Healthy controls were recruited from the employees of the National Taiwan University Hospital. After signing informed consent, the individuals underwent a screening interview followed by blood withdrawal. Exclusion criteria for the healthy controls were: under age 30; diagnosed with psychiatric disorder, especially schizophrenia; having a history of diabetes mellitus (DM), major systemic disorder, or neurological disorder (e.g., epilepsy); mental retardation; facial dysmorphism; and clinical evidence of brain, trunk or limb anomalies.

The control DNA pools were constructed by pooling equal amounts of DNA extracted from ten healthy men and ten healthy women. These normal genomic DNA pools were used as reference samples for array CGH analysis and real-time quantitative PCR.

Patients for array CGH analysis were enrolled from the outpatient clinics of the Department of Psychiatry, National Taiwan University Hospital. The inclusion criteria were: a diagnosis of schizophrenia according to the Diagnostic and Statistical Manual of Mental Disorders, 4^th edition (DSM-IV) 25 and confirmed by the Diagnostic Interview for Genetic Study (DIGS) and at least two siblings affected by schizophrenia in a given family. Patients affected with mental retardation, facial dysmorphysm, or clinical evidence of brain, trunk or limb anomalies were excluded. A total of ten pairs of schizophrenic siblings, a total of 20 cases, from ten unrelated Taiwanese families (A-J), were recruited. The mean age of the subjects was 30.6 years.

After the homozygous deletion of BU678720 was found in the initial 20 subjects, a different sample was recruited between 2003 and 2005 in Taiwan for genetic study. A total of 107 controls, 163 simplex schizophrenic patients (those from families with only one member affected with schizophrenia), and 72 multiplex schizophrenic patients (those from families having at least two affected siblings) were included in this study. All cases fulfilled the DSM-IV criteria for schizophrenia.

Clinical symptoms were rated using the scale for the assessment of negative symptoms (SANS) 26 and the scale for the assessment of positive symptoms (SAPS) 27 , both of which have demonstrated satisfactory reliability. According to SANS, the negative symptom score was the sum of scores for Affective Blunting, Alogia, Avolition-Apathy, and Anhedonia-Asociality. The Continuous Performance Test (CPT) 28 and Wisconsin Card Sorting Test (WCST) 29 were used for neuropsychological assessment of sustained attention and executive function, respectively. The assessment methods were described in our previous reports 28 29 .

Genomic DNA was isolated from peripheral blood lymphocytes with the PureGene DNA Purification Kit (Gentra Systems, Minneapolis, MN, USA) according to the manufacturer's instructions.

A commercial oligonucleotide array (Human Genome CGH microarray 44B, Agilent Technologies, Palo Alto, CA, USA) was used for array-CGH analysis. Genomic DNA fragmentation, labelling and array hybridization were performed as previously described 30 31 . Each array hybridization experiment was performed with differentially labelled gender-matched samples, one from the affected individual, and the other from the DNA control pool. To rule out probable CNPs in our ethnic group, two array hybridization experiments were performed using the male or female pooled control and commercial, normal, same-gender Caucasian samples (Promega, Madison, WI, USA).

Filtering procedures were applied to select qualified data sets for analysis. In total, 18 qualified arrays, exclusive of the samples for patients F2 and J2, were selected for further analysis. Aberrations were only considered if the aberration scores, automatically generated by Agilent CGH analysis software, were higher than 1.00 or lower than -1.00. The CNVs which exist within the control genome and are unlikely to be pathogenic were filtered out by comparison to the CNVs identified from the two reference arrays which were performed with normal male and female DNA pools in our laboratory. Additionally, the published CNVs listed in the Database of Genomic Variants 32 with relative high incidence in control subject were also excluded.

The data discussed in this publication have been deposited in NCBI's Gene Expression Omnibus (GEO) 33 and are accessible through GEO series accession number GSE16930.

In order to examine the copy number changes of those genes found in our array CGH analysis, we used the quantitative real-time PCR method. The detailed qPCR conditions were described by Lee 31 . The specific oligonucleotide primer pairs were selected from the Universal Probe Library (Roche Molecular Systems, Inc., Branchburg, NJ, USA). The ATPase, Ca++ transporting, plasma membrane 4 (ATP2B4, NM_001001396) was chosen as the reference. The fold change in gene copy number for a target gene is calculated by using the comparative ΔC_T method as described previously 34 . In each experiment the samples were analyzed in triplicate. The primer information is provided as Supplementary Table 1 (Additional file 1).

In order to investigate the prevalence of homozygous deletion of the specific UB678720 allele revealed in this array CGH study in a different sample, we used PCR followed by electrophoresis. PCR was carried out with the same primers as those used for qPCR. Details of PCR conditions were as described elsewhere 31 . Amplified products were analyzed by on-chip electrophoresis using Agilent 2100 bioanalyzer and Agilent DNA 1000 LabChip kit (Agilent Technologies). Homozygous deletion of BU678720 was readily distinguishable by the presence of the 85 bp amplified fragment.

For array CGH analysis, the hybridized arrays were scanned and analyzed as previously described 30 31 . Briefly, after washing, the hybridized arrays were immediately scanned at a resolution of 5 μ using an Agilent G2565BA DNA microarray scanner. The microarray images were analyzed using Agilent Feature Extraction software, version 8.1.1. Another custom analytical software package, Agilent CGH Analytics, version 3.4, was used for the subsequent data analysis. The locations of the copy number aberrations were calculated using the Aberration Detection Method 2 (ADM2) statistical algorithm. The ADM2 threshold was set at 9.0 to make an amplification or deletion determination. According to these settings, the aberration score was generated automatically for each altered locus.

The comparison of the negative symptoms between subgroups was calculated by using the Mann-Whitney U test. The comparison of the incidence of BU678720 homozygous deletion between multiplex and simplex families was analyzed by using the Genmod procedure with software SAS 9.1.

A total of 379 loci disrupted by CNVs were found. These included 343 losses, 10 gains, and 26 with both losses and gains. Summaries for each patient are presented in Table 1. There were great variations in the number of loci affected by CNVs (range 0-318). Thus we classified patients into subgroups based on the number of loci with CNVs. There were six or fewer scattered loci disrupted by CNVs in subgroup I (A1, B2, C1, E1, F1, G1, G2, and J1), while subgroup II had up to 318 loci with CNVs (A2, B1, C2, D1, D2, E2, H1, H2, I1 and I2). We also noted that the patterns of CNVs showing deletions were more common than those showing amplifications. In addition, we observed that four sib-pair subjects (50%) were in different subgroups, while there were four sib-pair subjects (50%) in the same subgroup (one in subgroup I, three in subgroup II) (Table 1).

We compared the clinical features between subgroup I and subgroup II patients and examined whether the array CGH profiles were correlated with the phenotypes defined by psychopathological parameters of clinical symptoms and neuropsychological performance. We found that the negative symptom score was significantly higher (Mann-Whitney U = 14, n1 = 8, n2 = 9, p = 0.033, two-tailed) for subgroup II when compared to subgroup I (Table 1). However, there was no significant difference in the scores for delusions/hallucinations and disorganized symptoms dimensions. There were also no differences in the performance of sustained attention and executive function between these two subgroups.

The genes related to CNS growth and development were referred to as CNS-related from a computer search of the relevant literature on PubMed 35 . For subgroup I patients, the three CNS-related aberrant loci including BTB (POZ) domain containing 8 (BTBD8, NM_183242, 1p22.1), paired-like homeobox 2b (PHOX2B, NM_003924, 4p12) and apoptosis-associated tyrosine kinase (AATK, NM_001080395, 17q25.3) were detected in patients A1, G2 and F1, respectively. It was noted that the copy number gain of AATK and copy number loss of PHOX2B were the only CNVs detected in patients F1 and G2, respectively.

Of the genes disrupted by CNVs identified in the subgroup II patients (10 subjects), a total of 16 genes with an incidence of at least 30% (3/10) were selected. The distributions of these highly recurrent CNVs are presented in Table 2. Four of these 16 target loci, genes encoding CCAAT/enhancer binding protein, delta (CEBPD, NM_005195), retinoid × receptor, alpha (RXRA, NM_002957), LIM homeobox protein 5 (LHX5, NM_022363), serine/threonine kinase 11 (STK11, NM_000455) are CNS-related. There are only two genes, PVR and BU678720, with familial incidence. Of these two genes, the BU678720 (BU678720) (function not yet known) had a concordant loss in both sib-pairs of families B and D (Figure 1) and accounted for the highest familial incidence (Additional file 2: Supplementary Table 2). In comparison with healthy controls determined by screening against the Database of Genomic Variants, the alteration incidences are rather high in schizophrenics for all 16 identified CNVs (Additional file 2: Supplementary Table 2, Table 2, Fig 1).

Quantitative PCR analysis was performed for the four recurrent CNS-related genes, CEBPD, RXRA, LHX5, STK11, and the gene of BU678720 with a high incidence (two out of 8 pairs: 25%) of concordant loss in both sib-pairs. We analyzed the siblings for whom CNS-related CNVs were detected in at least one of the sib pairs. Comparisons between results from arrays and those from qPCR are summarised in Supplementary Figure 1 (Additional file 3). Deletions of BU678720 could be confirmed by qPCR, while, for others, the qPCR patterns were not perfectly consistent for array CGH results. For BU678720, the low qPCR signals suggested that the gene copy may be completely lost in sib pairs from families B and D.

Since homozygous deletion may have a more profound influence on regulation of gene expression than single copy deletion, we performed the PCR experiments on the ten pairs of affected siblings to detect a possible homozygous deletion in the four CNS-related candidate genes (CEBPD, RXRA, LHX5 and STK11) and the gene BU678720. Homozygous deletion was detected in the gene BU678720 in six individuals A1, B1, B2, D1, D2, and J1 (Fig. 1b), while no homozygous deletion was detected in the four CNS-related candidate genes. To assess whether the homozygous deletion cosegregated with schizophrenia in the four families, we recruited the first-degree relatives of families A, B, D and J to examine for homozygous deletion of BU678720. The results are depicted in Figure 2. Homozygous loss of BU678720 cosegregated with schizophrenia in families B and D. However, the phenomenon of cosegregation was not observed in families A and J. To further clarify the association between BU678720 and schizophrenia, we performed the PCR assay in 72 and 163 affected individuals from multiplex and simplex families, respectively, as well as in 107 controls. The homozygous deletion of BU678720 was detected in 7.4% of patients from simplex families, 11.1% of patients from multiplex families, and 8.4% of controls. Though the prevalence in the patients from the multiplex families was higher than that in controls and patients for simplex families, the comparisons did not attain a level of statistical significance (p = 0.098, df = 1).