A Specific Enzyme-Linked Immunosorbent Assay for Measuring β-Amyloid Protein Oligomers in Human Plasma and Brain Tissue of Patients With Alzheimer Disease

Objective  To examine in vivo levels of β-amyloid (Aβ) oligomers (oAβ) vs monomeric Aβ in plasma and brain tissue of patients with sporadic and familial Alzheimer disease (AD) using a new enzyme-linked immunosorbent assay (ELISA) specific for oAβ.

Design  To establish the oAβ ELISA, the same N-terminal Aβ antibody was used for antigen capture and detection. Plasma and postmortem brain tissue from patients with AD and control subjects were systematically analyzed by conventional monomeric Aβ and new oAβ ELISAs.

Subjects  We measured oAβ species in plasma samples from 36 patients with clinically well-characterized AD and 10 control subjects. In addition, postmortem samples were obtained from brain autopsies of 9 patients with verified AD and 7 control subjects.

Main Outcome Measures  Oligomeric Aβ and 4 monomeric Aβ species in plasma samples from patients with AD and control subjects were measured by ELISA.

Results  The specificity of the oAβ ELISA was validated with a disulfide–crossed-linked, synthetic Aβ_1-40Ser26Cys dimer that was specifically detected before but not after the dissociation of the dimers in β-mercaptoethanol. Plasma assays showed that relative oAβ levels were closely associated with relative Aβ₄₂ monomer levels across all of the subjects. Analysis of sequential plasma samples from a subset of the patients with AD, including a patient with AD caused by a presenilin mutation, revealed decreases in both oAβ and Aβ₄₂ monomer levels over a 1- to 2-year period. In brain tissue from 9 patients with AD and 7 control subjects, both oAβ and monomeric Aβ₄₂ levels were consistently higher in the AD cases.

Conclusions  An oAβ-specific ELISA reveals a tight link between oAβ and Aβ₄₂ monomer levels in plasma and brain. Both forms can decline over time in plasma, presumably reflecting their increasing insolubility in the brain.

Alzheimer disease (AD) is characterized by the progressive accumulation of extracellular amyloid plaques and intracellular neurofibrillary tangles. The protein subunit of the amyloid plaques, β-amyloid (Aβ), does not occur as a single molecular species; many different Aβ-containing peptides have been detected in human cerebrospinal fluid (CSF) and/or brain tissue.^1,2 The most common Aβ isoform in vivo is Aβ_1-40, ie, a peptide that begins at Asp1 and terminates at Val40 of the Aβ region of amyloid precursor protein (APP). Increased accumulation of Aβ_1-42, a peptide that differs from Aβ_1-40 by the inclusion of Ile41 and Ala42, is particularly associated with development of AD. The extra 2 hydrophobic amino acids of Aβ₄₂ greatly enhance its aggregation propensity,³ leading to accelerated formation of small (low-n) Aβ oligomers (oAβ), larger intermediate assemblies like protofibrils, and eventually the typical approximately 8-nm amyloid fibrils found abundantly in neuritic plaques and amyloid-bearing microvessels. Small, soluble oligomers of Aβ have been linked to neuronal toxic effects and synaptic failure (for review, see the article by Walsh and Selkoe⁴). How the oligomeric assemblies reach a balance with monomeric Aβ and large protofibrils and fibrils in the human brain is under investigation.

Studies of circulating Aβ in blood have provided insights into Aβ equilibrium between the brain and the periphery. A few studies have associated increased levels of plasma Aβ₄₂ with AD at different stages. For example, an increase in the plasma Aβ₄₂ level was associated with conversion from normal cognition to mild cognitive impairment (MCI) and on to AD, albeit with unsatisfactory sensitivity and specificity.⁵ Another study found that patients with AD at baseline and those who developed AD later had significantly higher plasma Aβ₄₂ levels; some of the patients with AD showed elevated levels of Aβ₄₂ and Aβ₄₀ before and during the early stages of AD, but plasma levels declined thereafter.^6,7 Another study showed that nondemented subjects with high levels of plasma Aβ₄₂ were more than twice as likely to develop AD than those with low levels and that patients with AD showed higher Aβ₄₂ levels than control subjects without AD.⁸

Patients with familial AD (FAD) having mutations in the presenilins, the catalytic subunits of the γ-secretase complex that generates Aβ, have increased plasma levels of Aβ₄₂.⁹ Plasma Aβ₄₂ levels and the Aβ₄₂/Aβ₄₀ ratio were higher even in presymptomatic subjects carrying FAD mutations in presenilin 1 (PS1) or APP, and Aβ₄₂ levels may decrease with disease progression prior to symptom onset.¹⁰ Elevated plasma Aβ₄₂ levels have been linked to a locus on chromosome 10 in some typical (late-onset) AD cases,¹¹ and some first-degree relatives of patients with late-onset AD have elevated Aβ₄₂ levels. These suggest that an increased plasma Aβ level is a heritable trait.^12,13

Plasma Aβ is a potentially promising but understudied candidate marker for diagnosis and preclinical prediction. However, plasma Aβ₄₀ or Aβ₄₂ was found not to be an optimal candidate in unbiased proteomic searches for AD fluid biomarkers.^14,15 In one study,¹⁶ increased plasma Aβ₄₂ levels were detected in patients with MCI, but a significant association was only observed in women. In a cohort of men at age 70 years, plasma Aβ₄₀ and Aβ₄₂ were not associated with incident AD at follow-up, whereas low plasma Aβ₄₀ levels in another cohort of men at age 77 years were associated with a higher incidence of AD.¹⁷ Yet another study¹⁸ found that subjects with low plasma Aβ₄₂/Aβ₄₀ ratios had a higher risk of MCI or AD and greater cognitive decline. A correlation between increased plasma Aβ₄₀ levels and an increased risk of dementia has also been reported.¹⁹ A recent study of patients with MCI followed up for 7 years showed no significant difference of plasma Aβ species between patients with MCI who later developed AD and patients with stable MCI or healthy control subjects.²⁰

These often inconsistent reports on the association of plasma Aβ levels with AD may reflect the fact that measurements to date only represent the pools of monomeric Aβ and were measured by different Aβ enzyme-linked immunosorbent assays (ELISAs).²¹ To understand the relationship among different Aβ species in vivo, we developed an ELISA that can detect oAβ and simultaneously measured both Aβ monomers and oAβ in human plasma or postmortem brain tissue. Levels of human oAβ detected by our new oAβ-specific ELISA were closely associated with the levels of monomeric Aβ₄₂. Levels of both soluble and insoluble Aβ₄₂ and oAβ species were significantly higher in brain tissue of patients with AD compared with those in brain tissue of control subjects. We also observed decreases in plasma Aβ levels in follow-up samples from the same patients 1 to 2 years later. The development of an oligomer-specific ELISA applicable to plasma extends Aβ measurement to a highly relevant neurotoxic form of this pathogenic peptide.

Blood samples were collected in potassium-EDTA–containing collection tubes and centrifuged at 1600g for 15 minutes. The plasma supernatant was aliquoted and stored at −80°C until measured. Average ages for patients with AD and control subjects at the time of blood drawing were 72 years (n = 36) and 62 years (n = 10), respectively. Brain lysates from postmortem human brains (n = 16) were prepared as recently described.²²

Sandwich ELISAs for monomeric Aβ were performed as described.²³ The use of C-terminal capturing antibodies and N-terminal or midregion detecting antibodies has been a standard format for measuring monomeric Aβ species in many studies.^9,24-26 The capture antibodies 2G3 (to Aβ residues 33-40) and 21F12 (to Aβ residues 33-42) were used for Aβ₄₀ and Aβ₄₂ species, respectively. The detecting antibodies were biotinylated 3D6 (to Aβ residues 1-5) for Aβ_1-40/42 or biotinylated 266 (to Aβ residues 13-28) for Aβ_x-40/42 species. These antibodies were kindly provided by Peter Seubert, PhD, and Dale Schenk, PhD (Elan Corp, plc, South San Francisco, California).

To detect oAβ species, the same N-terminal antibody, either 82E1 (to Aβ residues 1-16; Immuno-Biological Laboratories, Inc, Minneapolis, Minnesota) or 3D6, was used for both capture and detection. A sandwich ELISA procedure identical to that for the traditional monomer assays was followed to measure relative oAβ levels, which were calculated using standard curves of synthetic Aβ_1-40 peptide captured by 2G3 antibody and measured by the same detecting antibodies (82E1 or 3D6).

Chinese hamster ovary cells stably expressing wild-type (wt) human APP751 plus either wt or mutant PS1 (resulting in lines PS1_WT-1, PS1_WT-2, PS1_WT-3, PS1_M146L-1, PS1_M146L-2, PS1_M146L-3 [M146L missense mutation], and PS1_C410Y-1, PS1_C410Y-2, PS1_C410Y-3 [C410Y missense mutation]) were maintained in 200-μg/mL G418 (Invitrogen Corp, Carlsbad, California) plus 25-μg/mL puromycin (for PS1). We also examined Chinese hamster ovary cell lines singly transfected with wt APP₇₅₁ or APP₇₅₁ bearing the V717F (“Indiana”) missense mutation. Cells were incubated in methionine-free, fetal bovine serum–free media for 45 minutes before labeling with 200-μCi/mL sulfur 35 ([³⁵S])–labeled methionine for 38 hours (to convert microcuries per milliliter to becquerels per milliliter, multiply by 37 000). Conditioned media were collected and immunoprecipitation was performed as described.²⁵

β-Amyloid dimers (dAβ) were generated by atmospheric oxidation of a 20μM solution of synthetic Aβ_1-40Ser26Cys in 20mM ammonium bicarbonate, pH 8.0, for 4 days at room temperature. To facilitate disassembly of aggregates formed during the oxidation reaction, the peptide solution was lyophilized and the lyophilate was incubated in 5M guanidine hydrochloride, 50mM TRIS hydrochloride, pH 8.0, for 4 hours. Disulfide– crossed-linked dAβ were isolated from unreacted monomer and higher aggregates by size exclusion chromatography using a Superdex 75 10/30 high-resolution column (GE Healthcare, Milwaukee, Wisconsin) eluted with 50mM ammonium acetate, pH 8.5, at a flow rate of 0.8 mL/min. Fractions (0.5 mL) were collected, an aliquot of each was electrophoresed on 16% TRIS-tricine polyacrylamide gels, and protein was detected by silver staining. Size exclusion chromatography fractions found to exclusively contain dAβ were pooled and used as the dimer stock. The concentration of peptides in stock solutions was determined by comparison with wt Aβ_1-40 of known concentration. Once collected, all of the samples were stored at −80°C until used. To disrupt the disulfide bonds linking the monomers, dAβ were treated with 3% β-mercaptoethanol (βME), followed by serial dilutions for quantification by ELISA. The diluted, residual amount of βME did not interfere with the capture and detecting antibodies. The Aβ_1-40Ser26Cys was synthesized by the Biopolymer Laboratory, Department of Neurology, UCLA Medical Center, Los Angeles, California, and the correct sequence and purity were confirmed by amino acid analysis, reverse-phase high-performance liquid chromatography, and mass spectrometry.

To analyze oAβ species in human samples, we established a sensitive and specific oAβ ELISA. The assay relies on the use of a single monoclonal antibody for both capture and detection. Thus, for oAβ to be detected by this sandwich ELISA, an Aβ assembly must contain at least 2 exposed copies of the same epitope that is accessible by the identical capturing and detecting antibody.^27,28 This means that this assay will recognize only Aβ assemblies that contain at least 2 Aβ molecules. Two monoclonal antibodies, 82E1 and 3D6, that each recognize the extreme N-terminus of human Aβ were tested.

To confirm the specificity of the ELISA, we used a synthetic Aβ peptide, Aβ_1-40Ser26Cys, that is capable of reversibly forming covalently cross-linked dAβ under oxidizing conditions.^22,29 Disulfide–crossed-linked dAβ were separated from unreacted monomer and higher aggregates by size exclusion chromatography. The concentration of dAβ in stock solutions was determined by comparison with wt synthetic Aβ_1-40 of known concentration; both dAβ and monomeric wt Aβ peptides were detected by sodium dodecyl sulfate–polyacrylamide gel electrophoresis and silver staining (Figure 1A). Based on the densitometric signal obtained from synthetic wt Aβ_1-40 peptide, we estimated that the concentration of our synthetic dAβ stock was 81μM. In the presence of 3% βME, the disulfide bonds linking the monomers that form dAβ were disrupted, producing a reduced monomer (Figure 1B). In the absence of βME, no monomeric Aβ was detected and the vast majority of Aβ species were dAβ (Figure 1B).

Serial dilutions of dAβ were made after the dAβ stock was treated with or without βME, and the diluted samples were assayed using either antibody 82E1 (Figure 1C) or antibody 3D6 (Figure 1D) in a sandwich format. When the relative levels of the diluted dAβ (800pM) of the dAβ stock were normalized to 1, either monoclonal antibody detected a clear linear reduction in the levels of dAβ when further diluted to 400pM and 200pM (Figure 1C). In the presence of βME, dissociation of the disulfide bond markedly reduced the amount of dAβ signal; the levels of remaining dAβ were less than 10% of the same fraction without βME, with antibody 82E1 showing even greater specificity than antibody 3D6 in this regard (Figure 1C and D). Therefore, our new ELISA accurately measured the levels of dAβ in a linear fashion and is highly sensitive to the dissociation of dAβ to monomers.

Using the oAβ-specific ELISA, we screened plasma samples obtained from 36 patients with well-characterized AD and 10 control subjects. The ages of the patients with AD ranged from 50 to 90 years at the time of blood sampling, with an average age of 72 years. The age of control subjects ranged from 52 to 68 years, with an average age of 62 years. Conventional Aβ ELISAs were applied to measure monomeric Aβ species.

We analyzed plasma levels of oAβ as a function of subject age but did not observe an age-dependent alteration of oAβ levels (Figure 2). There were a number of subjects who carried high plasma levels of oAβ-reactive species, with a wide age distribution between 60 and 90 years. Overall, we found that most control subjects (7 of 10 subjects) had plasma levels of oAβ below our detection limit, whereas more than half of the patients with AD (19 of 36 patients) had detectable oAβ levels (Figure 2).

From the same aliquot of plasma, we measured 4 monomeric Aβ species using 4 distinct ELISAs: Aβ_1-40, Aβ_1-42, and N-terminally heterogeneous Aβ₄₀ and Aβ₄₂ species (ie, Aβ_x-40 and Aβ_x-42). The levels of Aβ_1-40 and Aβ_x-40 did not reveal a clear separation of patients with AD from control subjects (Figure 3A and B). On average, the levels of Aβ_x-40 were higher than those of Aβ_1-40 (note scales on the ordinates of Figure 3A and B), indicating that a variable portion of plasma Aβ₄₀ is N-terminally truncated. When the levels of Aβ_x-42 (Figure 3C) and Aβ_1-42 (Figure 3D) were analyzed from the same aliquots of plasma, we found that the levels of Aβ_1-42 species were generally lower than those of Aβ_1-40 species and that the levels of Aβ_x-42 were likewise lower than those of Aβ_x-40 (compare Figure 3A and B with Figure 3C and D).

We calculated the average plasma Aβ levels for each of the 4 ELISAs in all of the patients with AD and control subjects. Plasma levels of Aβ₄₀ and Aβ_x-42 did not differ significantly between patients with AD and control subjects (Figure 4A). However, the average plasma levels of both Aβ_1-42 and oAβ were found to be significantly higher in patients with AD than in control subjects (Figure 4A). Furthermore, the relative levels of Aβ_1-42 and oAβ were closely associated. For this, we chose a threshold of 10pM for Aβ_1-42 (ie, readily detectable above baseline) and identified 7 subjects who carried levels of Aβ_1-42 greater than 10pM (Figure 4B). Next, we identified 7 subjects who carried higher levels of oAβ than the remaining subjects (Figure 4C). We found that the subjects who carried levels of Aβ_1-42 greater than 10pM (Figure 4B) were the same subjects who carried high levels of oAβ (Figure 4C), and the relative levels of Aβ_1-42 and oAβ for each subject were tightly linked. Except in 1 patient who had the highest oAβ level (Figure 4C) but had a relatively low Aβ_1-42 level (Figure 4B), our oAβ ELISA clearly detected oAβ-reactive species whose levels were tightly associated with those of monomeric Aβ_1-42.

To further validate the close association of monomeric Aβ₄₂ with oAβ detected by our oAβ ELISA, we analyzed these Aβ species generated from cell lines expressing FAD-causing mutations in either APP or PS1. These cell lines have elevated Aβ₄₂ monomer levels and secrete sodium dodecyl sulfate–stable, low-n Aβ oligomers into the media.^25,30,31

Similar to our AD and control cases that carry different levels of plasma Aβ, we have found variable levels of Aβ produced from multiple stable cell lines expressing wt or mutant (M146L or C410Y) PS1 and wt or mutant (V717F) APP. These lines were metabolically labeled with [³⁵S]methionine, and the conditioned media were immunoprecipitated with antibody 21F12 (to Aβ ending at residue 42) or 1282 (a pan-Aβ polyclonal antibody) and analyzed by gel fluorography. Antibody 21F12 immunoprecipitated both monomeric Aβ₄₂ and the p3₄₂ species (generated by sequential α- and γ-secretase cleavages of APP) from the conditioned media of all of the cell lines. In media from APP_V717F and PS1_M146L-2 cells, additional Aβ₄₂-immunoreactive bands migrating at approximately 5 kDa, 8 to 10 kDa, and 12 to 14 kDa (collectively designated sodium dodecyl sulfate–stable, low-n oligomers) were also detected (Figure 5A). Light exposure of blots separating Aβ₄₂ monomers from oAβ indicated that more Aβ₄₂ was produced from cells expressing APP_V717F or mutant PS1 (PS1_M146L-2) (Figure 5B). The levels of oAβ were correlated with the levels of monomeric Aβ₄₂, ie, the cell lines secreting elevated Aβ₄₂ also had higher levels of oligomers in their media. We further examined the levels of total Aβ precipitated by antiserum 1282 but did not observe an elevation in total Aβ signal corresponding to those of Aβ₄₂ and oAβ. Thus, the amounts of total Aβ generated by cells expressing APP_V717F or mutant PS1 (PS1_M146L-2) were comparable to those from the other cell lines (Figure 5C). Moreover, no oAβ species were detected in these immunoprecipitates of total Aβ, suggesting that the oAβ species made by the cells are principally composed of Aβ₄₂. These results in well-defined cultured cell lines support our finding in human plasma that levels of oAβ are closely associated with those of monomeric Aβ₄₂.

The close association of monomeric Aβ₄₂ and oAβ levels suggests a dynamic conversion between these 2 Aβ species. To further examine the relationship of Aβ₄₂ and oAβ in vivo, we obtained sequential blood samples from a patient with symptomatic FAD carrying a mutant PS1. Blood samples were drawn at ages 58 and 60 years. Levels of Aβ were quantified by separate ELISAs for oAβ and the 4 monomeric Aβ species (Table). For oAβ species, we observed a significant decrease such that oAβ was no longer detectable at age 60 years. For monomeric Aβ, the plasma concentrations of Aβ_1-40 and Aβ_x-40 were 48pM and 84pM, respectively, at age 58 years. The concentration of Aβ_1-42 was 7pM, and that of Aβ_x-42 was 36pM. The plasma Aβ₄₂/Aβ_total ratio was in the same range as those reported in patients carrying FAD-causing PS1 mutations.⁹ Plasma samples collected 22 months later had significantly lower Aβ concentrations: the level of Aβ_1-40 measured 36pM (25% reduction), and Aβ_1-42 was now undetectable (Table). Concentrations of Aβ_x-40 and Aβ_x-42 declined to 48pM (43% reduction) and 15pM (58% reduction), respectively. Thus, while the levels of both Aβ species decreased over 2 years, Aβ₄₂-ending species were reduced more dramatically than Aβ₄₀-ending species. The complete loss of the Aβ_1-42 and oAβ signals is consistent with our findings that the levels of these 2 species are closely associated.

We also obtained sequential blood samples from a subset of our patients with AD (n = 12, including the patient with FAD described earlier) and compared their Aβ levels over a 1- to 2-year span. All of the 5 Aβ species were again measured by ELISA and the changes in Aβ levels were calculated. A similar number of subjects showed increased or decreased levels of Aβ_1-40 over the course of 1 to 2 years (Figure 6A), failing to show a clear temporal trend in Aβ_1-40. Comparison of Aβ_x-40 levels revealed that 3 of 11 subjects had an increased Aβ_x-40 level and the remaining 8 subjects had a reduced Aβ_x-40 level (Figure 6B). In the 4 patients with AD who had detectable plasma Aβ_1-42 levels (Figure 6C), 3 subjects showed a dramatic reduction in the Aβ_1-42 level after 1 year but 1 subject had almost identical levels of Aβ_1-42 at both times. The latter subject also maintained similar levels of oAβ after 1 year, whereas 4 patients with AD with detectable oAβ levels had a reduction in the relative plasma oAβ level (Figure 6E). Thus, the decrease of Aβ_1-42 levels occurred in subjects who also showed a similar decrease in oAβ levels. Strikingly, all but 1 subject showed decreases in Aβ_x-42 in the second plasma samples (Figure 6D).

To investigate the association of monomeric Aβ₄₂ and oAβ directly in human brains, we obtained postmortem brain tissue from 9 patients with AD and 7 age-matched control subjects distinct from the subjects examined in the plasma Aβ studies described earlier. We performed sequential extractions of brain tissue to obtain “soluble” (TRIS-buffered saline [TBS]) and “insoluble” (guanidine hydrochloride) extracts and measured their Aβ contents by ELISA.

The soluble pool of brain Aβ extracted by TBS had relatively low levels of Aβ_1-40 (Figure 7A); an average of 1.1 pmol/g of Aβ_1-40 was detected in the patients with AD vs 0.7 pmol/g in the control subjects (P = .14). The average level of TBS-soluble Aβ_1-42 was significantly higher in patients with AD than in control subjects (Figure 7B). Likewise, the relative levels of TBS-soluble oAβ were significantly higher in patients with AD as most control subjects had almost undetectable levels of soluble oAβ (Figure 7C).

After the extraction of soluble Aβ by TBS, the resultant pellets were further extracted in guanidine hydrochloride to obtain the insoluble pools of Aβ. Overall, insoluble Aβ_1-42 levels (Figure 7E) were several hundred–fold higher than soluble Aβ_1-42 levels (Figure 7B), and they were almost 10-fold higher than insoluble Aβ_1-40 levels (Figure 7D). This is consistent with numerous previous reports that amyloid deposits in brain tissue of patients with AD are mainly composed of Aβ₄₂. The levels of both insoluble Aβ_1-40 and insoluble Aβ_1-42 were significantly higher in brain tissue from patients with AD than those in brain tissue from control subjects. Furthermore, a substantial and significant increase in the relative levels of insoluble oAβ was observed in brain tissue of patients with AD (Figure 7F). Importantly, brain tissue levels of insoluble Aβ_1-42 were closely associated with the levels of oAβ in both patients with AD and control subjects (for AD: correlation coefficient = 0.88; for control subjects, correlation coefficient = 0.98). For example, 1 patient with AD (patient 4) showed relatively low levels of Aβ_1-42, and accordingly the oAβ level was low. A control subject (subject 3) showed a relatively high level of Aβ_1-42, corresponding to a high level of oAβ (Figure 7F). On average, significantly higher levels of all of the 3 insoluble Aβ species (Aβ_1-40, Aβ_1-42, and oAβ) were found in brain tissue of patients with AD compared with those in brain tissue of control subjects (Figure 7D-F).

Given the burgeoning evidence that small oligomeric assemblies of Aβ may be principally responsible for neuronal dysfunction in AD, it is important to be able to specifically detect and quantify these species in a range of biological samples. Conventional approaches to measuring the levels of oAβ have focused on semiquantitative analysis by Western blotting or silver staining of sodium dodecyl sulfate–polyacrylamide gel. Recent advances in nanotechnology have enabled the measurement of subpicomolar concentrations of oligomers (also designated Aβ-derived diffusible ligands) in human CSF.³² For ELISA-based approaches, several groups have generated antibodies believed to be specific for aggregated Aβ. Among them, monoclonal antibody NAB61 specifically recognizes a complex conformational Aβ_1-11 epitope on peroxynitrite- or UV light–treated synthetic Aβ, with much less reactivity for monomeric Aβ. Using NAB61 as a capture antibody and another Aβ antibody (BA27) as a reporter, this sandwich ELISA can differentiate UV–cross-linked synthetic Aβ from monomeric synthetic Aβ with a relative sensitivity of 5:1.³³ Another antibody (158) recognizes 50- to 200-kDa protofibrillar aggregates of synthetic Aβ but not low-molecular-weight oligomers and monomers of synthetic Aβ.³⁴

In this study, we oxidized synthetic Aβ_1-40Ser26Cys peptide to generate disulfide–crossed-linked dimers to test our new oAβ ELISA. The Aβ_1-40Ser26Cys dimer preparation was treated with guanidine hydrochloride and therefore was free of any higher-molecular-weight aggregates (Figure 1B). Subsequent separation of the stable dimers from un–cross-linked monomers was achieved by size exclusion chromatography. Detection of the synthetic dimers only in the absence of βME clearly demonstrated the specificity of our oAβ ELISA, with far less ability to detect monomers.

Using this new oAβ ELISA and previously established monomeric Aβ ELISAs, we obtained several lines of evidence for a close association between the levels of monomeric Aβ₄₂ and oAβ species in human plasma and brain tissue. First, the close association of Aβ_1-42 and oAβ was observed in a subset of patients with AD who had readily detectable levels of plasma Aβ_1-42 (Figure 4). The average levels of these 2 Aβ species were significantly higher in the patients with AD than in the control subjects. Indeed, only 1 of 10 control subjects showed such readily detectable levels of Aβ_1-42. Second, our studies of brain tissue from patients with AD and control subjects showed the same pattern, ie, levels of insoluble Aβ_1-42 associated closely with the relative levels of insoluble oAβ across individual cases (Figure 7). Third, among all of the patients with AD and control subjects, higher levels of oAβ were not linked to lower levels of monomeric Aβ₄₂, suggesting that the conversion of monomeric Aβ into oAβ per se is not the major contributor to the observed reduction of monomeric Aβ₄₂. Overall, our oAβ-specific ELISA allowed us to establish a close quantitative relationship between the levels of Aβ_1-42 and oAβ in human plasma and brain tissue.

This association was also observed in cultured cells expressing human APP. We found that the appearance of soluble oAβ in the medium occurred exclusively in those clonal cell lines with significantly increased Aβ₄₂ monomer production (Figure 5). The FAD mutations in PS1 or APP lead to enhanced Aβ₄₂ generation, and the occurrence of oAβ is attributable to enhanced γ-secretase cleavage of APP at the 42nd residue of Aβ, which is facilitated by AD-causing mutations in either the substrate (APP) or the protease (PS1). Thus, levels of Aβ₄₂ are closely associated with the formation of oAβ.

Our longitudinal comparisons of individual patients with AD provide new insight into the dynamic changes in plasma Aβ levels without the problem of intersubject variation. Using monomeric Aβ and oAβ ELISAs, we have provided 2 snapshots of Aβ levels within a relatively short period. Whereas we saw no clear directional change of Aβ_1-40 levels, the levels of the remaining 4 Aβ species all showed a reduction over the course of 1 to 2 years. About three-quarters of cases showed a reduction in Aβ_x-40 levels. Among the cases with detectable plasma levels of Aβ_1-42 and oAβ, all but 1 showed a decrease in Aβ levels. In the case of Aβ_x-42, 9 patients showed a reduction over the 1- to 2-year period, whereas 1 patient showed an increase. Importantly, the individuals with decreasing Aβ_1-42 monomer levels were the same subjects who showed decreases in oAβ levels.

While our oAβ ELISA provides an accurate method to measure such species in human blood, our results do not yet validate any one Aβ species as a biomarker for AD. Currently, the CSF tau/Aβ₄₂ ratio has been the best predictor of the development of AD-type cognitive decline in still-nondemented subjects.^35,36 Changes in the levels of tau and Aβ₄₂ in CSF may reflect dysfunction in the cerebrum that can be measured by the electroencephalographic rhythm,³⁷ and low Aβ₄₂ levels in CSF correlate well with positive Pittsburgh Compound B uptake by positron emission tomographic scanning.³⁵ Precise measurement of Aβ in plasma and CSF and subsequent correlation with levels in postmortem brain tissue should yield a clearer picture of Aβ metabolism in vivo. Mathematic modeling of the equilibrium between monomeric Aβ and oAβ would also help elucidate the catabolic turnover of Aβ in the peripheral and central nervous systems.³⁸ Considering a wide range of factors that could contribute to variations in plasma Aβ levels, it is currently difficult to obtain a clear separation of patients with AD from control subjects simply by measuring the levels of plasma Aβ species.³⁹ However, our findings suggest that measuring plasma Aβ and oAβ over 1 to 2 years or more can reveal a significant reduction in plasma Aβ levels, especially Aβ₄₂ levels, and this finding raises the possibility of a direct relationship of plasma Aβ to brain amyloid formation.

Correspondence: Weiming Xia, PhD, Center for Neurologic Diseases, Harvard Institutes of Medicine, HIM 616, 77 Ave Louis Pasteur, Boston, MA 02115 (wxia@rics.bwh.harvard.edu).

Accepted for Publication: August 21, 2008.

Author Contributions: All of the authors had full access to all of the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis. Study concept and design: Xia, Yang, Shankar, Walsh, and Selkoe. Acquisition of data: Xia, Yang, Smith, Shen, and Selkoe. Analysis and interpretation of data: Xia, Yang, Shankar, Smith, and Selkoe. Drafting of the manuscript: Xia. Critical revision of the manuscript for important intellectual content: Xia, Yang, Shankar, Smith, Shen, Walsh, and Selkoe. Statistical analysis: Xia and Yang. Obtained funding: Xia and Selkoe. Administrative, technical, and material support: Xia, Yang, Shankar, Smith, and Selkoe. Study supervision: Xia, Walsh, and Selkoe.

Financial Disclosure: Dr Walsh is a shareholder and scientific advisory board member of Senexis Ltd. Dr Selkoe is a founding scientist and consultant of Elan Corp, plc.

Funding/Support: This work was supported by grants AG015379 from the National Institutes of Health (Drs Xia and Selkoe) and 067660 from the Wellcome Trust (Dr Walsh).

Additional Contributions: Peter Seubert, PhD, and Dale Schenk, PhD, provided Aβ antibodies 3D6, 21F12, 266, and 2G3.