Autoantibodies to Redox-modified Oligomeric Aβ Are Attenuated in the Plasma of Alzheimer's Disease Patients*

Accumulation of Aβ protein in β-amyloid deposits is a hallmark event in Alzheimer's disease (AD). Recent findings suggest anti-Aβ autoantibodies may have a role in AD pathology. However, a consensus has yet to emerge as to whether endogenous anti-Aβ autoantibodies are elevated, depressed, or unchanged in AD patients. Whereas experiments to date have used synthetic unmodified monomeric Aβ (Aβmon) to test autoimmunity, up to 40% of the Aβ pool inB AD brain consists of low molecular weight oligomeric cross-linked β-amyloid protein species (CAPS). Recent studies also suggest that CAPS may be the primary neurotoxic agent in AD. In the present study, AD and nondemented control plasma were analyzed for immunoreactivity to CAPS and Aβmon. Plasma of both nondemented and AD patients were found to contain autoantibodies specific for soluble CAPS. Nondemented control and AD plasmas demonstrated similar immunoreactivity to Aβmon. In contrast, anti-CAPS antibodies in AD plasma were found to be significantly reduced compared with nondemented controls (p = 0.018). Furthermore, age at onset for AD correlated significantly (p = 0.041) with plasma immunoreactivity to CAPS. These data suggest that autoantibodies to CAPS are depleted in AD patients and raise the prospect that immunization with anti-CAPS antibodies might provide therapeutic benefit for AD.

A convergence of histological, biochemical, and genetic evidence links the widespread neuronal loss characteristic of Alzheimer's disease (AD) 1 with deposits of ␤-amyloid that pervade the brains of AD patients. The principal component of extracellular ␤-amyloid is the ␤-amyloid protein (A␤). The A␤ peptide is not directly expressed as a functional protein entity (1) but is released by the processing of the much larger amyloid protein precursor (APP) protein (2,3). A␤ appears to be a normal product of cellular APP catabolism and is found as a soluble component of human cerebrospinal fluid (CSF) and plasma (4 -6). Whereas A␤ can contain between 39 and 43 amino acids, the predominant species in brain are A␤40 (40 residues) and A␤42 (42 residues) (7,8).
Analysis of material purified from human tissue suggest that up to 40% of the A␤ pool in AD brain consists of SDS-stable dimeric and low molecular weight oligomeric species (9). We have previously shown that incubation of A␤ with copper generates cross-linked oligomeric species that co-migrate on SDS-PAGE with A␤ oligomers purified from AD brain (10). Characterization studies demonstrated the redox modifications induced by incubation with copper are shared by oligomeric A␤ purified from AD brain, including covalent cross-linking (10,11). We have coined the term "CAPS" to describe low molecular mass (Ͻ100 kDa) cross-linked ␤-amyloid protein species. Covalent cross-linking of A␤ involves oxidation of the protein and is tied to the peptide's propensity to bind the redox active metals copper and iron (12). The mechanism of A␤ neurotoxicity remains controversial. However, evidence is mounting that the most neurotoxic forms of A␤ are not mature fibrils but prefibrillar oligomers or protofibrils (13), which would include CAPS. Notably, recent studies have demonstrated that the most toxic CAPS may be cross-linked dimeric species of A␤ (14,15). Despite the abundance and harmful bioactivity shown for CAPS, the vast majority of currently available data has focused on nonoxidized monomeric forms of the peptide.
Interest in autoimmunity to A␤ has been stimulated by recent findings that amyloid burden in transgenic animal models can be attenuated by circulating anti-A␤ antibodies (16 -18). ␤-Amyloid deposition can be inhibited by either peripheral infusion of exogenous anti-A␤ antibodies or autoimmunity induced by immunization with synthetic A␤ peptide. Initial studies suggested that anti-A␤ antibodies aid in the clearance of amyloid by crossing the blood-brain barrier and binding directly to plaques. However, subsequent studies have suggested that antibodies (19) and other A␤ binding agents (20) may not need to cross the blood-brain barrier to be effective in inhibiting cerebral A␤ plaque formation. In this model, A␤ is bound and sequestered in the periphery and prevented from crossing back into the brain, thus promoting a net flux out of neurological tissue (19). Whatever the mechanism, the use of circulating anti-A␤ antibodies is a therapeutic strategy being actively pursued. Unfortunately, dosing in the first clinical trial using A␤ vaccination to treat AD patients was terminated in phase II because of complications associated with inflammation of the central nervous system vasculature (21). Nonetheless, limited data suggest that amyloid load may have been attenuated in some trial subjects by autoantibodies specific for insoluble A␤ deposited as ␤-amyloid (22). Despite the earlier problems, clinical trials aimed at elevating anti-A␤ antibody levels in AD patients will most likely proceed. Therefore, it is imperative to advance our understanding of the autoimmune response to A␤ and its derivatives with greater alacrity.
The presence of anti-A␤ immunoreactivity in human serum and CSF was first reported in 1991 by Mönning et al. (23). Subsequently, Epstein-Barr virus-transformed B cells from AD patients have been shown to secrete anti-A␤ antibodies (24). More recently, several studies have used ELISA assays to compare anti-A␤ autoimmunoreactivity in control and AD plasma and CSF. However, a consensus has yet to emerge as to whether anti-A␤ autoantibodies are elevated (25), depressed (26,27), or unchanged (28) in AD patients compared with nondemented controls. Experiments to date have used synthetic unmodified monomeric A␤ peptides to test autoimmunity. The current study is the first to test human plasma for specific anti-CAPS antibodies. Moreover, findings are consistent with an association between AD pathology and autoantibodies specific to cross-linked A␤ species.
The future success of AD therapies based on anti-A␤ antibodies will require greater delineation of the naturally occurring autoantibodies to A␤. In this study, we present data suggesting that CAPS generated by exposing A␤ to mild redox conditions may be more immunogenic than the normal unmodified, monomeric A␤ species. We also show that plasma samples taken from AD patients exhibit significantly less immunoreactivity to CAPS than do samples drawn from nondemented controls. Moreover, lower signal for anti-CAPS antibodies correlated with earlier age-at-onset (AAO) of AD.

MATERIALS AND METHODS
AD and Control Cases-Plasma samples were collected from patients in the Memory and Movement Disorders Unit of Massachusetts General Hospital in Boston following informed consent. Samples were collected as part of a biomarker study approved by the Massachusetts General Hospital Institutional Review Board. Participants had a diagnoses of AD (n ϭ 59) by NINCDS-ADRDA (National Institute of Neurological and Communicative Disorders and Stroke-Alzheimers Disease and Related Disorders Association) criteria (29) or nondemented controls (n ϭ 59). Cognitive status of patients at the time of plasma collection was determined by the Blessed dementia scale informationmemory-concentration score (30). The Blessed dementia scale information-memory-concentration is scored from 0 to 37 mistakes, with 0 -3 mistakes considered within the normal range. The demographics of the population are shown in Table I.
Preparation of Stock Solutions and Buffers-Stock A␤ solutions were prepared by the addition of 30% trifluoroethanol in high pressure liquid chromatography grade water to ϳ1 mg of powdered A␤40 (synthesized by the W. Keck Laboratory, Yale University, New Haven, CT). Undissolved peptide was solubilized by light sonication of the stock solution for 3 min. Following sonication, the A␤ solution was centrifuged to remove undissolved material, and peptide concentration in the supernatant was determined by a bicinchoninic acid protein assay (Pierce). Zn(II)-histidine buffers were prepared by combining zinc standard solution (National Institute of Standards and Technology) with 150 mM NaCl in 50 mM Tris, pH 7.4 (TBS), containing histidine for a final Zn(II)/histidine molar ratio of 1:6.
Redox Treatment of A␤-The method first described by Galeazzi et al. (31) was used to generate cross-linked ␤-amyloid protein species (CAPS). Fresh A␤40 (100 g/ml) was incubated (2 days at 37°C) in TBS with 10 g/ml horseradish peroxidase (HRP) in the presence of hydrogen peroxide (100 M). Following incubation, HRP in the sample was inactivated by incubation (1 h at 37°C) with sodium azide (5%), an irreversible inhibitor of peroxidase activity. No peroxidase signal was detected in solutions of, or microplate wells coated with, HRP-treated A␤ (data not shown). A␤ oligomerization during incubation with HRP was monitored by Western blot using the polyclonal antibody (pAb) pan-A␤ raised against residues 15-30 of the A␤ peptide (Calbiochem).
Anti-A␤ Antibody ELISA-A␤ was first immobilized to the solid phase. Unmodified or redox-treated peptide (100 g/ml) was incubated in TBS (20 l/well) in the wells of a 384-well microplate. For some experiments, unmodified A␤ was incubated in wells containing 100 M Zn(II)-histidine buffer. Following the capture step, plates were blocked overnight at 4°C with BSA/TBS buffer (10% bovine serum albumin in TBS). Wells were then incubated with plasma samples diluted 1:50 in BSA/TBS buffer. After washing, wells were incubated with a 1:25,000 dilution of goat anti-IgG antibody conjugated to HRP (Calbiochem). The plate was washed, and luminescence was measured after the addition of 20 l/well Luminol solution (Pierce).
Immunoblotting (Western Blotting)-Samples were first resolved by electrophoresis on SDS-PAGE (4 -12% Bis-Tris gels) and then transferred to nitrocellulose membrane. Membranes were blocked overnight at 4°C with TBST (TBS containing 0.1% Tween) containing 5% each skimmed milk and BSA. For detection of A␤, membranes were first probed (2 h at room temperature) with 1:3000 dilution of pAb pan-A␤, then incubated with goat anti-rabbit IgG-coupled to HRP (1:10,000). For detection of A␤ immunoreactivity in human plasma, membranes blotted with A␤ were incubated (overnight at 4°C) with plasma samples diluted 1:100 in BSA/TBST. Membranes were then washed and probed with anti-human IgG-HRP conjugate. Both A␤ detection and A␤ immunoreactivity blots were developed for exposure to ECL film with super signal ultra (Pierce). Immunoreactive bands identified by pan A␤ Western blot of CAPS samples were equivalent in apparent molecular weight and relative intensity to species stained by Coomassie Blue on SDS-polyacrylamide gels (data not shown).
IgG ELISA-Plasma samples and standards of known IgG concentration were diluted (1:20) in BSA/TBS buffer and incubated (1 h at room temperature) in fresh untreated microplate wells. Following washing, wells were probed (1 h at room temperature) with anti-IgG-HRP-conjugated antibody (1:50,000) in BSA/TBS. Well chemiluminescence was then measured following the addition of Luminol.
A␤40 and A␤42 Plasma Levels-Plasma levels of A␤40 and A␤42 were determined by sandwich ELISA as described by Fukumoto et al. (32). Briefly, A␤ was captured to the solid phase using an antibody directed against residues 11-28 of the peptide (anti-A␤11-28). Bound A␤ was detected using anti-A␤ antibodies BA27 (A␤40 specific) or BC05 (A␤42-specific). The three antibodies were obtained from Takeda Chemical Industries (Osaka, Japan).

RESULTS
We first established an ELISA to measure titers of anti-A␤ autoantibodies in human plasma. Initially, we characterized three different A␤ species employed to capture autoantibodies from human plasma. To this end, wells were coated with either unmodified monomeric A␤ (A␤ mon ), redox-treated peptide containing cross-linked ␤-amyloid protein species (CAPS), or peptide assembled into noncovalent multimers by incubation with Zn(II)/histidine (A␤ Zn ). To demonstrate that A␤ mon and CAPS were captured to the solid phase, wells were extracted with SDS sample buffer, and the extracts were immunoblotted using a pAb raised against full-length A␤ peptide (pan-A␤). Analysis of immunoblot signal confirmed that the wells contained immobilized A␤ mon and CAPS (Fig. 1a). Whereas untreated A␤ and peptide immobilized in the presence of Zn(II)/histidine were monomeric, redox-treated peptide contained additional cross-linked oligomeric species. We next titrated pooled nonde- mented control plasma against immunoreactive signal for unmodified A␤ (Fig. 1b). Signal was blanked against signal from wells coated with BSA. Maximal signal and end point titers for A␤-coated wells were 1:50 and 1:21,000, respectively. Next we tested the specificity of the ELISA for anti-A␤ immunoreactivity. Pooled plasma from nondemented control subjects was preincubated (30 min) with BSA, reverse A␤ peptide, soluble A␤ mon , or CAPS (final peptide concentration in wells of 1 g/ml) before being assayed. Signal from plasma preabsorbed with reverse A␤ peptide was equivalent to plasma with BSA (Fig. 1c). However, consistent with specificity for anti-A␤ immunoreactivity, signal in samples preincubated with A␤ mon or CAPS was reduced for plates coated with A␤ mon . Notably, a third of the signal was not absorbed out by A␤ mon for wells coated with CAPS. These data suggest that plasma has specific immunoreactivity for species present in the heterogeneous CAPS preparations. No detectable attenuation of signal was observed for plasma spiked with Ͻ100 ng/ml exogenous A␤ (data not shown). Previous studies have suggested that the total pool of A␤ in undiluted human plasma is Ͻ5 ng/ml (32)(33)(34)(35)(36). In addition, plasma samples were diluted 50-fold prior to assay. Thus, endogenous A␤ would be unlikely to significantly reduce assay signal by competing with immobilized peptide for autoantibody binding.
The 42-residue isoform of A␤ was also tested in our ELISA. However, we were unable to reproducibly coat replicate wells with equivalent loadings of redox-treated A␤42 peptide. This was due to the greater propensity of preparations containing A␤42 to aggregate following redox treatment as compared with the less hydrophobic 40-amino acid isoform. Therefore, A␤42 could not be tested in the experiments in this study.
Following assay characterization experiments, plasma from nondemented control (n ϭ 59) and AD (n ϭ 59) subjects were compared for immunoreactivity to A␤ mon , A␤ Zn , or CAPS. Wells were coated with the various A␤ preparations, blocked, incubated with samples of diluted (1:50) plasma, and then probed for bound human IgG (Fig. 2). No significant differences were found between control and AD immunoreactivity to A␤ mon or A␤ Zn aggregates (Fig. 2, a and c). However, signal from control plasma incubated with CAPS was significantly elevated (p ϭ 0.018 by t test) as compared with AD samples (Fig. 2b). In addition, the immunoreactivity of plasma from nondemented patients was greater for CAPS than for A␤ mon .
In a control experiment, plasma from control and AD cases was also assayed for total IgG levels. Consistent with previous studies of plasma (27,37,38) and CSF (39), no significant contained SDS-stable oligomeric A␤ species consistent with redox modifications and covalent cross-linking. b, titration of anti-A␤ immunoreactivity in pooled human plasma. Wells coated with BSA or normal unmodified A␤ (A␤ mon ) were incubated with serial dilutions of pooled plasma from nondemented control subjects. Bound antibodies were then detected with anti-IgG-HRP conjugated antibody. Signal for A␤ was blanked against signal from wells coated with BSA. Maximal signal and end point titers for A␤-coated wells were 1:50 and 1:21,000, respectively. c, anti-A␤ signal in plasma is attenuated by preabsorption of endogenous antibodies with synthetic A␤ peptide or CAPS. Pooled plasma from nondemented control subjects was preincubated with BSA, reverse A␤ peptide (A␤ 40 -1 ), soluble unmodified A␤ (A␤ mon ), or CAPS (1 g/ml). Plasma incubants were then assayed by anti-A␤ ELISA for immunoreactivity against normal A␤ (A␤ plate) or CAPS (CAPS plate). All experiments used 384-well plates. Well signal was determined from luminescence following the addition of chemiluminescent reagent. Anti-A␤ ELISA data is shown as the average of eight replicates Ϯ S.E.

FIG. 2. AD plasma has significantly reduced autoimmunity for redox-cross-linked A␤ oligomers as compared with nondemented controls.
Wells were coated with BSA or A␤ that was unmodified (a), pretreated with HRP (b), or captured in Zn(II)/histidine buffer (c). Following blocking, wells were incubated with plasma from AD (n ϭ 59) or nondemented control (n ϭ 59) cases. Bound antibodies were detected by incubation with anti-IgG-HRP conjugate. Control plasma autoimmunity to redox-treated A␤ was significantly elevated (*, p ϭ 0.018 by two-tailed Student's t test) compared with AD samples. All assays used 384-well plates. Well signal was determined from luminescence following the addition of Luminol. Signal from A␤-containing wells was blanked on signal from wells preincubated with BSA. Data are shown as average sample signal (16 replicates for each plasma sample) for each test group Ϯ S.E. differences were found between the average plasma IgG concentrations in nondemented and AD (11.4 and 11.6 mg of IgG/ml, respectively) cohorts. Thus, the reduced levels of anti-CAPS antibodies we observed for AD plasma could not be attributed to a nonspecific decline in circulating IgG concentrations.
Next, immunoblotting techniques were used to characterize the immunogenicity of human plasma to CAPS. For these experiments, redox-treated A␤ was resolved by SDS-PAGE and transferred to nitrocellulose membrane. Membranes were then incubated with diluted plasma and probed with goat anti-IgG-HRP-conjugated antibody. Where immunoreactivity could be unambiguously detected in plasma from nondemented controls, the signal was highest for species with apparent molecular masses of 15-35 kDa (Fig. 3). However, the sensitivity of the immunoblot assay was insufficient for discrete detection of anti-A␤ immunoreactivity in most samples in our cohort. Notably, the two samples presented in Fig. 3 that generated clear immunoblot signals also possessed the highest anti-CAPS signal by ELISA.
Next, we compared signal for anti-CAPS antibodies to levels of soluble A␤ in plasma as determined by sandwich ELISA. Anti-CAPS antibody levels did not correlate significantly with either A␤40 or A␤42 concentrations or A␤42/A␤40 ratio in control, AD, or the combined plasma samples (Table II). However, plasma immunoreactivity to CAPS was found to correlate positively (r ϭ 0.279 with p ϭ 0.041) with the AAO of AD; the earlier the AAO, the lower the plasma immunoreactivity toward CAPS. AD plasma immunoreactivity for A␤ mon also demonstrated a trend toward positive correlation with AAO but did not reach statistical significance (p ϭ 0.136). We also tested whether plasma anti-CAPS immunoreactivity in AD patients decreased as a function of cognitive status as determined by Blessed dementia scale information-memory-concentration score (r ϭ Ϫ0.146 with p ϭ 0.297) or time since diagnoses (r ϭ Ϫ0.209 with p ϭ 0.111). Although the data are suggestive of a trend toward lower anti-CAPS signal with disease progression, correlations did not reach statistical significant with the cohort sizes used in this study. Our findings suggest that whereas lower levels of anti-CAPS antibodies are associated with diag-nosis of AD, they may not be strong markers for disease severity after onset. DISCUSSION Experiments with AD transgenic animal models suggest autoantibodies may play an important role in A␤ clearance (16 -18). However, although immunoreactivity of human plasma to A␤ mon has been previously investigated (25)(26)(27)(28)40), this is the first report of autoimmunity specifically to the subpopulation of cross-linked A␤ that we refer to as "CAPS." Our findings show that plasma from elderly nondemented control patients contain autoantibodies specific for CAPS in addition to immunoreactivity to unmodified A␤. Furthermore, we report that immunoreactivity to CAPS is significantly reduced in AD plasma as compared with nondemented controls, and this reduction correlates with AAO of the disease. The last several years has witnessed the emergence of data that increasingly suggest that CAPS play an important, perhaps central, role in AD neuropathogenesis. A␤ neurotoxicity appears to be greatly potentiated when the peptide self-associates into organized structures. Recent studies also suggest that the proximal effectors of A␤ neurotoxicity may be the intermediates of fibril assembly (14,15), particularly dimeric and trimeric CAPS (41). Consistent with the potentially important pathological role for CAPS, the protofibrils of other amyloid-forming proteins have also been shown to induce cell death, including ␣-synuclein, Huntingtin, and prion protein (13).
Antibodies to A␤ appear to develop in both nondemented and AD patients (23, 24, 26 -28, 40) and most likely occur as a part of normal aging. Circulating levels of autoantibodies are generally known to increase with aging in accord with at least two mechanisms (see review by Weksler et al. (42)). First, production of neoantigens increases with age in response to general increases in protein oxidation, the accumulation of aggregated proteinaceous material, and subtle shifts in posttranslational possessing, most notably glycosylation. Second, although levels of neo-antigens increase with age, the diversity of the general antibody repertoire steadily declines. This leads to an increase in the concentration of B-cell clonal idiotypes, eventually stimulating production of anti-idiotypic autoantibodies.
Several previous studies have evaluated AD and agematched control plasma for potential differences in the levels of anti-A␤ autoantibodies. However, the collective findings have been contradictory with an elevation (25), a decrease (26,27), and no change (28) reported for levels of anti-A␤ antibodies in the plasma of AD cases versus nondemented controls. Our data are consistent with previous reports, at least with regard to our observation of equivalent levels of autoantibodies to unmodified A␤ in AD and nondemented control plasma samples (28) (Fig. 2a). However, control plasma contained additional immunoreactivity to oxidized A␤ species (Fig. 1c). These data demonstrate that control plasma contains antibodies that recognize epitopes specific to oxidized forms of the peptide. AD plasma contained significantly less immunoreactivity to oxidized A␤ (Fig. 2b). Our analysis also showed that the signal for anti-CAPS antibodies correlated significantly with AAO for AD (Table II). Reduced anti-CAPS immunoreactivity in AD plasma suggests that autoantibodies to CAPS may be protective for AD. One possibility is that anti-CAPS antibodies may aid in the clearance of these oxidized forms of A␤ or attenuate their neurotoxicity by binding to the oligomer structures (43).
Our experiments employed heterogeneous CAPS preparations containing monomeric, dimeric, and multimeric crosslinked oligomers (Fig. 1a). Whereas Western blot assays were not sufficiently sensitive to conclusively quantitate the relative immunoreactivity of these different CAPS for most samples in our cohort, we were able to identify A␤ oligomers with apparent  2-4) were resolved on SDS-PAGE and transferred to nitrocellulose membrane. Blots were incubated with pAb pan-A␤ (lanes 1 and 2) or diluted (1:100) human plasma from two nondemented patients (lanes 3 and 4) previously identified by A␤ autoimmunity ELISA to give the strongest signal for antibodies to redoxtreated A␤. Blots were incubated with anti-rabbit or anti-human HRPconjugated antibodies and developed for exposure to ECL-film. Signal was highest for species with apparent molecular weights corresponding to A␤ oligomers containing 4 -8 cross-linked monomeric units. molecular masses  corresponding to 4 -8 crosslinked monomeric units as the species with the highest immunoreactivity in our CAPS preparations in control plasma (Fig.  3). Interestingly however, these were not the most abundant A␤ oligomers in our CAPS preparations (Fig. 1a). It is unclear what structural features render the (15-35-kDa) CAPS relatively high apparent immunogenicity. The relative toxicity of different CAPS species is also unclear. Recent studies have demonstrated that A␤ mon is substantially less neurotoxic than either cross-linked dimers (14,15) or SDS-stable oligomers of 4 -10 subunits (referred to as ADDLs) (15,44). Although it remains to be determined whether the autoimmunogenicity and neurotoxicity of CAPS are linked, our data are consistent with high immunoreactivity for the redox cross-linked oligomers, which to date have been identified as highly neurotoxic (15,44). Thus, further characterization of the immunoreactive groups of CAPS may be potentially useful for treatment strategies employing A␤ vaccination to reduce amyloid burden in AD patients. Our data are consistent with the prediction that an immunogen incorporating the autoimmunogenic structural features of 15-35-kDa redox-modified A␤ oligomers may increase the specificity of antibodies for pathologically relevant A␤ species and thus represent a more effective immunization based therapeutic strategy for treating and preventing AD.
At least two populations of autoantibodies are likely to react with the CAPS used in our assay. Redox-modified A␤ from brain and peptide oxidized in vitro contain a number of chemical modifications, including isomerization (45,46), carbonylation (47), and amino acid oxidation (10), whereas monomeric units in SDS-stable oligomeric species appear to be cross-linked by dityrosine bridges (11,31,48). The chemical modifications observed for CAPS are common to many oxidized proteins and are known to be epitopes for so-called natural autoantibodies (49,50). Natural autoantibodies are characterized by broad reactivity directed against very well conserved public epitopes (42,49,50). It is highly possible that a portion of anti-CAPS immunoreactivity is mediated by natural autoantibodies. Consistent with this theory, many of the anti-A␤ antibodies secreted by Epstein-Barr virus-transformed B cells are polyreactive (51). However, in addition to public epitopes, the secondary/tertiary conformation of cross-linked CAPS oligomers may also generate neoantigenic epitopes, and autoantibodies to these epitopes are likely to be much more specific for redox-modified A␤.
AD plasma possessed significantly less immunoreactivity to CAPS than did control samples. It is unclear if the increased immunoreactivity in control plasma is directed against specific or public epitopes on CAPS. Notably, control and AD plasma have exhibited equivalent immunoreactivity to zinc-generated A␤ assemblies (Fig. 2c). In the presence of zinc, A␤ self-associates into aggregates with an ordered structure that mimics many of the physiochemical properties of ␤-amyloid (52,53). However, zinc treatment does not oxidize or covalently crosslink A␤ monomers (Fig. 1a). Thus, whereas the identity of the epitopes remains unclear, our findings suggest that the elevated levels of anti-CAPS immunoreactivity in control plasma are most likely directed at oxidized cross-linked oligomers, as opposed to noncovalently bound assemblies of A␤ (e.g. aggregates that form in the presence of zinc).
It remains unclear whether AD patients have a low autoimmune response to CAPS prior to the onset of the disease, or AD pathogenesis involves attenuation of anti-CAPS antibody titers. AD pathogenesis could potentially lower anti-CAPS antibody titers via several mechanisms. AD patients may develop increased immunotolerance to CAPS after disease onset, possibly in response to elevated levels of redox-modified A␤ species. Elevated levels of circulating CAPS may also act to deplete the pool of anti-CAPS antibodies as antigen-antibody complexes are cleared from plasma. Unfortunately, direct determination of levels of specific forms of CAPS in human plasma must await the development of assays specific for various redox-modified forms of the peptide. It is also possible that the pool of circulating anti-CAPS antibodies in plasma may be depleted by absorption to insoluble ␤-amyloid deposits that line the cerebral vasculature of AD patients (54). Recent findings from human A␤ vaccination trials confirm that ␤-amyloid in cerebral vasculature can provide a peripheral sink for anti-A␤ antibodies (19). The ELISA used to measure anti-CAPS immunoreactivity may also report artifactually low signal if anti-CAPS antibodies were absorbed by elevated levels of plasma CAPS. However, our ELISA characterization experiments suggest that to significantly reduce anti-A␤ antibody capture under the conditions of our assay would require CAPS concentrations 500-fold higher than the levels previously reported for soluble A␤ species in plasma (32)(33)(34)(35)(36). Thus, it seems unlikely that elevated endogenous CAPS levels explain the attenuated signal observed for AD plasma.
Finally, it remains to be determined whether reduced levels of anti-CAPS antibodies represent a risk factor for the development or progression of AD pathology. It is possible that anti-CAPS autoantibodies may protect neurons by aiding in the clearance of these neurotoxic species. Alternatively, antibodies TABLE II Autoimmunoreactivity signal correlates with age at onset of AD but not plasma A␤ levels The nonparametric correlation coefficient (Spearman rank) and p value (two-tailed) was calculated for immunoreactivity to unmodified (A␤ mon ) or HRP-generated CAPS and corresponding plasma levels of A␤40 or A␤42 or A␤40/A␤42 ratio. Analysis tested control (n ϭ 59) and AD (n ϭ 59) cohorts separately and combined. Plasma A␤ isoform levels were determined by sandwich ELISA. For the AD cohort, plasma immunoreactivity was also tested against AAO, cognitive status as determined by the Blessed dementia scale information-memory-concentration score at sample collection, and duration (years from first positive diagnoses to sample collection) of AD. Consistent with previous analysis, AAO for AD was found to correlate (p ϭ 0.041) with plasma immunoreactivity to redox-treated A␤. may neutralize the bioactivity of CAPS. Consistent with this posit, studies have demonstrated that antibodies raised to A␤ preparations that include CAPS are effective at attenuating protofibril neurotoxcicity (43). If this proves to be the case, then CAPS may have potential utility in immunization therapies. Recent studies suggest that targeting specific epitopes on A␤ may generate antibodies that provide improved protection against AD-like neuropathology in transgenic mice (21). Furthermore, immunization with protofibrillar aggregates that contained CAPS has been reported to reduce undesirable inflammatory responses in mouse models of AD (55). It is possible that immunogen material that minimizes epitopes common to CAPS and other A␤ species while emphasizing chemical moieties specific to CAPS has the potential to enhance the efficacy of A␤ immunization and anti-A␤ infusion therapies.
In conclusion, our findings demonstrate that redox crosslinked oligomeric A␤ species are immunoreactive with human plasma. The immunoreactivity is specific for cross-linked oligomers and not directed at A␤ assemblies bound by noncovalent forces such as those found in zinc-induced A␤ aggregates. We also observed that AD plasma contained lower levels of anti-CAPS antibodies compared with nondemented control subjects and that immunoreactivity to CAPS correlated with AAO of the disease. These findings may be useful for diagnosis and facilitating future designs of reagents for A␤ vaccination and antibody perfusion therapies aimed at treating and preventing AD.