Neuroprotective natural antibodies to assemblies of amyloidogenic peptides decrease with normal aging and advancing Alzheimer's disease

Plasma Aβ Antibodies Predominantly Recognize High Molecular Mass Assemblies in Oligomeric Preparations of Aβ1-42.

To characterize the specificity of human plasma Aβ antibodies we isolated IgGs, and analyzed their reactivity to assemblies of Aβ1-42 peptides on Western blots. Some IgG samples reacted distinctively with a nonfibrillar Aβ assembly between 55 and 78 kDa, reminiscent of the previously described 56-kDa Aβ*56 (30) but the main reactivity in all samples was against entities >210 kDa (Fig. 1A). The lack of binding to fibrillar Aβ preparations (Fig. 1A) indicates that most of the IgG antibodies are specific to smaller, potentially oligomeric conformations. The Aβ1-5 specific antibody 3D6 detected various assemblies in the oligomeric and fibrillar Aβ preparations (Fig. 1B). Atomic force microscopy confirmed the presence of predominantly fibrillar structures in fibrillar preparations (Fig. 1C) or nonfibrillar Aβ species in oligomeric preparations of Aβ1-42 (Fig. 1D). Together these data demonstrate the existence of highly specific IgG antibodies in human plasma, recognizing conformational epitopes in high molecular mass Aβ1-42 assemblies.

Plasma Aβ IgG Antibodies Are Abundant and Directed Preferentially Against Oligomers and Posttranslationally Modified Aβ.

To characterize and quantify the repertoire of natural antibodies against Aβ in human plasma on a large scale, we developed peptide microarrays containing various synthetic Aβ peptide variants, either nonaggregated or in different assembly states, and non-Aβ amyloidogenic peptides and controls (Table S1 and Fig. S1). To mimic some effects of oxidative stress in vitro we generated di-tyrosine cross-linked Aβ peptide species (CAPS) (9, 23). In a first experiment we measured relative Aβ antibody reactivities of the IgG class against 25 different peptide preparations (Table S1) in plasma from 75 AD patients at various stages of disease and 36 healthy nondemented controls (NDC) (Table S2). Strikingly, antibodies recognizing higher-order Aβ assemblies, including oligomer and fibrillar preparations, were more prevalent and up to 12-fold higher than the average reactivity against presumably nonaggregated preparations of Aβ1-40 and Aβ1-42 (Fig. 1E). No significant differences were observed between plasma from this mixed group of AD patients and NDC. In addition to antibodies against conformational Aβ species we discovered antibody reactivities against pyroglutamate Aβ variants, and curiously, against the peptide residue ABri3–34 (Fig. 1E).

To replicate these findings in an independent sample set and extend the panel of peptides, we screened plasma samples from an additional sample set of 55 AD patients and 62 NDC subjects (Table S2) for antibodies against 74 different peptide preparations and controls (Table S1 and Fig. 1 F and G). We confirmed that IgG reactivities against Aβ assemblies were generally higher than reactivities against nonaggregated Aβ (Fig. 1F). Although IgG titers to oligomeric preparations of Aβ1-42 (Fig. 1H) and other peptides (Fig. S2A) are relatively low, they were 2- to 4-fold higher than IgM reactivities within the same plasma sample (Fig. 1I) and for certain peptides up to 86-fold higher (Fig. S2 A and B). These antibody titers are also consistent with peptide array-based measurements of IgGs reactive to Aβ and myelin proteins in patients with multiple sclerosis (31) and support the concept of a circulating pool of self-reactive IgG antibodies in the general population (32, 33).

Interestingly, most plasma samples had relatively high IgG antibody reactivity against posttranslationally modified peptides. For example, IgGs against pyroglutamate variants Aβ3(pE)-42 or Aβ11(pE)-42 were up to 12-fold higher than those against Aβ11-42 or the average reactivities against Aβ1-40 and Aβ1-42 (Fig. 1J). Similarly, the pyroglutamate form of Bri1(pE)-6 elicited a 26-fold higher reactivity than Bri1–6 (Fig. S2C). Oxidative modification of Aβ peptide as present in Aβ29-M35(MetSox) (Fig. 1K) or Aβ1-42 (Ox) (Fig. S2D) led to a prominent increase in IgG binding compared with the median IgG reactivity to nonaggregated Aβ1-40 and Aβ1-42. In contrast, antibody reactivities against control peptides APLP1 (residues 559–580), APLP2 (residues 662–686), or human Amylin1–37 were similar to reactivities against nonaggregated Aβ1-40 or Aβ1-42 peptides (Fig. 1F).

General Elder Population Carries Relatively High Antibody Reactivity Against Amyloidogenic Peptides Unique to Autosomal Dominant Dementias.

Of particular interest in the first experiment was the increased antibody reactivity against mutant peptides that are found exclusively in different familial dementias. Consistent with this finding, relative antibody reactivities to mutant forms of Aβ1-40, Aβ1-42 or Aβ1-55 were severalfold increased compared with median reactivity for nonaggregated preparations of Aβ1-40 or Aβ1-42 in the second experiment (Fig. 1G). We observed relatively high antibody reactivities against full-length ABri and measured similar IgG titers for ADan (Fig. 1G), both peptides being mutant amyloidogenic non-Aβ peptides present exclusively either in British (34) or Danish dementia (35). Oligomer preparations and posttranslationally modified residues or artificial variants of the same peptides typically elicited even higher reactivities, e.g., relative reactivity against an oligomeric preparation of artificial mutant Bri16–23(A16K/V17K) was the strongest observed in our study (Fig. 1G). Electron microscopy confirmed in this preparation the presence of structures reminiscent of previously described globular or annular assemblies of Aβ and other amyloidogenic peptides (36, 37) but no assemblies were detected in freshly dissolved peptide (Fig. 1 L and M). Together, these data from 2 independent sample sets demonstrate the abundance of a diverse IgG repertoire in human plasma recognizing higher-order assemblies of Aβ peptides, pyroglutamate Aβ, oxidized Aβ, and mutant Bri peptide residues. The presence of antibodies in individuals who are not carriers of a mutation for familial dementia supports the existence of cross-reactive conformation-specific antibodies.

Reduced Antibody Reactivities Against Oligomeric Aβ1-42 Assemblies in Advanced AD and in Normal Aging.

Although no overall differences in Aβ or other peptide antibody reactivities were observed between a diverse group of AD patients and NDC, patients in sample set 1 with moderate to severe AD [minimental state examination (MMSE) scores ≤19] had significantly lower levels of antibody reactivities against oligomeric preparations of Aβ1-42 than patients with mild disease (MMSE scores ≥20) (Fig. 2A). A similar group difference was found in sample set 2, although it did not reach significance (Fig. 2B), possibly because of the low number of patients with severe disease (Table S2). No group difference was detected for antibodies recognizing nonaggregated Aβ, control antigens “pneumococcal vaccine polyvalent” and Candida albicans extract, arguing against a general age-related decline in antibody titers in our sample sets. Also, ApoE4 carriers did not differ from noncarriers in their immunoreactivity with all tested antigens.

To determine antibody reactivities at different ages we screened plasma samples from a third set of samples collected from 66 healthy, nondemented females ages 21–85 years old (Table S2 and Fig. S3) and found 6 antibody reactivities that were on average 45–65% higher in plasma from the youngest (21–44 years) compared with the oldest (≥70 years) age groups (Fig. S4 A--F). Total IgG levels and other reactivities did not change (Fig. S4 G–L), and only reactivity to pneumococcal vaccine increased with age (Fig. S4M), again arguing against a general loss in overall immunity in the elderly studied here. Using linear regression methods we found 8 antibody reactivities that were strongest associated with age as illustrated in a node map generated by an unsupervised cluster algorithm to arrange all samples based on similarity of antibody reactivities (Fig. 2C). Antibody reactivities against some of these peptides showed significant inverse correlations with age (Fig. S4 N and O). Low pH treatment (38) increased antibody reactivities against oligomeric Aβ in all tested plasma samples 2- to 7.4-fold (Fig. 2D) compared with untreated samples. This finding is similar to reports in human serum and IvIg preparations (25, 26). The percentage of free antibodies out of the total pool recognizing Aβ1-42 oligomers ranged from 13% to 50% but was not associated with age or AD (Fig. S5). Together, these data indicate that relative plasma antibody reactivities against several amyloidogenic Aβ mutant and non-Aβ peptides and in particular to oligomeric Aβ1-42 are reduced with age.

Aβ IgG Antibodies Are Present in CSF and Plasma IgGs Protect Primary Neurons from Aβ Toxicity.

To investigate whether any of the antibodies detected in the periphery may have access to the CNS we measured antibody titers in CSF samples from 16 healthy aged normal controls and 6 patients with mild AD using peptide arrays. Average reactivities in CSF were ≈30–230 times lower in CSF than in plasma but a similar repertoire of IgG antibodies was seen. Highest titers were observed for oligomeric preparations of Aβ1-42, pyroglutamate-modified Aβ peptides, and mutant variants of Aβ, ABri, and ADan (Fig. 3A). There was no significant difference in reactivities in the analyzed sample groups and IgM was not detectable.

To assess the potential physiological relevance of natural Aβ antibodies we tested the protective potential of IgGs against Aβ toxicity. IgGs purified from 3 plasma samples and acidified to release bound peptides (38) were preincubated with oligomeric preparations of Aβ1-42 and this mix was added to primary hippocampal neurons from embryonic day (E) 16 mice. All 3 IgG preparations significantly reduced cell death (Fig. 3B) and increased cell viability (Fig. 3C) to an equal or greater extent than monoclonal antibodies to Aβ, consistent with similar experiments done with IvIg (25).

Immunization with Aβ Peptides Amplifies Immune Response to Aβ, ABri, and ADan Peptides in Nonhuman Primates.

To further study the in vivo relevance of natural antibodies to amyloidogenic peptides, we measured immunoreactivity in plasma of a subcohort of vervet monkeys from a previously published study (39). Aged vervets can develop features of AD including cerebral amyloid plaques that stain positive with anti-human Aβ antibodies (39), indicating high homology between human and vervet Aβ similar to other nonhuman primates (40). Immunization with a mixture of human Aβ1-40 and Aβ1-42 almost entirely cleared cerebral Aβ plaques (39). Although standard ELISA methods detected miniscule levels of Aβ1-40 antibodies in nonimmunized vervets (39), using our peptide arrays, we were able to measure reactivities to nonaggregated and oligomeric Aβ preparations and to Aβ3(pE)-42 species comparable to what we detected in human plasma (Fig. S6 A–C). At the end of the study, at 301 days, control animals had unchanged or decreased titers to most of the tested peptides (Fig. 3D and Fig. S6B), whereas immunized vervets developed a 370-fold median increase (range of 80- to 50,000-fold) in IgG antibodies to full-length Aβ preparations (Fig. 3E and Fig. S6D). In support of earlier findings that immunization with Aβ was mainly targeting the peptide's N terminus (39), immunized vervets failed to develop a comparable immune response to Aβ33–42 (0.2- to 22-fold). Most importantly, and in likely support of their physiological relevance, immunized vervets significantly increased antibody levels to pyroglutamate-modified Aβ [e.g., Aβ11(pE)-42], mutant Aβ [e.g., Aβ1-42(E22Q) Dutch], and the foreign peptides ABri and ADan (Fig. 3E and Fig. S6D). These data show that immunization with full-length Aβ not only increases immunity to select Aβ peptide species but also triggers the expansion of cross-reactive antibodies against other known toxic amyloidogenic peptides.

Plasma and CSF Samples.

Archived human EDTA plasma and CSF samples were obtained from academic centers specialized in neurological or neurodegenerative diseases (Table S2). Informed consent was obtained from all human subjects in accordance with Institutional Review Board-approved protocols. Vervet plasma samples came from a previously published Aβ immunization study (39) (Table S2). One control animal was not part of the original study.

Peptides and Arrays.

We obtained RP-HPLC-purified peptides from different sources (Table S1). Aβ assemblies were made as described (23, 42). Some preparations were analyzed by atomic force or electron microscopy and immunoblots. Antigen microarrays (Table S1 and Fig. S1) were printed in batches of up to 200 slides with a custom-built robotic microarrayer by contact printing onto reactive epoxide-coated glass slides. Glass slides were probed with diluted sample or specific antibodies and data were extracted as described (58).

Statistical Analyses.

Most statistical analyses were done in GraphPad Prism 5.0 with P ≤ 0.05 considered significant. Multivariate comparison analysis between the youngest and oldest age groups and linear regression analysis of age with antibody reactivities was done with the software package SAM (59) with 0% false discovery rate considered significant. Unsupervised cluster analysis was done in Cluster 3.0 (60), and a node map was generated in Java TreeView (61). Penalized linear regression models were calculated by the regularization and variable selection method elastic net (62).

ELISA and In Vitro Aβ Neurotoxicity Assay.

Dissociation of immune complexes in plasma at pH 3.5 and detection by indirect ELISA was done as described with modifications (25, 38). Hippocampal neuronal cultures from E15–16 CF1 mice were treated on day 6 with 2 μM Aβ oligomers in the presence or absence of antigen-dissociated and purified human IgGs or the indicated monoclonal antibodies. After 3 days, cell death or viability was measured with standard lactate dehydrogenase (LDH) release and tetrazolium salt-based assays.