Glial cells play important roles in the developing 11 as well as in the adult central nervous system (CNS). In the adult CNS, glia has a supportive, a protective, a regenerative, and an active regulatory role. Glia cells are sensors of infection and produce cytokines to limit viral replication. In adults, they induce neurogenesis in the hippocampus and the subventricular zone 22, influence neuronal activity and synaptic strength 23, and appear to be the third partner in synaptic transmission (tripartite synapse) 24. Synaptic strength and cellular growth depend on the synaptic and the general protein-synthesis rate 2526 which is influenced by growth factors such as neurotrophins and neuregulins 2728. Glial cells are part of a positive feedback loop between presynaptic neurons and their postsynaptic targets 29 involving neurotrophins and neuregulins (NRGs).

NRGs are synthesized by neurons 11 and promote the differentiation, survival and repair of the neuronal targets such as glial cells 11, acetylcholine receptors 21, and postsynaptic densities (PSD) in hippocampal neurons 30. Neuregulin-1 (NRG1) is concentrated at synaptic sites suggesting a role in synapse-specific gene expression 28. Furthermore, NRGs influence the growth of neural precursor cells, the radial migration of newborn neurons during neocortex genesis, the rate of migration in a dose-dependent manner 31, the interaction between pre- and postsynaptic neurons during synaptogenesis including neuromuscular synapse, activity-dependent maintenance of synaptic connections, synaptic plasticity, long-term potentiation, and the expression of ligand and voltage-gated channels in central neurons 9112021323334. NRGs are also known as glial growth factor (GGF), Neu differentiation factor (NDF), heregulin, sensory and motor neuron derived factor (SMDF), and acetylcholine receptor inducing activity (ARIA) 21.

Biochemically, NRGs are structurally related to what is perhaps the best studied trophic factor – epidermal growth factor (EGF) 11 and encode a large group of polypeptide growth, survival and differentiation factors 2021 that all contain an extracellular epidermal growth factor (EGF)-like domain, which is essential for their bioactivity 35. They are derived by alternative splicing from four genes: NRG1, NRG2 also known as Don-1 and NTAK, NRG3, and NRG4. NRG1-3 is expressed during neurodevelopment and in the adult CNS 21, whereas NRG4 transcripts have not been detected in neural tissue 21. The cytoplasmic tail of NRG interacts with the protein kinase LIM kinase 1 (LIMK1) 36.

NRGs act through a network of ErbB tyrosine kinase receptors consisting of ErbB1 (also termed EGF receptor or HER1), ErbB2 (Neu/HER2), ErbB3 (HER3) and ErbB4 (HER4) 11202137. NRGs bind to ErbB2-4, EGF and transforming-growth factor alpha (TGFA) to ErbB1 2137. ErbB receptors serve as docking sites for cytoplasmic signaling molecules such as Grb2-related adaptor protein 2 and phosphatidylinositol-3 kinase (PI3K) 2137. In addition, the ErbB receptor system integrates signaling events from other receptor classes, such as G-protein-coupled receptors and cytokine receptors 37.

NRGs induce the expression of growth factors, cholinergic, GABAergic and glutaminergic receptors (e.g., transforming growth factor beta 21, N-methyl D-aspartate (NMDA) receptor 2C subunit 21, gamma-amino butyric acid (GABA) receptor beta-2 subunit 21, acetylcholine receptor (AchR) alpha-5, alpha-7 38, beta-4, delta and epsilon subunits 2139. The promoter region of the nicotinic AchR epsilon subunit gene contains a NRG-responsive element 39. Furthermore, the NRG-dependent regulation of AChRs at the neuromuscular junction (NMJ) requires the serine/threonine cyclin-dependent kinase 5 regulatory subunit 1 (p35) 21.

NRGs appear to act synergistically with other growth factors such as insulin-like growth factor-I (IGF1) 4041, EGF 42, insulin 43, and growth arrest specific genes (GAS6) 444546. In turn, neurotrophic growth factors (e.g., BDNF and NT-3), glutamate, and neural activity increase the expression of NRG 3447 and of neuritin in hippocampal and cortical neurons, regions well characterized in regard to synaptic plasticity 48.

With regard to the latter, ErbB receptors are enriched in postsynaptic densities (PSD) and interact with other PSD proteins such beta-2 syntrophin, NMDA receptor subunit 2C and 2B, Ca2+-activated potassium channels 214950, protein kinase C interacting protein (PICK1) and glutamate (AMPA subtype) receptors 21.

The synaptic connections are maintenanted by a positive feedback loop between the presynaptic neuron and its targets which includes glial cells. The latter provide neurotrophins, e.g., GDNF, BDNF, NT-3, NT-4, hepatocyte growth factor 1134 that enhance neuronal survival, differentiation, plasticity, pruning, synaptic strength and stabilization, synaptic transmission, presynaptically by increased secretion of neurotransmitters and postsynaptically via NMDA receptors 9115152. The synthesis of neurotrophins is down regulated by inhibitory synaptic activity (e.g., GABA), up regulated by neuregulin, activation of metabotropic glutamate group I receptors (mGluRI), acetylcholine, physical exercise, physical and social stress, cytokines, and neurotrophin itself 952535455. Neurotrophins are in short supply in the CNS 9. Neurons and axons that do not receive adequate amounts of neurotrophic factors risk degeneration or synapse elimination 9. Neurotrophins are critical for long-term potentiation (LTP) in the hippocampus and activity-dependent synaptic plasticity (SP), which are thought to be cellular models for learning and memory 52.

GGFs such as NRG1, NGF, and EGF induce protein-synthesis 5657, which is required for cellular growth 25, for activity-dependent maintenance of synaptic connections 58, synaptic strength 26, and long-term memory 59.

A search for convergent loci revealed 41 genes (see Fig. 1) with related functions localized within significant or potential linkage regions of schizophrenia (chromosomal localization in parenthesis): neuregulin-1 (NRG1) (8), neuregulin-2 (NRG2) (5), neuregulin-3 (NRG3) (10), epidermal growth factor (EGF) (4), and neuregulin receptor ErbB3 (12), transforming-growth factor alpha (TGFA) (2), transforming-growth factor beta receptor II (TGFBR2) (3), LIMK1 (7), Grb2-related adaptor protein 2 (22), phosphoinositide-3 kinase class 2 beta polypeptide (PIK3C2B) (1), GABA receptor beta 2 (GABRB2) (5), acetylcholine receptor alpha 7 (CHRA7) (15) and epsilon (CHRNE) (17), growth factors such as glial cell line derived neurotrophic factor (GDNF) (5), GDNF receptor alpha 1 (GFRA1) (10), GFRA2 (8), GFRA4 (20), GDNF family related persephin (19), glutamate and its receptors (GRM1, metabotropic glutamate receptor 1) (6), GRM4 (6) GRM5 (11) and GRM6 (5), insulin-like growth factor I (IGF1) (12), insulin receptor (INSR) (19), insulin receptor substrate 1 (IRS1) (2), GAS6 (13) and GAS5 (1), neuritin (6), syntrophin B2 (16), protein kinase C interacting protein (PICK1) (22), NMDA receptor 2B (12) and 2 C (17), p35 (CDK5R1) (17). Furthermore, epiregulin (4) is a potent pan-ErbB ligand 60. Ephrins such as ephrin-B2 (13) and ephrin-A5 (5) are involved in neurodevelopment, and in the adult CNS in LTP, synaptic strength 61, and cell proliferation of the adult subventricular zone 62. Fibroblast growth factors such as FGF2 (4) and indirectly the FGF binding protein 1 (FGFBP1) (4) have a similar effect on development, adult neurogenesis, cell survival and synaptic transmission 6364. Cytokines are neurotoxic or neurotrophic 65. For example, interleukin-8 (IL8) (4) exerts neurotrophic effects on glial cells 65 and the latter constitutively release tumor necrosis factor alpha (TNF) (6), another cytokine, which markedly influences synaptic strength 66 and leads to impaired insulin signaling via ErbB2 and ErbB3 67. Moreover, interleukin-1 receptor accessory protein-like 1 (X), a member of the IL1 receptor family, is highly expressed in the adult hippocampus suggesting a role in learning, memory, and synaptic strength 68. Neurogranin (11) is concentrated in the post-synaptic terminals of the hippocampus and involved in learning, LTP, and synaptic plasticity 69.

The convergent loci of the genes described above are displayed in Fig. 1 along with genetic markers showing some evidence for linkage in schizophrenia. The inclusion of significant as well as non-significant linkage markers may well be the object of criticism if the intention were to prove, rather than to generate in the context of discovery, a hypothesis which has to be tested by other research programs. The convergences shown in Fig. 1 led us to propose the following hypothesis for schizophrenia.

A scientifically useful hypothesis should explain the main facts and be, at least, consistent with the rest of the facts 72. The GGF/SD hypothesis is able to explain major findings of schizophrenia research such as (1) genetic linkage, (2) neurodevelopmental disturbances, deviant neuronal migration, (3) prenatal timing, (4) expression studies, (5) multiorgan involvement, (6) growth deviations, (7) seasonality of birth, (8) age of onset, (9) neurodegeneration, (9) regeneration and neural remodelling, (11) memory disturbances, and (12) the maintenance of the disorder in the population despite the reduced fertility of the patients.

(1) Genetic linkage: The CHRNA7 gene on chromosome 15q shows evidence for linkage to schizophrenia 7374 and appears to be downstream of NRG because NRG leads to an increase of alpha7 nicotinic acetylcholine receptors (CHRNA7) and is highly expressed in cholinergic neurons of the CNS 38. The NRG1 gene itself is located within a linkage region of schizophrenia on 8p. Further convergences seem to exist (see Fig. 1). Moreover, reported associations of schizophrenia with 5-HT receptor 2A 75, 5-HT5A 76, NT-3 77, metabotropic glutamate receptor subtype 5 78, NOTCH4 79, possibly NRG1 80, potassium channel gene hKCa3 81, proline dehydrogenase (oxidase) PRODH2 8283, and regulator of G-protein signaling 4 (RGS4) 884 are all in agreement with the hypothesis. NRG1 is a glial growth factor which interacts with Ca2+-activated potassium channels 214950. Notch plays a role in gliogenesis 85. NT-3 is produced by glial cells 1134. 5-HT increases the release of glial glutamate 86. Proline oxidase appears to catalyze the first step of an alternative route for glutamate production in glial cells of the hippocampus 87. The receptors for EGF and neuregulins, the ErbB receptor system, integrates also the signal transduction from G-protein-coupled receptors 37, which is inhibited by RGS4 88. However, some of the findings cited in this paragraph and shown in Fig. 1 are very probably false positives.

(2) Neurodevelopment: evidence for glial cell abnormalities, neurodevelopmental disturbances, neuronal migration, and cognitive disturbances has been found in schizophrenia 61213. Glial cells play an important role in neurodevelopment and neuronal migration 1189, and the highest ratio of glia-to-neurons is found in the neocortex suggesting a key role for glial cells in higher cognitive functions 90.

(3) Prenatal timing: NRG is important for midgestation 2021, a prenatal period with evidence for neurodevelopmental disturbances in schizophrenia 6.

(4) Expression studies in postmortem brains are also in line with the postulated functional deficiency of GGFs. Since GGFs are involved in growth, they should influence the expression of many genes. Indeed, a large number of decreased mRNAs or proteins has been reported in schizophrenia (for review 7, e.g., synapsin, synaptophysin, glutamate receptors, somatostatin, glutamic acid decarboxylase, protein kinase C, adenylate cyclase, regulator of G-protein signaling 4 (RGS4) 19, MAP kinase phosphatase MKP2, postsynaptic density protein 95 growth associated protein-43, and alpha7-nicotinic receptor 91). A functional GGFs deficiency could also explain the results of a gene expression analysis suggesting a glial cell deficiency in schizophrenia 13.

(5) Multiorgan involvement has been observed in schizophrenia, e.g., brain, blood vessels (nailfold plexus), lung (cancer, TBC), immune system (lymphocytes, rheumatoid arthritis), and skin (dermatoglyphics) (for review 7). These findings are in line with the expression of NRG in neurons, glia, heart, liver, stomach, lung, kidney, immune system, and skin 1120, of GAS6 in an equal number of tissues including rheumatoid artheritis 92, and finally with the role of the ErbB receptors in growth regulation in a wide variety of cell types 20.

(6) Growth deviations have been observed in schizophrenia 5678. NRGs are growth factors and act synergistically on glial cells with the insulin-like growth factor I (IGF1) 93, a major determinant of fetal growth 94 as well as with GAS6, which is involved in bone formation and glial growth 4495. In addition, the reported negative association between schizophrenia and cancer 9697 agrees with the role of ErbB receptors in the genesis of carcinomas 20.

(7) Seasonality of birth: The increase of IGF1 in the summer 94 might account for the seasonality of birth in schizophrenia 98. IGF2 is imprinted in humans 99100. Imprinted genes often help to regulate the growth of the fetus 101. Growth regulatory signals lead to histone acetylation and hence epigenetic variation of gene expression 102.

(8) Age of onset: Brain growth might be used as an indirect indicator of the activity of GGFs. Fig. 3 shows that a decrease in head and brain growth is accompanied by a sharp increase in the risk for developing schizophrenia.

(9) Neurodegeneration: Disruption of the NRG1 gene causes neuronal degeneration after initial synapse formation, especially in motor neurons, 33103. Degeneration of the second motor neuron does also take place in schizophrenia causing atrophy of skeletal muscle fibres, a mechanism elucidated by Meltzer in the search for the cause of increased serum CK, a muscle-derived enzyme, in acute psychosis 104105. Furthermore, a GGFs deficiency would explain the increased levels of S100B, a marker of glial cell integrity, in acute schizophrenia 1415 and the evidence for a minor neurodegenerative process in chronic schizophrenic patients revealed by longitudinal MRI studies 106.

(10) Regeneration: NRG is important for neuronal regeneration following environmental insults 21. A genetically or epigenetically determined deficiency of GGFs is compatible with the limited regeneration of motor neurons observed in schizophrenia 104105 and with the incomplete remission of the disorder, which is the foundation of Kraepelin's schizophrenia (dementia praecox) concept.

(11) Memory disturbances: In the adult brain, NRGs and neurotrophins appear to be involved in activity-dependent synaptic plasticity 304851107, a mechanism relevant to protein-synthesis dependent long-term memory 59. They are expressed in brain regions well characterized in regard to synaptic plasticity (hippocampus and cortex). Memory impairment has been found in schizophrenia and appears to be responsible for the characteristic symptoms of the disorder 108.

(12) The maintenance of schizophrenia in the population despite the reduced fertility of the patients 109 is a partially unsolved puzzle 110 which might be explained by the postulated growth deficiency. Growth-dependent birth weight is the classical example for stabilizing selection in humans 111, a selection against both extremes.

We conclude that the postulated GGFs deficiency and synaptic destabilization in schizophrenia fulfils the first requisite for a scientific hypothesis, to explain the main facts. However, is the hypothesis also consistent with the rest of the facts? For such a test, an extensive list of 84 major findings of schizophrenia research has been compiled by one of us (HWM) from the literature using books and review articles for literature from 1900 to 1975, PubMed's and PsycLit's electronic databases for publications from 1966 to 2001 and from 1974 to 2001, respectively 7. A major finding was defined operationally as results obtained by at least two independent groups of researchers. The GGF/SD hypothesis appears to be consistent with the major findings in schizophrenia.

The second fundamental condition for a scientific hypothesis is to be susceptible to verification and refutation and to aid in the prediction of new facts and relationships 72.

The GGF/SD hypothesis can be verified or falsified by searching for concomitant variations between the postulated independent and the dependent variables, that is between concentration or signal transduction of GGFs such as NRG, IGF1, GAS6, TNF-alpha, neuritin etc. and schizophrenia or schizophrenia-like symptoms. The independent variables can be measured in animals or humans, and the dependent variable in animal models of the disease endophenotypes or in acute schizophrenic psychosis, preferably with elevated serum CK or S100B. The different parameters of GGFs can be investigated at different levels and by different methods, e.g., by studying mRNA, proteins or signal transduction and by using knock-out animals, antisense mRNA, cDNA microarrays, antibodies or signal transduction assays in immune cells, fibroblasts or muscle cells from patients. A confounding variable might be the large number of alternatively spliced NRGs.

Furthermore, deductive reasoning from the hypothesis leads to certain logical consequences that can be tested. The hypothesis predictions include, among others, (1) a reduction of neurotrophins and synaptic strength induced by virus infections of glial cells, (2) the involvement of risk genes for schizophrenia in the pathway of GGFs and/or synaptic strength (SS), (3) differences in the structure and function of organs influenced by NRG, (4) a decreased synaptic protein-synthesis rate, decreased SS, and synaptic destabilization as a common final pathway, and (5) synaptic stabilization via improvement of astroglia function as an antipsychotic drug mechanism.

(1) The predicted reduction of neurotrophins/SS by acute, latent, or reactivated virus infections can be investigated in animal models or glial cell cultures. Glial cells display genetically determined individual differences to virus infections 112 and are infected by a large number of viruses (for review 113114). Especially interesting in this regard is the human herpes virus-6 (HHV-6) and the JC virus (JCV). The HHV-6 infects worldwide, within the first two years of life, nearly 100% of the human population, infects predominantly glial cells with a low-level of viral production 115, persists latently lifelong, and has been detected in 13% – 74% of normal human brains 116117118. The JCV infects more than 70% of the world's population during early childhood 119, infects mainly glial cells (oligodendrocytes as well as astrocytes) 114120, remains in the latent state without apparent clinical symptoms but shows in vitro the ability to deregulate the cellular function of oligodendrocytes and perhaps astrocytes 121122. The JC virus is spread by urban sewage 119123 which might contribute to the urban factor observed in schizophrenia (for review 8).

(2) Another prediction is that some of the GGF- and SS-related genes increase the risk for schizophrenia. Such genes (e.g., Fig. 1) can be tested for allelic association (linkage disequilibrium, LD) with schizophrenia. Since LD studies in complex disorders have often identified regulatory regions influencing the disease 124, and since different parts of the same gene are not necessarily in LD with each other (e.g., 125), candidate genes for schizophrenia and their regulatory regions will have to be investigated by haplotypes consisting of densely spaced SNPs. Under optimal circumstances, sample sizes of more than 500 affecteds appear to be required to detect LD in complex disorders such as schizophrenia 126127. Optimal circumstances are present, when a single disease-causing mutation accounts for a large proportion of the phenotypic variance and has arisen recently on a relatively uncommon haplotype background 128. Such favorable circumstances rarely exist in schizophrenia. Locus and allelic heterogeneity are common in complex disorders, produce dramatic decreases in power 128, and an increase in sample size (e.g., 129 or by meta-analysis 130) further increases heterogeneity 124. Therefore, the frequent elusiveness of positional cloning results in complex disorders is hardly surprising 124131. Given the inefficiency of LD studies 124132 and the generally low repeatability of positive findings in complex disorders 127, negative results of LD studies are, unfortunately, unable to disprove anything. To falsify the hypothesis, it might be necessary to investigate protein concentrations and to perform functional assays of cells derived from schizophrenic patients and their relatives.

(3) The multiorgan expression of GGFs such as NRG, GAS6 and their receptors predicts differences in the structure and function of relevant organs such as liver, skin, heart, lung, kidney, bones, muscle, sensory and motor neurons, and acetycholine receptor density at the NMJ. For example, a decrease of muscle mass, of muscular-growth rate, of bone structure, and an increase of the risk for liver cirrhosis are predicted for schizophrenic patients, their family members and for introversion, an associated personality trait 7.

(4) A decreased rate of glial, neuronal or synaptic protein synthesis is predicted by the GGF/SD hypothesis based on the facts that NRGs, GAS6, and other GGFs are growth factors, that NRGs result in an increased synthesis of growth factors, that growth is associated with protein synthesis, and that NRG1 stimulates the protein-synthesis rate 756.

(5) Finally, the GGF/SD hypothesis predicts an antipsychotic effect by reducing the synaptic destabilization via the improvement of astroglial function. The prediction can be tested by stimulation of the synaptic protein synthesis rate, e.g., via glial growth factors such as NRGs, neurotrophins, EGF, IGF1, insulin, prolactin, activation of ErbB receptors or PI3K. Hyperprolactinemia via dopamine receptor blockade is a well known "side-effect" of traditional neuroleptics 133. Dopamine receptor blockade and/or elevation of prolactin levels have also been reported for atypical antipsychotics such as clozapine 133, olanzapine 133, risperdone 134, amisulpride 135136, ziprasidone 137, and zotepine 138. Furthermore, clozapine and olanzapine increase the level of insulin 139140141. This suggests that an increase of prolactin and/or insulin might be the working mechanism of neuroleptics. Hence, an augmentation of antipsychotic efficacy is predicted by combining neuroleptics with a predominant insulin profile with others showing a marked hyperprolactinemia. The latter prediction can be easily tested, e.g., by combining clozapine or olanzapine with amisulpride in the treatment of acute schizophrenic psychosis.