NOS1AP in Schizophrenia

Abstract

NOS1AP is an attractive candidate gene for schizophrenia susceptibility. Linkage and association studies from multiple samples drawn from different populations indicate that a schizophrenia susceptibility gene is located in the region of chromosome 1 containing NOS1AP. Increased expression of NOS1AP is observed in post-mortem samples from individuals with schizophrenia. NOS1AP binds to neuronal nitric oxide synthase and synapsin, other candidate genes for schizophrenia, and may disrupt signal transduction through the NMDA receptor complex, leading to hypofunctioning of that system. In this review I will present the evidence supporting NOS1AP as a schizophrenia susceptibility gene, with a focus on explaining the strengths and weaknesses of the evidence obtained from each type of study that has been conducted.

Introduction

Schizophrenia is a serious neuropsychiatric illness estimated to affect approximately 1% of the general population. Family, twin and adoption studies have demonstrated that schizophrenia is predominantly a genetic disorder, with a high heritability (reviewed in [1]). Multiple genetic and non-genetic factors are likely to be involved [2]. The goal of modern molecular genetic studies is to implicate specific genes in the etiology of schizophrenia, preferably with identification of specific DNA changes that alter risk. The current belief is that except in very rare cases these genetic changes do not, in isolation, cause illness. Rather they produce some biological change that increases risk. Such changes may be observed in certain individuals who will not become affected. Illness develops only when sufficient biological and environmental risk factors are combined in a given individual to surpass some threshold. The hope is that through a better understanding of the genetic mechanisms underlying risk we may develop insights in new treatment approaches or better methods to make a certain diagnosis earlier in the course of illness.

NOS1AP (nitric oxide synthase 1 [neuronal] adaptor protein) is an attractive candidate gene for schizophrenia susceptibility. NOS1AP (previously known as CAPON) was first identified in the rat as the C-terminal PDZ domain ligand of neuronal nitric oxide synthase (nNOS), an nNOS binding protein capable of disrupting the association of nNOS with the postsynaptic density scaffolding proteins PSD93 and PSD95 [3]. The interaction between nNOS and PSD93 and PSD95 is important in targeting nNOS to the postsynaptic N-methyl-D-aspartate receptor (NMDAR) complex, and facilitates the tight coupling between activation of the NMDAR and nNOS, producing NMDAR-mediated NO release into the synaptic structures [4, 5]. This places NOS1AP at the scene of NMDAR glutamate neurotransmission, long proposed to be involved in schizophrenia (reviewed in [6]). NOS1AP can also serve as an nNOS adaptor protein, with the N-terminus either binding to a direct target of NO-mediated activation by S-nitrosylation [7] or to synapsin [8], resulting in the localization of nNOS to the presynaptic terminals.

At least two variants, or isoforms, of NOS1AP exist in humans. The first is a 501 amino acid protein made from all 10 exons of the gene. This isoform contains two known functional domains, an N-terminal phosphotyrosine-binding (PTB) domain and a C-terminal PDZ-binding domain [9]. The second isoform is made from the last two exons of NOS1AP and produce a short protein of 210 amino acids that contains the PDZ-binding domain. Prior work has demonstrated that the terminal 125 amino acids of the full-length protein are sufficient to bind the PDZ-domain of nNOS and interfere with the binding between nNOS and PSD93 or PSD95 [3]. The terminal 125 amino acids also appear to be able to directly bind to the second PDZ domain of PSD95 [10], blocking this normal site of nNOS binding [4]. As the first 180 amino acids of NOS1AP have been previously demonstrated to contain the domain needed to bind to the N-terminal targets Dexras1 and Synapsin [7, 8], it would seem that only the full-length form of NOS1AP would be able to serve as an adaptor protein between nNOS and these targets. A physiological role of the short-form would likely be limited to the competitive inhibition of binding of other ligands to the PDZ domains of nNOS and PSD93/PSD95.

Recent functional studies on NOS1AP suggest that increased expression of the full length, but not the short, isoform can significantly reduce the number and branching of dendrites of cultured hippocampal neurons [11]. Interestingly, neither the PTB nor PDZ-binding domains seem to mediate this function, with the middle portion of the protein (amino acids 181–307) responsible for this effect. Decreased dendritic field size has been observed in schizophrenia [12], raising this as an additional potential mechanism of action of NOS1AP for increasing schizophrenia susceptibility.

Based on known functions, NOS1AP is an intriguing candidate for a schizophrenia susceptibility gene. It directly binds to nNOS and Synapsin, other candidate genes for schizophrenia. There is evidence that NOS1AP can disrupt signal transduction through the NMDAR complex, leading to hypofunctioning of that system, and lead to altered dendrite branching during neuronal development. Multiple theories about potential mechanisms of action could be proposed. But what is the actual evidence implicating NOS1AP as a genetic risk factor for schizophrenia?

Evidence from Linkage Studies

Genetic linkage studies are a method for localizing genetic susceptibility to particular regions of chromosomes (reviewed in [13]). In a linkage study, the inheritance of the disorder within families is compared with the inheritance of sites of naturally occurring genetic variation, or genetic markers. A consistent pattern of co-inheritance of particular genetic markers and the disorder across families can quickly localize the responsible gene to a relatively small region (10–20 million bases, or Mb, of DNA) of the genome. While 10 Mb is still large enough to contain, on average, about 75 genes, it represents a very small percentage (0.3%) of our total genome [14].

While different genetic linkage studies often give consistent results (i.e. implicating the same region of the genome) for disorders caused by a single gene, the situation is typically not so straightforward for disorders that are caused by multiple genes and environmental influences. Since different individuals may have the disease on the basis of different combinations of genetic and environmental factors, etiologic heterogeneity becomes a significant issue. Different studies usually collect subjects from different populations using different methods, leading to possibility that these samples may actually be genetically different. For disorders caused by multiple genes, it has been demonstrated that it is much easier to identify an initial significant linkage in the area of any one susceptibility gene than it is to subsequently detect significant linkage to the same area again [15]. Even with a modest number of genes (six) and when subsequent samples are selected from the same population as the original study, it is predicted that samples on the order of five to six times larger then the original sample will be needed to detect significant linkage to the same location again.

Unfortunately, simply using large samples is not an approach that guarantees a significant linkage result. A series of recently published studies that all used very large numbers of families (>400) for linkage analysis of schizophrenia or affective disorders failed to produce any statistically significant results [16–19]. To collect such larger samples in a timely fashion, studies typically use multiple recruitment sites. Despite the best efforts of the investigators, unknown differences may exist between samples recruited at different locations. Studies have demonstrated that pooling datasets with unappreciated heterogeneity can result in a loss of power to detect linkage under a variety of analysis methods [20, 21]. Simply increasing the total number of families in a dataset may not lead to an increase in the overall linkage signal if the overall proportion of families exhibiting linkage decreases. The current trend of failure to detect significant linkage in this recent series of very large studies may simply be the observation in real data sets of this predicted result (see [22] for further discussion of these issues).

So what is the linkage evidence supporting a role for NOS1AP in schizophrenia? We have published a highly significant (heterogeneity lod score of 6.5; p < 0.0002) linkage finding of schizophrenia to chromosome 1q22 in a modest sample of 22 medium-sized Canadian families, selected for study because multiple relatives were clinically diagnosed with schizophrenia [23]. This initial study suggested that the chromosome 1 susceptibility gene was located within a 6 Mb region of DNA. A subsequent fine-map linkage study of these same families further restricted this to an approximately 3 Mb interval [24], now known to contain approximately 50 protein coding genes. Several other independent studies have reported linkage of schizophrenia to this region [25–27], while many additional studies, including two recent very large studies, have not [16, 18]. These results are expected, given our understanding of the chances of replicating a linkage finding of a complex genetic disorders in new, even large, samples [15, 20, 21]. Of note is that a meta-analysis of 20 genome-wide studies of linkage concluded that there was significant evidence to support a schizophrenia susceptibility gene in this region of chromosome 1 [28]. However, as noted above, linkage studies typically implicate a region of DNA large enough to contain many genes. Additional analyses are needed to further narrow the list of potential candidates.

Evidence from Association Studies

Genetic association studies are a method for localizing genetic susceptibility to smaller regions of chromosomes (reviewed in [13]). In association studies, the frequency of specific DNA sequences at given locations is compared in affected and unaffected individuals. These individuals can be affected and unaffected members of families or unrelated cases and controls. Association studies are based on the premise that a DNA change increasing susceptibility occurred in some ancestor long ago enough that the change has had time to be widely disseminated among affected individuals within the population. As neighboring DNA sequences will tend to be inherited along with the actual susceptibility change, affected individuals will usually tend to have similar DNA sequences in a region surrounding the functional change. The physical extent of this region of similarity will depend on a number of factors, such as the number of generations ago that the change arose and the propensity of the DNA in the region to undergo recombination during meiosis. In general populations the region of DNA with increased sharing will typically be small enough to implicate only a single gene, often with sequences from only a portion of the gene exhibiting association to the disease. In isolated populations where the susceptibility sequence was introduced relatively few generations ago, association often extends for much greater physical distances, encompassing multiple genes.

Like linkage, the ability to detect association to a given gene in different samples may be limited by genuine genetic differences in the samples. The same difficulties in replicating real linkage findings apply to real association findings [29, 30]. But association studies are subject to additional obstacles. If, in the history of a given population, many different susceptibility changes have occurred in a given gene, then it will be very difficult (if not impossible) to detect association to that gene. Even if the same susceptibility change has independently arisen several times, it may be difficult to detect association unless the exact site of the susceptibility change is tested, as the associated flanking DNA will likely be different for each occurrence of the change. It is important to note that in these situations we may still be able to detect linkage to this region, as linkage only examines the co-inheritance of sequences within families and not within populations.

There is a final important difference between linkage and association that must be considered. Patterns of linkage in the human genome are generally predictable. If two DNA sequences exhibit linkage to each other, then a DNA sequence in between these two will exhibit linkage to both of the flanking sequences. The same is not true of association; two DNA sequences exhibiting allelic association with each other may fail to show any detectable association to a sequence located physically between them [31]. The recognition of the complexity of association in the human genome and the need to understand it was a major rationale for the International HapMap Project [32]. This has resulted in a map of the association relationships between 3.1 million single nucleotide polymorphisms (SNPs) [33] and statistical algorithms to aid in the efficient selection of sets of SNPs (tag SNPs) to test for association to clinical definitions of illness [34]. It should be noted that the number of SNPs required to carefully examine a given gene for association may be large; on the order of 50 to 100 tag SNPs are predicted to be needed to fully evaluate NOS1AP for association.

So what is the association evidence supporting a role for NOS1AP in schizophrenia? The first association study in this region investigated six short sequence repeat markers from 1p21.1-q23.3 in 80 Spanish nuclear families and reported significant association to schizophrenia only for D1S1679 [35], which is located within 25 kb of NOS1AP. Our initial study of 15 SNPs in the Canadian families exhibiting linkage to this area identified three SNPs within NOS1AP that were significantly associated with schizophrenia [36]. These study sparked several attempts to extend this association to other samples. Association to schizophrenia has been demonstrated in the Chinese Han population, in a study using eight SNPs within the gene and 664 cases and 941 controls [37], although association was detected to different specific SNPs in the Chinese then in the Canadian sample. Association has also been evaluated in a sample of 110 trios (individuals with schizophrenia and their parents) from Antioquia, Colombia. An initial association study using nine short sequence repeat markers from four different chromosomal regions suspected of harboring schizophrenia susceptibility genes produced significant association only to D1S1679 [38]. Further genotyping of 24 SNPs from within NOS1AP detected significant association to eight SNPs, including two SNPs that were associated in the Canadian sample (B. Kremeyer and A Ruiz-Linares, personal communication). To date only one study, testing eight SNPs in 450 British cases and 450 controls, has failed to detect evidence of association to NOS1AP [39], but given the large number of SNPs required to completely evaluate a gene the size of NOS1AP (50 to 100), this study was not comprehensive enough to rule out association of NOS1AP in this sample.

Finally, it is important to note that two other genes in the area of the 1q22 linkage peak have attracted attention as possible schizophrenia susceptibility genes. UHMK1, encoding a kinase interacting stathmin, is located 129 kb from NOS1AP and has been implicated in a single association study [40]. RGS4, encoding a regulator of G protein signaling, is 700 kb from NOS1AP and has been the subject of multiple association and functional studies (reviewed in [41]). As discussed for linkage studies, association studies implicate a region of DNA in disease susceptibility. Association can extend over genomic segments large enough to contain multiple genes, and regulatory regions of DNA have been documented to act at distances of up to a Mb, with the possibility of regulatory regions controlling expression of more then one gene (reviewed in [42]). These observations raise the possibility that disease-associated sequences in one gene could actually be altering expression of a different gene, or that such sequences might alter the expression of multiple genes in the region, with two or more genes potentially contributing to disease susceptibility. Therefore, localization of an association signal to within a given gene cannot be taken as proof of the involvement of that gene in a given disease. Further evidence implicating the gene from functional studies is necessary.

Evidence from Functional Studies

Functional studies are a method for gathering evidence that gene function is altered in an illness. These studies may take many specific forms, but two general approaches are common. In the first, the expression of the gene of interest is examined for any group differences between affected and unaffected individuals. Levels of mRNA or protein, structural changes in primary RNA or protein sequences, or changes in protein modifications may all be assessed. One challenge of this approach is obtaining suitable samples for these studies. Easily available material, such as lymphocytes from peripheral blood, may not recapitulate expression changes that are present in the brain or may not even express the gene of interest at all. Brain tissue, while the organ of interest, must be collected from post-mortem sources that may be limited in quality and quantity. A second issue with these studies is that they do not reveal causality; expression of a gene may be altered as a secondary reaction to some other abnormality of the illness or a reaction to treatment. Still, when taken in combination with positional information from linkage and association studies these types of functional studies can be very useful in furthering the evidence implicating a given candidate gene.

The second type of functional study seeks to correlate changes in function to specific alterations in DNA sequence. Assessments of function for these studies can take a wide variety of forms. As in the first type of functional study, correlations can be sought between altered expression in brain or other tissues and DNA sequence. The issue of causality is not problematic here, as each individual’s DNA sequence was determined prior to the onset of illness. Other studies may seek to correlate structural or functional imaging data or performance on neurocognitive tasks with DNA sequence. Changes in DNA sequence can also be directly assessed for impact on gene expression through in vitro studies. These experiments eliminate the variation that is inherent in studies conducted in human subjects. Here, clonal cells are used as the experimental system. The DNA of the basic regulatory element of the gene of interest, the promoter, is systematically combined with different DNA sequences thought to potentially alter gene expression. These different sequence combinations are individually placed into the clonal cells, and the resultant expression of a reporter gene (a gene that makes a product that is easily quantified) is measured to see which sequences do alter expression. These studies provide a highly controlled system where the effects on gene expression of even single base changes can be accurately assessed.

So what is the functional study evidence supporting a role for NOS1AP in schizophrenia? All the types of studies described above have been conducted on NOS1AP, producing consistent evidence for involvement in schizophrenia. We have reported significantly increased expression of NOS1AP in post-mortem samples from the dorsolateral prefrontal cortex of individuals with schizophrenia compared to psychiatrically normal controls [43]. This study was conducted using 105 samples from the Stanley Array Collection, which consists of samples collected from several locations within the United States, and so clearly represents an independent sample from our Canadian familial schizophrenia collection. While this result alone does not establish the causality between altered NOS1AP expression and schizophrenia, it is important corroborating functional evidence. (Of note is that RGS4, the other main positional candidate in this region, also exhibits altered expression in schizophrenia in post mortem studies, reviewed in [41]).

If increased NOS1AP expression is the mechanism by which DNA changes at this location increase risk for schizophrenia, then increased gene expression should be observed in individuals with these DNA changes, regardless of diagnostic status. Only the subset of individuals with increased NOS1AP and additional genetic and environmental risk factors would be expected to go on to develop illness. We therefore examined the NOS1AP expression data from our post-mortem study without regard to diagnostic category, searching for correlations to DNA sequence at the three SNPs previously reported associated to schizophrenia in our Canadian sample. All three SNPs produced a significant correlation with NOS1AP expression, and in each case the sequence variant associated with higher expression was the sequence variant associated with schizophrenia in our Canadian families [43].

An extension of this study design is to consider a sample of only unaffected individuals. Such control subjects are usually easier to acquire and do not have confounding issues of medication treatment and secondary sequela of illness. These individuals can be assessed for functions thought to be related to illness to see if any the variation present in the normal population is attributable to genetic changes in the gene under study. A recent functional imaging study was conducted on a sample of 125 healthy subjects separated subjects into groups based on DNA sequence at a NOS1AP SNP [44]. Healthy individuals carrying the schizophrenia-associated DNA sequence displayed significantly greater activation of the dorsolateral prefrontal cortex during a task of working memory. This implies that alterations in NOS1AP may lead to inefficient information processing in this region (even in healthy individuals), suggesting a mechanism by which NOS1AP may mediate risk for schizophrenia.

Motivated by a desire to conduct a more comprehensive survey of NOS1AP for association with schizophrenia in our Canadian families, we have recently extended our analysis of NOS1AP to 60 SNPs (45 new and 15 previously reported), and have functionally evaluated the three with the strongest evidence for association using a luciferase reporter assay in two human neuronal derived cell lines [45]. These three SNPs had nearly indistinguishable statistical support for association, due to the fact that the sequence present at any one of the three is very highly correlated with the sequence present at the other two. However, only one of the three SNPs exhibited the ability to alter gene expression. The observed change was also in the expected direction; the sequence variant associated with illness in the Canadian families caused higher expression in the cell lines [45]. Of note is that this specific SNP has not yet been evaluated in any other association or functional study of NOS1AP. It will be important to evaluate this functional SNP for association to schizophrenia and related behavioral characteristics in other samples. It will also be important to continue to search for additional functional sequence variants; identification of this functional sequence does not preclude the existence of others, particularly in more distantly related populations.

Conclusions

NOS1AP is an attractive candidate gene for schizophrenia susceptibility. Linkage and association studies from multiple samples drawn from different populations indicate that a schizophrenia susceptibility gene is located in this region of chromosome 1 [23–28, 35–38]. Increased expression of NOS1AP is associated with schizophrenia [43]. DNA sequence changes in the gene are associated with inefficient information processing in the dorsolateral prefrontal cortex [44]. A specific schizophrenia-associated single base change in DNA sequence leads to increased gene expression in vitro [45]. Overexpression of the gene has been demonstrated to lead to decreased dendrite branching in developing neurons in vitro [11] and is predicted to decrease signal transduction through the NMDAR complex [43]. Yet can we say that we have “proven” NOS1AP to be a schizophrenia susceptibility gene?

Direct evidence implicating a gene in disease susceptibility may be extremely difficult to obtain for a psychiatric disorder. Essentially, one would like to be able to introduce and remove the effect of the gene at will and see the illness appear and disappear. Introducing the gene effect and illness into a human would obviously be unethical, and it is clear from simpler medical disorders that genetically engineered animal models do not always recapitulate the human pathology. Removing the effect of the gene within humans strictly as a means for gathering proof of its involvement in disease would also be unethical; but even in cases where the risk-benefit ratio of such an attempt were acceptable, there is no guarantee that altering the gene in an ill individual would reverse the illness if the critical impact of the gene were earlier in development.

Thus, the case for NOS1AP, as for many complex disease candidate genes, hinges upon less direct evidence. While necessarily circumstantial, the evidence is obtained from multiple independent investigators using multiple experimental modalities, yet fits together into a cohesive and internally consistent story. Future studies of this gene will further elucidate its role in the etiology of schizophrenia.