Molecular analyses of circadian gene variants reveal sex-dependent links between depression and clocks

An extensive literature links circadian irregularities and/or sleep abnormalities to mood disorders. Despite the strong genetic component underlying many mood disorders, however, previous genetic associations between circadian clock gene variants and major depressive disorder (MDD) have been weak. We applied a combined molecular/functional and genetic association approach to circadian gene polymorphisms in sex-stratified populations of control subjects and case subjects suffering from MDD. This approach identified significant sex-dependent associations of common variants of the circadian clock genes hClock, hPer3 and hNpas2 with major depression and demonstrated functional effects of these polymorphisms on the expression or activity of the hCLOCK and hPER3 proteins, respectively. In addition, hCLOCK expression is affected by glucocorticoids, consistent with the sex-dependency of the genetic associations and the modulation of glucocorticoid-mediated stress response, providing a mechanism by which the circadian clock controls outputs that may affect psychiatric disorders. We conclude that genetic polymorphisms in circadian genes (especially hClock and hPer3, where functional assays could be tested) influence risk of developing depression in a sex- and stress-dependent manner. These studies support a genetic connection between circadian disruption and mood disorders, and confirm a key connection between circadian gene variation and major depression.

, A priori power calculations for various effect sizes. Sample size refers to number of cases. Effect size is measured as the odds ratio from a logistic regression analysis of the relationship between genotype at a SNP with minor allele frequency 0.2 and susceptibility to MDD. Estimated power for various effect sizes is indicated by colored lines. Methods are as in Figure 1. Levels of firefly luciferase activity were normalized with the Renilla luciferase reporter (pCMV-Rluc). The activity in samples transfected with the reporter constructs containing the common 3'UTR allele (pCl-Luc-T or pSV40-luc-T) was set as 1.0. Results are shown as mean±SEM of two independent experiments (left panel, total n = 12; right panel, total n = 8). *p < 0.05 by two-tail unpaired t-test. The P SV40 ::FLuc::3'UTR reporter was kindly provided by Dr. Malcolm von Schantz.   Figure 1. hClock expression is modulated by dexamethasone and the rs1801260 SNP. Schematic shows a glucocorticoid response element (GRE) in the SNP region of 3' UTR of the hClock gene. In this figure, the minor allele (rs1801260 SNP) is "C" (indicated by yellow font) and the major allele is "T" (indicated by green font). The consensus motifs of activated vs. repressed GREs are based on the analyses of Kuo and coworkers (1), and these portions of the figure are modified from that publication. Immunoblotting was used to assess the transfected expression of Flag-hPER3 in HEK 293 after CHX treatment (GAPDH served as loading control, WT = wild-type, MT = rs228697 variant). Two independent immunoblots are shown in panel A, and quantification of Flag-hP3 protein expression normalized with GAPDH for each experiment is shown in panel B (for both Flag-hPer3-WT and Flag-hPer3-MT, the expression at the first time-point {0.1h} after CHX treatment was set as 1).  1 Variable is not normally distributed (Shapiro--Wilk test P--value < 0.05). P--value presented was determined using the Mann--Whitney U test. 2 Mean (SD) for categorical variable 3 P--value presented was calculated using standard Two --Sample Test of Proportions (Z Test)  2 Minor Allele Frequency in cases or controls 3 Hardy--Weinberg Equilibrium P--value in cases or controls determined using (----hardy2) option in PLINK  2 Minor Allele Frequency in cases or controls 3 Hardy--Weinberg Equilibrium P--value in cases or controls determined using (----hardy2) option in PLINK  2 Minor Allele Frequency in cases or controls 3 Hardy--Weinberg Equilibrium P--value in cases or controls determined using (----hardy2) option in PLINK  GT  TT  11  rs70965441  ARNTL  CC  CA  AA  11  rs10830963  MTNRIB  GG  GC  CC  11  rs1562444  MTNRIB  AA  AG  GG  12  rs10548381  ARNTL2  TT  TA  AA  12 rs7137588 ARNTL2 CC CG GG 1 For Genotype distributions; 11--Homozygote (Minor Allele), 12--Heterozygotes, 22--Homozygote (Major Allele) Table S4. Summary of non-significant Chi-square analysis results for association with MDD in Combined Dataset rs7137588 ARNTL2 ALLELIC 0.682 ----1 Describes the type of Chi--Square test performed; genotypic or allelic 2 P--values are from Chi--Square test performed in the combined dataset *Fisher's Exact P--value or One--sided Fisher's Exact P--value shown

Molecular Analyses of Circadian Gene Variants Reveal Sex-dependent Links Between
Depression and Clocks Figure S1, A priori power calculations for various effect sizes. Sample size refers to number of cases. Effect size is measured as the odds ratio from a logistic regression analysis of the relationship between genotype at a SNP with minor allele frequency 0.2 and susceptibility to MDD. Estimated power for various effect sizes is indicated by colored lines.       Table S5.

Participants
In an effort to aid in the advancement of genetic research directed at severe neuropsychiatric disorders, the National Institute of Mental Health (NIMH) funded the Human Genetics Initiative (HGI, see acknowledgements below). 2 The HGI compiled a database of clinically diagnosed MDD pedigrees, including MDD cases and affected relatives (Depression 2.0, NIMH Major Depression Genetics Initiative). All participants were administered the Diagnostic Interview for Genetic Studies (DIGS v2.0) by trained interviewers. Probands who self-reported as Caucasian and carried a clinical diagnosis of subtype 1 MDD, also termed melancholia, were selected from this dataset for our study. We identified participants with past or current alcohol or drug abuse based on positive answers to specific questions from the DIGS v2.0: (i) the subject experienced at least 3 positive symptoms of abuse occurring in the same 12 month period; OR (ii) two of the positive symptoms persisted for over a month or occurred over a longer period of time. In the DIGS v2.0, positive symptoms are described for alcohol abuse in starred (*) questions 13-33 in the 'Alcohol Abuse and Dependence' section and for drug abuse in starred (*) questions 19-33 in the 'Drug Abuse and Dependence' section.
The HGI also collected a broad-based control sample to aid in case-control association studies focused on neuropsychiatric disorders. These controls were phenotyped using an online lifetime version of the Composite Instrument for Diagnostic Interviewing Short Form (CIDI-SF); an online clinical self-report assessment of MDD and several other psychiatric conditions that have been previously described. 3 Because this dataset was to function as a control database for a wide number of disorders, individuals who indicated the presence of mental disorders or substance abuse were not excluded from the HGI control dataset. Investigators were advised to screen the HGI control sample prior to analysis to ensure that none of the controls that they included in their sample possessed their disorder of interest. In our analyses all controls whose responses indicated the presence of lifetime MDD or self-reported as other than Caucasian were excluded. The presence of MDD in the HGI control dataset was identified by either (i) selfreport of clinical diagnosis of MDD or (ii) affirmation of being depressed for two weeks or more in a row (questions A1A, A1B) and experiencing a total of 5 symptoms in the MDD category (questions A1A, A1B, A2, A3, A4). Controls were also excluded if they declined to answer a survey question that was critical to determining MDD diagnosis. Participants with past or current alcohol or drug abuse were also identified in the HGI dataset. The presence of alcohol abuse was indicated by affirmation that (i) drinking interfered with work, school, or home life (question G2) at least 3-5 times (question G2A) and (ii) participants were either under the influence of alcohol in a situation where they could get hurt (question G3) or the participant had emotional or psychological problems from using alcohol (question G4). The presence of drug abuse was indicated by affirmation that (i) the participant used a drug either without a doctor's prescription, in larger amounts than prescribed, or for longer than prescribed (questions H1_1-H1_9) and (ii) drug use interfered with work, school, or home life (question H3) at least 3-5 times (question H3) and (iii) participants were either under the influence of drugs in a situation where they could get hurt (question H4) or the participant had emotional or psychological problems from using drugs (question H5).

Informed consent was obtained for all HGI dataset and NIMH Major Depression Genetics
Initiative participants prior to DNA collection.

Genotyping
DNA samples from the HGI and NIMH Major Depression Genetics datasets were genotyped for 32 polymorphisms within circadian clock genes that are common in the Caucasian population and have potential functional significance based on well-characterized roles in regulating circadian rhythms and/or sleep disorders. 4,5 Single Nucleotide polymorphism screening and genotyping. All SNPs (except the variable number of tandem repeat in exon 18 of hPer3, AB047536) were genotyped by Sequenom screening (Sequenom Inc. San Diego Ca). DNA samples were loaded in 384-well plates and subjected to PCR primer extension. The single base extension products were detected with a Mass Array MALDI-TOF mass spectrometer on a Sequenom iPLEX genotyping platform. The sequence difference at the single nucleotide level was detected as an allele-specific difference in mass between extension products. The difference in mass was correlated to a specific genotype. The call rate threshold for all the SNPs analyzed by Sequenom was at least 95%.
Variable Number of Tandem Repeat polymorphism screening and genotyping. PCR genotyping for the hPer3 VNTR (AB047536) utilized primers as previously described. 6 Platinum PCR Supermix (Invitrogen, Carlsbad, CA, USA) was used to produce the 475 vs. 550 bp amplicon from genomic DNA using the following parameters: 95°C for 90 sec, then 35 cycles of 95°C for 30 sec, 55°C for 30 sec and 72°C for 30 sec, followed by a final extension at 72°C for 7 min and a 4°C hold. After separation by electrophoresis in a 2% agarose gel and visualized by ethidium bromide staining, the PCR products distinguished between the 5-repeat allele (550 bp) and the 4-repeat allele (475 bp).

Association Analyses
After genotyping, the samples were assessed for genotyping efficiency (SNP and samplelevel) and deviation from Hardy-Weinberg Equilibrium (HWE) as measures of quality control (QC). The following QC exclusion criteria were applied to the our dataset: (1) SNP genotyping efficiency < 95%, and (2) sample genotyping efficiency < 95%, HWE test p-value < 0.001. All QC testing was performed using PLINK. 7 After quality control, 18 genetic variants remained for analysis in our dataset. Prior to beginning analysis, we performed power calculations using PS Power 8 to determine whether our study was adequately powered to detect common variants with modest effects on MDD. A priori power calculations indicated that our study was well powered ( ≥ 80%) to detect common variants (minor allele frequency ≥ 0.2) with modest effect sizes, and α= 0.5 ( Figure S1).
To assess statistical association between the analyzed SNPs and MDD, single locus analyses were performed in the combined dataset (n = 592 cases, 776 controls). Allelic and genotypic Chi-square analyses were performed, using PLINK. 7 Where appropriate, Fisher's Exact test of association was performed. To investigate the possibility of potentially differing effects of the analyzed SNPs on MDD between the genders, sex separated analyses were also performed (males, n= 94 cases, 253 controls; females, n = 498 cases, 523 controls) for all markers that showed a p-value < 0.2 in Chi-square genotypic or allelic tests in the full dataset. Correction for multiple testing of Chi-square p-values was performed using False Discovery Rate (FDR) (q = 0.1). 9 An alternative single locus analysis, Prevalence-based Association testing (PRAT), based on the principles of HWE has been shown in simulation studies to be more powerful than Chisquare testing under some genetic models and was also performed to provide independent evidence for association. 10 PRAT uses an estimated population allele frequency to generate expected genotype frequencies, taking into account the population prevalence of the selected disease. These expected genotype frequencies are then used to calculate expected numbers of individuals in each genotype class for the cases and controls separately, under the assumption of no association between SNP and phenotype, and Chi-square statistics are generated from these. The resulting PRAT p-value is then assessed for cases and controls independently. Significant PRAT p-values for either cases or controls, or both, are indicative of an association. 10 Prevalence of MDD was set at 7% for PRAT analysis as an approximation of the Caucasian population prevalence of MDD (NIMH statistic 2005), and 1000 permutations of affection status were performed to correct for multiple testing. Simulation data has demonstrated that minor misspecification of prevalence rates do not affect results significantly. 10 PRAT was performed in the combined and sex separated datasets as a part of the PLATO software package. Therefore, to confirm significant SNPs by a different and methodologically independent analysis, we evaluated the associations between MDD and our assayed genetic variants using Prevalence-based Association Testing (PRAT) 10 (Table S5). PRAT analyses in the full dataset provided additional evidence for the significant association between four variants previously identified using χ 2 analyses (rs228697 and rs17031614 in hPER3, and rs485133 and rs34705978 in hNPAS2) (Table S5). Additionally, PRAT analysis in the male subset revealed a novel association between MDD and a single variant in Arntl2, rs7137588 (p-value = 0.008) (Table S5). While none of the SNPs reached significance at the p ≤ 0.05 level in the male case-only dataset in the PRAT analysis, rs1801260 of hClock was the only SNP in the male caseonly subset that was marginally significant (p = 0.081, Table S5).
To assess effect size of statistically significant SNPs discovered using the previous two methods, logistic regression, and where appropriate Exact logistic regression, using an additive model was performed using the STATA version 10.0 software package (StataCorp, College Station, TX). In cases where sample size was < 5 in any genotypic category, but Exact logistic regression was computationally infeasible, standard logistic regression using a dominant model for the risk allele was employed.

Functional Analyses
Construction of luciferase reporters and expression vectors: The P SV40 ::FLuc::3'UTR firefly luciferase reporter for the 3'UTR of hClock (containing the major allele T) was provided by Dr. Malcolm von Schantz. 11 To create a minor allele reporter, a T to C mutation at position 3111 (rs1801260) was introduced by site-directed mutagenesis (Stratagene) and verified by sequencing. To create P Clock ::FLuc plasmids so that the luciferase expression is under the control of the endogenous hClock promoter, a fragment extending from the hClock upstream region (-1908 from the transcription start site) to the first intron (+101 from the transcription start site) was amplified by PCR from human cell genomic DNA and was used to replace the SV40 promoter of the P SV40 ::FLuc-T/C plasmids. The hPer3 expression plasmid under the control of the CMV promoter (P CMV ::hPer3) was provided by Dr. Joon-Kyu Lee. 12 A G to C mutation (rs228697) in the coding region of hPer3 was introduced to change the proline at position 856 to an alanine residue by site-directed mutagenesis (Stratagene) that was verified by sequencing.
Cell culture and functional assay experiments: BMAL1/CLOCK transactivation of E-box containing promoters (and PER3 repression thereof) was assessed by a transient transfection assay. 13,14 Cell lines {HEK-293T, HEK-293 (ATCC), HepG2 (a gift from Dr. Richard O'Brien, Vanderbilt Medical Center), COS-1 (a gift from Dr. Stephen Brandt, Vanderbilt Medical Center)} were maintained at 37°C, 5% CO2 for 24 h in a 24-well plate in Dulbecco's modified Eagle's medium supplemented with 10% fetal bovine serum and 1% penicillin, streptomycin and glutamine solution. All four cell lines have been authenticated within the past two years and tested for mycoplasma contamination.
Because our association analyses indicated sex-dependent effects, we also wanted to test cell culture lines that were originally derived from both males and females. The sex of the cell culture lines we tested is as follows: Female derivation: HEK293T, U2OS; Male derivation: HepG2, COS-1. Methodology: 1. Using the P Clock ::FLuc::3'UTR or P SV40 ::FLuc::3'UTR reporters to characterize the expression levels affected by the Clock SNP (rs1801260), the cells were transiently transfected with one of the reporters (T or C allele) and a control vector (P CMV ::Rluc) using Lipofectamine 2000 (Invitrogen). After transfection (24 h), the cells were harvested and the luciferase activity was measured using the Dual Luciferase Reporter Assay kit (Promega).

2.
To determine the sensitivity of the P Clock ::FLuc::3'UTR or P SV40 ::FLuc::3'UTR reporters containing the Clock SNP (rs1801260) in response to glucocorticoid stimuli, HepG2 cells were transfected with the same of amount of reporter plasmid and the P CMV ::Rluc control vector. After