Over 600 genes have been implicated in schizophrenia in association studies, supporting the contention that multiple genes of small effect contribute to this condition [1, 2] (see http://www.polygenicpathways.co.uk/schizgenesandfunc.htm for association references). These genes cluster together in clearly defined signalling networks related to the diverse subpathologies of schizophrenia [3–7]. Epistasis between genes within these same signalling networks markedly affects the degree of risk-promotion [8–10], in part, explaining the inconsistency in genetic association studies.
Schizophrenia has also been associated with prenatal complications including maternal rubella (German measles), influenza [12, 13], Varicella zoster (chicken pox), Herpes (HSV-2), common cold infection with fever, or poliovirus infection while in childhood or adulthood, coxsackie virus infection (in neonates) or Lyme disease (vectored by the Ixodes tick and Borrelia Burgdorferri) or Toxoplasmosis have been reported as risk factors [19, 20] (see Table 1). The human endogenous retrovirus, HERV-W, has also been implicated in schizophrenia. A number of schizophrenia-related genes are implicated in the life cycles of these pathogens, suggesting an interplay between genes and risk factors.
Many schizophrenia genes relate to the immune network [5, 6, 22, 23]. Immune activation is also observed in the schizophrenic brain [24, 25] or in lymphocytes [26–29]. A number of autoantigens/autoantibodies to key schizophrenia-related proteins have also been reported. These include dopamine, serotonin, acetylcholine, and NMDA receptors; inter alia (Table 2). Maternal immune activation in animal models has also been shown to generate phenotypes relevant to schizophrenia in the offspring.
As shown below, genes, risk factors, and immunity can be linked together forming a unifying pathway whose elements are interdependent. Dysfunction of this network which is conditional upon interactions between its three branches may be responsible for schizophrenia.
The human herpesvirus 2 genome (NC_001798) as well as those of the rhinovirus (NC_001490), rubella (NC_001545.1) and Varicella zoster (NC_001348.1) and HERV-W (NP_055405.3: env polyprotein) viruses, Borrelia Burgdorferri (NC_011728) and T. Gondii (NC_001799: Partial genome) were screened against the human proteome using the NCBI BLAST server and the Entrez query filter “schizophrenia”. The HERV-W, influenza, HSV-2 and rubella viruses were also screened unfiltered (Translated pathogen genome versus human proteins: BlastX). The BLAST algorithm detects overall homology between entire gene or protein sequences, and it is necessary to set parameters to low significance levels in order to detect short intraprotein consensus homology. The parameters used were: Expect 20,000, E value = 100,000; matrix PAM30. The original BLAST results are stocked at http://www.polygenicpathways.co.uk/blasts.htm. Information for all abbreviations is available at this site, provided by the NextBio highlighting service.
BLAST files were scanned by an online tag cloud generator producing tags sized according to gene word occurrences http://www.tagcloud-generator.com/generator.php#anker. Word occurrences were counted using a “Highlightall add-in” for Firefox https://addons.mozilla.org/en-US/firefox/addon/4240/.
Antigenicity (B-cell epitope prediction) was estimated using the BepiPred server http://www.cbs.dtu.dk/services/BepiPred/ (Table 4).
Kegg pathway analysis of 632 schizophrenia susceptibility gene candidates was performed using Kegg mapper http://www.genome.jp/kegg/tool/color_pathway.html. The results of this analysis are available at http://www.polygenicpathways.co.uk/keggszgenes.htm. Venn diagrams were constructed online at http://www.bioinformatics.org/gvenn/index.htm.
Genes and risk factors with at least one positive association are included in this study. Although certain genes and risk factors are clearly more important than others, and problems of replication in both gene and risk factor studies abound, gene, gene, and gene/environment interactions may explain some of the heterogeneity. For example many schizophrenia-related genes are involved in the life cycle of T. Gondii, but may be irrelevant if this pathogen is not encountered. Similarly T. Gondii infection may have little effect is such gene variants are not present. Pathway analyses of genome wide association data, and previous studies, are showing that the risk-promoting effects of many genes in similar pathways are better predictors of risk, than when treating each gene in isolation (see Section 1). Although some of these factors may be false positives, many genes and risk factors may have a role to play in certain conditions, but the greater import of genes such as DISC1 or neuregulin is recognised.
3.1. Autoantigens in Schizophrenia
Many autoantibodies have been reported in schizophrenia. The pathogens implicated in schizophrenia also express proteins that are homologous to these autoantigens. Again the profile of each autoantigen or pathogen is distinct as shown in Table 2.
DISC1 is a key “hub gene” in schizophrenia linked, via its interactome, to many other schizophrenia susceptibility gene products [3, 63–66]. Its viral homology is illustrated in Figure 2. The Varicella virus is homologous to DISC1 in several regions, over its entire length, many matches in regions of high immunogenicity. These figures illustrate the types of matches seen in other proteins and shows that the vatches are often part of larger gapped consensus sequences. Interestingly, Varicella infection also results in the production of antibodies to pericentrin, a DISC1 binding partner.
DISC1 is a highly immunogenic protein, as predicted by B-cell epitope prediction (Figure 4). Autoantibodies to DISC1 have not been reported in schizophrenia. However, the viral risk factors implicated in schizophrenia express proteins that are homologous to the highly antigenic regions of the DISC1 protein, as shown in Figure 2. These viral proteins are equally antigenic and antiviral antibodies might also thus be expected to target multiple regions of the DISC1 protein.
3.3. Viral Proteins Are Part of the DISC1 Interactome
DISC1 and many of its binding partners, or other members of its interactome, contain vatches that are homologous to proteins expressed by the Rubella virus (Figure 5). (Other viruses also display this property, although the interactome members targeted are distinct, and specific for each virus (see http://www.polygenicpathways.co.uk/vatches.htm). Upon infection, viruses might therefore be considered as extraneous spurs to these types of protein/protein networks, and are likely to markedly affect their integrity. Indeed, several viruses, including herpes simplex, hepatitis C, Epstein-Barr, the cytomegalovirus, adenovirus and Coxsackie virus are known to bind to DISC1 interaction partners (Table 4).
3.4. Viral DNA within the Human Genome
The insertion of viral DNA into the human genome had until recently been thought to be the preserve of retroviruses. However the incorporation of DNA into mammalian genomes has recently been demonstrated on a large scale for both RNA and DNA viruses. Viral integration may be mediated by nonhomologous recombination with chromosomal DNA or, in the case of RNA viruses, by interactions with host chromosomal retrotransposons [68, 69]. It has also been shown the herpes virus HHV-6 can be transmitted from parent to child via chromosomal integration. The BLAST analyses of the viruses detailed in this paper, and of others at http://www.polygenicpathways.co.uk/blasts.htm clearly show that viral DNA from many species is present within the human genome. This viral homology may well cover the entire human genome. For example, a Blast of human chromosome 10 against all viral genomes (almost 3,000 viral forms) yielded 119,857 hits with entire coverage of 135.5 million bases. Viral DNA is thus both inter and intragenic (Figure 1). It has been proposed that retroviral integration, into paternal and maternal gene lines, inserting several genes at once and effectively creating a new being, is responsible for evolutionary saccades. The fact that RNA and DNA nonretroviruses can also be so incorporated has important implications in this area.
The HSV-2 virus is homologous to several dopamine receptors and the BLAST pictogram shows how the same virus provokes repeating patterns in the human proteome (Figure 6). The same is true of the Herpes simplex virus (HSV-1) which is homologous to multiple lipoprotein receptors as well as to multiple kinases or of the cytomegalovirus which expresses proteins homologous to many chemokine receptors (see http://www.polygenicpathways.co.uk/blasts.htm). One interpretation of this, given the ability of chromosomal integration, is that repeated viral visits to the human genome over millions of years are responsible for the creation of gene families.
It is also possible that viral/human homology reflects convergent viral evolution, although this is difficult to reconcile with the presence of viral DNA in intergenic regions, for which there would be little evolutionary drive or selective pressure. It is also plausible that a bidirectional transfer of human and viral DNA could be at work.
For whatever reason, the result is that human proteins resemble those expressed by a multitude of today's viruses and other pathogens. Upon infection, these pathogens are thus able to interfere with the function of their human counterparts in a number of ways (see below).
3.5. Copy Number Variations and the Effects of Parental Age on Risk
Repeated viral insertion could well explain copy number variations, which are associated with a number of diseases, including schizophrenia [72, 73]. As their number increases, so will the number of matches to the same viral proteins, thus increasing the risk of viral interference and autoimmunity. As viral infection can be passed from parent to child via chromosomal integration, perhaps this is also why both paternal and maternal older age have been reported as risk factors in schizophrenia and other disorders [74, 75].
3.6. KEGG Pathway Analysis of Schizophrenia Susceptibility Genes
The color-coded pathways for this analysis are posted at http://www.polygenicpathways.co.uk/keggszgenes.htm. It confirmed the involvement of a number of polygenic pathways, including long-term potentiation and oxidative stress growth factor/neuregulin pathways, neuroactive ligand pathways (dopamine/serotonin/glutamate and others) as well as dopamine metabolism pathways. In the context of this review, a large number of immune-related pathways are traced out by these genes, together with many pathogen-related pathways, including toxoplasmosis, which heads the list (Table 5). The involvement of schizophrenia related genes in the life cycles of pathogens has been the subject of a previous review and this relationship is supported by this analysis. Other pathogen related pathways relating to amoebiasis, Staphylococcus aureus and Helicobacter pylori infection, might indicate the involvement of other pathogens in schizophrenia, although such pathways could also be considered as generic pathways related to many pathogens.
There is no specific viral life cycle pathway within the KEGG dataset. However, viruses use adhesion molecules as receptors, endocytosis for cellular entry and the intracellular actin and tubulin networks for migration to and from the nucleus, mediated via dynein and kinesin motors. They also subjugate intracellular vesicular trafficking pathways, and are able to subvert both lysosomal and phagosomal pathways. Their exit may depend upon exocytosis, or by apoptotic or other means of killing their host cell. These pathways are heavily represented within the schizophrenia gene analysis.
3.7. Mechanisms of Action
Individual proteins are homologous to multiple viral proteins, which nevertheless are specific for a spectrum of viruses, while individual viruses are homologous to a large but specific subset of human proteins.
Our proteomes therefore contain proteins with sequences exactly matching those in the current virome, and in the proteomes of bacteria and other pathogens, which are also subject to phage or viral infection. Pathogens' proteins are therefore homologous to receptors, transporters, peptide messengers, growth factors, and other protein products of diverse gene families. Upon infection, surrogate dopamine, NMDA serotonin and other receptors, as well as transporters and enzymes are made available, which in effect may steal the ligands of their human counterparts. It is already known that the dopaminergic ligand, amantadine, binds to the influenza virus, which expresses proteins homologous to dopamine receptors (Table 3). When homologous to peptide ligands, viral proteins may occupy and block or perhaps stimulate their cognate receptors, or use them for entry, as is the case with the AIDS virus and the CCR5 and CXCR4 chemokine receptors.
This is illustrated by the Norovirus (Norwalk) which causes vomiting sickness. The virus expresses proteins homologous to monoamine and other amine oxidases as well as to a number of dopamine and monoamine transporters (Table 6). Dopamine subversion by the viral homologues would be expected to increase dopamine levels resulting in emesis, thus explaining the recurrent vomiting produced by infection.
The potential interference by viruses within protein/protein networks is well illustrated by the homology of rubella proteins to DISC1 and other members of its interactome, and by the fact that many viruses have indeed been found to bind to these components (Table 4).
The homologous human proteins of the viral risk factors implicated in schizophrenia correspond to the genomic locations of 632 schizophrenia susceptibility genes (see Venn diagrams). Both negative and positive genetic association results have been reported for these many genes and it now seems plausible that, in some cases, this may be due to the presence or absence of active infection with these and other pathogens, and that DNA assays have been detecting pathogen as well as human DNA in the blood samples used for assay. There is evidently no way of discriminating viral or bacterial double-stranded DNA from human DNA.
This is not specific to schizophrenia, as the viruses implicated in Alzheimer's disease (HSV-1, HIV-1, HHV-6 and the cytomegalovirus) [125–127] are also homologous to proteins encoded by Alzheimer's disease susceptibility genes see http://www.polygenicpathways.co.uk/blasts.htm.
It seems that a viable interpretation, given the same phenomenon in these diseases, is that these genes are susceptibility genes precisely because they encode for proteins with homology to the viral risk factors. Infection and genetics therefore appear to be interdependent. The pathogens may promote disease if the human genes encode for homologous products, and the genes promote disease if the homologous pathogen is encountered. Such interdependence likely explains the heterogeneous data in both gene and risk factor association studies.
Other pathogens, including Borrelia Burgdorferri and T. Gondii have also been implicated in schizophrenia. These too express many homologous proteins to both viral and human proteomes. These parasites tend to be associated with schizophrenia in adulthood, while viral infections are predominantly prenatal risk factors. These may have primed the antibody network to respond to homologous antigens expressed by Borrelia or T. Gondii, suggesting that detection and elimination of these pathogens may be of therapeutic benefit in adult life.
Schizophrenia is a neurodevelopmental disorder [129, 130] and, as the risk-promoting effects of viruses are related to maternal infection, it is possible that knockdown or interference of foetal proteins by viral-induced antibodies targeting their human counterparts may contribute to the neurodevelopmental disturbances observed in schizophrenia. Indeed DISC1, neuregulin, ERBB4, FEZ1 or COMT knockout mice display many of the pathological and behavioural symptoms associated with schizophrenia [131–135]. Viral interference with these same proteins might be expected to promote the same effects, but on a massive scale, targeting many relevant proteins at once. It is also possible that such autoantibodies play a role in the comorbid conditions associated with schizophrenia, for example autoimmune disease such as Thyrotoxicosis, celiac disease, acquired haemolytic anaemia, interstitial cystitis, or Sjogren's syndrome.
Autoantibodies to several proteins have been reported in schizophrenia (muscarinic, nicotinic, dopaminergic and NMDA receptors, inter alia, (Table 2) and all are homologous to proteins expressed by the risk factors in schizophrenia. The effects of antibody knockdown have not been analysed for any schizophrenia related proteins, but have been reported for the microtubule-related protein tau, in relation to Alzheimer's disease. In mice, tau immunisation produces tau hyperphosphorylation, neurofibrillary tangles and axonal damage as seen in the human condition. Tau (MAPT) is homologous to Herpes simplex (HSV-1) and a number of other pathogens. Such effects are relevant to the autoantigens observed in schizophrenia.
Schizophrenia is also a degenerative disease in adolescence or adulthood, characterised by oligodendrocyte cell loss, impaired synaptic connectivity and pyramidal cell dendrite shrinkage [41, 138–140], In the light of the above homologies it seems likely that such degenerative changes may relate to autoimmune-related attack of these diverse compartments. Indeed there is evidence for microglial activation in the schizophrenic brain and several studies have reported changes in the cytokine profile in the brain, CSF or peripheral immune compartments [24, 142–146].
3.8. Clinical Implications in Schizophrenia and Other Conditions
These data suggest that susceptibility gene products are the vehicles enabling the risk-promoting effects of pathogenic risk factors, via the interactions described above, and that the two are indispensable for the genesis of schizophrenia. Pathogen detection and elimination or vaccination, particularly prior to pregnancy might be expected to reduce the incidence of schizophrenia and also to be of clinical benefit in adulthood. Interestingly, vitamin D is able to stunt the growth of T. Gondii and low levels of this vitamin, both prenatally and in adulthood, have been associated with schizophrenia risk, although abnormally high levels are also a risk factor. Pharmaceutical effort in this direction may also vastly improve the armoury and safety of drugs against parasites such as T. Gondii and Borrelia.
Autoimmunity, involving several key schizophrenia-related proteins may well be a consequence of pathogen infection, and related to viral/human protein homology. Antigen and antibody removal by immunoadsorption techniques might therefore also be if clinical benefit.
This scenario suggests a novel and probably common class of “pathogenetic” autoimmune disease caused by pathogens but dependent on our genes. Indeed, the same phenomenon has been observed in Alzheimer's disease where the risk factor herpes simplex expresses proteins containing peptide matches to the products of multiple susceptibility genes. Work from Kanduc's laboratory has also shown that 30 viral proteomes, including many nonretroviruses, contain multiple pentapeptide matches to many human proteins. This is corroborated by data posted at http://www.polygenicpathways.co.uk/blasts.htm which shows, inter alia, that Bornavirus proteins, a virus implicated in Bipolar disorder, display this type of homology in relation to Bipolar disorder susceptibility gene products, that the coronavirus implicated in Parkinson's disease expresses proteins homologous to the PARK7 gene product and to dopaminergic and oxidative stress-related proteins, and that multiple sclerosis autoantigens are homologous to the products of the Epstein-Barr virus which has been implicated in this disorder. Our genomes and polymorphisms determine which vatches we possess, which pathogens match these sequences and which pathogen-related disorder we might develop. Environmental variables, and vaccination, determine which pathogens we encounter and our immune system (HLA-antigens and immune background determined soon after birth) may determine how we deal with these pathogens. With the power of current day bioinformatics, it should be possible to rapidly identify all vatches in the human proteome and to pair them with the various pathogenic species and human diseases. This would greatly aid our understanding of the implication of pathogens in disease and may lead to radically new therapies and prevention strategies in many disorders.