Introduction

Panic disorder (PD) is an anxiety disorder characterized by recurrent and unexpected panic attacks, subsequent worry and phobic avoidance. Life prevalence of PD is 1–3% and twice as many women as men suffer from the disorder.1 The disorder frequently takes a chronic course, with many remissions and relapses.2 Genetic epidemiological studies including family and twin studies have shown that genetic as well as environmental factors have an important role in the pathogeneses of PD. First-degree relatives of proband with PD have an approximately fivefold increased risk of PD.3, 4 Twin studies show that about 40% of the liability towards PD consists of heritable factors5, 6, 7 However, the etiology of PD is currently unknown.

Linkage analyses of PD have implicated several chromosomal regions including 1q,8 2q,9 7p,10, 11 9q,12 12q,13 13q,14, 15 15q9 and 22q.15 In association analyses, candidate genes including HTR 1A,16 2A,17, 18 CCK,19 ADORA2A,20 MAOA21 and COMT22, 23 have been investigated. Most of the association studies were, however, of small sample size (N<200) and the results were controversial. PD is considered to have polygenic- or oligogenic-multifactorial etiology, and most studies to date may be underpowered to detect the modest effect size of susceptibility loci.

We previously conducted a genome-wide association study of PD in the Japanese population using a 500 K single nucleotide polymorphism (SNP) chip.24 In the first stage, 200 patients and 200 controls were studied using the 500 K SNP chip (the GeneChip Human Mapping 500 K Array Set). A portion of the results of the first stage was previously reported.26 In the second stage, replication analysis of the SNPs, which were suggested in a previous study, was conducted in a Japanese second sample of PD (558 patients and 566 controls) using the DigiTag2 assay.

Materials and methods

Subjects

All patients and control subjects in the first and second stages were ethnically Japanese and were recruited in the vicinity of Tokyo and Nagoya, Japan. Subjects in the first stage consisted of 200 PD patients (64 males and 136 females; age=39.5±9.4 years, mean±s.d.) and 200 controls (64 males and 136 females; 38.2±9.3 years). Details were described in a previous paper.24 Subjects for the second sample comprised 558 unrelated PD patients (175 males and 383 females; age=38.9±11.1 years, 252 from Tokyo and 306 from Nagoya) and 566 controls (272 males and 294 females; age=35.7±12.9 years), who were recruited around Tokyo (n=466) and the Nagoya-Tsu area (n=100). For the multimarker haplotype analyses in the most susceptible genes in replication, subjects comprised 563 PD patients (177 males and 386 females; age=38.5±10.8 years; second sample and other five PD patients) and 506 controls (246 males and 260 females; age=35.2±13.1 years; second sample). Most of the sample overlapped with the second sample. The diagnosis was confirmed according to the DSM-IV criteria,25 using the MINI26 and clinical records. The controls received a short interview by one of the authors and filled out questionnaires to exclude the history of major psychiatric illness. Objective of this study was clearly explained and written informed consent was obtained from all subjects. The study was approved by the Ethical Committee of the Graduate School of Medicine, University of Tokyo.

Genome-wide association analysis of PD

In the first stage, 200 Japanese PD cases and the same number of controls were genotyped using the GeneChip Human Mapping 500 K Array Set (Affymetrix, Santa Clara, CA, USA) below for the replication in the second stage.

SNP selection

In the first stage, we employed the following quality control criteria as in Miyagawa et al.27 : call rate ⩾95% in patients or controls, Hardy–Weinberg equilibrium (HWE) P-value ⩾0.1% in controls and minor allele frequency ⩾5% in all samples. As a result, we observed 48 SNPs on autosomal chromosomes and one SNP on X-chromosome in female subjects at the level of P-value < 10−4. For the second stage, 32 SNPs were selected out of 49 as follows. Among the 49 SNPs, 29 had a minor allele frequency of ⩾10% in all subjects. Out of the 29, rs12227600 was excluded because it was located in the same linkage disequilibrium (LD) block with another of the 28 SNPs. Rs16985077 on X-chromosome was also excluded because the sample size was relatively limited and the analysis by sex may be underpowered. Thus, 27 SNPs were selected for the replication. As the DigiTag2 system simultaneously genotypes 32 SNP on one plate, five others (rs3804375, rs12579350, rs941184, rs3816995 and rs1383290) with the lowest P-values in the rest of the 49 SNPs (minor allele frequency ⩾5%; P-value < 10−4) were also selected and studied in the replication.

Genotyping

Genomic DNA was extracted from leukocytes by using the standard phenol–chloroform method. Genotyping was conducted at the SNP-Typing Center (University of Tokyo) using the DigTag2 assay (Olympus, Tokyo, Japan) following the manufacturer's protocol. A multiplex SNP-typing method designated DigTag2 is suitable for genotyping an intermediate number of SNPs (tens to hundreds of sites) with a high conversion rate (>90%), high accuracy and low cost.28, 29 The conversion rate is defined as the proportion of successfully genotyped SNPs among the total number of SNPs examined. The reproducibility of this assay was examined by duplicate experiments, resulting 100% identical between duplicate experiments.

Haplotype analyses

We conducted additional haplotype-based analyses to augment the signals of the most associated SNPs in known genes. We examined five SNPs in CLU (APOJ) and four in APOL3 (Figure 1). Besides rs17466684 and rs132617 investigated in the second stage, four tag SNPs (rs2279591, rs9331949, rs2279590 and rs9331888) in CLU (APOJ) and three tag SNPs (rs3827346, rs1807672, rs132660) in APOL3 were selected from the HapMap database (www.hapmap.org). These tag SNPs were selected to include at least one SNP within each LD block, which was determined with the criteria D′ ⩾0.75 using SNPbrowser software ver.3.5 (Applied Biosystems, Foster city, CA, USA). All SNPs were genotyped using the ABI PRISM 7900HT Sequence Detection System (Applied Biosystems), according to the manufacturer's protocol (Applied Biosystems), with ABI PRISM 7900 SDS2 Software (Applied Biosystems). To check for typing errors, we re-genotyped rs9331949 in duplicated samples. The concordance rates between the genotypes using the TaqMan were 100%.

Figure 1
figure 1

The strength of linkage disequilibrium (LD) between pairs of polymorphisms in patients and controls. The values of D′ for patients are shown in the upper portion and those for controls in the lower portion. The heavy-line frames show suggested haplotype blocks. (a) CLU (or APOJ) (b) APOL3.

Full size image

Data analyses

Deviation from the HWE was tested using the χ2-test. Allele and genotype frequencies of the patients and control subjects were compared using the χ2-test. Significance for the result was set at P<0.05. Data analyses for replication and the TaqMan study were performed using the PLINK software package30 (http://pngu.mgh.harvard.edu/purcell/plink/). LD and haplotype were analyzed in patients and controls who were studied using TaqMan method for all nine SNPs. D′ was used to analyze pairwise LD.31 Haplotype block analysis was conducted using the Gabriel method.32 The HaploView 4.1 program was used to conduct the LD and haplotype block analyses.33 A power calculation was performed using the method as described elsewhere.34, 35

Results

Replication analysis

Three of 32 SNPs (SNP29, SNP30 and SNP31) were not successfully genotyped because of strong false-positive signals, which resulted from misligation in the encoding step. Therefore, 29 SNPs were investigated for the replication. Genotyping call rate was 99.57% in the 29 SNPs. Allele and genotype frequencies are summarized in Table 1. The genotype distributions of two SNPs were significantly deviated from the HWE (SNP16; P=0.035 in patients, SNP20; P=0.024 in controls, respectively). Other SNPs were within HWE (P>0.05) in both patients and controls. Significant differences were observed in the allele and genotype frequency of the SNP2 (rs860554; allele P=0.015, genotype P=0.043), SNP4 (rs6733840; allele P=0.021) and SNP32 (rs132617; allele P=0.025) between patients and controls (Tables 1 and 2), although the associations did not reach statistical significance after Bonferroni correction. Risk alleles of SNP4 and SNP32 were the same as the first sample whereas risk allele of SNP2 was different from the first sample. SNP32 (rs132617) is located within 3 kb of the APOL3 gene and SNP4 (rs6733840) is located in an unknown gene. Four SNPs (SNP8, SNP12, SNP16 and SNP25) showed allele P-value between 0.05 and 0.1. Risk alleles of three SNPs (SNP12, SNP16 and SNP25) were the same as in the first sample, whereas risk allele of SNP8 was different from the first sample. SNP12 (rs17466684) is located within 1 kb of the CLU (or APOJ) gene and SNP25 (rs8013992) is located in the STON2 gene, whereas SNP16 (rs12420498) is not located in or adjacent to any gene.

Table 1 Allele and genotype frequencies of 29 SNPs successfully genotyped in the second stage between patients and controls
Full size table
Table 2 Allele frequencies in the first and second samples of SNPs showing allele P<0.001 in the first sample and P<0.1 in the second sample
Full size table

Haplotype analyses

Among seven SNPs at the allele P-value of <0.1 in the second sample, risk alleles of five SNPs were the same as in the first sample. Among five SNPs, three SNPs are located in known genes including APOL3, CLU (or APOJ) and STON2. We selected APOL3 and CLU (or APOJ) in which SNPs showed significant or marginally significant allele P-values. Allele and genotype frequencies of SNPs in the APOL3 and CLU genes between patients and controls were shown in Table 3. The genotype distributions of rs9331949 in controls were nominally deviated from the HWE (P=0.045). Other SNPs were within HWE (P>0.05) in both patients and controls. Significant differences were observed in allele frequencies of rs17466684 and rs132617 (P=0.019 and 0.033, respectively), although the statistical level became insignificant after correction for multiple testing. No significant difference was observed in allele frequencies of other SNPs or genotype frequencies of all SNPs between patients and controls.

Table 3 Allele and genotype frequencies of SNPs in the CLU and APOL3 genes
Full size table

The strength of LD denoted D′ between pairs of polymorphisms is shown in Figure 1. The LD pattern was broadly similar in the patient and control groups. In haplotype block analysis, two blocks consisting of SNP2–SNP3 (rs2279591-rs9331949) and SNP4–SNP5 (rs2279590-rs9331888) were suggested in CLU; and a block consisting of SNP1–SNP2 (rs132617-rs3827346) in APOL3 was suggested to be in strong LD (D′⩾0.75). We then compared frequencies of the suggested three haplotypes between patients and controls in addition to two haplotypes consisting of all polymorphisms in each gene. A significant difference was observed in the frequency of the haplotype consisting of all polymorphisms between patients and controls in CLU (ATTCG: P=0.035; Table 4), whereas the statistical level became insignificant after correction for multiple testing. No significant difference was observed in frequencies of other haplotypes between both patients and controls (Table 4).

Table 4 Haplotype frequencies according to haplotypes in each blocks and all polymorphisms of the CLU and APOL3 genes
Full size table

Discussion

We conducted a two-stage association analysis of PD in the Japanese population. Genome-wide association analysis was conducted in the first stage and 32 candidate SNPs according to the first stage result were tested in the second stage using the DigiTag2 assay. Among 32 SNPs selected for replication, 29 SNPs were successfully genotyped. Although we did not find a significant association after Bonferroni correction, two SNPs (rs6733840 and rs132617) were associated with PD at 5% level with the same risk allele as in the first sample. We also conducted haplotype analyses of SNPs in the APOL3 and CLU genes. Our results failed to show any significant association with PD in either a single SNP or haplotypes in these genes. However, rs132617 in the APOL3 gene and rs17466684 in the CLU gene were associated with PD at 5% level. The haplotype ATTCG in the CLU gene was nominally associated with PD.

APOL proteins might be involved in lipid transport and metabolism. Six APOL genes (APOL1-6) are located near each other on the 22q12 locus. The cholesterol content of intracellular membranes of neurons might be involved in modulating gene transcription and signal transduction in the brain during neurodevelopment and in the adult.36 A postmortem study using the candidate gene cDNA array showed the upregulations of APOL1, APOL2 and APOL4 in the prefrontal cortex of schizophrenic patients,37 whereas recent association study showed no association of this gene cluster with schizophrenia.38 In this study, SNP rs132617 was nominally associated with PD. In haplotype analysis, we observed a trend of decrease in the frequency of the haplotype TA (s132617-rs3827346) in PD patients (P=0.07). The result suggested that this haplotype might be protective against PD.

Furthermore, Clusterin, which is a protein encoded by the CLU (or APOJ) gene located on chromosome 8p21, might function in membrane lipid recycling.39 It may also be involved with ongoing synapse turnover.40 APOJ might regulate the process of amyloid-β deposition in the brain of Alzheimer's disease.41 Some studies suggested a correlation between cholesterol levels and the intensity and frequency of panic attacks reported by PD patients.42, 43 In our analyses, trends of increase in the frequencies of the allele A (rs17466684) and the haplotype ATTCG (rs17466684-rs9331888) in PD patients were observed. The results suggested that these, SNPs and haplotypes, might be a risk factor of PD. Previous findings from animal studies suggested that Clusterin increases post-ischemic damage and may exert a negative effect on the structural integrity and functionality of hippocampal neurons.44

Several limitations may be noted in this study. First, sample size (558 patients and 566 controls) might not be adequate. The second sample has statistical power of >0.80 (alpha=0.05) for the detection of the role of the polymorphism with minor allele frequency of 0.1, when the genotype relative risk is >1.52. A larger sample is requested for the detection of genes with a smaller effect than this. Second, proportions of the subjects from Tokyo and those from the Nagoya-Tsu area were not same between patients and controls. However, all subjects were ethnically Japanese and minor allele frequencies of significant SNPs were not different between the subjects from Tokyo and those from the Nagoya-Tsu area (data not shown).

In conclusion, we conducted a replication analysis of a genome-wide association study of PD in the Japanese population. As a result, no significant association with PD was found after correction for multiple testing. However, the difference was observed at the nominal allele P-value <0.05 for two SNPs (rs6733840 and rs132617). We also conducted haplotype analyses of SNPs in the APOL3 and CLU genes. Our results failed to show any significant association with PD in these genes. Further studies on these variants with a larger sample size may be worth testing to confirm the results.