Community-based healthy young adults were recruited through social media. Of the 292 participants recruited, 187 were male and 105 were female. The age range of the subjects was 18–35 years. Participants were confirmed to be free of lifetime or current diagnoses of psychiatric disorders by an experienced psychiatrist using the Mini-International Neuropsychiatric Interview. Subjects with family histories of substance abuse or major psychiatric disorders (e.g., schizophrenia or mood disorders) were excluded. The subjects were unrelated Koreans who were free of major medical problems and lived in Seoul, Korea. All participants gave informed consent, and study protocols were approved by the Institutional Review Board of Korea University Hospital (2014AN0241).

Genomic DNA was extracted from leukocytes with a QIAamp Blood Kit (Qiagen, Hilden, Germany). Genotyping for the PER3 rs57875989 variable number tandem repeat (VNTR) was performed using polymerase chain reaction (PCR) (forward primer 5’-TGT CTT TTC ATG TGC CCT TAC TT-3’ and reverse primer 5’-TGT CTG GCA TTG GAG TTT GA-3’) with the following amplification conditions: 94°C/3 min, then 38 cycles of 94°C/45 sec, 58°C/45 sec, and 72°C/1 min. Agarose gel electrophoresis was used to determine whether individuals were heterozygotic or homozygotic for the PER3 rs57875989 VNTR allele.

Genotyping for five other circadian gene SNPs was performed by high-resolution melting curve analysis. PCR was performed in a volume of 20 μL per reaction with a 96-well Bio-Rad CFX 96 Real-Time PCR system (Bio-Rad, Hercules, CA, USA). Reaction mixtures included 1.5 μL of genomic DNA as template, 200 mM of each primer [NPAS2 (rs2305160), forward 5’-GAT TTG TGG GAA CTT CTT GAG-3’ and reverse 5’-ATG AGG ACT GTT TGA TGT GT-3’; ARNTL (rs2278749), forward 5’-CAT CAT TAC AGC ACC TCC TC-3’ and reverse 5’-CCA CTT TTC CTT CCC CAT AG-3’; CLOCK (rs1801260), forward 5’-GGG AAA GTT CCA GCA GTT-3’ and reverse 5’-ATC CAG GCA CCT AAA ACA-3’; PER2 (rs2304672), forward 5’-TGT TTG CCA GCT TCG TCC-3’ and reverse 5’-GTT GGT CTC ATG TCC ACT-3’; PER2 (rs934945), forward 5’-ATC TCA AGT TTG TTA CTG ATT TCT C-3’ and reverse 5’-CTC TTC GAT CCT GTG ATT CAA-3’], 1X SsoFast^TM EvaGreen^® Supermix (Bio-Rad Laboratories Inc., Hercules, CA, USA), and sterile H₂O. The amplification protocol started with 98°C/3 min followed by 39 cycles of 98°C/10 sec and 58°C/20 sec. After an initial step of 95°C/10 sec and 65°C/10 sec, melting curves were generated from 65°C to 95°C in increments of 0.3°C/cycle. Melting profiles were analyzed with Precision Melt Analysis^TM Software (Bio-Rad Laboratories Inc., Hercules, CA, USA).

Obsessive behaviors were measured by a single administration of a 30-item Korean translation of the Maudsley Obsessive-Compulsive Inventory (MOCI), one of the psychological tests most commonly used for assessing obsessive and compulsive symptoms [26]. MOCI is a self-reporting questionnaire with dichotomized scores (“present” or “not present”), yielding a maximum MOCI total score of thirty. Based on thirty items, the following four clinical subscales have been constructed: checking (9 items), cleaning (11 items), delay (7 items), and suspicion (7 items). Four of the items contribute to the score on two of the subscales. The Korean version has shown to have good internal consistency, good test-retest reliability, good discrimination, and convergent validity [27]. All participants completed the Korean translation of the MOCI and were classified into two groups based on their total MOCI score. We determined the cutoff of the MOCI total score for the obsessive group as 17, based on two standard deviations from the mean total score in healthy controls recruited from our study because the original study did not suggest it. Among a total of 292 participants, 274 were classified into the non-obsessive group and 18 into the obsessive group.

Statistical analyses were carried out with IBM SPSS 22.0 for Windows (IBM Corporation, Armonk, NY, USA) and SNPAlyze (version 7, DYNACOM, Chiba, Japan). The presence of Hardy-Weinberg equilibrium, linkage disequilibrium, and a case-control study using dominant, recessive, and genotypic hereditary models were conducted using SNPAlyze software. Comparison of MOCI scores between the 3 genotype groups was performed using analysis of variance (ANOVA), followed by pairwise comparison using Scheffe’s method. In case of nonparametric groups, the Kruskal-Wallis test was used instead. The results of analysis were considered at nominal significance (p<0.05).

Among the total 292 participants, 187 were male and 105 were female. The mean (mean±standard deviation) age and total MOCI score were 22.92±2.26 and 7.86±5.03, respectively. The non-obsessive group consisted of 178 men and 96 women. The mean clinical characteristics of the non-obsessive group were as follows: age 22.95±2.29, body mass index (BMI) 21.42±3.41, and years of education 15.6±1.47, respectively. Among the 18 participants in the obsessive group, 9 were male and 9 were female. Their mean clinical characteristics were as follows: age 22.50±1.76, BMI 21.73±2.90, and years of education 15.2±1.51. No significant difference was found in age, BMI, or years of education between the two groups. The scores of total MOCI and its subscales (checking, cleaning, suspicion, and delay) in the non-obsessive group are shown in Table 1. The total and subscale scores were significantly higher in the obsessive group than in the non-obsessive group (p<0.001).

The VNTR and all SNPs analyzed in this study were at Hardy-Weinberg equilibrium and had minor allele frequencies over 5%. In the present study, six circadian gene polymorphisms (PER3 rs57875989, NPAS2 rs2305160, ARNTL rs2278749, CLOCK rs1802160, PER2 rs2304672, PER2 rs934945) were analyzed in 292 participants. The distribution of allele and genotype frequencies of the VNTR and SNPs in the obsessive and non-obsessive groups is presented in Table 2.

The genotype frequencies of PER3 rs57875989 in the obsessive and non-obsessive groups are shown in Table 2. The association analysis results indicate a significant difference in PER3 rs57875989 genotype frequencies between the two groups under a recessive hereditary model [Akaike’s information criterion (AIC)=2.1818, p=0.005175] and a genotypic hereditary model (AIC=0.2473, p=0.0195). However, after multiple comparison correction, only the recessive model was significant [False Discovery Rate (FDR) q-value=0.0327] (Table 2).

The genotypic frequencies of ARNTL rs2278749 in the obsessive and non-obsessive groups are also shown in Table 2. The association analysis results indicate a significant difference of ARNTL rs2278749 genotype frequencies under a recessive hereditary model (AIC=5.7245, p=0.0196), but it is not significant after multiple comparison correction (FDR q-value=0.1239). For the other circadian gene SNPs (NPAS2 rs2305160, CLOCK rs1801260, PER2 rs2304672, PER2 rs934945), the frequencies of these SNPs indicate no significant difference between the obsessive and non-obsessive groups based on all three hereditary models (p<0.05).

For PER3 rs57875989 and ARNTL rs2278749, we analyzed the differences in MOCI total score and its subscales between the genotypes. The mean scores of total MOCI and its subscales (checking, cleaning, suspicion, and delay) in PER3 rs57875989 genotypes are shown in Table 3. No significant differences in MOCI total score and its subscales between the genotypes of PER3 rs57875989 were found.

In the case of ARNTL rs2278749, the mean scores of total MOCI and its subscales (checking, cleaning, suspicion, and delay) are also shown in Table 3. ANOVA shows significant differences in checking and cleaning subscales (p<0.05) between the genotypes of ARNTL rs2278749. In subgroup analysis, the checking subscale scored significantly higher in the CC genotype than the CT genotype.