Antibodies to Potato Virus Y Bind the Amyloid β Peptide

Studies in transgenic mice bearing mutated human Alzheimer disease (AD) genes show that active vaccination with the amyloid β (Aβ) protein or passive immunization with anti-Aβ antibodies has beneficial effects on the development of disease. Although a trial of Aβ vaccination in humans was halted because of autoimmune meningoencephalitis, favorable effects on Aβ deposition in the brain and on behavior were seen. Conflicting results have been observed concerning the relationship of circulating anti-Aβ antibodies and AD. Although these autoantibodies are thought to arise from exposure to Aβ, it is also possible that homologous proteins may induce antibody synthesis. We propose that the long-standing presence of anti-Aβ antibodies or antibodies to immunogens homologous to the Aβ protein may produce protective effects. The amino acid sequence of the potato virus Y (PVY) nuclear inclusion b protein is highly homologous to the immunogenic N-terminal region of Aβ. PVY infects potatoes and related crops worldwide. Here, we show through immunocytochemistry, enzyme-linked immunosorbent assay, and NMR studies that mice inoculated with PVY develop antibodies that bind to Aβ in both neuritic plaques and neurofibrillary tangles, whereas antibodies to material from uninfected potato leaf show only modest levels of background immunoreactivity. NMR data show that the anti-PVY antibody binds to Aβ within the Phe4–Ser8 and His13–Leu17 regions. Immune responses generated from dietary exposure to proteins homologous to Aβ may induce antibodies that could influence the normal physiological processing of the protein and the development or progression of AD.

Despite great advances in our understanding of the genetics and molecular biology of Alzheimer disease (AD), 2 we do not fully understand why ϳ99% of people with the disease are affected. Although familial early-onset AD is caused by well described mutations in the amyloid ␤ (A␤) precursor (chromosome 21) and presenilins 1 and 2 (chromosomes 1 and 14) (1), these mutations are responsible for only ϳ1-2% of the cases of the disease. The most important genetic risk factor for the more prevalent (so-called sporadic) disease is the ⑀4 allele of apoE, which is well described and is responsible for ϳ40 -60% of the inherited risk. However, the ⑀4 allele is likely not causative, as approximately one-third of people with the disease do not have the gene, and many people with the gene do not have the disease. (45% of apoE ⑀4 homozygotes do not get the disease by age 80 (2).) Immunization with the A␤ peptide produces behavioral and histopathological improvement in transgenic mice bearing genes for human AD (3). In these transgenic mice, the A␤ vaccination paradigm is effective when administered either early in life, before onset of behavioral or structural evidence of the disease, or later, after disease onset (3). Because both active vaccination with the A␤ peptide and passive immunization with anti-A␤ antibodies have beneficial effects (4), the potential for AD therapy is under active investigation (4). This vaccination approach has been thwarted by the development of autoimmune meningoencephalitis in both mouse studies (5) and human trials in the United States and Europe (6). However, subjects who developed anti-A␤ antibody responses had improved cognitive function and activities of daily living (7) as well as clearance of the A␤ deposits (8). Hock and Nitsch (9) have concluded that "in humans . . . antibodies against A␤-related epitopes are capable of slowing progression of AD." Currently ongoing Phase 3 clinical trials of A␤ immunotherapy must be completed before answers concerning the therapeutic value of this approach can be obtained.
We propose that the mechanisms demonstrated by the A␤ immunization paradigm may also be operating lifelong, without active or passive vaccination. Those individuals with higher levels of the presumed naturally occurring anti-A␤ antibodies may be protected from developing AD. Conflicting studies have been reported thus far on this possibility: increased (10 -12), decreased (13)(14)(15), or unchanged (16) levels of anti-A␤ autoantibodies have been noted in studies of AD patients and control subjects. Moir et al. (17) found that circulating autoantibodies specific for A␤ oligomers are decreased in AD. It is not clear whether the studies discussed above measured total circulating anti-A␤ antibodies or only those antibodies that were not bound to circulating A␤ (18). Also, the presence of circulating anti-A␤ antibodies may very well be modified by the presence of disease; anti-A␤ antibody studies have not yet been completed in longitudinal studies of as yet unaffected subjects. It is also not clear if the assays applied in these studies were sensitive to cross-reacting antibodies. Thus, the active and passive A␤ immunization paradigm suggests that the presence of circulating anti-A␤ antibodies may influence the development of AD. In the absence of A␤ vaccination, exposure to an immunogen that bears significant amino acid sequence homology to A␤ could result in antibody production that has either protective or detrimental consequences (as illustrated by the studies mentioned above).
To explore this hypothesis, we identified a naturally occurring protein that is highly homologous to the human A␤ peptide and that is a nuclear inclusion b protein from a plant virus, potato virus Y (PVY) strain N (tobacco veinal necrosis) (BLAST, NCBI, and National Institutes of Health), to which humans are commonly exposed. PVY is an RNA virus and a member of the genus Potyvirus in the family Potyviridae (19,20). It contains a single-stranded RNA molecule of 9 Ϯ 7 kb, which is translated into a large precursor protein that is cleaved into 10 mature proteins (21,22). PVY infects solanaceous crops (of the nightshade family) such as potatoes, peppers, tomatoes, and tobacco. Potatoes are the fourth largest food crop in the world. Infection with PVY limits crop yield but does not destroy all growth. PVY is found worldwide, and it is estimated that ϳ15% of potato crops are infected. It is likely that some potatoes consumed by humans are infected with PVY (23).
We report that antibodies to PVY bind to A␤ in solution and in tissue sections. Data are presented illustrating the biochemical nature of the binding of anti-PVY antibodies to the same region of A␤ as is bound by therapeutic antibodies to the A␤ protein.

EXPERIMENTAL PROCEDURES
Antibody Production-50 g of peptide (27 amino acids from positions 52 to 77 of PVY with cysteine on the N terminus) was emulsified with 1:1 (v/v) Freund's complete adjuvant for the initial intraperitoneal injection, followed by a boost in Freund's incomplete adjuvant 2 weeks later, with monthly boosters thereafter. Mice were also inoculated with A␤ and infected and uninfected potato leaves. The positive control for PVY (catalog no. LPC20001, Agdia, Inc., Elkhart, IN) was resuspended in 10 mM citrate buffer containing 1 M urea and 0.1% ␤-mercaptoethanol, spun at 14,000 ϫ g for 15 min to remove the particulates, and then dialyzed against phosphate-buffered saline. The first injection used Freund's complete adjuvant, followed by a booster 2 weeks later in Freund's incomplete adjuvant and then four more boosters at 1-month intervals with the latter adjuvant. Equal volumes of the sample and Freund's adjuvant were emulsified for the injections.
NMR Spectroscopy-Commercially prepared anti-potato virus IgG polyclonal antibody (Phyto Diagnostics, North Saanich, British Columbia, Canada) was obtained as a solution (5 mM) in 50% ammonium sulfate and 100 mM phosphate buffer. Uniformly 15 N-labeled A␤-(1-40) and A␤-(1-42) peptides (recombinant peptides) were prepared in monomeric form using procedures developed in our laboratory (25). In brief, this procedure involved disaggregation of the A␤ peptides (0.2-0.5 mg) by sonication in aqueous basic solution (pD 11, 0.2 ml, 10 mM NaOD), followed by mixing with cold (5°C) phosphate buffer solution (pH 7.5, 0.6 -1.0 ml, 5 mM) containing 0.50 mM perdeuterated Na 2 EDTA-d 12 and 0.05 mM NaN 3 . The amount of peptide and buffer varied in accordance with the desired peptide concentration (50 -200 M). To prevent aggregation, peptide solutions were kept cold (5°C), and NMR spectra were obtained within 30 min of the sample preparation and at 5°C. A Varian 600-MHz Inova spectrometer equipped with an HCN Bioprobe was used for data acquisition, and the twodimensional 1 H-15 N heteronuclear single quantum coherence experiments were recorded (on average) with 32 scans, 2048 complex points and the transmitter placed on the water signal (26). The sweep widths were 6373.5 and 2000.0 Hz in the F 1 and F 2 dimensions, respectively. Processing was done on PC or Octane-2 (Silicon Graphics) computers equipped with the Felix program (Accelrys).

RESULTS
Homology Sequencing-The amino acid sequences of A␤ and the PVY nuclear inclusion b protein, an RNA-directed RNA polymerase, are shown in Fig. 1. The N-terminal domain of PVY is exposed to the exterior of the virion particle, enhancing the likelihood that it is immunogenic (27,28). PVY has 6 amino acids (at positions 60 -65) that share a high homology to the N-terminal region of A␤. This N-terminal region of A␤ (residues 1-40) has been demonstrated to be therapeutic in A␤ precursor protein-overexpressing animal models (29,30). Also, it is this region (residues 4 -10) of A␤ that is most highly immunogenic for B cells (32). The three-dimensional structure of the PVY protein is not yet known.
Enzyme-linked Immunosorbent Assay-To determine whether antibodies generated following vaccination with the PVY synthetic peptide labeled A␤ as well as the synthetic peptide, enzyme-linked immunosorbent assay screening was performed, which showed that antibodies made against the synthetic peptide had a high affinity for both the synthetic peptide and A␤, whereas antibodies to the positive leaf control showed a weaker affinity than the synthetic peptide antibody for both A␤ and the synthetic peptide (Fig. 2). Lower levels of immunoreactivity were found using the antibodies to the control leaf material.
A␤ is associated with senile plaques, neurofibrillary tangles, and neurons in AD (31); therefore, we tested whether mice inoculated with the PVY synthetic peptide develop antibodies that label A␤ in neuritic plagues as well as neurofibrillary tangles (Fig. 3). The synthetic peptide antibody recognized senile plaques, neurofibrillary tangles, and neurons. The positive control leaf antibody recognized neurofibrillary tangles, granulovaculolar degeneration, and neurons in AD cases (Fig. 3). Neuronal staining was observed in control cases for both the synthetic peptide and control leaf antibodies.
NMR Spectroscopy-To explore the binding between the A␤ peptide and the anti-PVY polyclonal antibody, we undertook NMR spectroscopic studies. The NMR peak assignments cor-respond to monomeric A␤ peptide (25), and the sample preparation protocol ensured that the A␤ peptides were monomeric at the beginning the NMR experiments. Aggregation during NMR data acquisition, particularly by the more aggregationprone A␤-(1-42) peptide, was prevented by acquiring the data at reduced temperatures (5°C). Fig. 4 shows the heteronuclear single quantum coherence NMR spectra of uniformly 15 N-labeled A␤-(1-40) and A␤-(1-42) peptides. The spectra of the peptides alone are superimposed with those containing 1:50 molar eq of the anti-PVY antibody. Heteronuclear single quantum coherence spectroscopy, which detects 1 H atoms directly attached to 15 N atoms, is a standard NMR experiment for proteins and provides a fingerprint for the backbone. The narrow chemical shift dispersion in the 1 H dimension (8.7 to 8.1 ppm) demonstrates that the peptides adopt predominantly monomeric, random, extended chain structures, consistent with previous studies (25,33).
With the anti-PVY antibody, several amide-NH peaks have different chemical shifts that are more confined to the polar 1-28 N-terminal residues and not within the Graphical depictions show that the 1 H and 15 N chemical shift differences are localized within two regions, Phe 4 -Ser 8 and His 13 -Leu 17 , which may constitute a binding pocket associated within PVY (Figs. 5 and 6). Control studies showed that the chemical shift movement upon addition of the anti-PVY antibody to the A␤ peptide solution was not caused by other components (such as ammonium sulfate) present within the antibody solution. Because these studies utilized commercially prepared anti-PVY antibody solutions, we were unable to conduct titrations at antibody concentrations greater than 1:50

FIGURE 2. Enzyme-linked immunosorbent assay data for plated synthetic peptide (PVY-(52-77)) probed with each of the specified antisera (A) and plated A␤ probed with each of the different antisera (B)
. PBS, phosphatebuffered saline. Leaf material refers to leaf infected with PVY.

FIGURE 3. Antibodies raised against PVY-(52-77) (A) or the infected control (C) bind to neurofibrillary tangles, senile plaques, and neurons in AD (B) and to neurons in control cases (D).
molar eq relative to peptide. However, even at these low concentrations, the anti-PVY antibody induced significant chemical shift movements, indicative of binding to the monomeric A␤ peptide.

DISCUSSION
The immunological approach to AD treatment has received great attention in animal and human studies since the original observations of Schenk et al. in 1999 (3). However, the role of immunological processes operating over a lifetime in determining who gets the disease has not been widely considered. If anti-A␤ antibodies are beneficial in A␤-overexpressing transgenic mice and humans with AD, then the presence of antibodies possessing the ability to bind A␤ may prevent or delay the onset of disease. A␤-binding antibodies may develop through natural mechanisms, as autoantibodies often develop with aging. Alternatively, anti-A␤ antibodies may be effectively produced through immunological responses to immunogens bearing sequence homology to A␤, such as PVY.
The aggregation and assembly of the A␤ protein into amyloid deposits are major neuropathological hallmarks of AD. The two predominant forms of A␤ are X-40 and X-42, with the latter protein being more aggregation-prone and whose overproduction has been linked to many familial forms of AD. The A␤ peptide is a normal physiological constituent that, from age-related micro-environmental changes, can undergo a conformational conversion from soluble monomeric random structures into aggregated ␤-pleated sheet structures, with the latter forming neurotoxic soluble aggregates (such as AD diffusible ligands) and protofibrils and eventually precipitating as mature amyloid fibrils. It is now thought that methods for preventing the A␤ conformational conversions and fibril formation could ameliorate the effects associated with A␤-induced neurotoxicity in AD. Because monomeric and oligomeric species of A␤ exist in equilibrium in tissue culture medium (34) and because the soluble oligomers are now thought to be the major culprit and resistant to proteolysis (35)(36)(37), the A␤ monomer may be the best therapeutic target for binding by an amyloid inhibitor. Current FDA-approved AD drugs include acetylcholinesterase inhibitors and an N-methyl-D-aspartate antagonist that improves cognition and behavior but does not reduce amyloid burden or delay progression.
Our NMR data demonstrate that the anti-PVY antibody binds to monomeric A␤. With our NMR sample preparation  protocol (25), A␤ is monomeric, and the lack of any line width changes is consistent with the anti-PVY antibody binding to monomeric peptide. However, it is possible that the anti-PVY antibody could also be binding with small amounts of soluble A␤ aggregates, and further work to investigate this possibility is currently under way in our laboratory. In contrast, significant line width reductions were seen with binding to human serum albumin, due to binding with A␤ oligomers (38). This is exceptional given that the majority of proteins or small molecules that reportedly bind to the A␤ peptide target the soluble aggregates or early-stage amyloid fibrils (39,40). ApoE ⑀3 is an endogenous inhibitor of A␤ aggregation that binds to pre-nuclear A␤ oligomers and blocks production of the nucleation steps in amyloid formation (41). More recent NMR studies showed that nicotine (42), human serum albumin (38), and the A␤-binding alcohol dehydrogenase (43) bind with soluble A␤ oligomers but not the monomers. Because the binding we detected was promoted with substoichiometric amounts (1:50) of the anti-PVY antibody, a stronger binding may occur at higher antibody concentrations. The binding seems localized within the Phe 4 -Ser 8 and His 13 -Leu 17 A␤ peptide regions. The importance of the central hydrophobic region for ␤-aggregation has been previously noted (44 -47), and the core of the amyloid ␤-strand structure is composed of Leu 17 -Ala 21 (48). A major advantage of the NMR approach is that it provides atomic level details of protein structure and dynamics in solution that are not available with other low resolution techniques. NMR provides site-specific structural data that assist in the development of specific amyloid inhibitors that select for the monomeric form of the A␤ peptide. Recent work in mice demonstrated that a 56-kDa soluble A␤-(1-42) assembly may be the actual culprit for initiating neuronal loss and memory deficits (49); thus, an inhibitor with any therapeutic value must prevent formation of this or other toxic A␤ aggregates (39). It is generally though that inhibitors that select for monomers or dimers are good starting points.
These results show promise that the anti-PVY antibody may be an effective means of regulating A␤ behavior, particularly because such a small relative molar ratio of antibody showed significant interaction with the peptide. However, because a polyclonal antibody was used, we saw only a solution average of the various forms of PVY present. As such, further work is under way to obtain a monoclonal antibody that can be used to perform more conclusive and quantitative experiments, including determination of a binding constant and epitope-binding domains.
Tabira and co-workers (50) in Japan have developed an oral vaccine for AD using a recombinant adeno-associated viral vector carrying A␤ cDNA. The vaccine reduced A␤ deposits without causing lymphocytic infiltration in the brain. It was proposed that mucosal immunity leads to safer immunological reactions to the vaccine. The lifelong development of antibodies that cross-react with A␤ following dietary exposure (such as PVY) to enhance clearance and inhibit aggregation may be less likely to elicit an autoimmune condition than late-life active vaccination because of the chronic development of the antibody response and the involvement of the immune system in the intestine, which is less likely than parenteral administration to elicit a T cell response (32). The immune system of the intestine enhances Th2 responses and suppresses Th1 responses, leading to relatively less cellmediated immunity (32,51). It has been proposed that immune mechanisms involving Th2-dependent responses would be the safest for an A␤ immune response in the setting of AD because Th2-dependent mechanisms produce antibodies that are less likely than those produced by Th1 responses to produce inflammation (52). The oral route of vaccination has also been used in studies in transgenic AD mice using transgenic potatoes expressing five tandem repeats of A␤- . Mice immunized with A␤ with this edible vaccine made antibodies against A␤ and had reduced A␤ plaques in the brain (53).
The mechanisms by which anti-A␤ antibodies may have a therapeutic effect include the following: 1) entry into the brain and binding to oligomeric and fibrillar A␤ with microglial activation, eliciting Fc receptor-mediated phagocytic mechanisms of removal of antibody-antigen complexes (52); 2) antibodymediated solubilization of fibrillar A␤ (32,54); 3) stabilization of the A␤ monomer, thus preventing the subsequent association into the soluble aggregates; 4) binding of A␤ to antibody in the circulation, enhancing clearance of A␤ from the brain (the peripheral sink hypothesis) (55); 5) altered proteolysis of A␤ (the ability of proteases that degrade A␤ (angiotensin-converting enzyme, neprilysin, endothelin-converting enzyme, plasmin, and insulin-degrading enzyme) (56) may be altered by binding of A␤ to antibodies); and 6) hydrolysis of A␤ by circulating autocatalytic IgM antibodies, as reported recently in studies of AD cases and controls by Taguchi et al. (12). Lifelong exposure to cross-reacting antibodies (such as PVY) that bind to A␤ may have protective effects through all of these mechanisms. This work is in keeping with current efforts to develop a safe and effective vaccine for AD (52). Novel immunogens have been developed that include the B cell epitopes of A␤ (the N terminus) but lack T cell-reactive sequences (57). The absence of a cellular immune response may provide for a safer therapy.
Plant viruses are found throughout the world, frequently infect crops used for human consumption, and have no known effects on human health. We propose that the development of antibodies to PVY following oral exposure is protective against the development of AD because of the beneficial effects of binding of the antibody to the A␤ protein. A model for this interaction may be supplied by the relationship between vaccinia infection (related to cowpox) and the resultant immunity to variola (smallpox). There are naturally occurring proteins other than PVY that bear significant homology to A␤ and that may influence the development of AD. For example, several proteins of Enterococcus contain sequences homologous to A␤ (NCBI and National Institutes of Health). The mechanism we propose may influence the pathophysiology of other conditions as well. Antibodies developed in response to naturally occurring plant or animal viruses, bacteria, or other agents may interact with protein trafficking in the brain and blood to influence handling and deposition of pathological proteins. This approach may be valuable for AD immunotherapy because of the relatively low inflammatory potential with intestinal immunogen delivery and the efficacy of antibody binding to pathogenic A␤ monomers.
It is of interest to note as well that circulating antibodies against both unphosphorylated and phosphorylated Tau proteins have also been observed (58), and active immunization with a phosphorylated Tau epitope in P301L tangle model mice reduced brain aggregated Tau and slowed progression of behavioral deficits (59). Also, antibodies generated against soluble oligomeric A␤ have been shown to neutralize oligomers of the prion protein and ␣-synuclein, suggesting that shared epitopes of these pathogenic proteins may play a role in several neurodegenerative illnesses (52,60).