INTRODUCTION

Schizophrenia (OMIM 181500) is debilitating psychiatric disorder that affects more than 29 million individuals worldwide (Barbato, 1998). It is a heritable disorder with a complex transmission pattern as evidenced from family and twin studies (Berry et al, 2003). The disorder has been linked to multiple chromosomal regions and several findings now appear to have been widely replicated, including Neuregulin, COMT. and DISC1 (Stefansson et al, 2004; Shirts and Nimgaonkar, 2004; Hennah et al 2003; Hodgkinson et al, 2004).

Both Disrupted in Schizophrenia 1 and 2 (DISC1 and DISC2, OMIM 605210, and OMIM 606271) were observed to be disrupted by a balanced (1;11) (q42.1;q14.3) translocation in a large Scottish family displaying a broad spectrum of psychiatric disorders, in particular, schizophrenia and major depression (Millar et al, 2000, 2001; Blackwood et al, 2001). Although analysis of other independent Scottish families with schizophrenia and bipolar disorder failed to show significant association between disease and these two genes (Devon et al, 2001), association to schizophrenia was shown in a Finnish population (Hennah et al, 2003) and to schizophrenia, schizoaffective disorder and bipolar disorder in a North American Caucasian population (Hodgkinson et al, 2004). The region containing the two DISC genes also showed linkage to schizophrenia and schizoaffective disorder in Taiwanese families (Hwu et al, 2003).

The DISC1 gene codes for a large (854 amino acid) protein of unknown function, with no homology to known proteins (Millar et al, 2000). DISC1 has been shown to occupy a variety of subcellular compartments and to exist in multiple isoforms generated by alternative splicing of the mRNA transcript (Millar et al, 2003; Morris et al, 2003; Ozeki et al, 2003; Brandon et al, 2004; James et al, 2004). Two-hybrid experiments in yeast systems have shown the DISC1 protein to interact with a large number of different proteins from a wide variety of different protein (Morris et al, 2003; Ozeki et al, 2003; Brandon et al, 2004; Miyoshi et al, 2004). The DISC2 gene is thought to encode a nontranslated mRNA antisense to DISC1, and may have a regulatory function (Millar et al, 2000).

FEZ1 protein (OMIM 604825), was identified as a DISC1 interacting partner in a yeast 2-hybrid screen of an adult human brain library (Miyoshi et al, 2003). FEZ1 is the ortholog of the Caenorhabditis elegans unc-76 gene (Horvitz, 1997). Worms carrying mutations in the unc-76 gene show severe abnormalities in movement and in elongation of axons along other axonal surfaces (but not non-neuronal surfaces) during development. This process, known as fasciculation, enables axons to associate in specific bundles and is likely to play a major role in determination of neural structures.

Association between SNP polymorphisms in the FEZ1 gene and schizophrenia has been shown in a Japanese cohort (Yamada et al, 2004). This association was shown not to have arisen due to population admixture. In this study, we have looked at two ethnically distinct North American populations for association of schizophrenia to the FEZ1 locus. In both of these populations, linkage of the FEZ-interacting partner DISC1 to schizophrenia was observed (Hodgkinson et al, 2004 and unpublished results), data that replicated findings in a Finnish cohort (Hennah et al, 2003).

MATERIALS AND METHODS

Subject Recruitment and Diagnosis

Subjects were recruited from the clinical services of the Zücker Hillside Hospital, a division of the North Shore—Long Island Jewish Health System (NSLIJHS), in Glen Oaks, NY, USA, and from the Psychiatry Service, Department of Veterans Affairs, New Jersey Health Care System in East Orange, NJ, USA. These studies were performed in a manner that fully complies with the Code of Ethics of the World Medical Association (Declaration of Helsinki) and was approved by the relevant Institutional Review Boards. After providing written informed consent to participate in the study, each subject was assessed with the Structured Clinical Interview for DSM-IV Axis I Disorders using method previously described (Hodgkinson et al, 2004). The Caucasian sample of 521 individuals was comprised of 212 healthy controls, 178 schizophrenics, 79 bipolar disorder patients, and 58 patients with schizoaffective disorder. The African American sample, totaling 323 individuals, contained 133 healthy controls, 162 diagnosed schizophrenics, and 28 with schizoaffective disorder.

Genotyping

Nine SNP loci spanning the 61 kb FEZ1 gene region were genotyped by 5′ exonuclease assay, using the primer-probe sets available as Taqman® Assays-on-Demand (Applied Biosystems, Foster City, CA, USA). Genomic DNA (5 ng) was amplified on a 9700 thermocycler (ABi) using a program with an initial incubation at 95°C for 10 min followed by 40 cycles of 92°C (15 s) and 60°C (1 min). At the end point of amplification, genotypes were discriminated using SDS 2.0 software on an Applied Biosystems 7900 Analyzer. Genotype completion rates for all markers were ⩾98.9%. Genotyping accuracy was assessed by regenotyping one in six samples, randomly selected, revealing an overall accuracy >99%. Haplotype block structure was determined using the HAPLOVIEW program (Barrett et al, 2005). Blocks were defined according to the criteria of Gabriel et al (2002) but the criterion for pairwise LD between markers was relaxed from D′>95% to D′>85% within a block. Haplotypes were assigned to individuals using PHASE 2.02 (Stephens et al, 2001; Stephens and Donnelly, 2003).

Statistical Methods

Nonparametric testing for significance values for allelic and genotypic association was performed using a standard χ2-test. Haplotype association was evaluated using the more stringent Fisher Exact Test because observed haplotype frequencies sometimes equaled zero. Deviation from Hardy–Weinberg distribution of alleles was determined using the Haploview program and the p-values are shown in Table 1. Logistic regression was used to test for interaction between genotypes for each FEZ1 SNP and DISC1 risk/protective haplotypes as described in Cordell (2002) and Norton et al (2006). Individuals were assigned a score of 1, 0.5, or 0 according to the number of copies of DISC1 haplotypes significantly associated with schizophrenia and/or schizoaffective disorders in Hodgkinson et al (2004). Analysis were conducted separately for each diagnostic category and for each of the three DISC1 haploblocks previously linked to schizophrenia and/or schizoaffective disorders.

Table 1 Allelic and Genotypic Association of FEZ1 SNPs in Caucasians, Comparing Healthy Controls (HC), Schizophrenics (SZ), Schizoaffective Disorder (SA), and Bipolar Disorder (BP)
Full size table

RESULTS

No FEZ1 SNP showed association to either schizophrenia or schizoaffective disorder in Caucasians (Table 1) or to schizophrenia in African Americans (Table 2), at either the allelic or genotypic level. The markers tested include rs559668 and rs597570, both of which had shown genotypic association in the Japanese study. Also, none of these markers showed association to bipolar disorder in Caucasians. All nine FEZ1 markers are in strong linkage disequilibrium and reside in a single haplotype block (Figure 1). As expected there was increased haplotype diversity in the African American population compared to the Caucasian group. Moreover the minor allele frequencies differed greatly between the populations (Table 3). There was no significant association between schizophrenia and any haplotype in either population (Table 4). Nor was there any association to schizoaffective disorder or bipolar disorder in Caucasians. In the Japanese population, schizophrenia was associated with homozygosity of the A allele at the rs559668 and rs597570 loci. The frequency of these alleles was higher in Caucasians (18%) and African Americans (29 and 21%) than in the original Japanese cohort (2%) and in an American Asian group (2%). In the Japanese study, association to schizophrenia was found to homozygosity for the minor alleles of the two tightly linked markers which are separated by over 200 kb. In that study, seven affected individuals but no controls were homozygous for the two markers. In our African American cohort, we found nine of 178 diagnosed schizophrenics and 13 of 212 healthy controls homozygous for the same two A alleles. In the Caucasian group, we found eight of 178 diagnosed schizophrenics to be homozygous for the same two A alleles, seven of 212 healthy controls, one of 70 diagnosed with bipolar disorder and 0 of 58 diagnosed with schizoaffective disorder. Thus, we found no increase in homozygosity. Analysis of homozgosity at the diplotype level showed no association in either the Caucasian or African American populations by Fisher Exact test (data not shown). Analysis of a small group of African Americans diagnosed with schizoaffective disorder (n=28) did show a genotype association (p=0.038) at marker rs559668 and a haplotype association (p=0.025) for the 112222212 haplotype. Neither of these associations remained significant after correction for multiple testing.

Table 2 Allelic and Genotypic Association of FEZ1 SNPs in African Americans Comparing Healthy Controls (HC), and Schizophrenics (SZ)
Full size table
Figure 1
figure 1

(a) Diagram of the human FEZ1 gene showing the relative positions of exons and the SNP markers used (shaded circles). Marker rs3217293 from Yamada et al (Biol Psychiatry 2004; 56: 683–690.) shown with open circle. (b) Haplotype block structure of the FEZ1 locus in Caucasians and African Americans.

Full size image
Table 3 Minor Allele Frequencies in Three North American Populations
Full size table
Table 4 Haplotype Associations in African American and Caucasians Comparing Schizophrenics (SZ), Schizoaffective Disorder (SA), and Bipolar Disorder (BP)
Full size table

The failure to detect association of schizophrenia to FEZ1 haplotypes is unlikely to arise due to lack of power in the data set. For schizophrenics both the Caucasian and African American datasets have sufficient power to detect significance (p=0.05) for frequency differences of 1.2 at 90% confidence, and for frequency differences of 1.6 at 99% confidence. For schizoaffective disorder the Caucasian dataset has sufficient power to detect significance (p=0.05) for frequency differences of 2.1 at 90% confidence, and of 2.8 at 99% confidence.

Given the biological evidence of interaction between DISC1 and FEZ1, logistic regression was performed to test whether variation at FEZ1 contributed to the risk of schizophrenia and/or schizoaffective disorders depending on DISC1 disease-associated haplotypes. For schizophrenia no significant interaction was observed for any FEZ1 SNP to DISC1 haplotypes for Haploblocks 1, 2, or 3. Significant p-values were obtained for interaction of rs2702009 (p=0.001) and rs10893385 (p=0.03) with the common protective DISC1 haplotype previously shown to be under represented in Caucasian women with schizoaffective disorder. These data, however, should be treated with extreme caution due to the relatively low number of affected individuals in the analysis (n=57) and the elevated number of statistical tests performed (n=54). Based on these data it appears unlikely that any significant GXG interaction between FEZ1 and DISC1 exists.

DISCUSSION

An important genetic component in the etiology of schizophrenia is thought to be a neurodevelopmental defect (Lewis and Levitt, 2002) that either occurs in latent form early in development (Marenco and Weinberger, 2000) or at a later stage during brain maturation in adolescence (Feinberg, 1982), the time at which early symptoms begin to appear. FEZ1 when expressed in unc-76 mutant worms can partially rescue the locomotor phenotype suggestive of a shared developmental function (Horvitz, 1997) and suggesting that these sequence-similar proteins are orthologs. Studies in primates and rodents demonstrate that FEZ1 and DISC1 have overlapping spatial and temporal expression patterns (Austin et al, 2004; Honda et al, 2004; Inoue et al, 2004; Schurov et al, 2004). Both proteins are expressed in the pyramidal neurons of the developing hippocampus, the cerebral neocortex and the olfactory bulb. Moreover disruption of the DISC1/FEZ1 interaction inhibits DISC1-stimulated neurite outgrowth in PC12 cells (Miyoshi et al, 2003). Together these data make FEZ1 a compelling candidate gene. Our data, however, indicates no strong role of variation at the FEZ1 locus in schizophrenia in Caucasian or African American populations. The asp123 residue is conserved between humans and rodents. This residue lies within a region shown to be responsible for the axonal targeting of the unc-76 protein (Horvitz, 1997) and it has been suggested that the glutamate substitution affects the subcellular distribution of the FEZ1 protein (Miyoshi et al, 2003). The asp123 glu transition (rs597970) is a conservative substitution and this residue is not conserved in either Gallus gallus (his) or in C. elegans (thr) suggesting that it is unlikely to have a functional role in axonal targeting of the mature FEZ1 protein. Our findings do not exclude the possibility that a functional polymorphism unique to the Japanese population exists in strong LD with rs559668 and rs597570 which when present in the homozygous state results in an increased predisposition to the development of schizophrenia. Sequencing of the Japanese schizophrenics homozygous for the two associated markers may identify the relevant polymorphism, which could lie either in the axonal targeting N-terminus, or the DISC1-interacting C-terminus of FEZ1.

Although association has been shown between schizophrenia and DISC1 in two independent Caucasian populations including the Caucasians studied here, no association was found in a case/control study involving ⩾1000 individuals from Japan (Kockelkorn et al, 2004). This suggests that variation at the DISC1 locus may play no role in the etiology of schizophrenia in the Japanese population. It would, therefore, be hardly surprising that the converse were found for other loci. Schizophrenia and schizophrenia spectrum disorders are complex diseases showing variability in many features in terms of symptoms, age of onset and genetics. Certainly the clinical heterogeneity coupled to the number and diversity of genes reported to be associated with schizophrenia might support the idea that the clinical entity known as schizophrenia could arise from dysregulation of different molecular pathways or by disruption of a common pathway at multiple different points.