High-molecular weight Aβ oligomers and protofibrils are the predominant Aβ species in the native soluble protein fraction of the AD brain

Abstract Alzheimer’s disease (AD) is characterized by the aggregation and deposition of amyloid β protein (Aβ) in the brain. Soluble Aβ oligomers are thought to be toxic. To investigate the predominant species of Aβ protein that may play a role in AD pathogenesis, we performed biochemical analysis of AD and control brains. Sucrose buffer-soluble brain lysates were characterized in native form using blue native (BN)-PAGE and also in denatured form using SDS-PAGE followed by Western blot analysis. BN-PAGE analysis revealed a high-molecular weight smear (>1000 kD) of Aβ42-positive material in the AD brain, whereas low-molecular weight and monomeric Aβ species were not detected. SDS-PAGE analysis, on the other hand, allowed the detection of prominent Aβ monomer and dimer bands in AD cases but not in controls. Immunoelectron microscopy of immunoprecipitated oligomers and protofibrils/fibrils showed spherical and protofibrillar Aβ-positive material, thereby confirming the presence of high-molecular weight Aβ (hiMWAβ) aggregates in the AD brain. In vitro analysis of synthetic Aβ40- and Aβ42 preparations revealed Aβ fibrils, protofibrils, and hiMWAβ oligomers that were detectable at the electron microscopic level and after BN-PAGE. Further, BN-PAGE analysis exhibited a monomer band and less prominent low-molecular weight Aβ (loMWAβ) oligomers. In contrast, SDS-PAGE showed large amounts of loMWAβ but no hiMWAβ40 and strikingly reduced levels of hiMWAβ42. These results indicate that hiMWAβ aggregates, particularly Aβ42 species, are most prevalent in the soluble fraction of the AD brain. Thus, soluble hiMWAβ aggregates may play an important role in the pathogenesis of AD either independently or as a reservoir for release of loMWAβ oligomers.


Introduction
Alzheimer's disease (AD) is characterized by the extracellular deposition of amyloid ␤ protein (A␤) aggregates in the brain [1]. Although high-molecular weight A␤ (hiMWA␤) oligomers, A␤ protofibrils and fibrils, low-molecular weight A␤ (loMWA␤) oligomers, such as dimers, trimers or A␤*56, have been observed in human AD brain tissue or in mouse models of AD [1][2][3][4][5][6][7][8][9], it is not entirely clear which A␤ species are the most relevant ones for the development of AD and how these A␤ forms are related to one another in vivo. Some studies have used SDS-PAGE for protein analysis [3][4][5]9], which denatures and dissociates proteins into individual polypeptides before determining its molecular weight. By contrast, others have performed only dot blot analysis [6]. Currently, only size exclusion chromatography has been used to study oligomers in non-SDS-treated conditions [3,9]. However, it is unclear whether interactions with the stationary phase may impact the aggregation state of hiMWA␤ species. A detailed analysis of the native A␤ aggregates in the AD brain using blue native-PAGE (BN-PAGE) in comparison with SDS-PAGE analysis that focuses on the identification of the above-mentioned forms of A␤ aggregates is still unavailable.
Antibodies and antibody fragments have been developed to detect specific hiMWA␤ oligomers (A11) and protofibrillar/fibrillar conformations (B10AP) [2,6]. These antibodies and antibody fragments allow isolation of oligomers, protofibrils and fibrils from soluble native protein lysates by immunoprecipitation for further protein analysis. Here, we employed these antibodies and BN-PAGE analysis to clarify whether soluble hiMWA␤ oligomers and A␤ protofibrils/fibrils or A␤ dimers and other loMWA␤ species represent the predominant A␤ aggregates in the native soluble fraction of the AD brain. SDS-PAGE was used to study the effect of protein denaturation on the spectrum of loMWA␤ and hiMWA␤ species.

Neuropathology and human sample characterization
A sample including six AD and four control cases was studied (Table 1). All autopsy brains were collected from individuals who died in the University Hospitals of Bonn or Ulm (Germany). All human tissue was obtained and processed in compliance with German federal laws and with university ethics committee approval.
Demented as well as non-demented patients were examined 1-4 weeks prior to death using standardized protocols for routine clinical examination, including neurological status, upon admission to hospital. These data were used to determine whether individuals clinically fulfilled the DSM-IV criteria for dementia [11]. AD was diagnosed when dementia was observed and when the degree of AD-related neuropathology indicated at least a moderate likelihood for AD according to internationally acknowledged criteria [12].
After assessment of unfixed tissue from one hemisphere for biochemical studies, the brains were fixed in a 4% aqueous formaldehyde solution for at least 3 weeks before undergoing neuropathological screening. Presence or absence of gross infarction, haemorrhage, tumour and other findings were recorded. Tissue blocks from the medial temporal lobe (MTL) were excised at the levels of the (i) anterior limit of the dentate gyrus and (ii) lateral geniculate body [13]. These blocks together with tissue blocks from the occipital cortex (Brodmann areas [17][18][19] were embedded in paraffin. All sections were cut at 10 m. Neurofibrillary changes were detected by immunostaining with an antibody directed against abnormal phosphorylated protein (AT-8, Pierce, Rockford, IL, USA, 1/1000) [14]. Neuritic plaques were also diagnosed in sections immunostained with this same antibody. The presence of A␤ deposition was assessed using immunohistochemistry with an antibody raised against A␤17-24 (4G8 [15], Covance, Emeryville, CA, USA, 1/5000, formic acid pre-treatment).
Diagnosis of the stages in the development of neurofibrillary changes (Braak NFT stage) and the semi-quantitative assessment of neuritic plaques (CERAD score) were performed in accordance with published and recommended criteria [12,14,16,17]. For staging of A␤ pathology, we used a previously published protocol for four phases of ␤-amyloidosis in the MTL [18]. This hierarchically based procedure facilitates study of the topographic distribution pattern of A␤ deposition in additional brain regions [18,19]: phase 1 represents A␤ deposition that is restricted to the temporal neocortex. Phase 2 is characterized by the presence of additional A␤ plaques in the entorhinal cortex and/or in the hippocampal subiculum-CA1 region. The third phase is marked by the presence of A␤ plaques in the outer zone of the molecular layer of the fascia dentata, subpial band-like amyloid and/or presubicular 'lake-like' amyloid. The existence of further A␤ plaques in the hippocampal sector CA4 and/or the pre-␣ layer of the entorhinal cortex characterize the fourth and final phase of A␤ deposition in the MTL. Reference pathology for all cases was performed by one and the same neuropathologist (D.R.T.).

Biochemical analysis of human AD and control brains
Fresh frozen human brain tissue from the six AD and four control cases was used to assess the presence and types of native A␤ aggregates in AD and control brains (Table 1). Protein extraction from 30 mg of fresh frozen human occipital (Brodmann areas [17][18][19] and temporal cortex (Brodmann  [12], age in years, gender, AD type [10], the stage of neurofibrillary tangle (Braak-NFT stage) pathology according to Braak et al. [14,16], the A␤-phase representing the distribution of A␤ deposits in subfields of the MTL [18] and the Consortium to Establish a Registry for Alzheimer's Disease (CERAD) score for the frequency of neuritic plaques according to Mirra et al. [17] For SDS-PAGE, sucrose fractions (50 g total protein) and immunoprecipitation products were electrophoretically resolved in a precast NuPAGE 4-12% Bis-Tris gel system (Invitrogen). The protein load was controlled either by Ponceau S staining or ␤-actin (C4, 1/1000; Santa Cruz Biotechnology, Santa Cruz, CA, USA) immunoblotting. The proteins were transferred to nitrocellulose membranes and the membranes were boiled with PBS for 6 min. followed by blocking with 5% non-fat dry milk (Roth; diluted in antibody-dilution buffer) for 1 hr at room temperature.

Analysis of synthetic A␤ 42 and A␤ 40 aggregates in native state and after SDS denaturation
To determine whether synthetic A␤ aggregates primarily form loMWA␤and hiMWA␤ aggregates, we dissolved 15 mol synthetic A␤40-peptide (Peptides International, Louisville, KY, USA) in 1 ml cell culture medium (Quantum 263; PAA Laboratories, Pasching, Austria) for 30 min. at 4ЊC [23]. A␤42-peptide (Bachem, Bubendorf, Switzerland) was also dissolved in cell culture medium (RPMI1640; GIBCO, Invitrogen) [23]. Aggregation was permitted to occur for 4 hrs at 22ЊC. To identify oligomers, fibrils and protofibrils structurally we used electron microscopy. For this purpose, 5 l of the A␤40and A␤42 solutions were placed on a formvar-coated grid for 1 min. before wiping off the excess liquid. The protein-coated grids were block-stained with a 2% aqueous solution of uranyl acetate (Merck).
The protein aggregates were also analysed with BN-PAGE and SDS-PAGE as well as subsequent Western blot analysis using the MBC-40 and MBC-42 antibodies to detect A␤40 and A␤42, respectively. This experiment was repeated five times.

Results
High-molecular weight A␤ 42 aggregates predominate in native protein preparations of the soluble fraction from human brain lysates BN-PAGE with subsequent Western blot analysis of the soluble fraction of human AD brain lysates revealed a high-molecular weight anti-A␤42-positive smear Ͼ1000 kD in AD cases (Fig. 1A)

Fig. 1 Western blot analysis of sucrose soluble proteins from AD and control brains after BN-PAGE (A-D) and after SDS-PAGE (E-H).
All BN-PAGE blots were developed 2-3 hrs for chemiluminescence exposure. SDS-PAGE blots were exposed for 2-5 min. (A) The protein lysates from AD brains (cases no. [5][6][7][8][9][10] in BN-PAGE showed a high-molecular weight anti-A␤42-positive smear Ͼ1000 kD. Such smears were not observed in controls (cases no. [1][2][3][4]. Synthetic A␤42 and A␤40 were loaded as positive and negative controls, respectively. In A␤42 preparations, long chemiluminescence exposure led to the detection of additional dimer and ~50 kD bands that were not observed after 2-5 min. exposures, as shown in Figure 3C. (B) The A␤42-positive material seen in (A) was not detectable in AD (cases no. [5][6][7][8][9][10] or in the controls (cases no. [1][2][3][4] in the native gel blotted with anti-A␤40 antibodies. Synthetic A␤42 and A␤40 were loaded as positive and negative controls, respectively. After 3 hrs of chemiluminescence exposure, synthetic A␤40 blots display a dimer band at ~10 kD in addition to the monomer band and the hiMWA␤ smear already detected with shorter exposure times as depicted in Figure 3C.

(E)-(G) SDS-PAGE analysis of AD brain protein lysates from cases no. 5-10 exhibited A␤ monomer and dimer bands with MBC-42 (E), MBC-40 (F) and anti-A␤1-17 (G) that were not detected in control brains (cases no. 1-4). The MBC-42-dimer (E) and 6E10-dimer bands (G) were not seen in all AD cases, whereas anti-A␤40 consistently detected dimer bands (F). A high-molecular smear was found in most AD cases with all three antibodies directed against A␤. Interestingly, cases 6 and 10 exhibited nearly no SDS-stable hiMWA␤42 aggregates (E), whereas both cases showed high-molecular anti-A␤42-positive material in the BN-PAGE (A). (H) With the help of anti-A␤1-17 (6E10)-immunoprecipitation, monomer and dimer bands as well as loMWA␤ (4-20 kD) smears and hiMWA␤ (Ͼ160 kD) smears were visible in SDS-PAGE of AD brain lysates (cases no. 5-10) but not in those of controls (cases no. 1-4). The detection of the loMWA␤ oligomers required chemiluminescence exposure for 3 hrs (i.e. long exposure times).
that was not found in controls. A␤ monomers, dimers, or other loMWA␤ species were not observed in AD cases or in controls (Fig. 1A). The high-molecular weight smear was also seen with anti-A␤1-17 at the level of stacking gel in AD cases (Fig. 1C) but not with anti-A␤40 (Fig. 1B). However, the detection of synthetic A␤40 but not A␤42 indicated specific antibody function (Fig. 1B). Anti-A␤1-17 stained an additional 150-250 kD band (Fig. 1C) that was also observable with anti-APP antibodies, thereby indicating that this band represents APP-containing material (Fig. 1D). The APP-related band was also present in control cases, whereas the high-molecular weight anti-A␤1-17 smear was not seen (Fig. 1C). (Fig. 1E-G). A␤ aggregates with a molecular weight of Ͼ160 kD were observed in four of six cases with anti-A␤42 and a smear Ͼ260 kD was observed in all cases with anti-A␤1-17 (6E10) (Fig. 1E-G). After immunoprecipitation with anti-A␤1-17, a smear of hiMWA␤-(Ͼ160 kD) and loMWA␤ aggregates (8-20 kD) as well as dimer bands at ~10 kD were consistently seen in AD

Fig. 2 Electron microscopic analysis of immunoprecipitated protein aggregates from AD and control brain as precipitated with B10APantibody fragments (B10AP-IP) and A11 antibodies (A11-IP). (A), (B)
In the control case no. 3, protein aggregates were precipitated with B10AP, but anti-A␤1-17 did not show A␤ within these aggregates. There was also no non-specific labelling with anti-A␤1-17 because no gold particles were observed. The protein aggregates exhibited protofibrilaggregate-like architecture that is more evident at higher magnification (B). This indicates that B10AP does not specifically bind to A␤ protofibrils or fibrils but to proteins with a distinct protofibrillar/fibrillar conformation, as reported previously [2]. (C), (D) A11-IP from control cases resulted in detection of amorphous to spherical presumably oligomeric protein aggregates, as shown in control case no. 4, but did not exhibit A␤ as a component of these protein aggregates. There was also no non-specific labelling with anti-A␤1-17 because no gold particles were seen. The high magnification demonstrates the spherical shape of the precipitated proteins (D). Thus, A11 also binds spherical protein aggregates other than A␤ oligomers, as reported earlier by others [6]. (E), (F) B10AP-IP from AD brain lysate of case no. 7 showed protein aggregates of protofibril-like morphology. Immunogold labelling indicated A␤1-17-positive proteins. The frame in E indicates the areas enlarged in (F). At higher magnification, A␤1-17-positive material following B10AP-IP exhibited protofibril-like morphology (arrows in F) and less frequently amorphous structures (arrowheads in F). These types of A␤ aggregates prevailed in B10AP precipitates. (G) Only a few precipitated A␤positive protein aggregates exhibited fibrillar architecture (case no. 10) resembling synthetic A␤ fibrils (Fig. 3A, B) cases (Fig. 1H). LoMWA␤ and dimers were detected only after 3 hrs of chemiluminescence exposure but not in controls (Fig. 1H).

A11-antibody and B10AP-antibody fragments precipitate oligomeric and protofibrillar/fibrillar proteins including A␤ oligomers, A␤ protofibrils and A␤ fibrils in AD cases
In controls, immunoelectron microscopy of A11-and B10APprecipitated proteins revealed a high number of precipitated and aggregated proteins that did not contain A␤-positive material ( Fig. 2A-D). There was no nonspecific labelling with anti-A␤1-17 in controls. B10AP-precipitated material exhibited a pattern that showed fibril-/ protofibril-like architectures ( Fig. 2A, B) whereas the proteins precipitated by A11 displayed a spherical pattern (Fig. 2C, D).
In B10AP precipitates of the soluble fraction of AD brain homogenates, we observed protein aggregates with a fibril/protofibril-like pattern similar to that seen in controls. However, in AD brains a high number of A␤-positive protein aggregates were detected with anti-A␤1-17 (Fig. 2E-G). High-magnification analysis of anti-A␤1-17-labeled protein aggregates revealed a protofibril-like pattern (Fig. 2F, arrows). However, a few amorphous protein aggregates (Fig. 2F, arrowheads) as well as fibrillar aggregates (Fig. 2G) were seen as well. Spherical A␤ oligomers were not observed following B10AP immunoprecipitation. Amorphous and spherical protein aggregates were observed after A11-immunoprecipitation from AD brain lysates. Anti-A␤1-17 antibodies detected protein aggregates of spherical and amorphous morphology (Fig. 2H-J). Protofibril-like structures as seen in B10AP precipitates were not observed in A11 precipitates.

SDS treatment destroys native high-molecular weight A␤ 42 and A␤ 40 aggregates
Synthetic A␤42 and A␤40 formed oligomeric and protofibrillar aggregates as detectable by electron microscopy (Fig. 3A, B). With 2-5 min. chemiluminescence exposure time, BN-PAGE blots revealed a monomer band and an additional prominent smear of A␤42and A␤40 aggregates with a molecular weight Ͼ700 and 240 kD, respectively (Fig. 3C). However, after longer exposure time of 2-3 hrs, we observed smeary bands at ~10 and ~50 kD in A␤42 preparations, whereas A␤40 preparations exhibited an additional dimer band that was not detectable in short chemiluminescence exposure blots (Figs 1A, B, 3C). In SDS-PAGE, we observed few high-molecular weight aggregates above 240 kD in A␤42 preparations but not in A␤40 preparations, whereas very prominent A␤ monomer, dimer and trimer bands were observed for both synthetic A␤42and A␤40 preparations. In SDS-treated A␤40 preparations, tetramer and A␤*56 bands were present that were not seen in SDS-treated A␤42 preparations (Fig. 3D).

Discussion
Our results show that hiMWA␤42 oligomers and protofibrils with a molecular weight Ͼ1000 kD predominate in the soluble fraction of AD brain homogenates when these samples are analysed under native BN-PAGE conditions. A␤40 aggregates were not detected following BN-PAGE. Immunoelectron microscopy of immunoprecipitated oligomeric, fibrillar and protofibrillar proteins confirmed the presence of protofibrillar and spherical hiMWA␤ aggregates in the soluble fraction of AD brain homogenates. These hiMWA␤  Figure 1A and B, thereby indicating that these preparations contain low levels of these A␤ species as well. (D) In SDS-PAGE, hiMWA␤42 aggregates were strikingly reduced. Instead, the monomer, dimer and trimer bands displayed strong staining. Other loMWA␤42 oligomers were not evident. No hiMWA␤40 aggregates were seen after denaturing SDS-PAGE, but A␤40 monomer, dimer, trimer and tetramer bands as well as an A␤*56 band were detectable at 56 kD.
aggregates were not seen in controls, whereas analysis of control cases revealed that the A11 and B10AP antibodies/antibody fragments precipitate other proteins of similar morphology, and that only a portion of the precipitated proteins from AD cases were A␤ aggregates. Denaturation of the hiMWA␤ aggregates by SDS resulted in the detection of A␤ monomers, dimers, and hiMWA␤ with a molecular weight Ͼ160 kD in SDS-PAGE analysis of AD cases but not in controls. When using immunoprecipitation with anti-A␤1-17, a smear of loMWA␤ and hiMWA␤ was consistently seen following SDS-PAGE and subsequent Western blot analysis, thus indicating that dimers, trimers, tetramers and A␤*56 are not the only A␤ oligomers that can be detected with SDS-PAGE as shown previously by other groups as well [4,5]. These results lead us to conclude that, under native conditions, A␤ monomers and loMWA␤ aggregates, such as dimers, trimers and A␤*56, do not represent the major pool of A␤ aggregates in the human AD brain. More likely, loMWA␤ aggregates may occur transiently during aggregation or after denaturation of hiMWA␤. The strongest argument in favour of this hypothesis is our finding that hiMWA␤ oligomer preparations and A␤ protofibril preparations of synthetic A␤42and A␤40 peptides did not exhibit high levels of loMWA␤ oligomers in BN-PAGE but did so in SDS-PAGE. Moreover, subsequent to SDS-induced protein denaturation, hiMWA␤40 aggregates were no longer seen and synthetic hiMWA␤42 aggregates were remarkably reduced. A possible argument against the predominance of hiMWA␤ in the soluble fraction is that A␤ monomers tend to aggregate in the presence of oligomers [24] and that this occurs during protein preparation. Nevertheless, in synthetic A␤ preparations with high amounts of aggregated A␤, we detected a significant monomer band after BN-PAGE. This may indicate that A␤ monomers in the soluble brain lysates remained stable during the process of native protein preparation. As such, it is likely that the hiMWA␤42 aggregates observed in the native soluble fraction indeed represent the major form of soluble A␤ in the human AD brain.
Our finding that A␤40 was detected in AD cases in SDS-PAGE but not in BN-PAGE could be attributable either to a lower resolution of native gels in comparison to that of denaturing gels or to the fact that a potential smear of A␤40 aggregates falls far below detectable levels. Presumably, SDS treatment denatures all kinds of A␤40 oligomers and, in so doing, leads to the accumulation of A␤40 monomers in a single band. Thus, a non-detectable A␤40 smear in BN-PAGE might be converted into a detectable welldefined band in the SDS-PAGE. This hypothesis is supported by our finding of a detectable hiMWA␤40 smear in synthetic A␤40 preparation that disappeared after SDS treatment and converted into strongly stained monomer and loMWA␤ oligomer bands. In BN-PAGE, the spectrum of synthetic hiMWA␤40 oligomers was greater (Ͼ240 kD) than that of synthetic hiMWA␤42 oligomers (Ͼ700 kD). This suggests that the concentrations of distinct hiMWA␤40 oligomers are lower than those of distinct hiMWA␤42 oligomers because of the more widespread distribution of hiMWA␤40 aggregates in the gel. As a result, hiMWA␤40 oligomers and may be less easily detected in native brain lysates.
An alternative explanation, on the other hand, could be that A␤40 interacts with other proteins that hide its C-terminus. In addition, the predominance of hiMWA␤42 in the native soluble fraction of the AD brains investigated here confirms previous reports of a predominant occurrence of A␤42 in parenchymal soluble and insoluble A␤ aggregates in AD [25,26].
At first, the results reported here appear to contradict the findings of other authors, who argue that distinct loMWA␤ oligomers, such as dimers and A␤*56, are critical for the development of AD [3,9,27]. These authors provide evidence that loMWA␤ oligomer preparations received by size-exclusion chromatography are detectable in human as well as transgenic mouse brains, and are capable of inducing cognitive deficits in the rat [9] or altering longterm potentiation [3]. Given our in vitro and in vivo findings, however, one could also speculate that small amounts of loMWA␤ oligomers (possibly resulting from the denaturation of hiMWA␤ oligomers, protofibrils and fibrils) either are critical for the development of AD or that, upon their administration, loMWA␤ oligomers may spontaneously aggregate and form hiMWA␤ oligomers, as appears to be the case based upon our native gel analysis of A␤40and A␤42 preparations and the results of Nguyen et al. [24], who showed that A␤ oligomers accommodate added A␤ monomers. Thus, hiMWA␤ aggregates contribute to the pathogenesis of AD either on its own or do so indirectly by providing the reservoir of hiMWA␤ aggregates that denature and, in so doing, release loMWA␤ oligomers.
That A␤ aggregates, including A␤ plaques, dissociate during the pathogenesis of AD is corroborated by the finding that in latestage AD cases plaque frequency is lower than in earlier stages [28]. The relevance of soluble hiMWA␤ for the pathogenesis of AD may be further supported by the finding of neuritic degeneration near A␤ plaques, i.e. in areas with high levels of hiMWA␤ presumably dissolved from A␤ plaques, in the APP transgenic mouse brain [7,8,29], in aged rhesus monkeys [30] and in the AD brain [31], and also by the finding that dendritic degeneration in another APP-transgenic mouse model begins at the same time as the deposition of initial diffuse A␤-plaques, i.e. when hiMWA␤ aggregates begin to predominate in the cortex [32].
Here, the stability of hiMWA␤42 aggregates was greater than that of A␤40 aggregates in in vitro experiments, thus confirming previous reports that soluble A␤42 aggregates are more stable than A␤40 aggregates [33,34]. Taken together with our finding that hiMWA␤42 aggregates predominate in the native soluble fraction of the brain, it may be speculated that it is the stability of soluble A␤42 aggregates in the soluble compartment of the brain that accounts for its predominance in parenchymal A␤ plaque deposition [26].
In conclusion, the results of the present study strongly suggest that hiMWA␤ oligomers, protofibrils and fibrils are the predominant soluble A␤ aggregates in the AD brain. loMWA␤ oligomers in high concentrations are detectable only after denaturation of hiMWA␤ aggregates. In view of the denaturation of hiMWA␤ aggregates and fibrils into loMWA␤ oligomers, we propose that A␤ plaques consisting of both fibrillar A␤ as well as soluble hiMWA␤ aggregates may serve as reservoirs for the release of loMWA␤ oligomers.