Recruitment of multiplex families containing two or more affected individuals among second degree relatives was performed through proband screening at Samsung Medical Center, Yong-In Mental Hospital, National Chuncheon Hospital and other mental hospitals and psychiatric clinics in Korea. The clinical assessment procedures used were described previously.²¹²² We limited the diagnosis of the probands to schizophrenia according to the DSM-IV criteria. Affectation status was assigned as broad and narrow liability classes. For the broad class, the affected individuals other than the probands were diagnosed as having schizophrenia, schizoaffective disorder or schizotypal personality disorder according to DSM-IV. In the case of schizoaffective disorder and schizotypal personality disorder, only those cases in which long-term maintenance with antipsychotics was needed were included. The narrow class included those families in which all of the affected individuals met the DSM-IV criteria of schizophrenia. Following the exclusion of one family showing bilineal transmission of psychoses, the clinical data and DNA samples were analyzed from 47 families with 161 members including 105 affected individuals. Among these families, 40 (containing 138 members including 89 affected individuals) also belonged to the narrow class.

For the case-control association study, unrelated schizophrenia patients were recruited from Samsung Seoul Hospital, Yong-In Mental Hospital and National Chuncheon Hospital. Diagnostic Interview for Genetic Studies clinical interviews and best estimate diagnosis were performed by the same raters and psychiatrists who assessed the multiplex families. DSM-IV schizophrenia patients (n=227) were analyzed. Also, 292 normal control persons were recruited and screened for any psychiatric illnesses by means of a brief unstructured interview. The male-to-female ratio of the case and control group was 1.20 and 1.10, respectively (χ²=0.21, df=1, p=0.65). There was no significant difference between the mean age of the case and control groups (33.2±9.6 and 33.5±8.9 years, respectively; t=0.046, p=0.96). The average age at onset of patients with schizophrenia was 25.7±7.0 years (range, 13-49). This study was approved by the institutional review boards of Samsung Medical Center and Yong-In Mental Hospital. Informed consent was obtained from all of the subjects who participated in the interviews and blood sampling.

Three microsatellite markers spanning a region of 4.35 Mb between 22q11.1 and 22q11.22 were selected (Figure 1) based on their heterozygosities and distances from each other (Table 1). The markers have a 5.86 cM average interval according to the deCODE map. Their heterozygosities were calculated from the genotype data of one-hundred normal Koreans (generated by one of the authors; JWK) and from the unaffected founders of the families of the current study (Table 1). Genotypes were obtained for a broad class of 47 multiplex families with 161 members including 105 affected individuals. DNA was extracted from the white cells of peripheral blood by a Wizard™ Genomic DNA Purification kit (Promega, Madison, WI, USA). In the polymerase chain reaction (PCR), primer sets used were provided by the ABI^PRISM linkage mapping set v2.5 (Applied Biosystems, Foster City, CA, USA) for D22S420 and D22S539. For D22S427, the primer set was designed using Primer 3 primer-designing software (http://frodo.wi.mit.edu/cgi-bin/primer3/primer3-www.cgi). All of the fluorescently labeled PCR products, by each forward primer, were separated by capillary electrophoresis on an ABI3100 Genetic Analyzer (Applied Biosystems). The results of the automatic scoring of the genotype peaks provided by “Genotyper” program (Applied Biosystems) were manually checked by referring to the distribution patterns of the peaks. Quality control of genotyping utilized DNA sample of CEPH 1347-02 in the assays, with manual checking of the Mendelian transmission for all of the markers and families.

Initially, an association study using pooled genotyping was done for 39 SNPs (35 intergenic and four intragenic SNPs) with a mean inter-marker distance of 33 kb spanning a 1.4 Mb region of 22q11, which contains four candidate schizophrenia genes (DGCR6, PRODH, COMT, and ZDHHC8) (Figure 1). DNA pools were constructed separately for 227 patients and 292 control subjects as previously outlined.²³²⁴ Concentrations of all DNA samples used for pool construction were determined using the PicoGreen dsDNA Quantitation method (Molecular Probes, Eugene, OR, USA) in a Labsystems Ascent Fluoroskan (Life Sciences international, Basingstoke, UK). Three microliters of a DNA suspension (20 ng/uL) from each individual was used to construct DNA pools. Genotyping of DNA pools were performed using SNapShot primer extension method (Applied Biosystems) as previously described²³ for each SNP. PCR and extension primers were designed using MassARRAY assay design software (Sequenom, San Diego, CA, USA). Each pooled DNA sample was genotyped four times and average allele frequencies were calculated.

Individual genotyping with all samples included in the pools was performed for SNPs with a trend of association at p<0.1 and four more intra-genic SNPs (Table 2). Genotyping of all the SNPs, except for two on PRODH, were performed with the SNapShot assay (Applied Biosystems). PRODH^*1945 T→C and 1766 A→G SNPs were genotyped using the PCR-restriction fragment length polymorphism method. A 499 bp DNA fragment containing the PRODH^*1945 T→C polymorphism was amplified using the forward primer 5’-CTCCCTGGTGCGATGGGGTAC-3’ and reverse primer 5’-GGGCCCACACATTCG-AGGAG-3’, followed by digestion with Pvu II (Roche, Basel, Switzerland). The digestion products were scored as cut and uncut type (1945^*C allele and 1945^*T allele, respectively). A 263 bp DNA fragment containing the PRODH^*1766 A→G was amplified using the forward primer 5’-GGGACAG- TGTCCTGGAACTG-3’ and reverse primer 5’-CATCACGGGGCCATAGGGCAC-3’, followed by digestion with Nci I (New England Biolabs, Ipswich, MA, USA). The scored cut and uncut type was 1766^*G allele and 1766^*A allele, respectively.

The model-free linkage analysis was performed using Genehunter 2.1_r4 β.²⁵ Both the single point and multi-point NPL_all scores were calculated by an affected-only allele-sharing method. This analysis does not have the context of a genome-wide linkage mapping, and we adopted the pointwise p-value to evaluate the statistical significance. The marker distances obtained from the deCODE map were adopted. The marker allele frequencies were calculated from the Korean population data described above, except for D22S427. For this marker, allele frequencies were calculated from the genotypes of the unaffected founders of the families recruited for the current analysis.

The genotypic distribution of each marker tested for the association studies was not significantly different from that expected based on the Hardy-Weinberg equilibrium in either the case or control group (SAS Genetics v 8.02). In the initial screening with pooled DNA samples, peak heights of the SNaPshot products in a pool of case DNA and in a pool of control DNA samples were measured. The frequency f of allele A was calculated as f(A)=H_A/(H_A+kH_B), where H_A and H_B are the peak heights of the SNaPshot products representing alleles A and B. A correction factor (k), the ratio of the peak heights in the heterozygous sample (H_A/H_B), was used to correct for systematic unequal representation of alleles that occurred with most genotyping methods. This was estimated from 8-16 independent heterozygotes samples that were individually genotyped. The frequency f of allele B was calculated as 1-f(A). Estimated allele frequencies were converted to numbers and were tested for approximate statistical significance by χ² analysis. For individual genotyping methods, genetic associations were also tested using Pearson χ²-test with the SAS statistical software package version 8.2 (SAS, Cary, NC, USA).

The sample size of 80% power is used to avoid false negative associations in genetic association studies. The effective sample size and statistical power were computed assuming 1% prevalence rate of schizophrenia, 1 of D-prime (Linkage disequilibrium), odds ratio of 1.05-1.5 for a risk allele, 1:1 case-control ratio. User defined type I error rate was calculated for the 11 tested SNPs. Genetic Power Calculator software26 was used for the statistical power determination.

Mendelian inconsistency was detected in only one family. The family was excluded from the statistical analysis. Forty-six families with 157 members including 103 affected individuals were finally included in the statistical analysis. No significant linkage was found at any of the three markers in single and multi-point analyses (Table 1).

In the initial analysis with pooled DNA, five SNPs showed positive associations with schizophrenia: rs2073762 (p=0.01), rs2073767 (p=0.0041), rs4680 (p=0.0133), rs1153415 (p=0.0287) and rs701430 (p=0.0068) (Figure 2). Among these SNPs, rs4680 was located in the COMT gene and the others were inter-genic SNPs. Two more inter-genic SNPs showed a trend for association (p<0.1), i.e., rs1008241 (p=0.0656) and rs2073752 (p=0.0883). For those seven SNPs and four additional SNPs on PRODH and ZDHHC8, genotyping with individual DNA samples were performed. In the association analysis, none of the SNPs showed significant association with schizophrenia (Table 2). The statistical power analysis showed the sample size was too small to achieve adequate statistical power. For instance, the statistical power to identify the evidence for association between rs207362 and schizophrenia after the Bonferroni correction under the assumptions of odds ratio for TT genotypes assuming a recessive model (1.44), risk allele frequency (41% LD between the marker allele and the risk allele=1), disease prevalence (1%) was estimated as 11.69%. At least 1070 cases and 1391 controls were required to achieve 80% statistical power under the same assumption.