Association study of PDE4B gene variants in scandinavian schizophrenia and bipolar disorder multicenter case–control samples†
How to Cite this Article: Kähler AK, Otnæss MK, Wirgenes KV, Hansen T, Jönsson EG, Agartz I, Hall H, Werge T, Morken G, Mors O, Mellerup E, Dam H, Koefod P, Melle I, Steen VM, Andreassen OA, Djurovic S. 2010. Association Study of PDE4B Gene Variants in Scandinavian Schizophrenia and Bipolar Disorder Multicenter Case–Control Samples. Am J Med Genet Part B 153B:86–96.
Abstract
The phosphodiesterase 4B (PDE4B), which is involved in cognitive function in animal models, is a candidate susceptibility gene for schizophrenia (SZ) and bipolar disorder (BP). Variations in PDE4B have previously been associated with SZ, with a suggested gender-specific effect. We have genotyped and analyzed 40 and 72 tagging single nucleotide polymorphisms (tagSNPs) in SZ and BP multicenter samples, respectively, from the Scandinavian Collaboration on Psychiatric Etiology (SCOPE), involving 837 SZ cases and 1,473 controls plus 594 BP cases and 1,421 partly overlapping controls. Six and 16 tagSNPs were nominally associated (0.0005 ≤ P ≤ 0.05) with SZ and BP, respectively, in the combined samples or in gender-specific subgroups. None of these findings remained significant after correction for multiple testing. However, a number of tagSNPs found to be nominally associated with SZ and BP were located in a high LD region spanning the splice site of PDE4B3, an isoform with altered brain expression in BP patients. Four tagSNPs were associated with SZ in women, but none in men, in agreement with the previously reported gender-specific effect. Proxies of two nominally associated SNPs in the SZ sample were also associated with BP, but the genotypic effect (i.e., homozygosity for the minor allele), pointed in opposite directions. Finally, four SNPs were found to be associated with Positive And Negative Syndrome Scale (PANSS) positive symptom scores in a subgroup of SZ patients (n = 153) or SZ female patients (n = 70). Further studies are needed to evaluate the implicated PDE4B region of interest, for potential involvement in SZ and BP susceptibility. © 2009 Wiley-Liss, Inc.
INTRODUCTION
Phosphodiesterase 4B (PDE4B) belongs to a family of four PDE4 genes, all coding for phosphodiesterases that hydrolyze the second messenger cyclic adenosine monophosphate (cAMP) [Houslay and Adams, 2003]. The PDE4B gene encodes at least four different isoforms, each with a unique N-terminal region [Cheung et al., 2007; Murdoch et al., 2007].
PDE4B was first suggested as a risk factor for schizophrenia (SZ) through the study of a Scottish family with a balanced t(1;16) translocation that directly disrupts PDE4B on chromosome 1p31 [Millar et al., 2005]. This translocation was inherited by two cousins, one diagnosed with SZ and the other with a psychotic disorder. Subsequent case–control genetic association studies of Scottish [Pickard et al., 2007], Japanese [Numata et al., 2008b], and Caucasian and African American [Fatemi et al., 2008] samples have reported an association between PDE4B and SZ. In the study by Pickard et al. 2007, PDE4B variants were only associated with SZ in women.
The distinction of SZ and bipolar disorder (BP) as separate biological entities is currently debated, and an etiological overlap has been suggested [Moller, 2003; Owen et al., 2007]. PDE4B is an interesting candidate gene for both SZ and BP. PDE4 genes are orthologous to the dunce gene in Drosophila melanogaster, and dunce mutants show impaired learning and memory [Davis et al., 1995], which is among the most consistently reported neurocognitive deficits in both SZ [Barch, 2005] and BP [Martinez-Aran et al., 2004; Simonsen et al., 2008]. Also, the selective PDE4-inhibitor Rolipram, has been shown to have antidepressant effects in humans [Zhu et al., 2001], as well as antipsychotic-like behavioral effects in mice [Kanes et al., 2007] and rats [Siuciak et al., 2007]. The expression of PDE4B isoforms in postmortem brain tissue from patients with SZ or BP has been shown to differ compared with controls [Fatemi et al., 2008]. Furthermore, all four reported PDE4B isoforms have been demonstrated to interact with Disrupted-in-schizophrenia-1 (DISC1) [Millar et al., 2005; Murdoch et al., 2007], a protein encoded by DISC1 which has been identified as a susceptibility gene for SZ and BP in several studies [Chubb et al., 2008].
We investigated the potential involvement of PDE4B in SZ and BP, using gene-wide genotyping of tagging single nucleotide polymorphisms (tagSNPs) in Scandinavian multicenter case–control samples.
MATERIALS AND METHODS
Sample Description
The schizophrenia case–control sample
The SZ association study was based on three independent case–control samples from Norway, Sweden, and Denmark, included in the Scandinavian Collaboration on Psychiatric Etiology (SCOPE). A total of 837 SZ spectrum cases (SZ (n = 734), schizoaffective disorder (SZA) (n = 87), schizophreniform disorder (SZPH) (n = 16)), and 1,473 control subject samples were successfully genotyped. The Norwegian patients had been diagnosed with SZ (n = 124), SZA (n = 31), or SZPH (n = 8) disorder, according to DSM-IV using Structural Clinical Interview for DSM-IV (SCID), the Danish patients with SZ (n = 388) or SZA (n = 31) according to ICD-10, and the Swedish patients with SZ (n = 224), SZA (n = 25), or SZPH (n = 8), according to DSM-III-R/DSM-IV. There is high concordance between the ICD-10 and DSM systems (pairwise concordance rate (CR) > 0.70, κ > 0.70) [Jakobsen et al., 2006]. The patient and control samples are described in more detail elsewhere [Hansen et al., 2007; Kahler et al., 2008]. Since the vast majority of the patients included in the SZ spectrum sample were diagnosed with SZ, further analysis of diagnostic subgroups were not performed due to low statistical power, and for simplicity, we generally refer to schizophrenia/SZ as the clinical phenotype throughout the text.
The bipolar case–control sample
The BP association study was based on two independent case–control samples from Norway and Denmark. A total of 594 BP cases and 1,421 control samples were successfully genotyped. The Norwegian patients had been diagnosed with bipolar disorder type I (BPI) (n = 125), bipolar disorder type II (BPII) (n = 80), and BP not otherwise specified (NOS) (n = 13), according to DSM-IV using SCID. The Danish patients had been included all over Denmark (1996–1998) (n = 161), or in the Copenhagen area by the Danish Psychiatric Biobank (2002–2007) (n = 215). The first patient group had been diagnosed with SCAN [Wing et al., 1998] interviews fulfilling a best estimate diagnosis of bipolar affective disorder (n = 81) and BPI (n = 80), according to the ICD-10-DCR [WHO, 1993] and the DSM-IV, respectively. The latter group was clinically diagnosed with bipolar affective disorder according to ICD-10-DCR [WHO, 1993]. We generally refer to Bipolar disorder/BP as the clinical phenotype throughout the text. The Norwegian healthy controls (n = 220) are described in more detail elsewhere [Hansen et al., 2007; Kahler et al., 2008], and a subset (n = 152) are overlapping with controls in the SZ case–control sample. The Danish controls were distinct from those in the SZ case–control sample, but recruited as previously described (n = 1,133) [Hansen et al., 2007], or included as selected controls screened for psychiatric disease in a previous study (n = 68) [Mellerup et al., 2001].
The Norwegian Scientific-Ethical Committees, the Norwegian Data Protection Agency, the Danish Ethical Committees, the Danish Data Protection Agency, the Ethical Committee of the Karolinska Hospital, the Stockholm Regional Ethical Committee and the Swedish Data Inspection Board approved the respective parts of the study. All patients have given written informed consent prior to inclusion into the project.
SNP Selection and Genotyping
To evaluate if PDE4B variants are associated with SZ and BP, a structured gene-wide approach was used, by genotyping tagSNPs. The tagSNPs were selected at the HapMap website (www.hapmap.org), based on the CEU population, using pair-wise tagging, with r2 ≥ 0.8 [de Bakker et al., 2005] (www.hapmap.org; HapMap Data Release 21 for the SZ study, and Release 22 for the BP study). The assumed northern and western European ancestry of the CEU population has recently been genetically confirmed [Lao et al., 2008]. PDE4B (NM_001037341) is a large gene, spanning 582.1 kb, with ∼450 SNPs with minor allele frequency (MAF) ≥ 5% (HapMap Data Release 23a). As a first screen of the most common SNPs in the SZ sample, a MAF ≥ 20% was used as the cut-off when choosing tagSNPs. The tagSNPs genotyped in the BP sample were picked independently in a separate genotyping project, using MAF ≥ 5% to cover most of the common variation.
Genomic DNA was extracted from whole blood, and both the SZ and BP samples were genotyped as part of two larger genotyping projects, using the GoldenGate 1536plex assay (Illumina, Inc., San Diego, CA) on the Illumina BeadStation 500GX at the SNP Technology Platform, Uppsala University, Sweden (www.genotyping.se), accredited by the Swedish accreditation agency SWEDAC, and approved according to a quality system based on the international SS-EN ISO/IEC 17025 standard. There were only two duplicate errors in 85,674 duplicate genotype calls (reproducibility of 99.998%) and five duplicate errors in 124,684 duplicate genotypes calls (reproducibility of 99.996%) for the SZ and BP genotyping projects, respectively.
The actual tagging efficiency of successfully genotyped tagSNPs was calculated using HapMap Data Release 21 at the Tagger website (www.broad.mit.edu/mpg/tagger/server.html).
Statistical Analysis
All SNPs were tested for departure from Hardy Weinberg-equilibrium in cases and controls separately, using the exact chi-square test implemented in PLINK (version 1.04; http://pngu.mgh.harvard.edu/purcell/plink/) [Purcell et al., 2007]. Potential SNPs with P < 0.001 in controls were considered in Hardy Weinberg disequilibrium (HWD) and excluded.
To estimate the level of heterogeneity between the three Scandinavian subpopulations in the SZ case–control sample, an overall fixation index FST has previously been calculated for a larger SNP set, using the control samples from Norway, Denmark, and Sweden, showing no evidence of population stratification [Kahler et al., 2008]. In addition, for the present study we calculated the gene-based FST for PDE4B, in both the SZ and BP control sample sets, as implemented in Arlequin 3.1 [Excoffier et al., 2005].
Allelic and genotypic single SNP association tests, as well as a sliding-window haplotype analysis, were performed with UNPHASED (version 3.0.13) [Dudbridge, 2008]. To account for potential population stratification, the population status was included as a confounder with discrete levels. Pairwise LD (D' and r2) and LD blocks were estimated in Haploview 4.1 [Barrett et al., 2005], the latter using the solid spine definition for the most extensively genotyped BP sample, acknowledging that such an estimation is limited when based on tagSNPs. Haplotype effects were examined by global and individual haplotype association tests, including 2-, 3-, and 4-marker sliding window haplotypes. Estimated haplotypes with a frequency below 0.05 in both cases and controls were considered rare and excluded from the association tests. We set the nominal significance threshold to P = 0.05. For nominally associated SNPs, the P-values and odds ratios (ORs) for each individual genotype compared to the other genotypes pooled together, was calculated. Also, ORs for the risk allele and the individual genotypes with the largest effect size were determined in each population separately, using UNPHASED. Each test was corrected for the multiple SNPs or haplotypes assessed, using 10,000 permutations.
Because a gender-specific effect of PDE4B has previously been reported [Pickard et al., 2007], the above analyses were also performed on data subdivided on the basis of gender.
Association With Clinical Symptoms
The diagnosis of SZ is based on the presence of positive symptoms (e.g., delusions and hallucinations), also frequently observed in BP during manic episodes, and/or negative symptoms (e.g., flattening of affect and lack of volition and drive), commonly observed in BP during depressive episodes. A subset of Norwegian SZ patients (n = 153; 54.2% men, 45.8% women), and BP patients (n = 128; 39.8% men, 60.2% women), were symptomatically evaluated with the Positive And Negative Syndrome Scale (PANSS) [Kay et al., 1987]. Symptom scores were tested for potential association with the SNPs nominally associated with SZ or BP diagnosis. Assessments were interview based, and performed by experienced MDs or psychologists. The SZ sample was moderately symptomatic with PANSS positive and negative sum scores being 15.3 ± 5.8 and 15.1 ± 6.1, respectively. The BP sample was less symptomatic, with PANSS positive and negative sum scores being 9.7 ± 2.8 and 10.7 ± 4.0, respectively. Due to departure from normal distribution, the genotype–phenotype association analysis was performed using the non-parametric Kruskal–Wallis test, implemented in SPSS (version 16.0). The genotype distribution for each SNP was used as grouping variable, and PANSS positive and negative sum score as dependent variable. The nominal level of significance was set to P = 0.05, and P-values were Bonferroni corrected for the number of SNPs assessed in each tested group.
RESULTS
Genotyping and tagSNP Coverage
Forty out of 44 selected tagSNPs in PDE4B (NM_001037341) were included in the SZ case–control study, based on probability of successful assay design. All 40 PDE4B tagSNPs had a call rate >96.7%, and the total genotyping rate was 99.51%. Genotype counts for all SNPs are given in Supplementary Table I. No tagSNPs had genotype distributions in HWD in controls (P > 0.001); lowest P in cases was 0.001 for rs12136401. The 40 tagSNPs had 86% coverage with r2 ≥ 0.8 (mean r2 = 0.93 and minimum r2 = 0.40), of the 318 SNPs included in the Hapmap data at the Tagger website (MAF ≥ 20%).
Seventy-three out of 76 selected tagSNPs in PDE4B were successfully genotyped in the BP case–control sample. TagSNP rs6692281 was excluded because only one minor allele was present. The remaining tagSNPs had a call rate > 94.8%, and the total genotyping rate was 99.66%. Genotype counts for all SNPs are given in Supplementary Table II. None of the tagSNPs had genotype distributions in HWD in either cases or controls (P > 0.001). The 72 tagSNPs had 92% coverage with r2 ≥ 0.8 (mean r2 = 0.94 and minimum r2 = 0.26), of the 449 SNPs included in the Hapmap data at the Tagger website (MAF ≥ 5%).
Population Stratification
The gene-based FSTs were 0.00004 and −0.00008, for PDE4B in the SZ and BP sample, respectively, showing no evidence of stratification between the control populations in each of the two case–control samples.
Single tagSNP Association Data
SNPs nominally associated in the genotype- and/or allele-based test for the SZ and BP case–control sample, are presented in Table I. Gender-specific results are given for the tagSNPs associated with disease only in females or males. An overview of the ORs for the most associated individual genotypes for the tagSNPs in Table I are given in Table II.
| SNP | Total number cases/controls | Minor allele | Risk allele | HWEtotal sample | Case frequency | Control frequency | Genotype teststratifieda | Allele teststratifieda | Allele testsample-separated | ||||||
|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|---|
| Total sample | Denmark | Norway | Sweden | ||||||||||||
| P | OR | P | OR | P | OR | P | OR | ||||||||
| Total schizoprenia sample | |||||||||||||||
| rs1892346 | 835/1,473 | T | A | 0.63 | 0.588 | 0.556 | 0.045 | 0.014 | 1.17 (1.03–1.32) | 0.033 | 1.20 (1.01–1.41) | n.s. | 1.06 | n.s. | 1.18 |
| rs596662 | 832/1,465 | C | A | 0.08 | 0.654 | 0.624 | 0.036 | 0.027 | 1.16 (1.02–1.31) | 0.007 | 1.26 (1.07–1.50) | n.s. | 1.07 | n.s. | 1.00 |
| Schizophrenia females [males] vs. controlsb | |||||||||||||||
| rs12088813 | 348/623 | C | A | 0.09 | 0.772 | 0.724 | 0.019[n.s.] | 0.025[n.s.] | 1.28 (1.03–1.60) | n.s. | 1.17 | n.s. | 1.43 | 0.089 | 1.47 |
| rs3009872 | 348/624 | C | T | 0.03 | 0.611 | 0.571 | 0.043[n.s.] | 0.088[n.s.] | 1.18 | n.s. | 1.14 | n.s. | 1.35 | n.s. | 1.15 |
| rs1937450 | 348/623 | T | G | 0.27 | 0.586 | 0.539 | 0.046[n.s.] | 0.032[n.s.] | 1.23 (1.02–1.49) | 0.067 | 1.26 | n.s. | 1.29 | n.s. | 1.11 |
| rs2455032 | 347/625 | T | G | 0.91 | 0.667 | 0.610 | 0.018[n.s.] | 0.014[n.s.] | 1.28 (1.05–1.56) | n.s. | 1.21 | 0.063 | 1.54 | n.s. | 1.27 |
| Total bipolar sample | |||||||||||||||
| rs7552762 | 594/1403 | G | G | 0.13 | 0.116 | 0.098 | 0.020 | 0.059 | 1.24 (1.00–1.55) | n.s. | 1.25 | n.s. | 1.21 | ||
| rs12080701 | 594/1418 | G | G | 0.13 | 0.116 | 0.098 | 0.019 | 0.059 | 1.24 (1.00–1.55) | n.s. | 1.26 | n.s. | 1.20 | ||
| rs11208776 | 593/1415 | A | A | 0.28 | 0.473 | 0.437 | n.s. | 0.045 | 1.15 (1.00–1.33) | n.s. | 1.10 | 0.043 | 1.32 (1.02–1.71) | ||
| rs6421482 | 591/1414 | A | A | 0.35 | 0.450 | 0.416 | n.s. | 0.047 | 1.15 (1.00–1.33) | n.s. | 1.14 | n.s. | 1.20 | ||
| rs17452121 | 594/1419 | G | G | 0.61 | 0.101 | 0.085 | 0.005 | 0.049 | 1.27 (1.01–1.60) | 0.051 | 1.31 (1.02–1.69) | n.s. | 1.15 | ||
| rs2186122 | 588/1407 | A | A | 0.35 | 0.452 | 0.409 | 0.053 | 0.017 | 1.19 (1.03–1.37) | 0.053 | 1.18 | n.s. | 1.22 | ||
| rs1937451 | 594/1420 | T | T | 0.18 | 0.179 | 0.152 | 0.042 | 0.034 | 1.22 (1.02–1.47) | n.s. | 1.20 | n.s. | 1.30 | ||
| rs12140107 | 593/1421 | G | A | 0.29 | 0.875 | 0.852 | 0.008 | 0.023 | 1.27 (1.03–1.56) | 0.042 | 1.29 (1.00–1.66) | n.s. | 1.22 | ||
| rs12731764 | 590/1405 | G | A | 0.24 | 0.731 | 0.736 | 0.016 | n.s. | 1.02 | n.s. | 1.11 | n.s. | 1.17 | ||
| rs522037 | 593/1416 | G | G | 0.96 | 0.402 | 0.394 | 0.047 | n.s. | 1.07 | n.s. | 1.04 | n.s. | 1.15 | ||
| rs2144719 | 593/1419 | G | T | 0.04 | 0.608 | 0.600 | 0.044 | n.s. | 1.02 | n.s. | 1.03 | n.s. | 1.02 | ||
| Bipolar females [males] vs. controlsb | |||||||||||||||
| rs17424885 | 329/741 | A | G | 0.92 | 0.869 | 0.837 | 0.038[n.s.] | 0.038[n.s.] | 1.33 (1.01–1.75) | n.s. | 1.28 | n.s. | 1.45 | ||
| rs11208793 | 328/740 | T | T | 0.76 | 0.349 | 0.305 | 0.076[n.s.] | 0.046[n.s.] | 1.23 (1.00–1.50) | n.s. | 1.14 | 0.045 | 1.48 (1.01–2.18) | ||
| rs12142070 | [264/679] | C | T | 0.10 | 0.600 | 0.586 | n.s.[0.0049] | n.s.[n.s.] | 1.07 | n.s. | 1.20 | n.s. | 0.81 | ||
| rs7415930 | [265/678] | T | T | 0.32 | 0.168 | 0.134 | n.s.[0.040] | n.s.[0.084] | 1.29 | n.s. | 1.28 | n.s. | 1.32 | ||
| rs11208816 | [264/676] | T | T | 0.01 | 0.491 | 0.426 | n.s.[0.017] | n.s.[0.008] | 1.33 (1.08–1.63) | 0.006 | 1.41 (1.10–1.82) | n.s. | 1.13 | ||
- a Both genotype- and allele-based tests for the total Scandianvian sample are stratified by including country of sample origin as confounder.
- b Association data is given for males in brackets and for females without brackets. For those SNPs associated only in females in the total Scandinavian sample, the ORs and association results for each country separarately are given for females only and vice versa.
| SNP | Genotype frequenciesa | Genotype | Combined sampleb | Denmark | Norway | Sweden | |||||
|---|---|---|---|---|---|---|---|---|---|---|---|
| Cases | Controls | Pc | ORc | P | OR | P | OR | P | OR | ||
| Total schizophrenia sample | |||||||||||
| rs1892346 | 0.340/0.496/0.164 | 0.306/0.500/0.194 | T/T | 0.040 | 0.79 (0.63–0.99) | n.s. | 0.76 | n.s. | 0.81 | n.s. | 0.83 |
| rs596662 | 0.426/0.457/0.118 | 0.400/0.448/0.152 | C/C | 0.012 | 0.72 (0.55–0.93) | 0.005 | 0.59 (0.40–0.85) | n.s. | 0.63 | n.s. | 1.05 |
| Schizophrenia females vs. controls | |||||||||||
| rs12088813 | 0.598/0.348/0.055 | 0.506/0.437/0.0578 | A/A | 0.006 | 1.46 (1.11–1.91) | n.s. | 1.33 | n.s. | 1.41 | 0.028 | 1.86 (1.07–3.26) |
| rs3009872 | 0.376/0.468/0.155 | 0.298/0.547/0.155 | T/T | 0.015 | 1.42 (1.07–1.88) | 0.024 | 1.53 (1.06–2.21) | n.s. | 1.26 | n.s. | 1.29 |
| rs1937450 | 0.351/0.471/0.178 | 0.275/0.528/0.197 | G/G | 0.013 | 1.44 (1.08–1.92) | 0.016 | 1.58 (1.09–2.30) | n.s. | 1.14 | n.s. | 1.4 |
| rs2455032 | 0.461/0.412/0.127 | 0.368/0.485/0.147 | G/G | 0.004 | 1.48 (1.13–1.94) | n.s. | 1.36 | 0.043 | 1.87 (1.02–3.43) | n.s. | 1.49 |
| Total bipolar sample | |||||||||||
| rs7552762 | 0.789/0.192/0.020 | 0.810/0.184/0.0057 | G/G | 0.005 | 3.53 (1.40–8.90) | 0.004 | 4.26 (1.47–12.37) | n.s. | 2.04d | ||
| rs12080701 | 0.788/0.192/0.020 | 0.810/0.184/0.0056 | G/G | 0.005 | 3.56 (1.41–8.99) | 0.003 | 4.32 (1.49–12.54) | n.s. | 2.03d | ||
| rs11208776 | 0.282/0.491/0.228 | 0.310/0.507/0.184 | A/A | 0.047 | 1.28 (1.00–1.63) | n.s. | 1.21 | ||||
| rs17452121 | 0.818/0.162/0.020 | 0.835/0.159/0.0056 | G/G | 0.0005 | 4.43 (1.78–11.02) | 0.002 | 4.07 (1.59–10.38) | NA | NA | ||
| rs2186122 | 0.296/0.505/0.199 | 0.343/0.496/0.161 | A/A | 0.046 | 1.30 (1.00–1.67) | n.s. | 1.27 | ||||
| rs1937451 | 0.680/0.283/0.037 | 0.715/0.267/0.018 | T/T | 0.019 | 2.01 (1.11–3.65) | n.s. | 1.92 | n.s. | 2.32e | ||
| rs12140107 | 0.781/0.189/0.030 | 0.730/0.245/0.025 | A/G | 0.002 | 0.68 (0.53–0.87) | 0.034 | 0.73 (0.54–0.98) | 0.018 | 0.59 (0.38–0.92) | ||
| rs12731764 | 0.515/0.431/0.054 | 0.547/0.377/0.076 | G/G | 0.020 | 0.61 (0.40–0.93) | 0.002 | 0.38 (0.20–0.73) | n.s. | 1.06 | ||
| rs522037 | 0.332/0.531/0.137 | 0.367/0.479/0.155 | C/G | 0.014 | 1.28 (1.05–1.56) | 0.019 | 1.32 (1.05–1.67) | n.s. | 1.19 | ||
| rs2144719 | 0.352/0.511/0.137 | 0.374/0.453/0.173 | T/G | 0.020 | 1.26 (1.04–1.54) | n.s. | 1.19 | 0.04 | 1.48 (1.02–2.16) | ||
| Bipolar females [males] vs. controls | |||||||||||
| rs17424885 | 0.770/0.201/0.029 | 0.697/0.287/0.016 | G/A | 0.011 | 0.66 (0.48–0.91) | 0.014 | 0.62 (0.43–0.91) | n.s. | 0.78 | ||
| rs11208793 | 0.415/0.473/0.113 | 0.488/0.414/0.099 | C/C | 0.024 | 0.73 (0.56–0.96) | 0.097 | 0.77 (0.56–1.05) | ||||
| rs12142070 | 0.341/0.569/0.090 | 0.351/0.462/0.187 | T/C | [0.004] | 1.54 (1.15–2.08) | 0.015 | 1.54 (1.09–2.18) | n.s. | 1.56 | ||
| rs7415930 | 0.706/0.253/0.042 | 0.746/0.239/0.015 | T/T | [0.008] | 3.17 (1.30–7.74) | n.s. | 2.48 | NA | NA | ||
| rs11208816 | 0.261/0.496/0.242 | 0.343/0.462/0.195 | C/C | [0.006] | 0.63 (0.45–0.87) | 0.003 | 0.54 (0.36–0.81) | n.s. | 0.86 | ||
- NA, data not available due to no controls have the tested genotype.
- a Homozygote major allele/heterozygote/homozygote minor allele.
- b All patients and controls are analyzed in combination, with sample origin as a confounder.
- c P-values and ORs are given for the genotype giving the highest risk (as indicated by the lowest P-value, or if the same P-values, the largest effect size) when compared to the pooled other two genotypes.
- d Four cases, and two controls.
- e Nine cases, and four controls.
Schizophrenia sample
There were nominally significant associations between two (rs596662 and rs1892346; r2 = 0.04) out of 40 independent tagSNPs and SZ in the combined case–control sample, in both genotype- and allele-based tests. The major alleles conferred risk with an effect size larger (rs596662) or similar (rs1892346) in the separate χ2-tests for the Danish sample (Table I). However, for both SNPs, the homozygotes for the minor allele gave the largest individual genotypic effect (with OR < 1; Table II).
When genders were analyzed separately, there were nominally significant associations between rs12088813, rs3009872, rs1937450, and rs2455032 (previous ID: rs9436312) and SZ in females. All but one was nominally associated both in allele- and genotype-based tests (P < 0.046), but not in males (P > 0.61), with the major alleles conferring a risk effect in each of the three case–control samples, as well as in the combined sample (Table I). These SNPs are present in a high LD region, flanking the splice site for the PDE4B3 isoform (see Supplementary Figure 1). For all four SNPs the major allele homozygotes were nominally associated with an increased risk for SZ (ORs 1.42–1.48; Table II). However, none of the tagSNPs remained significantly associated with SZ after correction for multiple testing (10,000 permutations) within each sample set analyzed (P ≥ 0.17). No tagSNPs were associated when analyzed only in males.
Bipolar disorder sample
There were nominally significant associations between 11 out of the 72 successfully genotyped tagSNPs and BP, in allele- or genotype-based tests (0.0005 ≤ P ≤ 0.05). TagSNPs rs7552762 and rs12080701 are in complete LD (r2 = 1.0). For the strongest associated tagSNP, rs17452121, the best fitting model was recessive (homozygote G/G: OR(95%CI) = 4.43(1.78–11.02), P = 0.00052). Three tagSNPs nominally associated with BP (rs17452121, rs2186122, and rs1937451) are located close to the splice site for the PDE4B3 isoform, present in an estimated three-tagSNP LD block. Five additional tagSNPs were nominally associated only in the male (three) or female (two) subgroups (Table I). However, none of the tagSNPs remained significantly associated with BP after correction for multiple testing (10,000 permutations), within each sample set analyzed (P ≥ 0.069).
Overlapping association signals in the schizophrenia and bipolar disorder samples
All of the six nominally associated tagSNPs in the SZ sample have either been genotyped themselves (rs2455032) or by proxies in the BP sample. Two of these proxies, rs11208776 and rs2186122 (r2 = 0.93 and r2 = 1.0 for SZ tagSNPs rs1937450 and rs3009872, respectively, in HapMap CEU), were nominally associated with BP in the total sample (Table I). All four tagSNPs were present in a 48 kb region spanning the PDE4B3 splice site (see Supplementary Figure 1). However, increased risk for SZ was associated with being homozygous for the major allele, while in contrast the homozygotes for the minor allele displayed increased risk for BP.
Haplotype Association Data
Schizophrenia sample
In the total sample set, the association signal was not strengthened by consideration of 2-, 3-, and 4-SNP haplotypes. Two 2-SNP haplotypes, including either rs1892346 or rs596662 and a tagSNP in high LD (D' > 0.81), were nominally associated (P < 0.042).
In the female population, several haplotypes were nominally associated, with larger effect sizes when combining several alleles (Table III). The strongest overall and individual haplotype association results were attained when combining tagSNPs rs2455032-rs1354060-rs6588186 (Pglobal = 0.0032; Phaplotype G-A-T = 0.0080, OR = 1.62 (CI (95%): 1.12–2.35)). The G-A-T haplotype was present in 9.8% and 6.1% in cases and controls, respectively. None of these associations remained significant after correction for multiple testing (P ≥ 0.090).
| No. | SNP | Positiona | LD (r2/D') | Pglobal association | Pindividual haplotypeb | |||||
|---|---|---|---|---|---|---|---|---|---|---|
| Single | Two | Three | Four | Two | Three | Four | ||||
| 8 | rs11208769 | 66092413 | 0.34/0.96 | 0.32 | 0.10 | 0.15 | 0.063 | 0.031 | 0.026 | 0.018 |
| 9 | rs12088813 | 66119721 | 0.49/1.00 | 0.025 | 0.077 | 0.041 | 0.091 | 0.021 | 0.014 | 0.033 |
| 10 | rs3009872 | 66123421 | 0.82/0.96 | 0.087 | 0.050 | 0.10 | 0.054 | 0.035 | 0.033 | 0.031 |
| 11 | rs1937450 | 66190861 | 0.71/0.98 | 0.032 | 0.033 | 0.029 | 0.0056 | 0.030 | 0.015 | 0.023 |
| 12 | rs1392816 | 66193209 | 0.34/0.59 | 0.28 | 0.070 | 0.0092 | 0.0033 | 0.028 | 0.023 | 0.015 |
| 13 | rs2455032 | 66208160 | 0.66/0.92 | 0.014 | 0.0053 | 0.0032 | 0.0081 | 0.0090 | 0.0080 | 0.013 |
| 14 | rs1354060 | 66223425 | 0.21/0.99 | 0.25 | 0.25 | 0.32 | 0.11 | 0.13 | ||
| 15 | rs6588186 | 66259030 | 0.07/1.00 | 0.89 | 0.19 | 0.092 | ||||
| 16 | rs11208796 | 66259486 | 0.11 | |||||||
- Position, linkage equilibrium (LD) data, and P-values for global and individual association tests for nine SNPs located in a region of high LD (D'-based) flanking the PDE4B3 splice site are shown.
- a The PDE4B3 isoform splice site is located between marker 10 and 11.
- b Lowest P-value from a score test for a difference in risk between one haplotype and all the others pooled together.
Bipolar disorder sample
In the total sample set, the best 2-SNP result was obtained for haplotypes combining rs17452121 and rs2186122, although the association was similar to the single SNP results (Pglobal = 0.030, Pindividual,A-T = 0.010). The 3-SNP results did not strengthen the significance. The best overall 4-SNP result was obtained with SNPs (rs558325-rs1040716-rs11803904-rs12142015) that were not associated in the single tagSNP analysis (Pglobal = 0.0057), but none of the individual haplotypes displayed association with disease (0.15 ≤ P ≤ 0.96). In females, one 2-SNP haplotype was nominally associated, but none of the 3- or 4-SNP haplotypes. In males, four of the five nominally associated 2-SNP haplotypes are combinations of SNP rs11208816 and the four closest upstream SNPs, with the strongest individual finding for the rs937605-rs524897 combination (Pglobal = 0.035; Pindividual,C-T = 0.0056, OR(95% CI) = 1.93(1.24–2.99)).
Association With Clinical Symptoms
All tagSNPs nominally associated in the SZ- or the BP case–control sample were also analyzed for their possible association with symptom scores, except for SNPs with a genotype count ≤5 (five BP-SNPs and one SZ-SNP). SNPs with gender-specific association signals were only investigated for association with PANSS scores in the relevant gender subsample.
TagSNP rs596662 was associated with positive symptoms in the total sample (P = 0.003; Table IV), and all of the three SNPs that were analyzed in females were associated with positive symptoms (0.001 ≤ P ≤ 0.004). None of the tagSNPs nominally associated with BP were associated with positive or negative symptoms in the BP sample. However, when investigating PANSS associations for the two tagSNPs (rs2186122 and rs11208776) that serve as proxies for the female SZ single tagSNP associations (rs3009872 and rs1937450, respectively), we found that both were nominally associated with negative symptoms in the female BP subgroup (Table IV). This association was stronger in a subsample (n = 36) of BP women with a history of at least one psychotic event (P = 0.007 and P = 0.002, respectively).
| SNP | Sample | #Cases | PANSS score test results | ||
|---|---|---|---|---|---|
| PPositive score | Pnegative score | Padjusteda | |||
| rs596662 | SZ total | 153 | 0.003 | 0.28 | 0.02 |
| rs3009872 | SZ females | 70 | 0.001 | 0.67 | 0.005 |
| rs1937450 | SZ females | 70 | 0.004 | 0.94 | 0.02 |
| rs2455032 | SZ females | 70 | 0.004 | 0.97 | 0.02 |
| rs2186122 | BP females | 77 | 0.33 | 0.051 | — |
| rs11208776 | BP females | 77 | 0.14 | 0.017 | 0.24 |
- a The P-values were adjusted for the number of SNPs, and samples tested for each SNP: for the SZ sample 5 SNPs were tested; for the BP sample 12 SNPs were tested, two of these in two samples.
DISCUSSION
To investigate the potential involvement of PDE4B in the etiology of both SZ and BP, we have performed a gene-wide association study. We provide important additional genotyping data, from a homogenous Scandinavian sample, but did not find statistically significant associations after correction for multiple testing. However, the nominal associations found between PDE4B markers and SZ and BP in this study, are hypothesis generating and of interest for future studies. Firstly, there is a cluster of nominally associated tagSNPs flanking the PDE4B3 isoform splice site, indicating a region of interest, which might harbor functionally relevant variants. Secondly, we provide additional data in line with the previously reported potential gender-specific effects of PDE4B variation on disease susceptibility. Thirdly, we provide novel data suggesting an effect of PDE4B on specific SZ symptoms.
It has been suggested that there are common biological mechanisms and/or susceptibility genes for SZ and BP [Rzhetsky et al., 2007; Hennah et al., 2008]. The gene coding for the DISC1 protein, which biologically interacts with PDE4B, has been associated with both diseases [Mackie et al., 2007; Hennah et al., 2008]. Also, the binding of the dephosphorylated form of the PDE4B1 to DISC1, as well as the influence of drug-induction on their interaction [Millar et al., 2005], has contributed to the discussion whether these two genes link SZ and BP [Sawa and Snyder, 2005].
This is to our knowledge the first study reporting PDE4B tagSNPs nominally associated with both SZ and BP. All of the tagSNPs nominally associated with SZ in women, as well as eight of the tagSNPs nominally associated with BP in the total or gender subsamples are located in a high LD region (D'-based), flanking the start-site of the isoform PDE4B3 (Supplementary Figure 1). It is interesting to note that the three genotyped SZ and BP tagSNPs closest to this start-site, are nominally associated with SZ and BP, respectively.
A decrease in isoform PDE4B3 expression in cerebellum has previously been shown in postmortem tissue from patients with BP compared with controls. Furthermore, reduced expression of isoforms PDE4B2 and PDE4B4 was found in the cerebellar tissue from SZ patients [Fatemi et al., 2008]. A non-isoform-specific increase in PDE4B expression in monocytes has been reported in patients with BP compared with healthy controls [Padmos et al., 2008]. In the latter study, the expression was fourfold higher in patients treated with lithium or antipsychotics, compared with unmedicated patients. The differences in isoform-specific expression in BP and SZ might indicate a complex role of this enzyme in the susceptibility to psychiatric disease.
Among the SNPs nominally associated with SZ and located close to the PDE4B3 splice site, two were in complete or high LD in the hapmap CEU population with tagSNPs, which were genotyped and nominally associated with disease in the BP sample. However, the association showed opposing direction in SZ and BP. Specifically, an increased risk for SZ was nominally associated with being homozygous for the major allele, while in contrast the homozygotes for the minor allele were nominally associated with increased risk for BP. Possible reasons for this discrepancy could be either a true difference in PDE4B isoform related susceptibility, or false positive results.
The association results for PDE4B and SZ in the present study, as well as in a previous report [Pickard et al., 2007], contribute to the hypothesis that there are differences between men and women in the effect of this gene on SZ. This potential gender-effect is based on the nominal association of several tagSNPs with SZ in the female subgroup in both studies, and non-significant results for all tagSNPs in the male subgroups. However, these data should be interpreted with caution, since a sex-genotype-interaction has not been formally tested for [Patsopoulos et al., 2007]. A potential gender-effect for PDE4B in SZ susceptibility needs to be confirmed in a larger sample in order to properly detect possible interactions. In the BP case–control analysis the gender-specific associations were not as consistent as for SZ, with different tagSNPs being nominally associated with BP in either the female or male subgroup.
PDE4B is inhibited by Rolipram, which has antidepressant and potential antipsychotic effects [Zhu et al., 2001; Kanes et al., 2007; Siuciak et al., 2007]. In the present study, one PDE4B tagSNP was associated with positive symptom scores in the total SZ sample. Furthermore, three PDE4B SNPs were associated with positive symptoms scores among SZ women. These results withstand Bonferroni correction, based on the number of tests for association between selected tagSNPs and the clinical phenotypes. The SNPs nominally associated with SZ susceptibility might therefore be functionally linked to increased positive symptoms. The two BP tagSNPs which act as proxies for two of the SZ tagSNPs were nominally associated with negative symptom scores in the female subsample, although not significant after Bonferroni correction. Interestingly, when these two SNPs were tested for symptom score association in a smaller subsample (n = 36) of BP women with a history of at least one psychotic event, the association with negative symptom scores was stronger. Positive symptoms, but to a lesser degree negative symptoms, vary during the course of illness. Thus, the genetic association to PANSS scores could be spurious. However, the present study sample was reasonable stable receiving mostly outpatient treatment, which makes such type 1 errors less likely, and genetic associations to subgroups of SZ based on PANSS scores have been reported earlier [DeRosse et al., 2006].
At the time of choosing the tagSNPs in this study, the three other association studies investigating PDE4B and SZ had not been published [Pickard et al., 2007; Fatemi et al., 2008; Numata et al., 2008a]. Therefore, the overlap between SNPs investigated in the present and published PDE4B studies is limited. For an overview of the marker positions, and associated SNPs, for the previous and present studies, see supplementary figure 1. Two of the previous studies investigated Caucasian samples: 26 SNPs were analyzed in a Scottish sample (386 SZ cases, 368 BP cases, 455 controls) [Pickard et al., 2007], and 27 SNPs were analyzed in an American Caucasian sample (644 cases, 407 controls) [Fatemi et al., 2008]. The nominal associations in the study by Pickard et al. were only found in females, and do not overlap with the associated SNPs in either the present or the Fatemi et al. study. A hapmap-based perfect proxy (r2 = 1 in the CEU population) for the most nominally significant SNP by Pickard et al., was investigated by us, but failed to replicate. We found one SNP previously associated with SZ by Fatemi et al. to be nominally associated with BP, in the genotype- but not allele-based test. In contrast, this SNP was not associated with SZ either in our study or in the study by Pickard et al. Four of the seven associated SNPs in the study by Fatemi et al., were genotyped and not associated to SZ in our study. The two most associated SNPs were linked to the same extent in both studies, but single and haplotype test results were non-significant.
There are several possible explanations for the conflicting association findings for PDE4B and psychiatric disorders. First, the present study is larger than the previously studied SZ and BP samples, reducing the risk of type II error in comparison with previous reports. Still, the power of the allelic test in our study, calculated using the Genetic Power Calculator (pngu.mgh.harvard.edu/∼purcell/gpc; settings: D' = 1 between disease and tagSNP, an additive model with OR (homozygote risk allele) = 1.5; α = 0.05; MAF for disease and tagSNP = 5–25%), is limited to 33–84% and 28–75% for the SZ and BP sample, respectively. Second, our study is based on homogenous samples (as measured by the FST), originating from Norway, Denmark, and Sweden, which makes them well suited for genetic studies, with lower risk of type I error risk due to population stratification. Third, even when comparing Caucasian samples, suggested locus heterogeneity has previously been reported for candidate genes within psychiatric genetics, such as DISC1 [Hennah et al., 2008]. Therefore significant associations to several tagSNPs in the same gene in several samples might impose involvement of the investigated gene in disease susceptibility, despite tagSNP heterogeneity.
We cannot exclude that the present results might be due to type I error, and we have limited power to detect signals in the lower frequency range. However, the nominally associated tagSNPs in our study are not randomly distributed over the large gene. Rather, several SNPs are located in a high LD region flanking the PDE4B3 splice site. Further studies should therefore examine if there are variants in the PDE4B gene region close to the isoform PDE4B3 splice site that are involved in SZ and BP susceptibility. Furthermore, investigations for potential epistatic interaction with DISC1 would be desirable.
Acknowledgements
We thank patients and controls for their participation in the study, and the health professionals who facilitated our work. We also wish to thank Morten Mattingsdal, Marie J Skogstad, Knut-Erik Gylder, Thomas Bjella, Eivind Bakken, and Bente G Bennike, for skilful technical and administrative assistance. We also thank Tomas Axelsson and Per Lundmark (SNP Technology Platform, Uppsala University and Uppsala University Hospital, Sweden), who were in charge of the Illumina-based genotyping at the platform in Uppsala, and the University of Oslo Bioportal for providing a platform for running the statistical software Unphased. The study was supported by grants from: the Research Council of Norway (#167153/V50,#163070/V50, #175345/V50), Eastern and Western Norway Health Authority (#123–2004), Ullevål University Hospital and the University of Oslo to support the Thematic Organized Psychosis Research (TOP) Study group and the Bergen group; the Copenhagen Hospital Corporation Research Fond, the Danish National Psychiatric Research Foundation, the Danish Agency for Science, Technology and Innovation (Centre for Pharmacogenomics) and the Danish Medical Research Council, the Lundbeck Foundation; the Stanley Medical Research Institute; and the Wallenberg Foundation, the HUBIN project and the Swedish Research Council (K2007-62X-15078-04-1, K2007-62X-15078-04-3, K2008-62P-20597-01-3).
