Soluble Amyloid-beta Aggregates from Human Alzheimer’s Disease Brains

Soluble amyloid-beta (Aβ) aggregates likely contribute substantially to the dementia that characterizes Alzheimer’s disease. However, despite intensive study of in vitro preparations and animal models, little is known about the characteristics of soluble Aβ aggregates in the human Alzheimer’s disease brain. Here we present a new method for extracting soluble Aβ aggregates from human brains, separating them from insoluble aggregates and Aβ monomers using differential ultracentrifugation, and purifying them >6000 fold by dual antibody immunoprecipitation. The method resulted in <40% loss of starting material, no detectible ex vivo aggregation of monomeric Aβ, and no apparent ex vivo alterations in soluble aggregate sizes. By immunoelectron microscopy, soluble Aβ aggregates typically appear as clusters of 10–20 nanometer diameter ovoid structures with 2-3 amino-terminal Aβ antibody binding sites, distinct from previously characterized structures. This approach may facilitate investigation into the characteristics of native soluble Aβ aggregates, and deepen our understanding of Alzheimer’s dementia.

cortical tissue with PBS + 0.45% CHAPS. Soluble Aβ aggregates were found primarily during the first and to a lesser extent the second extraction. Even after 5 extractions, further extraction using 0.5M Guanidine, a chaotrope known to solubilize Aβ plaques, yielded additional Aβ aggregates. This result indicated that the serial extractions in PBS + 0.45% CHAPS extracted a finite pool of Aβ but left a relatively insoluble pool of Aβ behind which could still be extracted with Guandine. D. PBS+0.45% CHAPS extraction yielded more soluble Aβ aggregates after initial homogenization (dark bars) and two-step homogenization (light bars) than PBS without CHAPS. Titration of concentrations for (A) SDS, (B) Triton™ X-100, and (C) Tween® 20, was used to homogenize cognitively normal cortical tissue spiked with 2 ng/mL of synthetic Aβ 1-42 . The resulting homogenates were immediately centrifuged at 100,000 x g RCF and assessed by ELISA for soluble Aβ aggregates. D. Soluble ADbrain derived Aβ monomers at 2 ng/mL were spiked into cognitively normal human cortical control tissue and homogenized in TBS, TBS+1% Triton X-100, or TBS+ 2% SDS. Ex vivo aggregation of Aβ monomers was observed with 1% Triton X-100 and 2% SDS detergents. Microcentrifuge tubes blocked with 1% BSA demonstrate a minimal loss of Aβ aggregate in a 100,000 x g RCF clarified cortical homogenate on ice or at room temperature. Conversely, unblocked tubes display a rapid loss of Aβ during incubation with moderate temperature dependence. Approximately 30% of the injected soluble Aβ aggregates from human AD brain lysate were lost during a single pass over the Superdex 200 10/300 column as demonstrated by the PBS only sample. This loss was greatly reduced with the addition of 0.05% BSA to the mobile phase buffer. No changes in the size distribution were observed. With SEC including bovine serum albumin in the mobile phase, the recovery of soluble Aβ aggregates was 86%. The recovery was calculated by measuring the concentration of soluble Aβ aggregates in the initial lysates and comparing it with the concentrations in each of the fractions. C. Soluble Aβ aggregates were found in the retentate without apparent change in size, but with loss >30% during concentration.  The correlations between area and number of immunogold labels for the insoluble Aβ aggregates were stronger than those between area and number of immunogold labels for the soluble Aβ aggregates (difference tests, p<0.05). The slopes of the correlations were between 1400 and 2400 nm 2 per gold particle.     As noted in the main manuscript, there are multiple differences between our method for isolating and purifying soluble Aβ aggregates from human brain compared with previous approaches. 1) We used an ELISA-based method to quantify soluble Aβ aggregates that does not cause ex vivo aggregation. In contrast, methods involving SDS such as gel electrophoresis likely cause both ex vivo aggregation and disaggregation, making them inaccurate reflections of the quantity and size of the aggregates. 2) We quantitatively tracked the loss of soluble Aβ aggregates at each step in the procedure, and used albumin blocking to prevent nonspecific loss. Previous preparative approaches that did not involve quantitative bookkeeping or systematic albumin blocking likely resulted in substantial and unknown amounts of nonspecific loss of soluble Aβ aggregates. Thus, the properties of the soluble Aβ aggregates previously characterized may have reflected those of minority species that were especially resistant to nonspecific binding. 3) We have detected bona fide full length Aβ in the soluble aggregates, based on mass spectrometry. In contrast, previous methods used antibodies with incomplete specificity (e.g. 6E10 which binds to amyloid precursor protein fragments, A11 which binds to oligomeric forms of other proteins). Previous mass spectrometric characterization was limited to mid-domain and C-terminal digested peptides, which could also have arisen from amyloid precursor protein fragments.
There are many unanswered questions arising from the data presented: Are there other classes of soluble Aβ aggregates without canonical N-termini exposed? We can envision in theory several fundamentally distinct types of soluble Aβ aggregates: A) Aggregates with two or more canonical Aβ N-termini exposed, which would have been detected by the methods used here; B) Aggregates with only one canonical Aβ N-terminal exposed, which would not have been quantified by our ELISA but still could have been immunoprecipitated and detected by electron microscopy. Indeed, this could explain why some of the soluble Aβ aggregates in our preparation had only 1 immunogold label (Fig. 6E); C) Aggregates with no canonical Aβ N-termini exposed, which could have been immunoprecipitated if the HJ5.1 epitope (mid domain) were exposed but would not have been detected by any of the methods used here. Thus, a priority for future research will be a broader exploration of the spectrum of Aβ aggregates present in the human AD brain using other detection reagents. We are in the process of characterizing antibodies to other Aβ epitopes (including truncated and post-translationally modified forms), and we plan to repeat the immunoprecipitations and immunoelectron microscopic characterizations using additional appropriate antibodies. A recent report indicates that the peri-plaque Aβ species in transgenic mouse brains appear to be quite heterogeneous with regard to antibody binding, with only partially overlapping subsets binding several aggregate-selective and epitope specific antibodies 24 .
Why does the low molecular weight fraction contain a few particles which bind 2 anti-Aβ antibodies on electron microscopy, while this fraction yielded no detectible signal on the HJ3.4-HJ3.4 soluble Aβ aggregate ELISA? This could represent an incomplete sensitivity of the ELISA, or alternatively could indicate that these species have both Aβ epitopes locked in a relatively rigid conformation with both facing the same direction. A (hypothetical) species with 2 epitopes held in a relatively rigid parallel orientation during ELISA could be bound to the capture antibody but not by the detection antibody, accounting for a potential false negative result on our ELISA. Does the CHAPS used in our extraction procedure alter structural or functional properties of the soluble Aβ aggregates? Additional preparations using larger amounts of tissue but excluding CHAPS will be required to address this concern.
Why are the electron microscopic-based sizes of the soluble Aβ aggregates seen in our preparation larger than those seen in Noguchi et al 9 ? It is possible that sub-aggregates were broken apart Noguchi et al's preparation or clustered together during ours.
Why are portions of the surfaces of the soluble Aβ aggregates not immunoreactive with the anti-Aβ antibody? It is possible that there are other protein constituents, or Aβ species that do not have intact N-termini and therefore would not be immunoreactive with HJ3.4. Further investigations will be needed using a panel of monoclonal antibodies directed at different Aβ epitopes as well as other protein constituents once they are identified.
What is the relationship between soluble Aβ aggregates and tau pathology? Increasing evidence implicates tau pathology as a close correlate of neurodegeneration in Alzheimer's disease [25][26][27] http://www.alzforum.org/news/conference-coverage/tau-pet-studies-agree-tangles-follow-amyloid-precedeatrophy. Tau pathology may be downstream of Aβ 28,29 in transgenic mice, where tau is required for APP transgene-related behavioral deficits 30 likely via tau coupling to the kinase Fyn 31 . However, investigations of the mechanisms underlying the relationship between Aβ aggregates and tau pathology have mainly relied on synthetic Aβ aggregate preparations and transgenic mice. The effects of human AD brain-derived soluble Aβ aggregates on tau pathology remain to be determined and represents an important future direction.