Background
Amyloid-beta (Aβ), the major plaque-associated protein in the Alzheimer's disease (AD) brain, has become the main target for AD therapy since the formulation of the "amyloid hypothesis" [
1]. The significance of serum antibodies to Aβ in AD is unclear, because these antibodies have been reported to be decreased [
2‐
7], unaltered [
8‐
12], or increased [
13‐
17] in this disorder. These studies are summarized in Table
1. Some investigators have suggested that reduced levels of anti-Aβ antibodies may contribute to the pathogenesis of AD [
18,
19].
Table 1
Summary of previous studies
Hyman et al., 2001 | Plasma: 82 AD, 271 NCI | No differences between groups (ELISA) |
Weksler et al., 2002 | Serum: 19 AD, 33 NCI | Decreased AD anti-Aβ levels (ELISA) |
Nath et al., 2003 | Serum: 16 AD, 31 NCI | Anti-Aβ higher in AD patients |
Gruden et al., 2004 | Serum: 17 AD, 15 NCI | Increased anti-Aβ25-35 oligomer antibodies in AD patients (ELISA) |
Baril et al., 2004 | Serum: 36 AD, 34 NCI | No differences between groups (ELISA) |
Mruthinti et al., 2004 | Plasma: 33 AD, 42 NCI | Anti-Aβ antibodies significantly (4-fold) increased in AD plasma (ELISA) |
Moir et al., 2005 | Plasma: 59 AD, 59 NCI | No differences for anti-Aβ monomer antibodies; decreased AD levels for anti-Aβ oligomer levels (ELISA) |
Brettschneider et al., 2005 | Serum: 96 AD, 30 NCI | Anti-Aβ levels decreased in AD (immunoprecipitation assay) |
Jianping et al., 2006 | Serum: 20 AD, 20 NCI | Decreased AD anti-Aβ levels (ELISA) and avidity |
Song et al., 2007 | Serum: 153 AD, 193 NCI | Decreased AD anti-Aβ levels (ELISA) |
Gruden et al., 2007 | Serum: 48 AD, 28 NCI | Increased anti-Aβ25-35 oligomer antibodies in AD patients (ELISA, dot blot) |
Gustaw et al., 2008 | Serum: 23 or 35 AD (assays performed in two laboratories), 35 NCI | Anti-Aβ levels consistently increased in AD vs. controls only after dissociation |
Xu et al., 2008 | Plasma: 113 AD, 205 NCI | No differences between groups (plaque immunoreactivity) |
Britschgi et al., 2009 | Plasma: 75 AD, 36 NCI | No differences between groups (Aβ microarrays) |
Sohn et al., 2009 | Serum: 136 AD, 210 NCI | Anti-Aβ decreased in AD patients (ELISA) |
Gustaw-Rothenberg et al., 2010 | Serum: 25 AD < 1 year, 18 NCI, 27 AD > 1 year | Anti-Aβ increased in both AD groups (ELISA) vs. NCI, before and after dissociation |
In previous studies [
20,
21] we used enzyme-linked immunosorbent assay (ELISA) to measure antibodies to Aβ1-42 monomer and soluble oligomers in intravenous immunoglobulin (IvIg) preparations. IvIg preparations consist of pooled and purified plasma immunoglobulins (> 95% IgG) from thousands of clinically normal individuals. These drugs are being evaluated as a possible treatment for AD; encouraging results were obtained in two clinical trials in which IvIg was administered to AD patients [
22,
23] and a multi-site phase 3 trial is in progress. In our ELISA studies we found that in addition to IvIg's binding to Aβ-coated wells, it also bound extensively to wells coated with buffer or with an irrelevant protein, bovine serum albumin (BSA). We referred to this as nonspecific binding [
20,
21] and concluded that it should be subtracted from IvIg's binding to Aβ-coated wells to accurately calculate specific anti-Aβ antibody concentrations. A subsequent study [
24] found this binding to be mediated by IgG's Fab fragments and therefore referred to it as "polyvalent." Among previous studies comparing serum anti-Aβ levels between AD patients and aged normal controls, in only one study [
3] was this binding subtracted from total antibody binding to Aβ. The conflicting results for anti-Aβ serum antibodies in AD may be due in part to failure to account for this binding. Other reasons could include binding of anti-Aβ antibodies by serum Aβ (antibody "masking"), which could reduce ELISA detection of these antibodies [
25], incorrect diagnosis of some study subjects (clinical diagnosis of AD is about 88-90% accurate [
26,
27]), differences in preparation of the Aβ conformations used to detect antibody binding and/or other methodological differences, and the small sample sizes used in some studies. In previous ELISA studies comparing these antibodies in AD subjects vs. normal controls, only Moir et al. [
3], Gruden et al. [
14,
15], and Nath et al. [
13] measured antibodies to Aβ soluble oligomers, which are thought to initiate AD-type pathology [
28], and only Gustaw et al. [
16] and Gustaw-Rothenberg et al. [
17] performed antibody-antigen complex dissociation. None of the studies performed both subtraction of polyvalent binding and dissociation of antibody-antigen complexes, nor did any of the studies confirm clinical diagnoses with post-mortem examinations or perform power analyses.
The objectives of this pilot study were therefore to compare serum antibody levels to Aβ1-42 soluble conformations between AD patients, subjects with mild cognitive impairment (MCI), and aged noncognitively impaired (NCI) individuals, incorporating all of these procedures, and to perform power analyses on the resulting data to obtain estimates of appropriate group sizes for future studies using this approach. Our findings suggest that relatively similar levels of specific, non-antigen-bound antibodies to soluble Aβ1-42 conformations are present in AD, MCI, and NCI sera. Large numbers of samples (estimated group sizes: 328 and 150 for anti-Aβ monomer and oligomer antibodies, respectively) would be required for a high probability of achieving statistical significance for the between-group differences with this approach.
Discussion
This study used ELISA, with subtraction of polyvalent antibody binding and dissociation of antibody-antigen complexes, to compare the concentrations of serum antibodies to soluble Aβ1-42 conformations between AD, MCI, and NCI subjects who were grouped on the basis of post-mortem clinical review. The between-group differences for serum anti-Aβ levels were not statistically significant. Although the mean levels of these antibodies tended to be increased in AD vs. NCI specimens, large group sizes (estimated at 328 for anti-Aβ monomer antibodies and 150 for anti-Aβ oligomer antibodies) would have been required for a high likelihood that differences of this magnitude would be statistically significant. These sample sizes are considered to be approximate values because they are based on variability estimates from small numbers of samples. Previous studies have suggested that anti-Aβ antibodies may play a protective role in AD, by preventing Aβ's neurotoxicity [
32,
33], inhibiting development of Aβ soluble oligomers [
21], increasing phagocytic clearance of fibrillar Aβ [
34], preventing Aβ fibril development [
35], and degrading preformed Aβ fibrils [
34]. Using procedures to measure specific, non-antigen-bound anti-Aβ antibodies, no evidence was found in the present study for altered levels of these antibodies in AD patients. Because the secondary antibody used to detect anti-Aβ antibodies in the serum samples, biotinylated goat anti-human IgG (H + L), was not IgG-specific, the measurements in the present study represent total serum anti-Aβ antibodies rather than IgG. Our results do not support the hypothesis that decreased concentrations of serum anti-Aβ antibodies may contribute to the pathogenesis of AD.
Some studies have suggested that human anti-Aβ antibodies may recognize conformational epitopes on aggregated forms of Aβ, while not recognizing linear epitopes on monomeric Aβ [
12,
33,
36,
37]. However, our IvIg study [
20] and the study of Moir et al. with AD and control plasma [
3] suggested that these antibodies do include those to Aβ monomer as well as to Aβ oligomers. In the present study, specific antibodies were found in AD, MCI, and NCI sera to both Aβ monomer and oligomer preparations. In an earlier study [
30] we evaluated our monomer preparation by western blot after electrophoresis on native gels, immediately after preparation and after storage at 4°C for more than two months. Only one band was seen in each blot, suggesting little, if any, oligomer contamination. The TEM images in the present study also showed clear differences between the 10 nm structures seen in the monomer preparation and the 50 - 100 nm structures observed in the oligomer preparation. These findings suggest that the antibodies measured in the present study to the Aβ monomer preparation were directed to monomer rather than to Aβ oligomers. However, because Aβ monomer may exist in equilibrium with low-order Aβ oligomers [
38], the possibility is not ruled out that some of the antibody binding to the Aβ monomer preparation could have been to Aβ oligomers whose concentrations were below the level of detection of western blot.
A further difficulty with regard to differentiating between antibodies to Aβ monomer and oligomers is that anti-monomer antibodies could also recognize Aβ oligomers. The strong association between anti-monomer and anti-oligomer antibody levels in the serum samples in this study raised the issue of whether the two antibody measures may essentially be the same. Depleting the samples of anti-monomer antibodies would not necessarily resolve this issue because this might also remove some anti-oligomer reactivity, if some of the anti-Aβ antibodies bind to both monomers and oligomers. If, in fact, most of the anti-monomer antibodies also recognize oligomers, then after subtracting the ~30% of antibody reactivity to the oligomer preparation which is likely to be due to binding to monomers, little or no reactivity should remain. However, substantial reactivity was still detected. This suggests that at least some of the reactivity was likely to be oligomer-specific.
Previous studies reported that antibody-antigen complex dissociation may allow detection of increased levels of serum anti-Aβ antibodies [
16,
17,
39]. The Aβ conformation to which antibodies were measured in those studies was not stated. In the present study, dissociation increased the measured concentrations of antibodies to Aβ monomer but not to Aβ oligomers. The dissociation procedure used pH 3.5 dissociation buffer to separate antibody-antigen complexes, followed by passage through a 30 kDa molecular weight cutoff filter to remove unbound Aβ. Unlike antibody-antigen dissociation with lower pH (2.5), dissociation at pH 3.5 should not produce artifactual increases in anti-Aβ antibodies or inactivate authentic antibody binding [
25]. This procedure should allow removal of Aβ monomer (molecular weight 4.5 kDa) and Aβ oligomers no larger than hexamers (27 kDa), while larger oligomers should be retained. A possible explanation for the lack of an increase in detectable anti-Aβ oligomer antibodies after dissociation is that complexes between anti-Aβ antibodies and larger Aβ aggregates may have re-formed after dissociation, although whether Aβ oligomers are present in serum is unclear. Detection of plasma Aβ oligomers by ELISA was reported by Xia et al. [
40], but heterophilic antibodies may have resulted in a false positive signal in that study by crosslinking capture and reporter antibodies, as noted by Sehlin et al. [
41]. We found similar false positive results (revealed as such when samples were diluted 1:1 with ELISA Diluent from Mabtech, Inc. [Mariemont, OH], stated by the manufacturer to prevent heterophilic antibody-related false positives) when we attempted to measure total Aβ1-42 in plasma samples from the subjects in this study (data not shown).
Surprisingly, the actual concentrations of specific anti-Aβ antibodies in serum and plasma are unclear. These antibodies have been reported as OD units [
5,
13,
16,
24], titers [
2,
6,
9,
10,
15], and as relative or arbitrary units [
3,
4,
14]. An exception is the study by Storace et al. [
39] which reported anti-Aβ antibody levels from dissociated plasma samples from MCI patients and normal controls as both concentrations and OD values. The levels reported in that study ranged from 8.0 to 9.5 μg/ml, higher than the range of 0.4 - 0.6 μg/ml in the present study. The reasons for these differences are unclear. One possibility for this discrepancy is that the concentrations for anti-Aβ antibody concentrations in our study were calculated on the basis of a standard curve using mouse anti-Aβ antibody, whereas Storace et al. used a purified human IgG reference standard. In addition, Storace et al. did not subtract polyvalent antibody binding.
Conclusions
We report that when specific antibodies to Aβ1-42 monomer and soluble oligomers were measured by ELISA in serum specimens from subjects with post-mortem clinical review diagnoses of AD, MCI, or NCI, no significant differences in these antibody levels were found between groups even after dissociation of antibody-antigen complexes to allow measurement of "free" (non-antigen-bound) antibodies. Further, power analyses on the data indicated that large group sizes (estimated at 328 and 150 for measurements of anti-Aβ monomer and oligomer antibodies, respectively) would have been necessary to achieve a high probability for the between-group differences in these antibody concentrations to achieve statistical significance. These results do not support the hypothesis that decreased levels of these antibodies may contribute to AD pathogenesis.
Competing interests
The authors declare that they have no competing interests.
Authors' contributions
ACK and LMS performed the experimental procedures, collected the data, and assisted in manuscript preparation. MPC performed the data analyses and assisted with manuscript preparation. DAB provided the serum samples and assisted with manuscript preparation. JMF provided guidance with Aβ monomer and oligomer preparation and assisted with manuscript preparation. LD performed the transmission electron microscope studies. DAL directed the research and wrote the manuscript. All authors read and approved the final manuscript.