Large Aggregates Are the Major Soluble Aβ Species in AD Brain Fractionated with Density Gradient Ultracentrifugation

Soluble amyloid-β (Aβ) aggregates of various sizes, ranging from dimers to large protofibrils, have been associated with neurotoxicity and synaptic dysfunction in Alzheimer's Disease (AD). To investigate the properties of biologically relevant Aβ species, brain extracts from amyloid β protein precursor (AβPP) transgenic mice and AD patients as well as synthetic Aβ preparations were separated by size under native conditions with density gradient ultracentrifugation. The fractionated samples were then analyzed with atomic force microscopy (AFM), ELISA, and MTT cell viability assay. Based on AFM appearance and immunoreactivity to our protofibril selective antibody mAb158, synthetic Aβ42 was divided in four fractions, with large aggregates in fraction 1 and the smallest species in fraction 4. Synthetic Aβ aggregates from fractions 2 and 3 proved to be most toxic in an MTT assay. In AβPP transgenic mouse brain, the most abundant soluble Aβ species were found in fraction 2 and consisted mainly of Aβ40. Also in AD brains, Aβ was mainly found in fraction 2 but primarily as Aβ42. All biologically derived Aβ from fraction 2 was immunologically discriminated from smaller species with mAb158. Thus, the predominant species of biologically derived soluble Aβ, natively separated by density gradient ultracentrifugation, were found to match the size of the neurotoxic, 80–500 kDa synthetic Aβ protofibrils and were equally detected with mAb158.


Introduction
Soluble aggregates of the amyloid-b (Ab) peptide have become the focus of Alzheimer's disease (AD) research as they are neurotoxic and inhibit synapse function [1,2,3,4,5,6,7,8]. While CSF levels of Ab42 declines during the presymptomatic stages of AD [9], elevated levels of soluble Ab in the brain has been demonstrated to correlate with AD progression [10,11,12] and to predict synaptic degeneration [13]. In addition, an increase in soluble brain Ab precedes plaque formation in Down syndrome brain [14]. Several different oligomeric Ab species have been identified both in vitro and in vivo and the Ab species responsible for neurodegeneration and synapse dysfunction has been suggested to be everything from Ab dimers up to large protofibrils [15,16,17,18,19,20,21,22,23,24,25,26,27]. The potential importance of these Ab species as targets for immunotherapy and biomarker assays emphasizes the need to study them in closer detail in vivo.
In this study the aim was to characterize the soluble pool of synthetic Ab as well as Ab derived from different biological samples under conditions as native as possible. Density gradient ultracentrifugation [28], unlike SDS-PAGE, size exclusion chromatography and ultra filtration, is a method where molecules are separated based on their size in a non solid matrix without any detergents or other denaturing agents. This approach is more likely to keep the Ab aggregates intact during the analyses. ELISA quantification of different Ab species in our centrifuged samples followed by structure analysis with atomic force microscopy (AFM) allowed us to divide the samples into four distinct fractions containing Ab of different size and appearance. From these analyses we could conclude that large Ab aggregates are the major Ab species in soluble extracts from AbPP transgenic and AD patient brains.

Characterization of fractionated synthetic Ab
Two different preparations of synthetic Ab were centrifuged in an optiprep density gradient and collected in four fractions (Fig. 1).
Synthetic Ab1-42, incubated for 30 min at 37uC, was used to obtain a wide range of soluble Ab aggregates of different sizes in contrast to freshly dissolved synthetic Ab1-40, which was used to represent monomeric and low-molecular weight Ab species. These Ab preparations were used to analyze how soluble Ab in different aggregation states separated in the optiprep gradient. Most of the freshly dissolved Ab1-40 was found in fractions 3 and 4, containing the smallest molecules ( Fig. 2A), confirming that the Ab1-40 preparation consists of monomers and smaller oligomers. Ab protofibril ELISA analysis of the fractionated freshly dissolved Ab1-40 revealed tiny amounts of Ab aggregates in fraction 2 but nothing in fraction 3 and 4, suggesting that aggregates present in fraction 3 are too small to be detected by the conformation dependent mAb158 of the protofibril specific ELISA. As seen in figure 2B, all four fractions of Ab1-42 contained Ab, most of which was found in fraction 2. Most Ab found in the two larger fractions was detected by the Ab protofibril specific ELISA, whereas the molecules in fraction 3 and 4 also in the Ab42 preparation were too small to be detected by the protofibril specific ELISA. Freezing of samples prior to analysis did not affect the results (data not shown).

Different fractions' influence on cell viability
After adjusting each fraction to an optiprep density of 1.5% and an Ab concentration of 0.1 mM, the four fractions of synthetic Ab1-42 were analyzed with an MTT assay to test the fractions' influence on PC12 cell viability. Interestingly, the two middle fractions, containing mAb158 positive aggregates (fraction 2) and smaller mAb158 negative oligomers (fraction 3), showed a significantly higher toxicity than the other two fractions (Fig. 2C).

Atomic force microscopy
To get an idea of the appearance of the fractionated Ab, the four fractions of synthetic Ab1-42 were analyzed with AFM. This analysis revealed the largest aggregates in fraction 1 with a height of 6 nm and a diameter of approximately 50 nm (Fig. 2D). The aggregates were then gradually smaller, both in height and diameter, in fraction 2 to 4 ( Fig. 2E-G). Estimated molecular weights, Mw, of the aggregates in the different fractions were calculated with equation 1 [29] (Table 1), based on the dimensions obtained from AFM analyses.
In this equation, r is the protein density [30], h is the height and r is the radius of the visualized structure, measured at half the height, to compensate for tip broadening effects [29]. N A is Avogadro's constant.
The smallest structures, with an estimated size corresponding to Ab monomers-tetramers, were to some extent visible in all fractions, possibly due to diffusion of small molecules over the gradient, caused by the high Ab concentration in the sample. For reference, a non centrifuged sample of the Ab1-42 preparation, aggregated for 30 min at 37uC, is displayed in Figure 2H, containing all types of Ab species present in the fractionated samples. Examples of these structures are marked with arrows and numbers referring to the fraction in question. As a reference for size, AbPP (approx. 100 kDa) and IgG (150 kDa) both ended up in fraction 2 (data not shown).

Fractionation of AbPP ArcSwe transgenic mouse brains
After verifying that the optiprep gradient is able to separate Ab aggregates according to their size, we continued to study how a biologically derived soluble Ab pool was separated in the same gradient. Homogenized brains from 10 month old AbPP ArcSwe transgenic mice, at this age displaying robust Ab plaque pathology [31], and non-transgenic mice were centrifuged and fractionated in the same way as synthetic Ab. Ab1-42 and Ab1-40 ELISA analyses revealed that the soluble Ab pool of transgenic mouse brain was primarily found to consist of larger aggregates ending up in fraction 2 and to some extent in fraction 3 with an Ab42:40 ratio of approximately 1:8 ( Fig. 3A-B). No Ab was detected in the non-transgenic mice (Fig. 3). The Ab in fraction 2 of the mouse brain homogenate was readily detectable with the Ab protofibril specific ELISA, whereas no Ab protofibrils were detected in fraction 3 (Fig. 3C). To further analyze the composition of the large Ab aggregates, immunoprecipitation was performed with the conformation dependent Ab protofibril selective antibody mAb158 [20], covalently bound to superparamagnetic Dynabeads. The immunoprecipitate was then analyzed with highly sensitive Ab1-40 and Ab1-42 ELISAs. As seen in figure 3D, immunoprecipitated material from a pool of fraction 2 from five mouse brain homogenates primarily contained Ab40, though with a lower Ab40:Ab42 ratio than the starting material. Still, this result confirms that aggregates in fraction 2 of transgenic mouse brain were mainly composed of Ab40. No Ab was immunoprecipitated from fraction 3 (data not shown), which is in agreement with the Ab protofibril ELISA analysis and implies that small oligomers have a different conformation than the larger protofibrillar Ab species and that mAb158 only detects the larger aggregates.

Fractionation of human brains
To study the soluble Ab pool in human brain, post-mortem temporal cortex tissue was taken from eleven individuals. Seven of these were post mortem confirmed AD, two of which were carriers of the Swedish or the Arctic mutation, respectively. In addition, three frontotemporal dementia (FTD) cases and one neurologically healthy control were included (table 2). Patients were diagnosed according to the NINCDS-ADRDA criteria [32]. Samples were homogenized, separated on the density gradient and fractioned as above. For the analyses, the FTD brains were included in the control group, since they were free of Ab pathology. Accordingly, the levels of soluble Ab40 and Ab42 in non-AD brains were close to zero ( Fig. 4A-D). All AD cases had high levels of Abx-42, especially in fraction 2, indicating a high content of large Ab aggregates (Fig. 4A). Also Abx-40 levels were elevated in fraction 2 of some AD cases, especially in the brain homogenate from the Swedish mutation carrier, but there was a large variability (Fig. 4B). In line with earlier observations [33], detection with the N-terminal 82E1 antibody revealed substantially lower levels of the full length Ab1-42 and Ab1-40 ( Fig. 4C-D), caused by a substantial degree of N-terminal truncation of soluble Ab in AD brain, with an average of 90% truncated Ab42 in AD brain and around 15% in non-AD brain (Fig. 4E).
Thus, although high levels of Ab42 were found in fraction 2 of the AD brains, no reliable Ab protofibril ELISA analysis could be carried out. Since the Ab protofibril ELISA is dependent on at least two epitopes with intact Ab N-termini to generate a signal, this truncation lead to a considerable underestimation of protofibril levels. Instead, the Ab aggregates in pooled material from fraction 2 of AD and non-AD brain were analyzed with mAb158 immunoprecipitation, requiring fewer intact N-terminal epitopes, followed by ultra sensitive Ab1-40 and Ab1-42 ELISA analysis. As displayed in figure 4F, the starting material contained around 40% Ab1-40 and 60% Ab1-42, whereas the immunoprecipitate contained 95% Ab1-42. This was also reflected in the supernatant, where the Ab42 level had dropped significantly, whereas Ab40 remained more or less the same. When performing the same analysis on fraction 3 of AD and non-AD brain, no Ab was immunoprecipitated (data not shown).

Discussion
Many different oligomeric Ab species have been described in the literature, but there is no consensus about which species actually exist and exert neurotoxic activity in the human brain. Small, naturally derived Ab oligomers have been suggested to potently cause synaptic failure [26] and neuritic degeneration [34], possibly via aggregation into large protofibrils [35], and large Ab aggregates in brain extracts and CSF have recently been associated with AD [27,36]. Immunotherapy with antibodies able to neutralize the toxicity of oligomeric Ab species have been  suggested as a future therapy of AD [37,38,39,40] and therefore, it is important to characterize the soluble Ab pool of i.e. brain extracts from AD patients.
Here, density gradient ultracentrifugation was used to investigate the size distribution and structure of soluble Ab from synthetic preparations and to compare them to different biological samples. This is a native and gentle method, which is important in order to maintain the structure of the Ab aggregates. Based on observations of centrifuged synthetic Ab, we could divide our samples into four distinct fractions, all containing Ab species of different size and with different appearances in AFM. We have previously reported that Ab protofibrils, recognized by our conformation dependent antibody mAb158, have an elongated structure when visualized by cryo-TEM [20]. Such protofibrils end up in fraction 2 when centrifuged on the same gradient as presented here (data not shown). Thus, the rounded shape of the larger aggregates seen in figure 2 was, although reported by others [41], somewhat unexpected. This could, to a certain degree, be an artifact caused by the attachment of samples to the mica surface, as this was not done in solution. Synthetic Ab aggregates found in fraction 1 and 2 were mAb158 positive, whereas Ab from fraction 3 and 4 were not, implying that the Ab aggregates are conformationally different. In agreement with previous observations using the same method [28], we found that synthetic Ab aggregates of intermediate size, found in fraction 2 and 3, exerted the highest toxicity to PC12 cells. Hence, the synthetic Ab preparation appears to contain two pools of neurotoxic Ab aggregates: a mAb158 positive pool of slightly larger aggregates and a mAb158 negative pool of smaller oligomers.
The different preparation of mouse and human brain tissue before density gradient centrifugation could potentially result in different relative amounts of Ab being present in these samples. Despite that, there was a striking resemblance between the biologically derived Ab, originating from AbPP ArcSwe transgenic mouse brain and from AD brain, with respect to size distribution and structure. Most Ab from these samples was found in fraction 2 and had an estimated size of 80-500 kDa. Ab from fraction 2 of the biological samples was recognized by mAb158, implying a common structure that can be immunologically discriminated from smaller oligomers, as shown by others [42,43,44]. Since mAb158 does not discriminate between Ab40 and Ab42 aggregates in vitro [45,46], this structure seems to be independent of which Ab peptide is the main constituent of the aggregates. As previously reported, Ab40 was the predominant Ab peptide in AbPP ArcSwe transgenic mouse brain [47] and this peptide was abundantly detected as mAb158 positive aggregates in fraction 2. The large amount of Ab40 incorporated in aggregates may be a result of the massive overproduction of Ab in general, caused by the Swedish mutation, in combination with the aggregation enhancing Arctic mutation. AD brains, on the other hand, contained mostly Ab42 as described earlier [12] although a subset of them, notably the Swedish mutation carrier, had high levels of Ab40, similar to the transgenic mice. Interestingly, Ab aggregates, immunoprecipitated with mAb158 from fraction 2 of pooled AD brain material, contained almost exclusively Ab42 despite a considerable amount of Ab40 in the starting material. This result suggests that while Ab42 in the AD brain is incorporated in large mAb158 positive aggregates, Ab40 may have ended up in fraction 2 as monomers or small, mAb158 negative aggregates, possibly in complex with other large molecules.
Although synthetic and biologically derived Ab may not be directly comparable, mainly because of differences in sample matrices, it is intriguing that most Ab found in our biological samples ends up in fraction 2, where some of the most toxic synthetic Ab species are found. This observation is in line with previous reports about size and function of Ab aggregates, both synthetic [35,41,46] and from human brain tissue [27,48].
In conclusion, although different in Ab peptide composition and amount of N-terminal truncation, soluble Ab aggregates from AbPP ArcSwe transgenic mouse brain and AD brain, mainly found in fraction 2, appear to be similar in both structure and size, as indicated by their binding to the protofibril selective, conformation dependent antibody. Furthermore, biologically derived soluble Ab aggregates resemble the neurotoxic aggregates found in fraction 2 of a synthetic Ab preparation. These insights may be of relevance for the further development of immunotherapy directed against soluble species of aggregated Ab.

Ethics statement
The use of human and mouse brain material was approved by the Regional ethical committee in Uppsala (decision numbers C223/8 and 2005-103). Written informed consent was obtained from all subjects (or their relatives) involved in the study.

Density gradient ultracentrifugation
A density gradient, for fractionation of samples by ultracentrifugation, was obtained using optiprep (Sigma-Aldrich, Stockholm, Sweden) diluted in phosphate buffered saline (PBS). The gradient was prepared by layering 0.65 ml of 50% optiprep at the bottom of a 4.9 ml OptiSeal TM Polyallomer centrifuge tube (Beckman Coulter, Bromma, Sweden) followed by 0.65 ml of 40%, 1.95 ml of 30%, 0.65 ml of 20% and 0.65 ml of 10% to give a final discontinuous gradient of 4.55 ml. A 0.25 ml aliquot of sample was carefully loaded onto the top of the gradient before centrifugation. An NVT65.2 rotor (Beckman Coulter) was used, and synthetic as well as biological samples were centrifuged at 384 0006g for 3 h at +4uC. Fractions (1-1.65 ml, 2-0.9 ml, 3-0.9 ml, 4-1.35 ml), were collected from the bottom of the tube, aliquoted and stored at 220uC until analysis.

Synthetic Ab samples
Synthetic Ab1-42 (American Peptide Company Inc., Sunnyvale, CA, USA), dissolved in 10 mM NaOH, diluted in 106 PBS to 443 mM (2 mg/ml), was incubated for 30 min at 37uC and centrifuged for 5 min at 17 9006g to remove any insoluble aggregates. It was then immediately applied to the density gradient for ultracentrifugation (The high concentration [443 mM] was required to perform the toxicity studies, where all fractions were diluted to the same final Ab concentration.). Synthetic Ab1-40 (American Peptide Company Inc.), dissolved in 10 mM NaOH, was diluted in 26 PBS to 50 mM immediately prior to centrifugation.

Human and mouse brain homogenates
Saline perfused brain hemispheres, with a weight of approximately 150 mg, from AbPP ArcSwe transgenic mice [31] (n = 5) and nontransgenic littermates (n = 5) were homogenized using a tissue grinder with teflon pestle (2610 strokes on ice) in tris buffered saline (TBS) (20 mM tris, 137 mM NaCl, pH 7.6 and Complete protease inhibitor cocktail (Roche, Bromma, Sweden)) in a 1:10 (tissue weight:extraction volume) ratio. Homogenates were centrifuged at 100 0006g at 4uC for 1 h to obtain a preparation of TBS-soluble extracellular and cytosolic proteins. Supernatants were aliquoted and stored at 280uC until analysis. Human brain samples of approximately 500 mg, obtained from temporal cortex, were homogenized 1:5 (tissue weight:extraction volume) in TBS (as above) and clear homogenate supernatants were obtained by centrifugation at 16 0006g for 1 h at 4uC.
Ab40 and Ab42 ELISA 96-well plates were coated with polyclonal antibodies specific for the C-terminal 40 or 42 Ab neo-epitopes respectively [12] and blocked with 1% BSA in PBS. Samples were denatured by boiling 5 min in 0.5% sodium dodecyl sulphate (SDS) to avoid impaired Ab quantification due to presence of aggregates [49]. After dilution 1:5 in ELISA incubation buffer (PBS, 0.1% BSA, 0.05% Tween) samples were added to the ELISA plates and incubated for 2 h. A one hour incubation with biotinylated 4G8 (Nordic Biosite, Tä by, Sweden), a monoclonal antibody binding to amino acid residue 18-22 within the Ab sequence, or 82E1 (IBL, Hamburg, Germany), specific for the N-terminal Ab neo-epitope (Ab amino acid residue 1-5), generated by BACE cleavage of AbPP, followed by streptavidin-HRP (Mabtech AB, Nacka Strand, Sweden), and tetramethyl-benzidine (TMB) (ANL produkter, Ä lvsjö, Sweden) were used for detection and the optical density was measured at 450 nm. Alternatively, samples were treated as above but analyzed with Ab1-40 ELISA kit Wako II or Ab1-42 ELISA kit Wako, High sensitivity (Wako Chemicals GmbH, Neuss, Germany) according to the manufacturer's instructions. The optiprep matrix or the SDS (0.1%) did not disturb ELISA quantifications (data not shown).

Ab protofibril specific ELISA
The protofibril specific sandwich ELISA, based on mAb158 both as capturing and detecting antibody and thus excluding the risk of detecting Ab monomers, is thoroughly described by Englund et al. [20]. In short, 96-well plates were coated with mAb158 and blocked with 1% BSA in PBS. Synthetic Ab and mouse brain samples were diluted in ELISA incubation buffer (PBS, 0.1% BSA, 0.05% Tween), whereas human samples were diluted in ELISA plasma diluent (Mabtech AB) to avoid interference from heterophilic antibodies [50]. Samples were incubated for 2 h at room temperature (RT) before addition of biotinylated mAb158. Streptavidin coupled HRP (Mabtech AB) and TMB (ANL produkter) were used for detection and the optical density was measured at 450 nm.
MTT cell toxicity assay PC12 cells [51] (obtained from A-L Svensson, Uppsala university [52]) were plated at a density of 10 000 cells/well in a 96 well plate (cell+, Sarstedts, Sweden) and incubated for 18 h at 37uC in RPMI 1640 medium (Invitrogen, Stockholm, Sweden) supplemented with 10% dialyzed FBS (fetal bovine serum, Invitrogen). Ab concentration in the different fractions was determined by Ab42 ELISA after SDS-denaturation (see above). New aliquots of the synthetic Ab42 fractions were thawed and added to the wells immediately after dilution to a final Ab concentration of 0.1 mM and an optiprep concentration of 1.5%. A set of optiprep samples (1.5%) without Ab was used as control and subtracted from the Ab toxicity. PBS and 0.005% H 2 O 2 were used as negative and positive controls respectively. Plates were incubated with samples and controls for 4 h at 37uC followed by addition of 3-(4,5-dimethylthiazol-2-yl)-2,5-diphenyltetrazolium bromide (MTT) at a final concentration of 0.5 mg/ml. Solubilization of the formazan product was achieved by addition of 100 ml/well of solubilization buffer (20% SDS in dimethylformamide, pH 4.8) followed by 24 h of incubation at 37uC. The formazan product was quantified by absorbance measurement at 570 nm. Three independent experiments were performed and standardized with an internal Ab control (Ab42, incubated 30 min at 37uC and centrifuged 5 min at 17 9006g).

Atomic Force Microscopy
AFM analyses were carried out with an XE-150 large sample AFM system (Park Systems Corp., Santa Clara, CA, USA) equipped with a 150 mm6150 mm XY scanner. All measurements were performed at ambient temperature in true non-contact mode using silicon based AFM probes (ACTA, AppNano, Santa Clara, CA). Fractions of ultracentrifuged synthetic Ab42 were diluted in PBS to a final concentration of 250 nM. A 10 ml aliquot of each sample was adsorbed to a freshly cleaved mica surface (Veeco, Cambridge, UK) over night at RT and then washed with distilled water and air dried. All analyses were performed with a scan rate of 1 Hz, with a set point between 0.93 and 1.79 mm and a Z servo gain between 1.07 and 1.83. Images were flattened and Sobel processed with XEI image analysis program (Park Systems Corp.). Sizes of the visualized structures were estimated using Equation 1, as explained in the results section.

mAb158 immunoprecipitation
The monoclonal Ab protofibril selective antibody mAb158 [20,45] was covalently attached to superparamagnetic beads with the Dynabeads coupling kit (Invitrogen). mAb158-coupled beads (5 ml) were added to pooled material of fraction 2 from AD brain (n = 7), non-AD brain (n = 4) or AbPP ArcSwe transgenic mouse brain (n = 3) in the presence of 0.05% Tween 20 and incubated 1 h on a shaker. After three washes in PBS with 0.05% Tween 20, beads as well as supernatant and the starting material were boiled 5 min in 0.5% SDS. The beads were removed with a magnet and samples were diluted 5 times in standard diluent and analyzed with Ab1-40 ELISA kit Wako II or Ab1-42 ELISA kit Wako, High sensitivity (Wako Chemicals GmbH) according to the manufacturer's instructions.