Familial and late-onset Alzheimer's disease: Autoimmune disorders triggered by viral, microbial and allergen mimics of beta-amyloid and APP mutants ?

Beta-amyloid (A[beta]) autoantibodies are common in Alzheimer's disease. Some may be derived from Herpes simplex and 68 other virus proteins matching an immunogenic and fibrillogenic VGGVV A[beta] sequence.Many other viral, microbial and allergenic proteins (particularly from dust mites) align with A[beta] as do proteins from _C.Neoformans_ which has been associated with a rare but curable form of the disorder. Immune-related proteins are present in amyloid plaques and the complement membrane attack complex in neurones in Alzheimer's disease brains. Alzheimer's disease may thus be an autoimmune disorder triggered by pathogenic antigens homologous to A[beta], whose antibodies target and kill A[beta] containing neurones, via immune activation and complement-related lysis. This scenario explains many epidemiological observations in Alzheimer's disease, which is more common in women and Afro-Americans, as is HSV-2 seroprevalence; related to the number of pregnancies (exposure to childhood infections) and less severe in nuns (low exposure to sexually transmitted diseases). Atopy and autoimmune disorders are common in Alzheimer's disease in accord with allergen homology to A[beta], and anti-inflammatory agents reduce Alzheimer's disease risk. Cancer-causing, and plant viruses from Mediterranean diet components align with the A[beta] sequence targeted by catalytic autoantibodies, perhaps explaining the inverse association of diet and cancer with Alzheimer's disease. As a papillomavirus vaccine already exists, it may have a role to play in Alzheimer's disease. This scenario is also relevant to familial Alzheimer's disease. Mutant forms of APP~717~ and the Swedish mutant convert the surrounding peptide to matches with commensal bacteria (_E.Coli_, _E. Faecalis_, _P.Gingivalis_) and to viruses with very high seroprevalence (HHV-6, norovirus, influenza and the common cold). Late-onset and Familial Alzheimer's disease may both be autoimmune disorders caused by diverse common pathogens and allergens homologous to A[beta] or mutant APP fragments. Immunosuppressants, vaccination and pathogen elimination may be of benefit in both conditions.


Introduction
Autoantibodies to beta-amyloid are common in the ageing population and in Alzheimer's disease and may exert a beneficial role in adsorbing the toxic peptide or catalysing its destruction. They may also mount an immune attack against beta-amyloid, activating inflammatory pathways and complement cascades that kill the neurones in which the peptide resides 54,79,95 . The membrane attack complex of the complement pathway is present in Alzheimer's disease neurones 48,72 , supporting a role for aberrant immune/complement activation within the brain. The source of these autoantibodies is not clear. They could be derived as a response to abnormal forms of beta-amyloid or from antibodies to other antigens that cross-react with the peptide.
It has already been noted that glycoprotein B of the Herpes simplex virus shows marked homology with beta-amyloid, particularly matching a VGGVV C-terminus pentapeptide 81 (Fig 1). The VGGVG epitope has been used to label beta-amyloid 1-40 in extracellular neurofibrillary tangles 93 . This pentapeptide, per se, forms aggregates characterised by twisted ropes and banded fibrils 75 . This is a characteristic of both beta-amyloid and of HSV-1 glycoprotein B peptide fragments containing this sequence. The viral glycoprotein B fragments form thioflavine T positive fibrils which accelerate beta-amyloid fibril formation, and are neurotoxic in cell culture 23 .Herpes simplex infection (HSV-1) has been shown to be a risk factor in Alzheimer's disease, acting in synergy with possession of the APOE4 allele 49 . HSV-1 infection in mice or neuroblastoma cells increases betaamyloid deposition and phosphorylation of the microtubule protein tau 112,113 .Viral infection in mice also results in hippocampal and entorhinal cortex neuronal degeneration and memory loss, all as found in Alzheimer's disease 5 . A recent study has also shown that anti-HSV-1 immunoglobulin M seropositivity, a marker of primary viral infection or reactivation, in a cohort of healthy patients, was significantly associated with the subsequent development of Alzheimer's disease. Anti-HSV-1 IgG, a marker of lifelong infection showed no association with subsequent Alzheimer's disease development 63 . All of these factors support a viral influence on the development of Alzheimer's disease. Antibodies to the Potato virus Y, which is highly homologous to beta-amyloid (Fig 1), are also able to label beta-amyloid containing plaques in the Alzheimer's disease brain 39 . It is therefore possible that beta-amyloid autoantibodies are derived from such homologous antigens. Homology searches within viral and microbial proteomes and allergenic proteins showed that many are highly homologous to immunogenic regions of the beta-amyloid peptide and also that APP mutations in Familial Alzheimer's disease convert the surrounding peptides to matches to very common viruses and bacteria.

Methods
Homology searches against viral, bacterial or fungal proteomes were performed via the NCBI or Uniprot BLAST servers and sequence alignments via the CLUSTAL server at UniProt 4 . Homology with allergen sequences was determined by interrogation of the Structural database of Allergenic Proteins 50 http://fermi.utmb.edu/SDAP/index.html and from homology searches at the AllergenOnLine database http://www.allergenonline.org/ 41 . Epitopes containing viral/beta amyloid matching sequences were found using the Human Immune epitope database. www.immuneepitope.org 89 . B-Cell epitopes were identified using the BepiPred server http://www.cbs.dtu.dk/services/BepiPred/ 60 .

Results
The homology between HSV-1 glycoprotein B and beta-amyloid 1-42 is shown in Fig 1. A homology search against viral proteomes showed that a number of other viruses contain this VGGVV sequence (Table 1). These include adenovirus 8, Dengue virus, Herpes Simplex (HSV-1, 2 and 6) hepatitis C, Lactate dehydrogenase-elevating virus, the polyoma virus and HIV-1, viruses infecting family pets (cats, guinea pigs and goldfish) and farmyard animals (cattle, horses, pigs and poultry) and viruses infecting certain foodstuffs (cherries, radish, strawberries and raspberries, tomatoes, potatoes, watermelon, oysters, salmon and shrimp). The VVGGV sequence is also present in a number of phages infecting bacteria that cause common childhood and adult diseases (Gastroenteritis, food poisoning, hospital infections (nococosmial), and wound infections). It should be noted that the VGGVV sequence was restricted to 69 viruses and phages (Table 1) out of 2463 viral genomes in the NCBI database.
A further homology search against the viral proteome database revealed that many other common viruses express proteins with marked homology (pentapeptides or more) to other sequences within the beta-amyloid peptide ( Table 2). These viruses include Adenovirus D,

Immune activation in the Alzheimer's disease brain
A number of immune-system related proteins are found in amyloid plaques or neurofibrillary tangles. Interleukin 1 alpha, interleukin 6, and tumor necrosis factor are all been localised within plaques 103 and acute phase proteins involved in inflammation, such as amyloid P, alpha-1 antichymotrypsin and C-reactive protein are also plaque components 31 while Immunoglobulin G is located in the plaque corona 30 . Large increases in IgG levels have been recorded in the brain parenchyma, in apoptotic dying neurones and in cerebral blood vessels in the Alzheimer's disease brain 25 . Complement component C3 is found in Alzheimer's disease amyloid plaques along with complement C4 105 . Complement components Clq, C3d, and C4d are present in plaques, dystrophic neurites and neurofibrillary tangles 72 .
The membrane attack complex (MAC), composed of complement proteins C5 to C9, forms a channel that is inserted into the membranes of pathogens, destroying them by lysis. These components cannot be detected in temporal cortex amyloid plaques in Alzheimer's disease 104,105 . However the MAC complex is present in dystrophic neurites and neurofibrillary tangles 72 and others have detected this complex in neuritic plaques and tangles, along with deposition of C1q, C3 and clusterin 116 . The membrane attack complex has also been detected in the neuronal cytoplasm in AD brains and associated with neurofibrillary tangles and lysosomes 48 .The presence of the MAC complex in neurones might suggest that neuronal lysis by the MAC complex could contribute to neuronal cell death 72 .

Pathogen seropositivity in Alzheimer's disease
Increased seropositivity to IgM anti HSV-1 antibodies has been reported to predict the risk of subsequently developing Alzheimer's disease 63 . Increased seropositivity in Alzheimer's disease has also been reported for Helicobacter pylori 67 and increased HHV-6 immunoreactivity has been observed in Alzheimer's disease CSF samples 114 . No differences in the seropositivity of Adenovirus, Chlamydia Group B, Coxiella burnettii, Cytomegalovirus, Herpes simplex virus, Influenza A, Influenza B, Measles and Mycoplasma pneumoniae were found in a study of 33 Alzheimer's disease patients and 28 controls 84 . However, there are a large number of diverse pathogenic antigens with homology to beta-amyloid, any of which may have been present, potentially provoking immune response-related neuronal loss, many years prior to seropositivity testing.

Antibodies and antigenicity
The tenet of the autoimmune hypothesis is that viral and other pathogens and environmental allergens with homology to beta-amyloid will trigger the production of antibodies that also target the beta-amyloid peptide. Those provoking a robust immune, inflammatory and complement response risk killing beta-amyloid containing neurones. Antibodies to the Potato virus 39 and to Borrelia antigens 74 ) have already been shown to label amyloid plaques in the Alzheimer's disease brain and antibodies to a phage epitope (AEFRH) also label beta-amyloid 35 .
Potential cross-reactivity can partly be tested in silico but ultimately requires the characterisation of the autoimmune beta-amyloid epitopes present in Alzheimer's disease and crossreactivity testing between pathogenic and environmental antigens and beta-amyloid. Thus far, the precise epitopes labelled by beta-amyloid autoantibodies have not been fully characterised.
The initial rendezvous of antigens with the immune system is with B-cells which bind to, engulf and digest the antigen. The B-cell epitope antigenicity prediction 60 for beta-amyloid is illustrated in Figs 2 and 3, over which are laid the pathogens and allergens that match particular sequences within the beta-amyloid peptide. As can be seen, Coxsackie, HIV-1, HSV-1, Hepatitis C, influenza, mumps, the respiratory syncitial virus, rhinoviruses (common cold) and clostridia, enterobacteria and vibrio phages match sequences that lie within predicted antigenic regions of beta-amyloid (Fig 2). An increase in immunogenicity is also observed within the VGGVV sequence, which has been used as an epitope to label beta-amyloid 93 . Allergenic proteins whose sequences match those of antigenic portions of beta-amyloid include those from fungal, nut, pollen, insect venom and house dust mites ( Table 2, Fig 3).
As a further theoretical test, all consecutive tetrapeptide sequences within beta-amyloid were screened against the human epitope and the HIV-1 molecular Immunology databases to determine which antigens contain these epitopes. As shown in Tables 2 to 7 many viral, fungal, bacterial, parasite and allergen antibodies contain tetrapeptide sequences from the beta-amyloid peptide. In addition, apart from beta-amyloid isoforms and a small number of mammalian proteins, the major matching epitopes were concentrated in viral, bacterial, fungal and allergen classes (i.e. not other species or other mammalian proteins). These databases contain over 73,000 epitopes, so again the results returned are highly specific to these classes. This at least shows that antibodies that have been used to label these viruses, pathogens and allergens, contain beta-amyloid sequences, and that potential cross-reacting antibodies are concentrated in these classes.
The ability of autoantibodies to adsorb beta-amyloid can be considered as useful, and certain of these appear to be associated with reduced plaque volume 54 . Other autoantibodies, present in normal and Alzheimer's disease sera, also possess catalytic properties and are capable of destroying the toxic peptide. Such antibodies do not form stable immune complexes and are less likely to activate immune inflammatory and complement related cell lysis 79 . This is where the danger lies, as antibodies capable of mounting a full-blown immune response against beta-amyloid are likely to kill the cells in which the peptide resides. It is relatively common to find Alzheimer's disease pathology at autopsy (plaques and tangles) in patients who were cognitively normal shortly before death 47 . In such patients, it is tempting to suggest that immune-activating antibodies are rare.
Catalytic beta-amyloid auto-antibodies cleave the beta-amyloid peptide primarily between (H/Q) at Abeta [13][14] , and to a lesser extent at other positions as shown in Figs 2 and 3 99 . The major cleavage site is within a non-immunogenic region of the peptide. Interestingly, several plant viruses (Amaranthus, Pepper, Pistachio, Potato, Rice, Soybean, Sunflower and Zucchini (courgette) and allergens (Rice, Pistaccio, Soybean as well as the cat allergen and insect venoms) express proteins homologous to this particular region. Rather intriguingly, several of these plants/vegetables are constituents of the Mediterranean diet which has been reported to reduce the risk of developing Alzheimer's disease 90 . Such diets impact on cholesterol/lipoprotein function and on atherosclerosis, a major contributory factor in Alzheimer's disease 28,101 but could also conceivably be related to beneficial antibodies generated by the ingestion of common plant viruses, a possibility that also suggests potential immunisation strategies.
Certain proteins from three cancer-causing viruses, Epstein-Barr (HSV-4), Hepatitis B and the human Papillomavirus virus are also homologous to this catalytic antibody target region, suggesting a plausible explanation for the inverse association between cancer and Alzheimer's disease 85

Vaccinations and Alzheimer's disease
It has been noted that the risk of developing Alzheimer's disease was reduced following vaccination against diphtheria, influenza, polio or tetanus 106 . Again, sequences within proteins of these pathogens are homologous to beta-amyloid as shown in Fig 2, (which represents the viral proteins rather than the epitopes used for vaccination). This is rather encouraging as it suggests that, if the effect of vaccination is due to beta-amyloid homology, other vaccines might similarly reduce the incidence of Alzheimer's disease. The papillomavirus vaccine, already used to prevent cervical cancer 40 , is a prime candidate as a viral protein matches the beta-amyloid region targeted by catalytic autoantibodies. However, as there is an evident danger of creating potentially toxic beta-amyloid autoantibodies, characterisation of the epitopes targeted by such vaccines must be of prime concern.
Epidemiological studies These observations suggest that Alzheimer's disease may have an autoimmune component, triggered by antigenic proteins with homology to beta-amyloid. Such a scenario helps to explain several, indeed most, epidemiological features of Alzheimer's disease. For example, Atopy and autoimmune diseases are common in Alzheimer's disease 33,38 , and could relate to allergen homology with beta-amyloid. Alzheimer's disease prevalence is higher in women and in Afro-Americans 1 , as is HSV-2 prevalence 2 .
. Herpes simplex infection is a risk factor in Alzheimer's disease acting in synergy with the APOE4 allele 65 and dementia, with Alzheimer's disease pathology is common in HIV-1 infected patients 34 , a situation perhaps explained by viral protein homology with beta-amyloid. Such homology with pathogen proteins could also explain the association of Borrelia Burgdorferri, Helicobacter pylori, C.Pneumoniae 21,74 and by inference P.Gingivalis and Streptococcus mutans (oral pathogens causing periodontitis and tooth loss 55,97 ) with Alzheimer's disease. The incidence of Alzheimer's disease is also related to the number of pregnancies 20 , which might be explained by greater exposure to common childhood illnesses (cf phage/bacterial homology with beta-amyloid). It has also been noted that nuns, who are less exposed to sexually transmitted pathogens, show some resistance to the ravages of Alzheimer's disease 47 . An autoimmune scenario may also explain why the use of non-steroidal anti-inflammatories reduces the risk of developing Alzheimer's disease 73 as do fish consumption and the Mediterranean diet 76,90 .Such diets are rich in N-3 polyunsaturated fatty acids, which also possess anti-inflammatory properties 61 . The beneficial effects of these diets and also of statins 111 are likely to be related to multiple problems in cholesterol and lipoprotein homoeostasis in Alzheimer's disease 13 but may also be related to pathogens and the immune system. For example viral entry is often lipid dependent a a factor related to fusion of the viral lipid envelope with cell membranes, and the cellular entry of HIV-1 and herpes simplex is cholesterol and lipid raft -dependent and blocked by nystatin 7,16 . Statins also have immunosuppressant properties 108 . The potential role of viruses as constituents of the Mediterranean diet has been discussed above in relation to their homology with beta-amyloid regions targeted by beneficial catalytic autoantibodies. The inverse association between cancer and Alzheimer's disease might also be explained in these terms (see above) and the beneficial effects of vaccination can also be related to pathogen homology with beta-amyloid (diphtheria, influenza, tetanus and polio). The risk promoting effects of aluminium in Alzheimer's disease 36 might also be explained by its common self-prescribed use as an antacid in gastrointestinal disturbances 71 often caused by phage/bacterial pathogens with homology to beta-amyloid.
In short almost all epidemiological observations related to Alzheimer's disease can be explained in terms of pathogens and/or autoimmunity.

Late-onset Alzheimer's disease susceptibility genes
Four genes, Apolipoprotein E, clusterin, complement receptor 1 and PICALM, are the main suspects in Alzheimer's disease 22,44,59 each of which can be implicated in viral life cycles. For example, APOE binds to HSV-1 46 and Hepatitis C viruses 8 and the Alzheimer's disease risk allele, APOE4, is also related to HSV-1 and HIV-1 infection, although, curiously, it protects against Hepatitis C infection 58 . Complement receptor 1 and clusterin are both involved in the complement cascades that play a crucial role in pathogen defence 68 . In addition the influenza virus and HSV-1 both bind to complement receptor 1 on erythrocytes 80 . One of the receptors used by Herpes simplex for cellular entry is the Mannose-6-phosphate receptor, M6PR 11 . This receptor binds to clusterin 62 and its traffic through the endosomal compartments is controlled by PICALM, whose overexpression reduces M6PR localisation in endosomes, suggesting blockade of its transport from the plasma membrane or the trans-golgi network 100 . The Herpes simplex and Influenza viruses also uses exportin (Crm1) dependent pathways for nuclear egress 32,110 . PICALM and other endocyticregulatory proteins bind to Crm1 102 . Many minor genes (see www.polygenicpathways.co.uk/alzpolys for references) are also implicated in viral life cycles. For example lipoprotein receptors implicated in Alzheimer's disease (LRP1, LDLR, VLDLR) are used by human rhinoviruses to gain cellular entry 82 . The nectin receptor PVRL2 is a Herpes viral receptor 69 , as is insulin degrading enzyme 64 . In addition over 30 immune related genes (chemokines, cytokines, complement related, toll receptors HLA-antigens) have been implicated as risk factors. All of these might be expected to modulate immune defence.
Given the plethora of viral and environmental beta-amyloid homologues, and the presence of autoantibodies to beta-amyloid in healthy aged subjects, it seems likely that the functional gene variants in the control group, rather than the Alzheimer's disease risk promoting genes, may be more important as these are presumably those that prevent the ravages of autoimmunity.

Familial Alzheimer's disease
Familial early-onset Alzheimer's disease is caused by a number of different mutations in presenilins and the APP gene, including several mutations as APP 717 (V/I, V/M, V/G, V/L) (London mutation and others) and at KM670/671NL (Swedish mutation) (inter alia :see http://www.molgen.ua.ac.be/ADMutations/) 24 . These are not within the beta-amyloid sequence. Both mutations modify APP processing and increase the generation of Abeta 1-42 18,98 . In so doing, they may thus be able to increase the probability of Abeta encountering autoantibodies, both in the brain and in the periphery resulting in a feed forward further generation of autoantibodies by the immune system.
The APP 717 mutant is within two gamma-secretase cleavage sites generating an undecapeptide as shown in Fig 1. The Swedish mutation is immediately prior to the beta-secretase cleavage site that, with gamma-secretase, generates beta-amyloid and frees the native or mutant APP 670/671 terminus 17 . Homology searches against viral and bacterial proteomes were performed using the native and mutant forms of the APP 717 undecapeptide and the nonapaptide upstream of the Swedish mutation sites (Fig 1). The APP 717G mutant increased the predicted B-type immunogenicity of the peptide, but other mutants were without effect (not shown).
Native APP 717 is homologous to several phages, viruses and bacteria, many of which are common; for example Enterobacteria, lactococcus and staphylococcus phages, adenovirus, cytomegalovirus, the parainfluenza virus and rotavirus and the less common rabies and coronaviruses (Table 4). This native form also shows homology with several bacterial species, which for the most part, with the exception of Lactobacilli, are pathological rather than commensal ( Table 5).
The various APP 717 mutations alter this matching profile in a number of ways. For example the V/F mutation creates matching peptides to Herpes viruses HSV-2, 3 and 8 and markedly increases the number of hits in relation to homology, particularly within bacterial flora (mainly due do many strains of the commensal E.Faecalis) ( Tables 4 and 5).
The APP 717 V/G mutant creates a peptide region homologous to proteins from human herpes virus 6, a pathogen with a seroprevalence approaching 100% 12 and to the JC and BK polyomaviruses which also display high seroprevalence in the normal population ( 39% and 82% respectively 52 ) . This mutant also creates regions homologous to the endemic soil bacterium B.Cereus, to a bacterium causing verrucas, and to P.Gingivalis a constituent of the oral flora that plays a role in periodontitis and tooth loss (tables 4 and 5).
The V/I (London) mutant creates homologues to proteins from over 30 Rhinococcal strains causing the common cold, to the mumps virus and to the common Norovirus responsible for vomiting sickness, as well as to the endemic soil bacteria B.Cereus and S.Aureus, and the commensal E.Coli and P.Gingivalis (Table 4,5).
The V/L mutant creates regions homologous to proteins from Influenza A and B viruses and to Hepatitis C and Herpes viruses 4 and 5, as well as to the commensal E.Coli and Streptococcus Hominis (Tabke 4,5).
The native APP 670/671 peptide is homologous to a number of influenza viruses, which are almost exclusively avine. The Swedish mutant increase the number of viral matches to the upstream peptide, particularly to enterobacterial phages, and creates homologous regions to several species of human adenoviruses, which also show high seroprevalence 70 , and to the respiratory syncitial virus as well as to human rhinoviruses and again to E.Coli. Several of these mutant peptides are homologous to mycobacterial phages (see Tables 6,7 ), perhaps explaining why tuberculosis (as well as possession of the HLA-DR3 allele) have been reported as risk factors in familial Alzheimer's disease 38 . The epitopes generated by these mutants are again concentrated within viral and pathogen proteins, particularly the Vaccinia Virus (Tables 4-7). There do not appear to have been any studies on autoantibodies in familial Alzheimer's disease but the large number of viral and bacterial peptides perfectly matching the mutant surrounds suggests that autoimmunity may also play a major role in this disorder. Although clearly genetic, its pathology may be autoimmune related, as well as to, or perhaps rather than aberrant beta-amyloid processing.
In short, these mutants, in different ways, all increase the number and variety of viral and bacterial matches, to commensal (Particularly E.Coli, E.Faecalis and P.Gingivalis ) as well as to pathological species (Clostridia, Mycobacteria, Vibrio, inter alia) and create homologous regions to very common pathogens causing the common cold, influenza, herpes infections and periodontitis.

APP transgenic mice
If autoimmunity rather than problems in APP processing is relevant, this could explain why APP transgenic models do not faithfully mimic Alzheimer's disease (i.e. extensive cholinergic neuronal loss, loss of hippocampal afferents and efferents and massive cortical degeneration) 56 . There have been two studies assessing the effects of infection in APP transgenic mice. Repeated Streptococcus Pneumoniae infection had no effect on pathology or behaviour in APP Tg2576 transgenic mice (Swedish mutant) 29 . S.Pneumoniae displays homology to both the native and mutant forms of the relevant peptide (Table). Borna virus infection resulted in a reduction in betaamyloid immunoreactivity in the brains of infected Tg2576 transgenic mice 96 . This virus also displays homology to both the native and mutant forms of the peptide (Table). S.Pneumoniae also expresses proteins containing the VGGVV sequence, as do a large number of other bacteria (not shown), while a homology search for Borna virus proteins containing VGGVV yielded no hits. However, in relation to autoimmunity, it is not the infection itself, but the antibodies generated in response to the infection that could cause the problem. A more suitable test might be repeated challenge with specific antigens in APP transgenic models.
Other models capable of producing cerebral beta-amyloid deposition in mice include C.Pneumoniae 66 or HSV-1 infection 113 , the latter also producing hippocampal and entorhinal cortex neuronal loss and memory deficits, perhaps more faithfully reproducing the pathology of Alzheimer's disease 5 . Such models may be useful in APP transgenic mice.

Beta-amyloid and other Vaccines and therapies. .
The F,G and I APP 717 mutants all generate peptides homologous to the related Vaccinia and Variola (Smallpox) viruses (Table 4, 6) . While vaccination against diphtheria, flu, tetanus and polio have been reported to reduce the risk of late-onset Alzheimer's disease (see above) there is also a possibility of vaccine cross-reactivity with endogenous peptides such as beta-amyloid. Many of the epitopes formed by these mutant peptides are contained within Vaccinia viral proteins. There is thus a possibility that smallpox vaccination may be a contributory factor in familial Alzheimer's disease, although this requires characterisation of the epitopes present in the vaccine, which was generated from the virus rather than from peptide components. The vaccinia virus has also been used as a primary vaccination against smallpox 94 following on from Edward Jenner's pioneering experiments with cowpox over 200 years ago 51 . This is a contentious area, but the possibility that vaccination might trigger nefarious autoantibody generation does need to be addressed, particularly in future studies.
The potential use of beta-amyloid antibodies is based on their ability to reduce plaque burden and neurite dystrophy in APP transgenic mice 91 Several studies have demonstrated that beta-amyloid antibodies reduce plaque burden in APP transgenic models and that they can also improve cognitive performance 87 . However amyloid antibodies extracted from the serum of old APP transgenic mice potentiate the toxicity of betaamyloid and Alzheimer's disease patients display an enhanced immune response to the peptide 77 . Again in transgenic mice, different immune backgrounds can influence the type of immune responses elicited by beta-amyloid. For example B-and T-cell responses to beta-amyloid can be modified in HLA-DR3, -DR4, -DQ6 or -DQ8 transgenic mice 27 . HLA-antigen diversity in Man is also likely to determine the outcome of beta-amyloid/antibody interactions.
The results of this survey suggest that different beta-amyloid autoantibodies might cause as well as mitigate against Alzheimer's disease pathology. Beta-amyloid vaccination in Alzheimer's disease (against Abeta 1-42 ) has so far not been successful and sadly resulted in meningoencephalitis and the death of a patient 37 . While certain beta-amyloid antibodies may reduce plaque burden, there is an evident risk that they may also trigger an auto-immune response, potentially killing betaamyloid containing neurones. Catalytic autoantibodies are less able to form stable immune complexes and likely represent the safest way forward in this area 79 . It is intriguing that the betaamyloid region targeted by catalytic autoantibodies matches peptide sequences of viruses that are constituents of the Mediterranean diet and of cancer-inducing viruses (Epstein-Barr, Hepatitis B, and the papillomavirus) as both cancer and the Mediterranean diet are inversely associated with Alzheimer's disease risk (see above). Vaccines to the human papillomavirus already exist 40 and could perhaps be considered as a therapeutic or preventive option in Alzheimer's disease, after due consideration of the epitope matches.
If beta-amyloid autoantibodies are the culprits, techniques such as immunoadsorption, which has proved to be of benefit in Myasthenia gravis (autoimmunity to acetylcholine receptors 107 or immunosuppressant use might be considered as potential therapeutic options. There is indirect evidence in support of such treatment. Natural immunoadsorbants include silica, which is however toxic, 6 , tryptophan and phenylalanine 115 . Levels of silica in drinking water are inversely related to Alzheimer's disease risk 86 and serum tryptophan levels are markedly depleted in Alzheimer's disease patients, a marker of immune activation 109 .Phenylalanine plasma levels are in contrast increased in Alzheimer's disease 83 . Fish oil (see above) also suppresses T-Lymphocyte activation 88 and statins, which have also been reported to reduce Alzheimer's disease risk 111 are also immunosuppressant 108 .

Heterogeneity in genetic and epidemiological studies
If there is one factor common to polygenic disorder research it is the discordance of genetic and environmental association data 9 . However, if, as suggested by this survey, there are dozens of potential Alzheimer's disease triggers, all funnelling towards a common cause, this heterogeneity becomes part of the answer and not part of the problem. Different gene products are related not only to human physiology, but also to pathogen life cycles (e.g. rhinoviruses and lipoprotein receptors, APOE and Herpes Simplex or influenza and complement receptor 1) , a situation that has been discussed in relation to Alzheimer's disease and Schizophrenia susceptibility genes and pathogens implicated in these disorders 14,15 . Viruses and bacteria are not uniformly distributed worldwide, nor is antibody seroprevalence, and different environmental risk factors may similarly vary from region to region. We also possess different different immune backgrounds (HLA-etc) and diverse panels of antibodies deppending on the pathogens we have encountered. The heterogeneity observed in these studies thus has a rational basis, reflecting the heterogeneity of cause, and need not necessarily be considered as a statistical artefact Relevance to other autoimmune and genetic diseases and evolutionary aspects .
Phages and viruses are the simplest form of "life", as defined by the possession of DNA/RNA and a proteinaceous structure, and were long ago proposed as the origin of higher cellular organisms 26,42 . There are currently 2463 viral genomes in the NCBI database, probably representing but a small proportion of those existing on the planet. While perhaps responsible for our origin these predecessors may have donated a legacy of human viral-derived proteins that closely match antigenic proteins in the currently existing virome that could be responsible for many human diseases. Viruses and infection have been implicated in many diseases including Bipolar disorder and Schizophrenia, multiple sclerosis and many others where this phenomenon of viral mimicry is likely to be relevant. This evidently has enormous implications for therapy and prevention of many human diseases. Vaccination, using epitopes against the non homologous human protein regions of the phages and viruses could perhaps destroy the pathogen and prevent the associated problems related to autoimmunity.

Summary and conclusions
Many common viruses, phages, bacteria, fungi, parasites and allergens express proteins with marked homology to antigenic and other regions of the beta-amyloid peptide. Epitope similarity with beta-amyloid is concentrated within these classes and it seems likely that these antigens provide a source of the beta-amyloid autoantibodies found in the ageing population and in Alzheimer's disease. While some of these antibodies may be benevolent, others may stimulate immune, inflammatory and other defensive measures, including complement mediated cell lysis that could kill the neurones in which the peptide resides. Activation of the immune system is supported by the presence of many immune-related proteins in Alzheimer's disease amyloid plaques, and by the presence of the complement membrane attack complex in Alzheimer's disease neurones. Reduced serum tryptophan levels, and an increased immune response to beta-amyloid also suggest immune activation. Familial Alzheimer's disease may also have a strong autoimmune component, as the various APP mutants convert the surrounding peptides to matches to commensal bacteria (E.Coli, E.Faecalis and P.Gingivalis) and to viruses with a high seroprevalence (HHV-6, polyoma viruses, influenza and the common cold rhinoviruses). The categorisation of Alzheimer's disease as an autoimmune disorder explains most of the epidemiological observations in Alzheimer's disease and the major genes implicated in Alzheimer's disease are all related to viral life cycles and/or to the complement arm of the immune defence network.
Diseases caused by these viruses, fungi or the bacteria infected with the phage/beta-amyloid homologues are very common and often recurrent (e.g. colds, influenza gastroenteritis or food poisoning) as is exposure to certain allergens (e.g. dust mites , cat, cow, horse allergens, pollen and food allergens, insect stings, marine algae etc) Over time, and with increasing age, the major risk factor in Alzheimer's disease 53 , antibodies that may also target beta-amyloid are more likely to be produced, gradually increasing the probability of a self-immune attack on neurones containing the beta-amyloid peptide. There may thus be dozens, if not hundreds of triggers promoting Alzheimer's disease risk, all funnelling towards an autoimmune scenario. If this is the case, then Alzheimer's disease might be considered as an autoimmune disorder and immunosuppressants or antibody adsorption might have a role to play in its therapy, once diagnosed. C.Neoformans eradication can result in a complete recovery, in very rare cases of diagnosed dementia 3,45 , and Helicobacter elimination has been reported to ameliorate cognitive function in infected Alzheimer's disease patients 57 . Aggressive targeting of opportunistic pathogens capable of mimicking the beta-amyloid peptide might thus be considered as a therapeutic option. Because so many are homologous to beta-amyloid, such therapy might well have to be tailored to individual pathogens, depending on the species identified by serum assay.
Vaccination against common diseases has already been shown to reduce the risk of developing Alzheimer's disease 106 and, in the long term, vaccination against other common viruses and bacteria might also be of benefit, although it is evident that potential vaccine antibody crossreactivity with beta-amyloid must be a prime concern. One such vaccine may already exist in the form of that for the human papillomavirus.
. In summary, the close homology of diverse viral, fungal bacterial and allergenic antigens with beta-amyloid and to peptides generated by APP mutations suggests an autoimmune component to both familial and late-onset Alzheimer's disease, triggered by these antigenic homologues. The autoimmune scenario explains many epidemiological observations and genetic studies also implicate the immune system and viral life-cycles. If autoimmunity is important, vaccination, viral and pathogen elimination and immunosuppressant therapy might be expected to play a role in the prevention and therapy of Alzheimer's disease and perhaps provide new rationales for developing a cure. This type of viral matching is also likely to be relevant to other autoimmune and genetic disorders and may be a near universal phenomenon reflecting our viral ancestral roots. This generality could open new therapeutic avenues in many human diseases. Non-stinging nettle family growing on rubbish and walls Table 2: Viral and allergen proteins matching tetrapeptide sequences, or more, within the beta-amyloid peptide (1-42). Shaded dark grey blocks represent viral matches to pentapeptides, or more, as shown by the numbers in each block. Allergens are boxed in light grey and protein accession numbers are provided. Species with protein epitopes containing these sequences are also illustrated. Short beta-amyloid epitopes (antibodies raised to peptides of 4-6 amino acids) are illustrated by the grey boxes with dashed surrounds. Matches to these short immunogenic peptides are perhaps more likely to cross-react with beta-amyloid.     Table 5 The effects of the APP 717 mutations on homology to bacterial proteins. Protein accession numbers are provided and amino acid matches are indicated by the asterisks or by the red letter of the mutant amino acid. Commensal or common soil species are highlighted in bold. * * L * * * * Table 6 The effects of the Swedish APP mutation on homology to viral proteins. Protein accession numbers are provided and amino acid matches are indicated by the asterisks or by the red letter of the mutant amino acid. Species with proteins containing epitopes matching those of the peptde amino acids are also shown (Marked by E) Phages infecting commensal bacteria and common viruses (eg Rhinoviruses and influenza) are highlighted in bold.  Table 7 The effects of the Swedish mutations on homology to bacterial proteins. Protein accession numbers are provided and amino acid matches are indicated by the asterisks or by the red letter of the mutant amino acid. Commensal or common soil bacteria are highlighted in bold.

T E E I S E V K M
Atopobium Parvilum C8WAI5 * * * * * * * * Viral protein matches to different regions of the beta-amyloid peptide in relation to the predicted B Cell Epitope antigenicity. Predicted epitopes are marked with an asterisk and the Y-axis is an index of antigenicity. The VGGVV epitope has been used to label beta-amyloid and is marked + as are other short epitopes used to label beta-amyloid (QKLV, FFAE, IIGL) Other short epitopes used to label beta-amyloid include MGGVV, VGGVV, MVGGVV, VGGVV and GGVVIA). The viruses and bacteria in black boxes represent those where vaccination was reported to reduce the incidence of Alzheimer's disease. These alignments are to viral proteins rather than to epitopes within the vaccine. The arrows represent the beta-amyloid cleavage sites of the catalytic beta-amyloid autoantibodies isolated from Alzheimer's disease sera. H*↑Q is the major site of catalysis. Note that cancer and plant viruses overlap this region. Burk= Burkholderia Phage; Camp Phage = Campylobacer Phage; HumRespSyn = Human respiratory syncitial Virus; HepC = Hepatitis C; HIV = Human Immunodeficiency Virus; HSV-1 = Herpes simplex (Human Herpesvirus 1); Mycobact = Mycobacteria phage; Strep = Streptococcus phage; C.Tet = Clostridium Tetani, C.Dip = Corynebacterium diphtheriae (See Table 2 for Accession numbers)

Figure 3
Allergenic protein matches to different regions of the beta-amyloid peptide in relation to the predicted B Cell Epitope antigenicity. Predicted epitopes are marked with an asterisk and the Yaxis is an index of antigenicity .Allergens expressing proteins that match different regions of the beta-amyloid peptide are aligned with their respective matches. The arrows represent the betaamyloid cleavage sites of the catalytic beta-amyloid autoantibodies isolated from Alzheimer's disease sera. H*↑Q is the major site of catalysis (see Table 2 for Accession numbers).