Focusing the amyloid cascade hypothesis on N-truncated Abeta peptides as drug targets against Alzheimer’s disease

Alzheimer’s disease (AD) is a progressive neurodegenerative disorder characterized by the presence of extracellular amyloid plaques composed of amyloid-β (Aβ) surrounded by dystrophic neurites and neurofibrillary tangles. The discovery that certain early-onset familial forms of AD may be caused by an enhanced production of Aβ peptides, led to the hypothesis that amyloidogenic Aβ is intimately involved in the AD pathogenic process [88]. Aβ is derived by proteolytic cleavage of the β-amyloid precursor protein (APP) [89].

As early as 1992, Haass and colleagues [30] reported the unexpected identification of full-length Aβ and the p3 fragment in media from cultures of primary cells and APP-transfected cell lines grown under normal conditions. In addition, using in vivo micro-dialysis in mice, Kang et al. [43] found that the amount of Aβ in interstitial fluid (ISF) correlated with wakefulness. ISF Aβ was assessed in Tg2576 mice at 3 months of age, several months earlier than initial deposition of Aβ. They found diurnal variation of ISF Aβ levels with significant increases (+75 %) during the dark period compared to the light period. Despite fluctuations in ISF, total tissue hippocampus homogenates levels of Aβ, full-length APP, APP C-terminal fragments, Aβ_1–40 and Aβ_1–42 were not significantly altered between dark and light periods. This indicates that the pool of ISF Aβ is likely to be regulated independently from total intracellular and membrane-associated Aβ. The amount of ISF Aβ also significantly increased during acute sleep deprivation and during infusion of orexin, a neurotransmitter regulating arousal and wakefulness, but decreased with infusion of an orexin receptor antagonist. Moreover, cerebrospinal fluid (CSF) levels of Aβ were studied in ten young healthy male volunteers via lumbar catheters over a 33 h period and illustrated clear evidence of diurnal fluctuation of Aβ in the CSF. Aβ levels increased throughout the day with a peak in the evening that decreased overnight. The group also reported that the APPswe/PS1ΔE9 mouse model of AD showed normal sleep–wake cycle and diurnal fluctuation in ISF Aβ in the brain before Aβ plaque formation [75].

A physiological effect of the observed diurnal variation of ISF Aβ could be the known overall increase in synaptic strength during the day and synaptic depression during periods of sleep [20, 24, 100]. These findings are corroborated by results showing that neuronal activity modulates the formation and secretion of Aβ peptides in hippocampal slice cultures bearing neurons that overexpress APP. In addition, Aβ depressed excitatory synaptic transmission in neurons expressing APP, as well as nearby neurons that did not, leading to the assumption that activity-dependent modulation of endogenous Aβ production may normally participate in a negative feedback loop that could keep neuronal hyperactivity under control [42].

Long-term depression (LTD) represents an activity-dependent reduction in the efficacy of neuronal synapses and has been described in a variety of neurons. It has been shown that Aβ is capable of regulating the amount of surface NMDA-type glutamate receptors [92]. In addition, several parallels between LTD and Aβ-induced synaptic changes have been described. Aβ-induced synaptic depression partially mimicked metabotropic glutamate receptor LTD synaptic transmission. It has been hypothesized that this could be a normal physiological role of full-length Aβ [36]. Recent results suggest that conformational changes of the NMDA receptor (NMDAR), and not ion flow through the channel, are required for Aβ to produce synaptic depression and a switch in NMDAR composition [45].

It has been suggested that N-terminally truncated Aβ_5–x peptides are preferentially formed by an alternative cleavage of APP involving caspase activity [67]. These N-truncations were detected in Aβ deposits of sporadic and familial AD [29, 62, 63, 72, 90, 94]. Using mass spectrometry and Western blot analysis of sporadic AD and familial AD cases (M146V PS1 or KM670/671NL APP), Aβ_5–40/42 was one of the detected N-truncated species. Regarding transgenic mouse lines, mass spectrometry of immunoprecipitated Aβ peptides also provided evidence of the presence of Aβ_5–42, e.g. in APP/PS1KI mice [11]. Our group detected Aβ_5–42 peptides in the 5XFAD mouse model [110] and using immunohistochemistry in APP/PS1KI, 5XFAD and 3xTG transgenic mouse models [29]. Interestingly, neither 5XFAD nor APP/PS1KI showed any evidence for intraneuronal Aβ_5–x, which is in good agreement with the observation that this peptide is readily secreted [31]. One should note that all of the above-mentioned transgenic mouse models express the Swedish APP mutant prone to BACE cleavage at Asp-1 of Aβ. It is therefore unclear whether or not the Arg-5 truncation may possess neurotoxic properties, a matter of concern for therapeutic strategies involving BACE inhibitor treatment.

The precise enzymatic activities leading to the generation of the diverse N-terminally truncated Aβ peptides are in most cases not known in detail; however, several candidates have been proposed (Fig. 1; Table 1).

Aβ molecules secreted by MDCK cells exhibit extensive amino-terminal heterogeneity with >80 % of molecules containing an amino-terminus at the Arg-5 residue and only a minority of fragments initiating at Asp-1 [31]. In contrast to the results obtained in cells expressing wild-type APP, the cells expressing APP-695swe showed that the majority of Aβ peptides began at Asp-1, while only ~10 % began at Arg-5. The results indicated that Asp-1 is a preferred site for the β-secretase cleavage of APP-695swe [54].

N-truncation of Aβ_3–40 and Aβ_5–40 is facilitated by reduced endocytosis of APP in vitro, a requirement for BACE cleavage [12]. The generation of Aβ was analyzed in human embryonic kidney (HEK) 293 cell lines stably expressing wild type and non-internalizing mutants of human APP [12]. APP lacking the entire cytoplasmic domain or with both tyrosine residues of the motif GYENPTY mutated to Ala showed at least fivefold reduced endocytosis. In these cell lines, the production of Aβ_1–40 was substantially reduced but accompanied by the appearance of two prominent alternative Aβ peptides differing at the amino termini which were identified as Aβ_3–40 and Aβ_5–40.

Portelius et al. [73] studied the Chinese hamster ovary cell line 7PA2 stably transfected with the 751 amino acid APP isoform harboring mutant V717F. Treatment of the cells with a BACE1 inhibitor decreased the abundance of the Aβ monomer band and resulted in lower levels of Aβ_1–40, Aβ_1–42 and secreted APP. Western blot bands thought to represent oligomers of Aβ increased in response to BACE1 inhibition. This increase was paralleled by the emergence of N-terminally truncated Aβ_5–40 in particular. Treatment of cell cultures and dogs with BACE1 inhibitors significantly reduced Aβ peptides starting at Asp-1, while amino-terminally truncated variants such as Aβ_5–40 increased [59, 83, 94]. Based on data from treatment of human neuronal and non-neuronal cells expressing wild-type APP with inhibitors of BACE and α-secretase in vitro, it has been proposed that Aβ_5–40/42 might be derived from alternative β-cleavage of APP by α-secretase-like protease(s) [94]. One has to consider though that the expression of a mutation within the APP gene and/or the cell types used can influence the variant and quality of the N-terminally truncated Aβ.

Incubation of HEK293 cells overexpressing APP containing the Swedish mutation with the aminopeptidase inhibitor amastatin revealed significantly increased levels of full-length Aβ in the supernatant. This led to the identification of aminopeptidase A as a candidate enzyme cleaving the N-terminal Asp-1 residue [91]. Very recently, meprin-β has been proposed as another enzyme with the ability to process Aβ peptides [6]. It has been demonstrated that this enzyme cleaves full-length APP in a β-secretase manner, leading to the generation of Aβ_2–x peptides, which have been previously described in AD patients [56, 104]. One of the major Aβ-cleaving proteases is the zinc-metalloprotease neutral endopeptidase or neprilysin (NEP) [38]. High-performance liquid chromatography in combination with mass spectrometry analysis identified several cleavage sites when Aβ_1–40 peptides were incubated with NEP. Among other truncations, NEP generates N-terminal truncated Aβ peptides by cleavage between Arg-2 and Glu-3 or between Glu-3 and Phe-4 but leaves full-length APP unaffected [35, 47]. In vitro experiments have demonstrated that exposure of cultured primary neurons to aggregated full-length Aβ leads to increased mRNA-levels of tissue plasminogen activator (tPA) and urokinase-type plasminogen activator (uPA), implying a role for the plasmin system in Aβ clearance. It has been shown that purified plasmin degrades Aβ with physiologically relevant efficiency, leading to the identification of different cleavage sites, e.g. after Arg-5, which is consistent with the known specificity of plasmin to cleave after basic amino acids [96, 97]. Genetic studies have established a relationship between angiotensin-converting enzyme (ACE) and AD [21]. Hu and colleagues [37] provided the first evidence that ACE is able to significantly counteract the aggregation, deposition and cytotoxicity of Aβ in vitro by cleavage of Aβ at Asp-7. Purified myelin basic protein (MBP) is another candidate protein that possesses endogenous serine protease activity and that, at least in vitro, has been demonstrated to degrade Aβ peptides. Mass spectrometry identified several cleavage sites in fibrillar and soluble Aβ42 preparations, including between Phe-4 and Arg-5 in the N-terminus of the Aβ-sequence [52]. Finally, the major protease responsible for the liberation of Aβ_1–x peptides in AD, BACE 1, is also capable of cleaving between Tyr-10 and Glu-11, leading to the release of Aβ_11–x peptides [98]. The identification of this β′-cleavage site matched previous observations in primary neurons and cell lines suggesting that many of the Aβ_{x–40/x–42} peptides start with Glu-11 [95, 103]. Whereas N-terminally truncated and post-translationally modified Aβ_pE11–42 peptides have been demonstrated predominantly in mature plaque cores in AD brains, both unmodified Aβ_11–40, as well as Aβ_pE11–40 peptides have been detected in vascular deposits by immunohistochemistry [53].

Pike et al. [71] demonstrated that N-terminal deletions are neurotoxic. Interestingly, the N-truncated Aβ_x–40 peptides exhibited an enhanced neurotoxicity in vitro, while no difference was reported between full-length Aβ_1–42 and the other truncated Aβ_x–42 peptides. The authors compared the aggregation characteristics and biophysical properties of Aβ starting with Asp-1, Phe-4, Ser-8, Val-12 and Lys-17. Peptides with N-terminal deletions exhibited enhanced peptide aggregation relative to full-length species, as quantitatively assessed by sedimentation analyses. Full-length and truncated peptides showed circular dichroism spectra consistent with predominant β-sheet conformation, fibrillar morphology under transmission electron microscopy, as well as significant toxicity in cultures of rat hippocampal neurons. The authors concluded that N-terminal deletions enhance aggregation of β-amyloid into neurotoxic, β-sheet fibrils and suggested that such peptides may initiate and/or nucleate the pathological deposition of Aβ into plaques. Others reported that pyroGlu-3 was found to be more neurotoxic as compared to full-length Aβ [79]. In addition, it has been demonstrated that irrespective of the C-terminus of Aβ, i.e., Aβ40 or 42, pyroGlu-3 modified Aβ peptides displayed dramatically accelerated initial formation of aggregates compared to unmodified full-length Aβ. The accelerated seed formation was accompanied by a change in the oligomerization kinetics [85]. The N-terminal pyroGlu-3 and pyroGlu-11 modifications in comparison to their non-pyroglutaminylated counterparts Glu-3 and Glu-11 or Asp-1 (only Aβ_x–40 was investigated), revealed a decrease of solubility in the physiological pH range which was accompanied by an increase in hydrophobicity [87].

Nussbaum et al. [69] reported that Aβ_pE3–42 and Aβ_1–42 form metastable, cytotoxic, hybrid oligomers possessing a prion-like activity. The authors compared the cytotoxicity of the peptides in cultured neurons or glia cells and found that 12 h of 5 µM Aβ_1–42 exposure had little effect on cell viability on wild-type or tau-knockout neurons, or wild-type glial cells. In contrast, most wild-type neurons died and detached from the substrate after exposure to 5 µM Aβ_pE3–42 or a mixture of 5 % Aβ_pE3–42 and 95 % Aβ_1–42 (5 µM peptides in total). Tau-knockout neurons and wild-type glia, which express little tau protein, were resistant to Aβ_pE3–42 and the mixture containing 5 % Aβ_pE3–42 and 95 % Aβ_1–42.

We have recently extended these observations showing that soluble aggregates of Aβ_4–42 and pyroGlu Aβ_pE3–42 have specific structural features that might carry their neurotoxic activity [7]. We demonstrated that Aβ_4–42, Aβ_1–42 and Aβ_pE3–42 are unstructured in the monomeric state. However, upon heating the Aβ variants showed a high propensity to form folded structures. Monomeric Aβ_4–42 and Aβ_pE3–42 were rapidly converted to soluble aggregated species, whereas Aβ_1–42 stayed in equilibrium between monomers and soluble oligomers. The soluble aggregates were capable of converting to fibrillar aggregates with Aβ_4–42 and Aβ_pE3–42 showing significant thioflavin-T-reactivity already during the nucleation phase of aggregation [7]. The observation that the propensity of Aβ_4–42 and Aβ_pE3–42 to form aggregates is more pronounced than that of the N-terminally intact Aβ_1–42 peptide suggests that Aβ_4–42 and Aβ_pE3–42 aggregation may precede Aβ_1–42 aggregation in vivo.

Using far-UV CD spectroscopy, NMR spectroscopy and dynamic light scattering, we also have demonstrated that Aβ_4–42 and Aβ_pE3–42, and to a lesser extent Aβ_1–42, had a remarkable tendency to form stable aggregates [7]. The aggregates formed by Aβ_4–42 and Aβ_pE3–42 were distinct in size and different from Aβ_1–42. In addition, the fibrillar structure of Aβ aggregates was studied using transmission electron microscopy. The observation that all peptides except for Aβ_1–42 formed clumps of fibrils pointed to the importance of the N-terminal residues pyroGlu-3 and Phe-4 for aggregate morphology [7].

In recent years, N-terminal truncated Aβ peptides have been described not only in human samples, but also in a variety of transgenic AD mouse models. A thorough analysis in the APP/PS1KI mouse model using two-dimensional gel electrophoresis in combination with mass spectrometry at different time points, revealed the presence of a variety of N-truncated Aβ species [11]. In addition to full-length Aβ_1–42 peptides, additional spots representing Aβ_4/5–42 or Aβ_{8/9/10/11–42} were detected as early as 2.5 months of age, followed by Aβ_2/3–42 being detectable at 4 months of age. In the respective 2D-gels, the spot corresponding to Aβ_{8/9/10/11–42} allows no further discrimination, making assumptions about the presence of Aβ_11–42 difficult. This is an important issue, as previous in vitro data has indicated species specificity for BACE1, which is reported to be due to an amino acid difference in the murine and human Aβ-sequence at position 10 (Tyr in human and Phe in mouse). In conditioned media of mouse primary neurons transfected with human wild-type APP, only murine Aβ_11–40 could be recovered and only co-transfection with human BACE1 led to considerable amounts of secreted human Aβ_11–40 [10]. This might lead to significant bias in the assessment of N-truncated Aβ variants in transgenic mouse models and the fact that most available models harbor the Swedish APP mutation favoring the generation of full-length Aβ peptides may skew results even further. In the APP/PS1KI mouse model pyroGlu-modified Aβ_3–X becomes detectable at 6 months and increases in abundance with aging [11]. Subsequent immunohistochemical studies using the APP/PS1KI mouse model revealed the presence of plaque-associated and intraneuronal pyroglutamate Aβ_3–x [8] or in spinal cord motor neurons [105]. Pyroglutamate Aβ_3–x-positive plaques increase significantly in abundance but at the expense of plaques containing full-length-Aβ (starting with Asp-1) which show a corresponding decrease in abundance [107]. This suggests that in the parenchyma, pyroglutamate Aβ-formation might represent a later step in plaque maturation which might depend on remodeling of existing extracellular deposits. The presence of pyroglutamate Aβ deposits in transgenic mouse models has been confirmed in a variety of studies demonstrating that pyroglutamate Aβ-immunoreactivity is mainly confined to the amyloid core [23, 33, 40, 55]. In order to verify the in vivo toxicity of pyroglutamate Aβ, mouse models expressing solely the respective peptide but not the entire human APP molecule have been developed. These models made use of constructs starting with an N-terminal glutamine residue at position 3, which has been demonstrated to represent a better substrate for enzymatic conversion to pyroGlu-3 [13]. Abundant intracellular Aβ_pE3–42, followed by subsequent loss of Aβ_pE3–42-accumulating neurons could be demonstrated [1, 106]. This cell loss was rescued by crossing to a Tau knock-out background [69].

In order to study a potential seeding effect of Aβ_pE3–42 on full-length Aβ in transgenic mice, Aβ_pE3–42 expressing mice (TBA42 model) were crossed with 5XFAD mice [110]. The resulting bigenic model FAD42 was examined at 6 months of age. FAD42 mice showed an aggravated behavioral phenotype compared with the single transgenic parental 5XFAD or TBA42 lines. ELISA and plaque load measurements revealed that Aβ_pE3–x levels were elevated in FAD42 mice; however, no change in Aβ_x–42 or other Aβ isoforms was detected by ELISA or mass spectrometry. As Aβ_1–42 is the most abundant peptide in 5XFAD and FAD mice, these observations point to a drastic effect of Aβ_pE3–42.

Mass spectrometric analysis of 5XFAD mouse brain following immunoprecipitation with pan-Aβ or pyroGlu-specific antibodies also revealed the occurrence of Aβ_1–42, Aβ_1–40, Aβ_pE3–40, Aβ_pE3–42, Aβ_3–42, Aβ_4–42 and Aβ_5–42. Aβ_4–42 was the most abundant species among the N-truncated forms, but Aβ_1–42 clearly had the highest levels of all peptides [110]. Using NT4X-167, an antibody recognizing the N-terminus of N-truncated Aβ species with a preference for Aβ_4–x, strong intracellular staining could be detected in young 5XFAD transgenic mice [3]. Very recently, a transgenic mouse model overexpressing Aβ_4–42 without any mutations under the control of the murine neuron-specific Thy1-promotor has been described. These mice develop a massive age-dependent CA1 pyramidal neuron loss which correlates with the transgene expression pattern in the hippocampus. In addition, age-dependent spatial reference memory deficits were detected using the Morris water maze paradigm, underscoring the in vivo toxicity of Aβ_4–42 peptides [7].

However, in relative amounts, N-terminally truncated Aβ peptides, and in particular Aβ_pE3–42, in transgenic mouse models are much less abundant compared to human brain samples [76, 84]. In very old Tg2576 mice (21–23 months), only 5 % of the total insoluble Aβ is N-terminally truncated, whereas the corresponding percentage in human brain is ~70–85 % [44]. The relative solubility of human and APP transgenic mouse amyloid is strikingly different. Whereas, e.g. amyloid cores in Tg2576 and APP23 mice are completely soluble in SDS solutions with EDTA, human amyloid deposits are much more stable and do not dissociate in the presence of ionic or nonionic detergents or strong denaturing agents like guanidine hydrochloride. Therefore, the increased solubility of transgenic mouse amyloid might be directly related to the relative absence of N-terminal truncations and other post-translational modifications [41].

Together with Glu-3 of Aβ, the N-terminus of monocyte chemoattractant protein 1 (CCL2 or MCP-1) is modified to a pyroglutamate residue protecting against degradation in vivo. Cynis et al. [14] showed that the pyroGlu-formation of MCP-1 depends on glutaminyl cyclase (QC) activity. The same group has also provided strong evidence that Glu-3 of Aβ is pyroglutamated by QC [86]. Genetic ablation of the glutaminyl cyclase iso-enzymes QC or isoQC revealed a major role of isoQC for pyroGlu-MCP-1 formation and monocyte infiltration [14]. As neuroinflammatory processes around amyloid plaques represent a major hallmark of AD, it is likely that glial activation leads to enhanced QC activity and subsequent pyroGlu-3 formation in AD plaques. In line with that notion, we have observed that during plaque maturation the amount of Aβ peptides with intact N-terminus starting with Asp-1 declines whilst pyroGlu-3 increases [107].

While it is clear that transgenic mice expressing only N-truncated Aβ peptides Aβ_pE3–42 and Aβ_4–42 do develop massive neuron loss in CA1 [1, 7], the mechanisms of cell death have not been elucidated. Palop and Mucke [70] discussed that AD is associated with cognitive decline and increased incidence of seizures. Sporadic cases are known to exhibit seizure activity, as well as many pedigrees with autosomal dominant early-onset AD, including those with mutations in presenilin-1, presenilin-2, or APP, or with duplications of wild-type APP. Moreover, high levels of Aβ in the brain of APP transgenic mouse models can cause epileptiform activity.

Due to variations in the methods and tools used to extract and identify different pools of Aβ, drawing conclusions on the exact levels of the various N-truncated Aβ variants is challenging. Evaluation of data is difficult as far as the exact levels of different Aβ peptides in post-mortem brains are concerned. Many factors can influence an analysis ranging from antibody specificities and sensitivities in applications like immunostaining, Western blotting or immunoprecipitation, as well as extraction protocols and brain areas studied. Most consistently, there is general agreement that plaque-born peptides harbor high amounts of N-truncated Aβ especially Phe-4, but also pyroGlu-3 and pyroGlu-11. In presymptomatic AD cases, Phe-4 seems to be the N-truncated variant most consistently reported. Besides plaque-associated Aβ, intraneuronal Aβ can be N-truncated in AD brain [27]. Immunohistochemical studies in cases with Down syndrome demonstrated plaque-Aβ starting at Asp-1 or pyroGlu-3 [48]. A transient accumulation of intraneuronal Aβ_x–42 was also evident [66]. Analyzing FAD patients, Ancolio and colleagues [2] firstly showed a selective and drastic increase of N-truncated Aβ_x–42 species triggered by the mutation APP V715M. In contrast to the N-terminus, there is common agreement that plaque-associated Aβ peptides mainly terminate at position 42 with Ala-42.

In the following paragraphs, we endeavor to shed light on what is known regarding the role of N-terminal truncated Aβ peptides in AD.

In 1985, ragged Aβ peptides were described to precipitate in AD plaques, including a major species beginning with phenylalanine at position 4 of Aβ (Phe-4; Aβ_4–x) [58]. A majority of 64 % of the peptides in amyloid plaques of the two sporadic AD cases and of 45 % in the patients with Down syndrome studied started with a Phe-4 residue. At the same time, Glenner and Wong [26] demonstrated full-length Aβ beginning with Asp-1 to be the main species detected in cerebrovascular deposits. A scheme of the amino acid residue numbering of N-terminal Aβ is shown in Fig. 1.

Miller et al. [62] compared the peptide compositions of the cerebrovascular and senile plaque core amyloid deposits in AD. Matrix-assisted, laser-desorption-time-of-flight (MALDI-TOF) mass spectrometry of plaque-Aβ revealed an array of peptides ending with Ala-42 of that sequence, while cerebrovascular Aβ began with Arg-1 ending at Val-40. They verified that Phe-4 is the main component in plaques, but cautioned that their MALDI-TOF spectral data suggests the presence of two pyroglutamyl amino termini (pyroGlu-3 and pyroGlu-11) that might escape detection by other methods. Other N-termini reported were Asp-1, Ala-2, Arg-5, Asp-7, Ser-8, Gly-9.

Surface-enhanced laser desorption/ionization mass spectrometry was performed comparing AD and vascular dementia patients [51]. In AD, the authors found Aβ starting with Asp-1, Ala-2, pyroGlu-3, Phe-4 and Arg-5 in senile plaque extractions with Phe-4 to be the most prevalent one.

The presence of pyroGlu-3 peptides as an important component of plaque depositions in patients with AD was further substantiated [77, 80]. MALDI-TOF mass spectrometry of Aβ peptides isolated from sporadic and familial AD (APP V717I and several PS1 mutations) brains indicated that besides full-length Aβ_1–40/42; pyroglutamylated Aβ_3–42 (Aβ_pE3–42) and Aβ_pE11–42 as well as Aβ_4–42 were detected in these cases [78]. Analysis of sporadic and familial AD cases by electrospray–ionization mass spectrometry even showed that Aβ_11–42/Aβ_pE11–42 represent the second most abundant species following Aβ_1–40 [68]. Further analysis of FAD cases revealed that N-terminally truncated Aβ peptide species ending at residues 42 and 43 are the main Aβ peptides deposited in brain parenchyma in association with the PS1 V261I mutation. MALDI-TOF mass spectrometry following immunoprecipitation using a mixture of Aβ antibodies showed that most intense signals corresponded to pyroGlu-11, pyroGlu-3, but also non-pyroglutamylated Glu-3 peptides, whereas the signals corresponding to Glu-11 and Asp-1 were less intense [63].

The Aβ isoform pattern was studied in the cerebellum, cortex and hippocampus in AD, including subjects with mutations in PS1 (M146V) or APP (KM670/671NL) genes, sporadic AD subjects and non-demented controls [72]. Using immunoprecipitation in combination with mass spectrometric analysis, the dominating Aβ isoforms in the three different brain regions analyzed from control, sporadic and familial AD were described as Aβ_1–42, Aβ_pE3–42, Aβ_4–42 and Aβ_1–40, with Aβ_1–42 and Aβ_4–42 being the dominant isoforms in hippocampus and cortex in all groups analyzed [72].

The question whether N-truncations of Aβ are a post-mortem artefact or might even precede the symptomatology of AD was addressed by Sergeant and co-workers [90]. They have adapted a proteomic method in combination with Western blotting and mass spectrometry for the characterization of insoluble Aβ extracted in formic acid. Full-length Aβ peptides represented 37 % of all Aβ species, while 17 % corresponded to N-truncated species starting at residues Phe-4, Arg-5 and 20 % with Ser-8, Gly-9 and Tyr-10. They also demonstrated that the first stage of amyloid deposition in non-demented individuals comprise N-terminal truncated variants starting at positions 4-, 5-, 8- and 9–42, or with a pyroglutamyl residue at position 3. At this stage, Aβ oligomers were exclusively made of Aβ_x–42 species.

N-terminal truncations of Aβ, especially pyroGlu-3 were reported to be more frequently found in plaques of sporadic AD cases as compared to the PS2APP mouse model [28].

CNS and the cerebrospinal fluid from APP23 transgenic mice were assessed using one- and two-dimensional gel electrophoresis, immunoblotting and mass spectrometry [84]. Significant differences between APP23 mice and brain samples from sporadic AD cases (Braak stage V–VI) were observed in their relative abundance of specific variants of Aβ peptides, such as pyroGlu-3, Aβ_1–42 and N-terminally truncated Aβ_2/3–42.

In a recent report, phosphorylated Aβ at Ser-8 (pSer-8) and pyroGlu-3 in soluble, dispersible, membrane-associated and plaque-associated amyloid-β aggregates in brains from 21 cases with symptomatic AD, 33 pathologically preclinical AD cases, and 20 control cases was compared [74]. Plaques containing pSer-8 were detected in all symptomatic cases with AD, but only in a few non-demented control subjects. The deposition occurred in a hierarchical sequence with pyroGlu-3 appearing early in the amyloid cascade corroborating earlier findings in preclinical AD cases [90].

Moore et al. [64] employed sequential pull-down with antibodies Ab9 (epitope Aβ_1–16) and 4G8 followed by mass spectrometry using brain samples from the prefrontal cortex. 16 brains from AD, eight brains from subjects without clinical evidence of dementia and seven brains with rare or no AD lesions from elderly individuals without clinical evidence of a neurological illness were studied. In the membrane-associated SDS extracted lysates Aβ_1–42, Aβ_4–42 and Aβ_1–40 were the most prevalent peptides. In plaque-associated formic acid lysates the spectrum became more diverse. The two pyroglutamylated peptides Aβ_pE3–42 and Aβ_pE11–42 were showing up, albeit at low levels compared to Aβ_4–42, Aβ_8–40, Aβ_8–42, Aβ_9–40 and the full-length Aβ_1–42, Aβ_1–43 and Aβ_1–40. The presymptomatic group revealed elevated Aβ_4–42, Aβ_pE3–42 and Aβ_1–42 levels compared to the control group.

A comparison of antibody staining profiles in the brain of a patient with sporadic AD against Asp-1, pyroGlu-3 (pan- and oligomer-specific), Phe-4- and Arg-5 is shown in Fig. 2. Antibodies against pan-Aβ, Asp-1 and pan-pyroGlu-3 stained amyloid plaques very strongly, whereas the oligomer-specific antibodies (9D5 against pyroGlu-3 and NT4X-167 against Phe-4) labeled predominantly cerebrovascular deposits. AB5-3 against position 5 of Aβ detected both hallmarks with a tendency towards stronger labeling of cerebrovascular deposits. The cerebrovascular localization indicates that these N-truncated Aβ peptides are forming preferentially soluble aggregates with a reduced tendency to aggregate in plaques.

In 1999, Schenk et al. [82] pioneered the AD field by introducing immunization as a causal therapeutic option. They immunized transgenic APP mice with pre-aggregated synthetic Aβ_1–42, either before or after onset of plaque deposition. Immunization of young animals essentially prevented the development of plaque formation and astrogliosis. Treatment of older animals also markedly reduced the extent and progression of these AD-like neuropathologies. These results implied that immunization with pre-aggregated Aβ_1–42 may be effective in preventing and treating AD. Moreover, vaccination with Aβ not only reduced plaque load, but also protected transgenic mice from the learning and age-related memory deficits [65]. Several mechanisms have been suggested since then for the significant therapeutic effects of immunotherapy in AD mouse models, which will be discussed below.

Using peptide microarrays, the presence of natural antibodies against Aβ in plasma samples and cerebrospinal fluid of AD patients and healthy controls aged 21–89 years was reported [9]. Antibody reactivity was most prominent against oligomeric Aβ and pyroGlu or oxidized residues. Interestingly, IgG levels specific for oligomeric preparations of Aβ_1–42 declined with age and AD progression. In good agreement, we have observed that the levels of pyroGlu-IgM autoantibodies significantly decreased in AD patients as compared to non-demented controls [57]. In the group of mild cognitive impaired patients there was a significant positive correlation between pyroGlu-IgM autoantibodies and cognitive performance, i.e. individuals with high levels of pyroGlu-IgM autoantibodies obtained higher scores in the Mini Mental State Examination test battery.

In contrast to Aβ_1–42, N-truncated pyroglutamate Aβ_3–42 and Aβ_4–42 peptides are not produced under normal, non-disease conditions. Pyroglutamate Aβ_3–42 and Aβ_4–42 form soluble aggregates and are toxic in vitro and in vivo. On the basis of these empirical data, we formulated a novel hypothesis on the role of soluble aggregates of pyroglutamate Aβ_3–42 and Aβ_4–42 (Fig. 3).