aβ accumulation, plaque maturation, seeding, and propagation in the pathogenesis of AD and for its conversion from the preclinical to symptomatic stage. Accordingly, the maturation step of Aβ aggregates while converting from p-preAD to symptomatic AD, i.e., the additional accumulation of pSer8Aβ, may be an attractive target to inhibit the progression of p-preAD to the symptomatic stage. The simple view on Aβ plaque pathology at autopsy or by amyloid PET does not automatically permit conclusions on Aβ aggregate maturation status or on the concentration of soluble and dispersible Aβ aggregates. Accordingly, for monitoring Aβ pathology in clinical studies amyloid PET and postmortem neuropathological assessments might be supplemented by maturation stage and Aβ concentration as additional outcome measures.