Neurotransmission and bipolar disorder: A systematic family-based association study†
Please cite this article as follows: Shi J, Badner JA, Hattori E, Potash JB, Willour VL, McMahon FJ, Gershon ES, Liu C. 2008. Neurotransmission and Bipolar Disorder: A Systematic Family-Based Association Study. Am J Med Genet Part B 147B:1270–1277.
Abstract
Neurotransmission pathways/systems have been proposed to be involved in the pathophysiology and treatment of bipolar disorder for over 40 years. In order to test the hypothesis that common variants of genes in one or more of five neurotransmission systems confer risk for bipolar disorder, we analyzed 1,005 tag single nucleotide polymorphisms in 90 genes from dopaminergic, serotonergic, noradrenergic, GABAergic, and glutamatergic neurotransmitter systems in 101 trios and 203 quads from Caucasian bipolar families. Our sample has 80% power to detect ORs ≥ 1.82 and ≥1.57 for minor allele frequencies of 0.1 and 0.5, respectively. Nominally significant allelic and haplotypic associations were found for genes from each neurotransmission system, with several reaching gene-wide significance (allelic: GRIA1, GRIN2D, and QDPR; haplotypic: GRIN2C, QDPR, and SLC6A3). However, none of these associations survived correction for multiple testing in an individual system, or in all systems considered together. Significant single nucleotide polymorphism associations were not found with sub-phenotypes (alcoholism, psychosis, substance abuse, and suicide attempts) or significant gene–gene interactions. These results suggest that, within the detectable odds ratios of this study, common variants of the selected genes in the five neurotransmission systems do not play major roles in influencing the risk for bipolar disorder or comorbid sub-phenotypes. © 2008 Wiley-Liss, Inc.
INTRODUCTION
Bipolar disorder (BP) is a complex psychiatric condition with a significant genetic component [Craddock and Jones, 1999; Taylor et al., 2002]. However, the search for specific susceptibility gene(s) in the last several decades has produced promising but inconsistent results [Craddock and Forty, 2006; Hayden and Nurnberger, 2006; Farmer et al., 2007]. Genes in several neurotransmission pathways/systems have long been a focus of candidate association studies in BP [Craddock et al., 2001], due to the effects of antidepressants on monoamine neurotransmission and of anti-manic drugs on glutamate and γ-amino butyric acid (GABA) transmission as well as on monoaminergic transmission [Hattori et al., 2005].
The Serotonin and Norepinephrine Systems
The original monoamine hypothesis of depression postulated a deficiency of monoamines 5-hydroxytryptamine (5-HT, serotonin) or norepinephrine (NE) in the pathophysiology of depression [Bunney and Davis, 1965; Schildkraut, 1965; Coppen, 1969]. Current evidence is based on the continued effect of most antidepressant drugs on these systems [Nutt, 2002; Berton and Nestler, 2006], brain imaging studies with specific ligands [Bremner et al., 2003; Cannon et al., 2007; Zipursky et al., 2007], transgenic mouse models with key 5HT or NE genes knocked out [Urani et al., 2005; Lesch and Mossner, 2006; El and Vaugeois, 2007], and genetic association studies [Craddock et al., 2001; Levinson, 2006; Kato, 2007]. In addition, gene expression data from the Stanley Medical Research Institute's (SMRI) brain collections reveal abnormal expression of several NE- and 5-HT-related genes in BP patients in contrast to mentally healthy controls (Supplementary Table I). However, some experimental data do not support the role of serotonergic or noradrenergic systems in mood disorders [Craddock and Forty, 2006; Hayden and Nurnberger, 2006; Levinson, 2006].
The Dopamine (DA) System
Dopaminergic dysfunction has also been proposed to contribute to the etiology of mood disorders [Nutt, 2006; Brugue and Vieta, 2007; Dunlop and Nemeroff, 2007; Gershon et al., 2007]. Some drugs targeting DA neurotransmission been shown to induce manic or major depressive episodes, as well as alleviate manic or depressive symptoms [Diehl and Gershon, 1992; Kapur and Mann, 1992; Gao et al., 2005]. Gene expression analysis of postmortem brains of BP patients (Supplementary Table I), neuroimaging studies, and genetic association studies provide suggestive evidence for a dopaminergic disturbance in mood disorders (see reviews [Stoll et al., 2000; Craddock et al., 2001; Dunlop and Nemeroff, 2007]).
The GABA System
Evidence for a GABAergic deficit in the pathophysiology and treatment of mood disorders includes stress-induced GABAergic functional changes in animal studies, the role of GABA receptor agonists and antagonists in improving the behavioral deficits of animal models of depression, the effects of available antidepressants and mood stabilizers on GABA function, genetic, neurobiochemical, and neuroimaging abnormalities of GABA system in patients with mood disorders, and the mood-modulating efficacy of drugs targeting the GABA system [Krystal et al., 2002; Brambilla et al., 2003; Sanacora and Saricicek, 2007].
The Glutamate System
Findings from pharmacological studies [Krystal et al., 2002; Paul and Skolnick, 2003; Berton and Nestler, 2006], neuroimaging studies [Krystal et al., 2002; Dager et al., 2004], gene expression analysis of postmortem brains of BP patients ([McCullumsmith and Meador-Woodruff, 2002; Knable et al., 2004; Woo et al., 2004; McCullumsmith et al., 2007] and Supplementary Table I), and genetic association studies [Fallin et al., 2005; Pickard et al., 2006; Blackwood et al., 2007] also support the glutamate deficit hypothesis of mood disorders [Krystal et al., 2002; Stewart and Reid, 2002; Paul and Skolnick, 2003; Kugaya and Sanacora, 2005].
Other neurotransmission systems (e.g., cholinergic) may also be involved in the pathophysiology and treatment of mood disorders [Belzung et al., 2006; Berton and Nestler, 2006; Hauger et al., 2006; Young, 2006], but have been studied separately (cholinergic system in Shi et al. 2007). It has been hypothesized that imbalance of the interacting neurotransmission systems (e.g., monoamine and amino acids) may result in neuropsychiatric diseases such as Parkinson disease, schizophrenia and depression [Carlsson et al., 2001; Pralong et al., 2002; Charney and Manji, 2004; Carlsson, 2006]. Variants in genes from one or several of the neurotransmission systems described here may contribute to BP susceptibility [Hattori et al., 2005]. We tested this hypothesis in three monoamine (NE, DA, and 5-HT) and two amino acid systems (glutamate and GABA) in a familial BP sample. We also investigated the relationship between variants in these neurotransmission genes and several comorbid conditions: alcoholism, psychosis, substance abuse, and suicide attempts [Baldassano, 2006; Potash et al., 2007].
MATERIALS AND METHODS
Subjects
The sample was composed of 1,115 individuals, including six trios (parents plus one affected child) and seven quads (parents plus two affected children) from the Clinical Neurogenetics (CNG) collections [Berrettini et al., 1991], 53 trios and 170 quads from the National Institute of Mental Health (NIMH) Genetics Initiative Bipolar Disorder collections waves 1–4 (http://www.nimhgenetics.org/) [Nurnberger et al., 1997; Dick et al., 2003], and 42 trios and 26 quads from the Chicago-Hopkins-Intramural Program (CHIP) collections [Potash et al., 2007]. Ascertainment and diagnostic methods for participants from the CNG, the NIMH waves 1–4 and the CHIP collections have been described in detail elsewhere [Berrettini et al., 1991; Potash et al., 2007]. All individuals are of European ancestry based on self-reported ancestry and further confirmed by STRUCTURE [Pritchard et al., 2000] analysis using 254 unlinked single nucleotide polymorphisms (SNPs) (data not shown). As BPII has been postulated to be an etiologically distinct mood disorder [Heun and Maier, 1993; McMahon et al., 2001; Hadjipavlou et al., 2004; Berk and Dodd, 2005; Nwulia et al., 2007], we analyzed only those families with affected offspring of either BPI or schizoaffective disorder bipolar type (SAB). Of 507 affected offspring from 101 trios and 203 quads, 481 met DSM-III-R or DSM-IV criteria for BPI and 26 for SAB. Three hundred were females and 207 were males.
Candidate Gene Selection
Gene selection from the dopaminergic, serotonergic, noradrenergic, glutamatergic, and GABAergic neurotransmission systems have been previously described [Hattori et al., 2005], where genes of various neurobiological pathways were culled by literature mining along with use of two databases, the Kyoto Encyclopedia of Genes and Genomes (KEGG) [Kanehisa et al., 2006] and the Gene Ontology database (http://www.geneontology.org/). The selection focused on metabolic enzymes, receptors, transporters, and intracellular signaling molecules. Five genes (ADCY1, ARRB2, GRIN3B, GRINA, and PAWR) were added following further literature review [Chatterton et al., 2002; Hardingham et al., 2002; Grange-Midroit et al., 2003; Lu et al., 2003; Park et al., 2005]. A total of 90 genes were chosen for this study (Supplementary Table II).
Single Nucleotide Polymorphism Selection and Genotyping
We used a series of bioinformatics tools (http://bioinfo.bsd.uchicago.edu/index.html) to select tag SNPs (tSNPs), as described previously [Shi et al., 2007]. Briefly, we first downloaded the 30 Centre d'Etude du Polymorphisme Humain (CEPH) trios' HapMap genotype data for SNPs within each genic region and within 10 kb sequences flanking each gene (Rel16c.1/phaseI June 05, on NCBI B34 assembly, dbSNP b124). Linkage disequilibrium (LD) was estimated for SNPs with minor allele frequencies (MAF) greater than 0.1. Next, we used the ldSelect algorithm[Carlson et al., 2004] to select tSNPs, which are SNPs that capture most of the genetic variation in a region due to high LD with other SNPs. A criterion of r2 ≥ 0.85 was set to define the bins. We selected tSNPs and singleton SNPs (only one SNP in each bin) for genotyping. Some originally selected SNPs that failed the Illumina check system (Illumina, Inc., San Diego, CA) were replaced with alternative tSNPs residing in the same LD bins. Finally, 1,099 SNPs in neurotransmission system genes were chosen for genotyping. Genotyping was performed using the Illumina BeadArray technology [Oliphant et al., 2002].
For quality control of genotyping, DNA from 46 HapMap European ancestry subjects, 15 never mentally ill individuals (11 from the NIMH collections, one from the CNG collections, and three from the CHIP collections), and 45 blank controls (using water instead of DNA) were used. Of 1,099 SNPs tested, 94 failed in genotyping or appeared as monomorphic. We used PedCheck1.1 to detect any Mendelian inconsistencies [O'Connell and Weeks, 1998], and Merlin to detect any unlikely recombinants [Abecasis et al., 2002]. All genotype errors were manually resolved by assigning problematic genotypes as missing prior to statistical analysis. The genotype data finally included had an average genotyping success rate of >99.85% for 1,005 SNPs.
Statistical Analysis
Pairwise LD, using the standard LD coefficient D′, was estimated between every pair of markers in each gene using Haploview (version 4.0) [Barrett et al., 2005]. Haplotype blocks were identified using the solid spine of LD method implemented in Haploview [Barrett et al., 2005]. This method searches for strong LD (a solid spine of D′ > 0.80 was set in our analysis), such that the first and last markers in a block are in strong LD with all intermediate markers, but the intermediate markers are not necessarily in LD with each other.
All the following analyses were accomplished using the whole genome data analysis tool set PLINK [Purcell et al., 2007], except where noted.
Allelic and haplotypic transmission disequilibrium tests (TDTs) were carried out for the disease trait. Permutations correcting for multiple testing were performed for individual SNPs and for all the SNPs and haplotypes in each gene, for each neurotransmission system, and for the five systems as a whole. To incorporate haplotypes into the permutation analysis, the most likely haplotypes were imputed using the hap-impute option in PLINK. These data were permuted to obtain analysis-wide P values. Haplotypic imputations and association analyses could not be performed for the five genes on the X-chromosome.
We used two different approaches to test potential effects of sub-phenotypes including alcoholism, psychosis, substance abuse and suicide attempts, on the SNP associations. The first method was the logistic regression analysis, using the sub-phenotypes as covariates. The second method was the heterogeneity test. For each covariate (e.g., psychosis), the sample was divided into two groups (strata). Then, association was tested within each stratum and across the two strata. The odds ratios from the groups with and without the covariate were compared (test of heterogeneity).
We also performed epistasis analysis to detect potential gene–gene interactions. First, we used the “focused interaction testing framework” (FITF) prescreening approach [Millstein et al., 2006] to test deviations in genotype distributions of pairs of SNPs from what is expected. This step helps to reduce the number of tests, which may enhance power. We selected the top 5% of pairs from this prescreening. We also analyzed SNP pairs in which both had an association with P-value < 0.05. This combined prescreening picks up interactions that may have no main order effects (from FITF prescreening) as well as those with main order effects (from individually nominally significant SNPs). A false discovery rate (FDR) of <0.025 (for two sets of epistatic analyses) in either analysis would be considered statistically significant. We estimated the FDR using the Statistical Analysis System (SAS Institute, Inc., Cary, NC).
Both the covariate and epistasis analyses were accomplished by taking one case per family and creating a pseudo-control of the untransmitted alleles and performing case-control analyses.
Comparison With Genome-Wide Association Study (GWAS) Data
We also reviewed the summary data from the GWAS on BP using an independent sample with 2,000 cases and 3,000 controls [Wellcome Trust Case Control Consortium, 2007], and compared these results with our data to see whether there were overlapping genes or SNPs showing nominally significant allelic association. We did not compare our data with another GWA study [Baum et al., 2008] because of overlap between our samples and those in that study.
RESULTS
Results of SNP and haplotype association analyses of 1,005 SNPs in selected genes from five neurotransmission systems are summarized in Supplementary Table III. Single nucleotide polymorphisms showing nominal association with BP are presented in Table I. Of them, rs2926835 in the GRIA1 gene, rs1799281 in the GRIN2D gene and rs744731 in the QDPR gene reached gene-wide association (permutation P < 0.05 for all SNPs of each gene). None of these associations survived multiple testing for its own system (e.g., glutamatergic or serotoninergic) or all the five systems taken together (permutation adjusted P > 0.05) (Supplementary Table III). Table II shows the gene-wide significant haplotypic associations with BP. Again, none of these associations survived correction for multiple testing (Supplementary Table III).
| SNP | Gene | CHR | Positiona | System | Alleles (1/2)b | T/UTc | TDT P-value | Gene-wide significanced |
|---|---|---|---|---|---|---|---|---|
| rs2453737 | DRD1 | 5 | 174791508 | Dopamine | A/G | 158/130 | 0.0233 | 0.0989 |
| rs2168631 | DRD1 | 5 | 174808608 | Dopamine | A/G | 84/105 | 0.0475 | 0.1898 |
| rs37022 | SLC6A3 | 5 | 1468629 | Dopamine | A/T | 88/104 | 0.0371 | 0.2108 |
| rs2036109 | ADRA1A | 8 | 26672701 | Norepinephrine | A/C | 157/126 | 0.0268 | 0.1728 |
| rs10102186 | ADRA1A | 8 | 26680568 | Norepinephrine | A/G | 142/120 | 0.0139 | 0.1728 |
| rs1048101 | ADRA1A | 8 | 26683945 | Norepinephrine | A/G | 166/140 | 0.0306 | 0.1888 |
| rs915841 | ABCG1 | 21 | 42552507 | 5-HT | A/G | 104/70 | 0.0494 | 0.5205 |
| rs9450594 | HTR1E | 6 | 87695058 | 5-HT | A/C | 44/72 | 0.0248 | 0.0579 |
| rs5028114 | HTR4 | 5 | 147991932 | 5-HT | A/G | 137/170 | 0.0432 | 0.2378 |
| rs744731 | QDPR | 4 | 17169112 | 5-HT | A/G | 94/138 | 0.0055 | 0.0140 |
| rs623580 | TPH1 | 11 | 18020553 | 5-HT | A/T | 138/119 | 0.0413 | 0.0809 |
| rs10743152 | TH | 11 | 2152557 | Metabolic enzyme | A/G | 149/126 | 0.0490 | 0.2058 |
| rs1042009 | ADCY1 | 7 | 45477211 | cAMP signaling | A/G | 109/145 | 0.0122 | 0.0829 |
| rs12754 | ADCY1 | 7 | 45534434 | cAMP signaling | A/C | 71/99 | 0.0070 | 0.1099 |
| rs4790694 | ARRB2 | 17 | 4573103 | cAMP signaling | A/C | 89/66 | 0.0251 | 0.0669 |
| rs9456 | ABAT | 16 | 8785092 | GABA | A/T | 75/62 | 0.0049 | 0.0859 |
| rs1273394 | ABAT | 16 | 8720788 | GABA | A/C | 110/131 | 0.0345 | 0.6224 |
| rs5925155 | GABRA3 | X | 151128486 | GABA | A/G | 65/80 | 0.0359 | 0.2428 |
| rs252952 | GABRB2 | 5 | 160646912 | GABA | A/C | 125/103 | 0.0066 | 0.1039 |
| rs592403 | GABRB2 | 5 | 160652562 | GABA | A/C | 159/133 | 0.0198 | 0.2517 |
| rs10515827 | GABRB2 | 5 | 160687535 | GABA | A/G | 88/110 | 0.0206 | 0.1409 |
| rs967771 | GABRB2 | 5 | 160694752 | GABA | A/G | 98/121 | 0.0278 | 0.2058 |
| rs6580038 | GRIA1 | 5 | 153113034 | Glutamate | A/G | 162/120 | 0.0264 | 0.3586 |
| rs2926837 | GRIA1 | 5 | 153151993 | Glutamate | C/G | 148/103 | 0.0294 | 0.3127 |
| rs1461225 | GRIA1 | 5 | 153162918 | Glutamate | A/C | 102/136 | 0.0056 | 0.0769 |
| rs2926835 | GRIA1 | 5 | 153169123 | Glutamate | A/T | 71/107 | 0.0036 | 0.0290 |
| rs592807 | GRIA3 | X | 122317191 | Glutamate | A/G | 44/58 | 0.0267 | 0.4036 |
| rs220592 | GRIN2B | 12 | 13856575 | Glutamate | A/C | 168/140 | 0.0449 | 0.7223 |
| rs2268115 | GRIN2B | 12 | 13760992 | Glutamate | A/C | 159/141 | 0.0352 | 0.6743 |
| rs1799281 | GRIN2D | 19 | 53640582 | Glutamate | A/G | 143/112 | 0.0029 | 0.0260 |
| rs362853 | GRM1 | 6 | 146754791 | Glutamate | A/C | 148/123 | 0.0443 | 0.3766 |
| rs733457 | GRM4 | 6 | 34176386 | Glutamate | A/C | 122/159 | 0.0249 | 0.1399 |
| rs6443074 | GRM7 | 3 | 6882193 | Glutamate | A/G | 42/49 | 0.0300 | 0.5674 |
| rs329037 | GRM7 | 3 | 7738386 | Glutamate | A/C | 176/131 | 0.0059 | 0.3766 |
| rs162724 | GRM7 | 3 | 7744282 | Glutamate | A/G | 65/96 | 0.0205 | 0.3037 |
| rs3818275 | SLC1A2 | 11 | 35265359 | Glutamate | A/G | 172/131 | 0.0325 | 0.3556 |
| rs2731877 | SLC1A3 | 5 | 36637612 | Glutamate | A/G | 150/166 | 0.0224 | 0.1948 |
| rs1645660 | SLC1A3 | 5 | 36680057 | Glutamate | A/G | 147/147 | 0.0292 | 0.2927 |
- SNP, single nucleotide polymorphism; CHR, chromosome; T, transmitted; UT, untransmitted; TDT, transmission/disequilibrium test.
- a Chromosome position is based on the human genome assembly B35 (hg17).
- b Minor allele for each SNP in unrelated parents is underlined.
- c Numbers of Transmitted and Untransmitted alleles from heterozygous parents to the first affected child in each family are presented. The T and UT counts for all affected offspring are not presented because they will not give accurate estimates of odds ratios when there is correlation between sibs due to any linkage.
- d P values were based on permutation for all SNPs in each gene. P values in bold (<0.05) indicate significant associations at the gene-wide level.
| SNPs/haplotype | Gene | System | T/UT | TDT P-value | Gene-wide significancea |
|---|---|---|---|---|---|
| rs37022|rs2042449/TG | SLC6A3 | Dopamine | 159/118 | 0.0068 | 0.0340 |
| rs1568447|rs690533|rs516035/GGA | GRIN2C | Glutamate | 30/42 | 0.0110 | 0.0410 |
| rs2597758|rs1031326|rs744731|rs2597778/AGGA | QDPR | 5-HT | 93/134 | 0.0064 | 0.0499 |
- a P values in bold, which reached gene-wide significance (P < 0.05), were based on permutation for multiple tests for all the single nucleotide polymorphisms and haplotypes in each gene.
Logistic regression analyses revealed that none of the covariates/sub-phenotypes (alcoholism, psychosis, substance abuse, and suicide attempts) produced significant associations for any SNP (data not shown). For the sub-phenotypes analyzed, none of the SNPs had a P-value less than the lowest P-value in the TDT test for the disease trait, implying that none of the associations would be significant after correction for multiple testing. While some sub-phenotype or covariate associations appeared to be present in one stratum but not the other in the heterogeneity tests, the difference was not significant, indicating no influence of these covariates on SNP associations.
We did not detect any significant gene–gene interaction, as shown by FDR correction for multiple testing of the results generated by FITF analysis (data not shown). Several genes from neurotransmission systems tested showed nominal allelic association with BP in this study or in the WTCCC GWA data (0.0001 < P < 0.05, Supplementary Table IV). However, only one SNP was nominally significant in both studies (rs623580 in TPH1, this study P < 0.0413; WTCCC P < 0.049 (allelic)), and would not be considered significant when multiple tests are corrected for.
DISCUSSION
This is the first association study to test five pathway/system disturbance hypotheses of BP, by systematically interrogating tSNPs in genes from each neurotransmission system. Several genes (GRIA1, GRIN2C, GRIN2D, QDPR, and SLC6A3) may have weak effects on risk for bipolar disorder (Tables I and II). The allelic or haplotypic associations shown in these gene region could survive correction for multiple tests for each gene but neither for corresponding individual system nor for all the five systems as a whole (Supplementary Table III). We did not detect significant SNP association with sub-phenotypes/covariates including alcoholism, psychosis, substance abuse, and suicide attempts or gene–gene interactions in influencing such disease risk. These results suggest that, within the detectable odds ratios (ORs) of this study, common variants of the selected genes in the five neurotransmission systems do not have major effects on etiology of BP or comorbid conditions.
Although multiple genes from the monoamine and amino-acid systems have been frequently tested in BP during the last two decades, no consistent association has yet been discovered [Craddock et al., 2001, 2006; Cho et al., 2005; Lasky-Su et al., 2005; Levinson, 2005; Preisig et al., 2005]. For example, a meta-analysis of 13 studies including 2,319 cases and 7,810 controls did not identify significant association of the functional Val158Met polymorphism (rs4680) in the catechol-O-methyltransferase (COMT) gene with BP [Craddock et al., 2006]. Our analysis of rs4680 in a sample of 376 trios overlapping with the samples used in the current study (unpublished data) did not show significant results. Two recent GWA studies [Wellcome Trust Case Control Consortium, 2007; Baum et al., 2008] did not detect any significant allelic association with BP either. As another example, the intron-2 variable number tandem repeats and gene-linked polymorphic region polymorphisms in the serotonin transporter gene (5-HTT, also known as SLC6A4) have shown significant association with BP in a meta-analysis (both with OR of 1.12) [Cho et al., 2005]. However, we did not identify any disease association in this gene region (Supplementary Table III), and only one SNP (rs7224199) in the 5-HTT gene showed a nominal association (OR = 1.10, P = 0.031) in the WTCCC GWA study [Wellcome Trust Case Control Consortium, 2007]. These results also imply that the effects of the neurotransmission system genes on BP susceptibility are relatively weak (e.g., OR of 1.1), if any exist.
Several genes with gene-wide significant association in the present study may deserve further investigation. For example, QDPR, encoding quinoid dihydropteridine reductase, is an essential cofactor for the metabolism of phenylalanine, tyrosine, and tryptophan hydroxylases of the serotonin system. Decreased gene expression has been shown in postmortem brain tissues of BP patients (Supplementary Table I). In addition, QDPR is located on 4p15, a BP linked chromosomal region [Detera-Wadleigh et al., 1997]. GRIA1 encodes an ionotropic AMPA glutamate receptor, which has been shown to play an important role in controlling synaptic plasticity [Zamanillo et al., 1999; Lee et al., 2003; Du et al., 2004]. An impairment of synaptic plasticity has been implicated in BP pathophysiology [Duman, 2002; Kempermann and Kronenberg, 2003; Carlson et al., 2006]. SLC6A3, encoding the dopamine transporter, is mapped to 5p15, which has been linked to BP [Kelsoe et al., 2001; Shink et al., 2002]. The dopamine transporter, an essential regulator of dopamine transmission, is the target of several psychostimulants [Iversen, 2006; Williams and Galli, 2006]. Furthermore, SLC6A3 has been associated to BP in several studies [Waldman et al., 1997; Keikhaee et al., 2005; Greenwood et al., 2006; Stober et al., 2006; Ohadi et al., 2007], but not in others [Gomez-Casero et al., 1996; Souery et al., 1996; Bocchetta et al., 1999; Georgieva et al., 2002].
Three serotonergic genes, ABCG1, HTR4, and TPH1, demonstrate nominally significant allelic association with BP in both the WTCCC GWA study and our family-based study (Table I and Supplementary Table IV). These three genes are also implicated by their location in BP-linked chromosomal regions [Aita et al., 1999; Zandi et al., 2003; Hong et al., 2004], by previous association evidence [Bellivier et al., 1998; Kirov et al., 2001; Ohtsuki et al., 2002], and by abnormal expression in postmortem brains of patients (Supplementary Table I).
The present study has several limitations. First, our sample has modest statistical power to detect associations with relatively weak effects at the study-wide significance level. For example, using PBAT (http://www.biostat.harvard.edu/∼clange/default.htm), under a multiplicative model and at P < 0.00005 (Bonferronni adjustment for 1,005 SNPs), our sample of 101 trios and 203 quads has 80% power to detect ORs of ≥1.82 and ≥1.57 for minor allele frequencies of 0.1 and 0.5, respectively. Second, SNP coverage for each gene is modest (on average, 9.3 kb/SNP) because we selected tSNPs from HapMap Phase I genotype data, which was the only data available at the time. Thus, we may have overlooked potential signals in the genes tested, although no strong association was identified by the WTCCC GWA study, which had denser coverage in some of the genes we studied [Wellcome Trust Case Control Consortium, 2007] (Supplementary Table IV). Lastly, we did not test all functionally interesting gene systems. For example, intracellular signaling molecules [Tanis and Duman, 2007], such as those encoding dystrobrevin-binding protein 1 (DTNBP1), neuregulin 1 (NRG1) and regulator of G-protein signaling 4 (RGS4) have been reported to influence NMDAR-mediated glutamatergic transmission and influence susceptibility to schizophrenia [Harrison and Weinberger, 2005; Carter, 2006], as well as to BP [Cordeiro et al., 2005; Fallin et al., 2005; Breen et al., 2006; Joo et al., 2007; Pae et al., 2007; Thomson et al., 2007]. Pathway analysis using tools such as KEGG, Protein ANalysis THrough Evolutionary Relationships (PANTHER) [Mi et al., 2007], and Ingenuity.com (http://www.ingenuity.com/), together with other lines of evidence may help to select such potential candidate genes for future studies.
Acknowledgements
The authors thank all the families participated in this study. We are grateful to the members of the NIMH Genetics Initiative for Bipolar Disorder consortium and the Chicago-Hopkins-Intramural Project for their efforts in family ascertainment, DNA sample collection and clinical diagnosis. We appreciate that the Stanley Medical Research Institute and its Collaborators, and Dr. Michael Elashoff and Dr. Fuller Torrey generously permitted us to publish gene expression data. This work was supported by NARSAD Young Investigator Awards (to E Hattori, C Liu, and J Shi), the Brain Research Foundation at the University of Chicago (to C Liu), NIH MH065560-02, MH61613-05A1 (to ES Gershon), NIH 5R01MH080425 (to C Liu) and the NIMH Intramural Research Program (to FJ McMahon). Support from the Geraldi Norton Foundation and the Eklund Family are also gratefully acknowledged. This study makes use of data generated by the Wellcome Trust Case-Control Consortium. A full list of the investigators who contributed to the generation of the data is available from http://www.wtccc.org.uk/. Funding for the project was provided by the Wellcome Trust under award 076113.
