Toxoplasmosis and Polygenic Disease Susceptibility Genes: Extensive Toxoplasma gondii Host/Pathogen Interactome Enrichment in Nine Psychiatric or Neurological Disorders

4.1. KEGG Pathway Analysis of the T. gondii/Host Interactome

2792 proteins or mRNAs are involved in the host/pathogen interactome, approximating to 10% of the human genome. A summary of the KEGG pathway analysis of the human arm of this interactome is provided in Tables 1 and 2.

As might be expected, a high proportion of genes are involved in the immune system and in pathogen defence pathways. Many are also involved in the life cycle pathways of a number of viruses, bacteria, and other parasites (Table 1). These stem in part from the common immune and defence mechanisms not only related to the pathogens (chemokine and cytokine activation, etc.), but also related to common signalling networks. The involvement of dedicated bacterial and viral defence pathways in the interactome (NOD, RIG1, and cytosolic DNA-sensing pathways) is likely to impact upon viral defence, although in which direction is impossible to determine. Interestingly, T. gondii produces an interferon-like substance with antiviral activity [50]. The host intestinal microbiome also influences T. gondii and is also able to act as an adjuvant in response to T. gondii infection by stimulating dendritic cells that provide the immunostimulation necessary to combat the parasite [51]. Such effects and the shared pathways between pathogens highlight an important potential cross talk between elements of the microbiome.

Diverse pathogens are implicated in all of the diseases in this study, and many of the pathways traced out by the disease susceptibility genes, per se, (posted on the PolygenicPathways website) also involve multiple viral and pathogen life cycle and immune-related pathways.

A number of cancer-related pathways are highly represented in the T. gondii interactome (Table 1). While a recent study has suggested its involvement in brain cancer, based on a correlation between cancer mortality and T. gondii seroprevalence [52], the parasite is able to arrest the growth of other cancerous cells via stimulation of the immune response and inhibition of angiogenesis. Antitumour effects have been observed in relation to spontaneous mammary tumours, leukaemia, lung cancer, and carcinogen-induced tumours following injections of Toxoplasma antigen or viable parasites in laboratory animals or cells [53].

Several autoimmune and atopic disease networks are involved in the parasite interactome. A high T. gondii antibody seroprevalence (as well as to the cytomegalovirus and the Epstein-Barr virus) has been observed in systemic lupus erythematosus, and it has been suggested that antibodies raised to the pathogen may contribute to the autoimmunity characteristic of this condition via pathogen/host protein mimicry [16, 54, 55]. Conversely, T. gondii infection has been shown to prevent the development of lupus-related nephritis in rabbits [56], a factor perhaps related to the immunosuppressant properties of parasitic infection. Toxoplasmosis has been reported to decrease leukocyte, natural killer cell, and monocyte counts in men, while increasing the same in women, with reduced B-cell counts in both [57]. No references were found for relationships between toxoplasmosis and Type 1 diabetes, a pathway also figuring in the interactome. Prior T. gondii infection has been associated with poor outcome in heart transplant patients (allograft rejection) [58]. Toxoplasmosis and other infectious agents have also been linked to cardiac myopathy [59–62], and diverse pathways of which were concentrated in the T. gondii interactome. In relation to asthma, the hygiene hypothesis, linking a reduced incidence of childhood infections (in general) to the worldwide increase in asthma and other allergic conditions, may be related to the concentration of T. gondii interactome genes within the asthma pathway, although a positive correlation of T. gondii infection and asthma has also been noted in Sweden [63–65]. The parasite clearly has multiple effects on diverse immune-related networks as noted above, and such effects are likely to be both beneficial and nefarious. For example parasite-related immunosuppression may well be useful (but perhaps not advisable) in autoimmune diseases such as multiple sclerosis but might also be expected to favour other infections.

Many of the more specific signalling networks within the interactome (Table 2) can be related to the general processes described above. While the MAP kinase pathway is involved in a multitude of functions, the JAK/STAT pathway is involved in cytokine signalling, also bridging cytokine activation to cancer pathways [66]. The calcium signalling pathway is also activated by many processes and more specifically by voltage or receptor-gated ion channels (and is relevant to the “channelopathies” implicated in autism, depression, bipolar disorder and schizophrenia, and in neurological disorders [67, 68]) or by processes modulating intracellular stores, while the phosphatidylinositol signalling system is also involved in the actions of multiple messengers. TGF beta regulates proliferation, apoptosis, differentiation, and migration (definition from KEGG). Calcium channel blockers, calmodulin antagonism, or extracellular calcium depletion diminish cellular invasion by the parasite [69, 70]. The P53 and growth factor signalling networks (ErbB, VEGF) can be cancer related, while insulin signalling is evidently related to diabetes. PPAR receptors control the transcription of many genes especially those related to fatty acid metabolism, but also those involved in cell proliferation and differentiation [71]. These and other pathways control a host of processes from embryonic differentiation to cellular death and apoptosis, and many metabolic pathways that are too numerous to individually review.

In relation to the diseases that are the object of this study, the Alzheimer’s and Parkinson’s disease pathways were both represented, as were the complement, PPAR, and terpenoid (cholesterol synthesis) pathways relevant to Alzheimer’s disease [72], and the ubiquitin pathway relevant to Parkinson’s disease and other degenerative disorders [73]. Erbb signalling is highly relevant to the control of peripheral and central myelination [74], and thus to multiple sclerosis and Alzheimer’s disease, but also to a range of psychiatric disorders including autism, anorexia, ADHD, bipolar disorder, depression, and schizophrenia [75]. Myelin is exquisitely sensitive to oxidative stress and glutathione depletion (c.f. glutathione pathways), and the glutathione precursor N-acetylcysteine has been shown to be of benefit in a number of psychiatric disorders [76–80]. The diverse neurotransmitter pathways and many signalling networks are also relevant to most of these conditions. Rather than single out any particular pathway from this extensive dataset (Tables 1 and 2), suffice it to say that parasitic infection has massive effects upon a variety of host signalling networks, metabolic pathways, and processes. These are nevertheless relatively selective, in the sense that certain pathways are more affected than others. In addition, within each disease dataset, the spectrum of pathways within the overlapping datasets is distinct and biologically relevant, as detailed below.

4.2. Enrichment of Interactome Genes within Susceptibility Gene Datasets (Table 3)

T. gondii interactome genes were significantly enriched in the susceptibly gene datasets for all diseases with the exception of anorexia and chronic fatigue and represented from ~13% (autism) to 33% (multiple sclerosis) of the total number of susceptibility genes analysed, with enrichment values from 1.08 to 2.83 fold the expected number (Table 3). For schizophrenia, the fold enrichment (interactome genes in susceptibility gene dataset) of 2.03 compares with a recent meta-analysis of T. gondii seroprevalence studies providing an odds ratio (OR) of 2.71 [81]. A further meta-analysis showed significant associations of schizophrenia with infections by human herpesvirus 2 (OR = 1.34), Borna Disease Virus (OR = 2.03), human endogenous retrovirus W (OR = 19.31), Chlamydophila pneumoniae (OR = 6.34), and Chlamydophila psittaci (OR = 29.05), including values far in excess of those for any gene [82]. For schizophrenia at least, these data and ample evidence from epidemiological and animal behaviour studies [83–85] firmly advocate toxoplasmosis as a significant cause of the disease, in those with a particular genetic constitution. The ability of the parasite to manipulate dopaminergic metabolism (via its own tyrosine hydroxylase) [86] and the involvement of NMDA receptor (e.g., glutamatergic signalling and long-term potentiation), serotonin, or cannabinoid-related signalling networks within the interactome is relevant to the drug-induced psychosis associated with the amphetamines, LSD, cannabis, or phencyclidine (see [87]). Dopamine also increases the number of T. gondii tachyzoites in cultured fibroblasts suggesting that neurotransmitters may also be able to manipulate the parasite [88].

For each disease, and across diseases, the types of susceptibility genes influenced were distinct and relatively selective for each disease. This was assessed in two ways: firstly by statistical analysis of the enrichment of KEGG pathways in each overlapping T. gondii interactome/disease dataset and secondly by a comparison of individual shared and specific overlapping interactome/disease genes across four diseases (the maximum possible using the Venny tool). The diseases analysed in this way were Alzheimer’s disease and multiple sclerosis, bipolar disorder, and schizophrenia.

4.3. Overlapping Interactome/Susceptibility Genes Common and Specific to Four Diseases (Table 4, Figure 1)

The permutations of genes common or specific to the various chosen diseases (Alzheimer’s disease, bipolar disorder, schizophrenia, and multiple sclerosis) are shown by the Venn diagram Figure 1 summarised in Table 4. All of these genes are members of the host/pathogen interactome. Several immune/cytokine and oxidative stress related genes, with different identities, but similar roles, appear as common risk factors across various permutations of diseases, which are all characterised by immune activation [89–91] and oxidative stress [92, 93].

Bipolar disorder and schizophrenia share many common genes, risk factors, endophenotypes, and subpathologies, and interactome genes relevant to certain of these are related to circadian rhythm, dopaminergic and glutamatergic neurotransmission, growth factors, and signalling networks as highlighted in previous reviews [75, 94, 95].

After sifting through these common subsets, the overlapping T. gondii interactome/susceptibility genes specific to each disease are remarkably relevant to the key primary pathologies in each. They include APP processing, cholesterol and lipoprotein function, complement and immune related genes, and oxidative stress, apoptosis and ubiquitin genes in Alzheimer’s disease [96–100]. In bipolar disorder, monoamine/GABA, signalling, adhesion, and ion transport genes are highlighted (see above and [101–103]) while in schizophrenia, monoamine/glutamate/neuregulin neuronal development and associated signalling related genes figure prominently, along with those related to adhesion, oxidative stress, and immune activation (see above). In multiple sclerosis, almost the entire common dataset is related to immune function and associated signalling pathways, that are relevant to the autoimmune aspects of the disease [104, 105], with a limited number of genes related to oxidative stress and apoptosis.

While the evidence for an involvement of toxoplasmosis in psychiatric disorders is relatively strong, there is less work either in the human condition or in animal models in the case of neurological disorders, such as Alzheimer’s or Parkinson’s diseases or multiple sclerosis. Toxoplasmosis has, however, been associated with a loss of grey matter density in schizophrenic patients, but not in controls, suggesting an influence on degenerative components [8]. T. gondii infection may not always be deleterious. For example, it inhibits the development of arthritis in mice deficient in the interleukin receptor antagonist (IL1RN) [106]. T. gondii infection is also able to reduce infarct size in focal cerebral ischaemia in mice, an effect attributed to the ability of infection to increase the expression of nerve growth factor, as well as that of anti-inflammatory cytokines and of glutathione and oxidative stress protective genes, while reducing the expression of proinflammatory cytokines [107].

Parasites have learnt to live with us for many millennia, and their immunosuppressant effects appear to be a relatively common defence mechanism. Indeed, the use of helminths (parasitic worms) has been suggested in a number of autoimmune settings including irritable bowel disease and multiple sclerosis [108]. A clinical trial with helminth egg infection (Trichuris Suis Ova) in autism is also listed at http://clinicaltrials.gov/, based on anecdotal reports of effectiveness in relation to certain symptoms. The preponderance of immune related host/pathogen genes in the multiple sclerosis dataset (and to a lesser extent within other datasets) may be related to these potentially beneficial effects, although the clinical use of T. gondii would be contraindicated by its malevolence directed elsewhere.

4.4. KEGG Pathway Analysis of the Overlapping Datasets Specific to Each Disease (Tables 5 and 6)

4.5. Neurodegenerative Disorders

5.1. Autoimmunity and Host/Pathogen Protein Homology

Several studies have recently shown that the entire human proteome contains short sequences (pentapeptides to heptapeptides or longer gapped consensi) that are identical to those within proteins expressed by numerous viruses, bacteria and other pathogens. For diverse pathogens, these human homologues appear to be concentrated within networks that are relevant to diseases in which the pathogen is implicated [35, 37, 38, 179, 180]. This problem is extensive and concerns all human proteins, along their entire length. For example, there are 18,000 pentapeptide overlaps between the poliovirus and the human proteome [181] while a single immunogenic pentapeptide (VGGVV) within the beta-amyloid peptide is identical to that within proteins from the herpes simplex virus and from 68 other viral species [19]. The extensive host/pathogen interactomes of numerous viruses, bacteria, and parasites no doubt result from this homology which enables pathogen proteins to mimic particular motifs within their human counterparts and to compete for their usual binding partners. Such homology must presumably relate to our evolutionary decent from monocellular organisms and to horizontal gene transfer, a process that applies to all living matter [182]. It is now also appreciated that DNA derived from both DNA and RNA viruses (and not only from retroviruses) has been extensively incorporated into the human genome, and it seems likely that this has also played a role in our evolution, and evidently in the generation of this protein homology [182–185]. Host parasite interactions have also contributed to this gene transfer, and genes from the Chagas disease vector, Rhodnius prolixus, have been found within the genomes of its tetrapod hosts [186]. Peptide homology is more extensive than genetic homology, due to the fact that a number of amino acids can be coded for by several triplet DNA codons (6 for arginine leucine and serine, 4 for alanine, glycine, proline, threonine, and valine, 3 for isoleucine, and 2 for asparagine, aspartate, glutamate, glutamine, cysteine, histidine, lysine, and phenylalanine) (see http://en.wikipedia.org/wiki/DNA_codon). These essentially correspond to single nucleotide polymorphisms that do not modify the translated amino acid. For short peptide sequences, numerous different DNA sequences can thus code for identical peptides.

This extensive homology, and more particularly slightly differing rather than identical peptides (which are more likely to be recognised as nonself) [36, 187], may well also contribute to autoimmunity problems that are evident in many diseases. For example, in Alzheimer’s disease, multiple sclerosis, schizophrenia, and AIDS, antigenic regions of several autoantigens particular to each disease are homologous to proteins expressed by the pathogens implicated in the same disease (including T. gondii and schizophrenia) [18–21, 39].

Diseases currently classified as autoimmune include celiac disease, multiple sclerosis, myasthenia gravis, lupus, rheumatoid arthritis, and inter alia (see Medline Plus article at http://www.nlm.nih.gov/medlineplus/ency/article/000816.htm. However, the autoimmune problem appears to be much more extensive than currently appreciated. For example, using a protein array of 9,486 unique human protein antigens, even control blood samples averaged over 1000 autoantibodies, although with extreme intersample variation. As only ~30% of the human proteome was used in this experiment, we may each eventually accumulate over 3000 autoantibodies, irrespective of any particular disease. However, in both Parkinson’s and Alzheimer’s disease the target profile of the autoantibodies is distinct and can be reliably used as a diagnostic and predictive tool [188, 189]. Autoimmune signatures, with diagnostic predictive value have also been reported in multiple sclerosis [190], breast cancer [191], and nonsmall cell lung cancer [192]. Such data, (in diseases generally not regarded as autoimmune) and the recognition that so many diseases are characterised by immune activation and inflammation suggest that further research in this area would be fruitful in relation to the understanding of the pathologies and eventual treatment of many diseases.

The immune system is trained, in early life, not to recognize the body’s own proteins as self [193]. These bioinformatics data suggest that the multiple autoantibodies seen in man (even in the absence of disease) may stem not from some inherent malfunction of the immune system itself, but from antibodies raised to the numerous pathogens that we randomly encounter during the course of our lifetime. Because of this extensive host/pathogen homology, such antibodies are also likely to target human proteins, and even if the pathogen is eliminated, continued encounter of these human homologues would sustain an autoimmune response. In this way, pathogens might be able to influence disease processes, even when no longer present, perhaps accounting for numerous studies that have failed to find pathogen DNA or protein within diseased tissue, a finding often cited as evidence against pathogen involvement, as recently applied to the controversial implication of the XMRV virus with chronic fatigue or prostate cancer [194–197]. The prospect that autoimmunity is pathogen related suggests that such agents may be able to punch far above their weight and influence biological processes even after their successful removal. This entails a revision of Koch’s postulate as already discussed in a recent review on autoimmunity and the metagenome [177]. This autoimmune scenario might also explain why the antiparasitic agent artemisinin failed to influence psychotic symptoms (as add-on therapy) in schizophrenia [156], as destruction of the parasite needs not to affect the behaviour of antibodies raised to it.

Antibodies to pathogens are clearly cross-reactive with cerebral tissue, although the precise targets remain to be identified. For example 14/25 antibodies to 17 neurotropic pathogens, including Borrelia burgdorferi, T. gondii, and various DNA and RNA viruses were found to bind to western blots of human nervous tissue [198]. It is impossible to verify cross-reactivity solely from sequence analysis, but the ability of pathogen antibodies to react with human proteins could perhaps be tested in bulk using the protein arrays described above. It is now known that antibodies can enter cells, transported by the pathogens to which they bind, [199], and are also able to traverse the blood brain barrier [200]. Antibodies to receptors can also enter cells using the receptor endocytosis apparatus [201]. Antibodies can have devastating pathological consequences. For example, in transgenic mice engineered to express nerve growth factor antibodies only in lymphocytes, the blood brain barrier is soon disrupted, with cerebral antibody entry provoking extensive cortical degeneration, cholinergic neuronal loss, tau hyperphosphorylation, and beta-amyloid deposition (i.e., the cardinal pathology of Alzheimer’s disease) [202]. This phenomenon is applicable to human diseases, including Sydenham’s chorea, believed to be caused by streptococcus induced antibodies which cross-react with basal ganglia antigens [203]. The same streptococcal pathogens (and likely a similar mechanism) have been implicated in paediatric autoimmune neuropsychiatric disorders (PANDA’s) whose diverse symptoms include tics, and dystonias, Tourette syndrome, and obsessive-compulsive disorder [204].

If autoantibodies do indeed play a key role in the pathogenesis of many diseases, then it is likely that their removal may be of benefit. However, given the large number of autoantibodies, some of which may well be beneficial and also required for pathogen defence, this may be no easy task. However, the research so far suggests that the number of autoantibodies specific to a particular disease may be more limited, allowing scope for analysis of their pathological or redemptive properties.

5.2. Population Genetics and a Proposed Gene/Environment Interaction Model (Figure 2)

The mechanisms described above provide a general example of multiple gene/environment interactions in relation to a single pathogen interactome, where several thousand genes (human and protozoan) are involved. Even for a simple population genetics model, with two genes, two risk factors, and a single cause, varying permutations can dramatically influence the eventual outcome. For example, the light and dark coloured genes of the peppered moth, or the light and dark colours of the clean or polluted trees on which they alight, can all be either risk promoting or protective depending on the varying permutations (the light gene “kills” the moth alighting on dark trees but is protective on the lighter trees, etc.) [205]. Neither gene, nor risk factor is relevant if there are no hungry birds or at night time. If one splits a complex disease into its component parts and gives the number of interacting processes involved in the T. gondii interactome, even this single pathogen could act either as a cause, a risk promoter, or as a protective agent, depending upon the pathways that it influences the most. For example, its effects on dopamine could promote psychosis or synuclein polymerisation, and its cholesterol scavenging may have beneficial effects in atherosclerosis, but deleterious effects on myelination, immunosuppressive effects might well protect against autoimmunity, but favour other infections, while the host’s inflammatory reaction or associated fever might contribute to inappropriate collateral damage.

These complex interactions are nevertheless based on a relatively simple concept; that each interaction has an effect on the processes and pathways regulated by the human protein concerned. This suggests a model that may have general application to the many other pathogens and environmental agents implicated in these diseases.

If one imagines the T. gondii proteins as a number of spheres, each with particular affinity for certain human genes or proteins, and their human interactome partners as a further series of spheres perched on a genetic ledge whose characteristics and apertures are regulated by polymorphisms, mutations, deletions, translocations, or copy number variations, then the trajectory of each, dropped through this genetic sieve or knocked off the ledge and falling through the apertures, will be influenced both by the strain of pathogen with different host/pathogen affinities, the dropping point, the timing, and localisation of infection, when and where different human genes are expressed and by the polymorphic genes themselves (for both the host and the pathogen).

Each of these human genes controls a particular element of one or many signalling networks, metabolic pathways, structural elements, developmental processes, and so forth, each represented by reception bins at different positions beneath the sieve. Depending upon varying permutations of these factors, the eventual number of spheres in each bin will vary, resulting in a diverse spectrum of pathway disturbance. Each pathway may be affected either positively or negatively, and the eventual assembly of this pathway mosaic leads to particular endophenotypes or subpathologies, which together combine to assemble into a particular disease. In this way, the same pathogen can produce diverse effects ranging from cause to prevention depending on a permutation of circumstance.

The genes, risk factors, and the immune system thus work together to determine the final outcome, while neither per se are likely to provoke a particular disease. While gene/environment interactions are appreciated in both genetic and epidemiological studies, most, particularly in relation to GWAS, are performed without data partitioning in relation to other variables [206]. Many other pathogens (each no doubt with extensive host pathogen interactomes) and many other risk factors are implicated in these and other diseases, and many are able to influence several relevant aspects of pathology (see Section 1). A clearer understanding of these complex effects could perhaps result in a metamorphosis from multiple genes of small effect in large populations to more restricted numbers of greater effect in particular conditions. It is likely that many disease phenotypes have several “causes,” that subsets of overlapping genes are relevant to each, and that despite the mass of data collection and processing entailed, a dissection of these relationships could eventually lead to disease prevention and cure in multiple conditions.

By their very nature, polygenic diseases are complex, with several underlying pathologies and endophenotypes, hundreds of interacting genes, and dozens of environmental risk factors. The failure to replicate either genetic or epidemiological data is a situation peculiar to these diseases, not seen in many other fields. However, the effects of genes and risk factors are clearly conditional and, as illustrated above, may well depend upon each other. Replication inconsistency may well be part of the answer and not part of the problem.