Abstract In this review, all papers relevant to the molecular genetics of bipolar disorder published from 2004 to the present (mid 2006) are reviewed, and major results on depression are summarized. Several candidate genes for schizophrenia may also be associated with bipolar disorder: G72, DISC1, NRG1, RGS4, NCAM1, DAO, GRM3, GRM4, GRIN2B, MLC1, SYNGR1, and SLC12A6. Of these, association with G72 may be most robust. However, G72 haplotypes and polymorphisms associated with bipolar disorder are not consistent with each other. The positional candidate approach showed an association between bipolar disorder and TRPM2 (21q22.3), GPR50 (Xq28), Citron (12q24), CHMP1.5 (18p11.2), GCHI (14q22-24), MLC1 (22q13), GABRA5 (15q11-q13), BCR (22q11), CUX2, FLJ32356 (12q23-q24), and NAPG (18p11). Studies that focused on mood disorder comorbid with somatic symptoms, suggested roles for the mitochondrial DNA (mtDNA) 3644 mutation and the POLG mutation. From gene expression analysis, PDLIM5, somatostatin, and the mtDNA 3243 mutation were found to be related to bipolar disorder. Whereas most previous positive findings were not supported by subsequent studies, DRD1 and IMPA2 have been implicated in follow-up studies. Several candidate genes in the circadian rhythm pathway, BmaL1, TIMELESS, and PERIOD3, are reported to be associated with bipolar disorder. Linkage studies show many new linkage loci. In depression, the previously reported positive finding of a gene–environmental interaction between HTTLPR (insertion/deletion polymorphism in the promoter of a serotonin transporter) and stress was not replicated. Although the role of the TPH2 mutation in depression had drawn attention previously, this has not been replicated either. Pharmacogenetic studies show a relationship between antidepressant response and HTR2A or FKBP5. New technologies for comprehensive genomic analysis have already been applied. HTTLPR and BDNF promoter polymorphisms are now found to be more complex than previously thought, and previous papers on these polymorphisms should be treated with caution. Finally, this report addresses some possible causes for the lack of replication in this field.
Bipolar disorder (manic depressive illness) and major depression are two major mood disorders. Bipolar disorder is sometimes accompanied by psychotic features such as delusion and hallucination. The lifetime prevalence of bipolar disorder is estimated to be 0.8–2.6%, while that for depression is approximately 15%. Mood stabilizers such as lithium are used for maintenance treatment of bipolar disorder, and antidepressants are the most common pharmacological treatment used for major depression. Whereas the contribution of genetic factors to bipolar disorder is well established from twin, adoption, and family studies,1 the role of genetic factors appears to be smaller in depression. In the case of depression, stressful life events and maltreatment during childhood are major risk factors.
Approximately 2 years ago, our article entitled ‘Genetics of Bipolar Disorder’1 concluded that, ‘many linkage loci and association with candidate genes were reported, but the results are not consistent, and there is no established causative gene or genetic risk factor for bipolar disorder.’ Unfortunately, this situation remains largely unchanged.
This review provides an overview of the progress made by molecular genetic studies over the past 2–3 years, for both bipolar disorder and major depression. A literature search was performed with the key words ‘bipolar disorder’ and ‘genetics’, and all papers relevant to molecular genetics were scrutinized. Association studies with a small sample size (fewer than 100 patients) were not cited. If the frequency of a certain allele is 30% in controls and it is 50% in patients (odds ratio 2.3), the power to detect the difference with alpha of 0.05 exceeds 80% when the sample number is 100 or more for each group. Thus, it was regarded that a study of 100 or more patients may have enough power to detect clinically significant findings. In the case of major depression, it was difficult to select relevant papers by similar literature search, because the strategy of molecular genetic study is more diverse. Some studies focus on comorbid depression in general medical diseases, while others focus on personality trait or stress. Thus, the topics to discuss were arbitrary defined as linkage analysis, gene–environmental interaction in serotonin transporter, BDNF, tryptophan hydroxylase 2, and pharmacogenetics, and relevant papers were selected.
To date, no causative gene or genetic risk factor has been identified for bipolar disorder or depression. Thus, the results of the literature search are summarized here in the context of research strategies.
Commonality with schizophrenia
Bipolar disorder and schizophrenia are typically distinguishable. Although psychotic symptoms such as delusion and hallucination can be seen in both of them, the clinical features are not the same: delusion and hallucination in schizophrenia are related to impairment of self awareness, while those in bipolar disorder are related to mood disturbance. However, some patients with bipolar disorder have mood-incongruent psychotic features resembling those of schizophrenia, and thus differential diagnosis is not always easy. Indeed, family studies suggest the presence of a common genetic background between these two disorders.2 Recently, promising candidate genes were identified in molecular genetic studies of schizophrenia, and some of them were found to be associated also with bipolar disorder (Table 1): G72,22–25DISC1,6NRG1,14RGS4,4,5NCAM1,17DAO,5,20GRM3, GRM4, GRIN2B,5MLC1,35SYNGR1,36 and SLC12A6.28
Table 1. Recent candidate genes for bipolar disorder
| PERIOD3||1p36.23||Circadian rhythm||–||–||3|
| RGS4||1q23.3||Regulator of G-Protein Signaling 4||+||–||4,5|
| DISC1||1q42.1||Cloned from translocation break point||+||+||6|
| IL-1 cluster||2q13||Immune system||+||–||7|
| WFS1||4p16.1||Wolfram syndrome 1||–||+||9|
| PDLIM5||4q22||Adaptor protein (PKC/Ca2+ Channel)||+||+||10|
| DRD1||5q35.1||Dopamine D1 receptor||–||+||12,13|
| GRM4||6p21.3||Metabotropic glutamate receptor 4||+||–||5|
| GRM3||7q21.1-q21.2||Metabotropic glutamate receptor 3||+||+||5|
| NRG1||8p21-p12||Neuregulin 1||+||+||14|
| HSPA5||9q34||Endoplasmic reticulum chaperone||–||+||15|
| ARNTL(BmaL1)||11p15||Circadian rhythm||–||+||3,16|
| NCAM1||11q23.1||Cell adhesion molecule, neural, 1||+||+||17|
| HTR3B||11q23.1||Serotonin 3B receptor||–||+||18|
| GRIN2B||12p12||N-Methyl-d-aspartate receptor subunit 2B||+||–||5|
| TIMELESS||12q12-q13||Circadian rhythm||–||–||16|
| CUX2||12q23-q24||Regulator of NCAM expression||–||+||19|
| DAO||12q24||d-Amino acid oxydase||+||+||5,20|
| Citron||12q24||Serine/threonine protein kinase 21||–||+||21|
| G72 (DAOA)||13q34||d-Amino acid oxydase activator||+||+||22–25|
| GCHI||14q22-24||GTP cyclohydrolase I||–||+||26|
| GABRA5||15q11-q13||GABA-A receptor alpha 5 Subunit||–||–||27|
| SLC12A6||15q13-q15||KCC3; potassium chloride co-transporter||+||–||28|
| NAPG||18p11||SNAP (Soluble N-ethylmaleimide-sensitive fusion (NSF)-attachment proteins)||–||+||29|
| PIK3C3||18q12.3||Phosphatydilinositol kinase 3C3||–||+||31|
| AD4D2||18q21.1||A gene with triplet repeat||–||+||32|
| TRPM2||21q22.3||Store operated Ca2+ channel||–||+||33|
| BCR||22q11||Breakpoint cluster region||–||+||34|
| MLC1||22q13||WKL1, cation channel||+||+||35|
| SYNGR1||22q13.1||Synaptogyrin 1||+||+||36|
| GPR50||Xq28||G protein-coupled receptor 50||–||+||37|
| mtDNA 3644||mtDNA||Mitochondrial complex I (ND1)||–||(+)||38|
| mtDNA 3243*||mtDNA||Mitochondrial tRNALeu(UUR)||–||(+)||39|
| Nogo||2p13-p13||Neurite outgrowth inhibitor||–||+||40|
| GAD1||2q31||Glutamate decarboxylase||–||–||41|
| DRD3||3q13||Dopamine D3 receptor||+||–||42|
| GSK3b||3q13.3||Glycogen synthase kinase 3-β||–||–||43–45|
| PHOX2B (PMX2B)||4p12||Transcription factor for DA neurons||+||–||46|
| ADRA2C||4p16.1||Adrenoceptor 2C||–||+||47|
| DAT1||5q14.3||Dopamine transporter||+||–||48|
| GABAAg2||5q31.1-q33.1||GABA-A receptor gamma 2 subunit||–||+||49|
| GABAAb2||5q34-q35||GABA-A receptor beta 2 subunit||–||+||49|
| NOTCH4||6p21.3||Notch signaling||+||–||50|
| TAAR6||6q23.2||Trace amine-associated receptor 6||+||+||52|
| ADRA1C||8p21||Adrenoceptor 1C||–||+||47|
| CHRNA2||8p21-22||Nicotinic acetylcholine receptor alpha 2||+||+||53|
| PIP5K2A||10p12.2||Phosphatydilinositol 5-phosphate kinase 2A||–||+||54|
| ADRA2A||10q24-q26||Adrenoceptor 2A||–||+||47|
| BDNF||11p13||Brain-derived neurotrophic factor||+||+||55–57|
| TPH1||11p15.3-q14||Tryptophan hydroxylase||+||+||58|
| DRD4||11p15.5||Dopamine D4 receptor||+||+||42|
| DRD2||11q23||Dopamine D2 receptor||+||+||42|
| FEZ1||11q24.2||Interaction with DISC1||+||–||59|
| NTF3||12p13||Neurotrophin 3||+||–||60|
| TPH2||12q21.1||Tryptophan hydroxylase 2||–||+||153|
| NOS1||12q24||Nitric oxide synthase 1||+||+||61|
| HTR2A||13q14-q21||Serotonin 2 A receptor||+||+||62|
| HTT||17q11.1||Serotonin transporter||+||+||63|
| ERDA1||17q21.3||A gene with triplet repeat||–||+||64|
| CTG18.1||18q21.1||A gene with triplet repeat||–||+||64|
| SYNJ1||21q22||Synaptojanin 1||+||+||65|
| ZDHHC8||22q11.21||Deleted in VCFS||+||+||66|
| XBP1||22q12||X-Box Binding Protein 1||+||+||67,68|
| MAOA||Xp11.23||Monoamine oxydase A||+||+||69|
| PCDH11Y||Yp11.2||Protocadherin 11, Y-Linked||–||–||70|
| NDUFV2||18p11||Mitochondrial complex I||+||+||71|
| IMPA2||18p11.2||Inositol monophosphatase 2||+||+||72,73|
No significant association was found for other genes examined in this context: ZDHHC8,66TAAR6 (6q23.2),52DTNBP1,51PHOX2B[PMX2B],46FEZ1,59NOTCH4,50 and NTF3.60
There is current consensus among researchers in psychiatric genetics that, among these genes, the association with G72 (13q34) is most robust. G72 is cloned from 13q34, the common linkage locus in bipolar disorder and schizophrenia, and it encodes d-amino acid oxidase activator (DAOA). There is another transcript encoding a non-coding RNA named G30, which is transcribed in the opposite direction to G72.74d-Serin has been identified in the brain as an endogenous modulator of the glycine site of N-methyl-D-aspartate-type glutamate receptors by Nishikawa’s group,75 and the role of d-serin in the pathophysiology of schizophrenia has been extensively studied. d-amino acid oxidase (DAO) is one candidate enzyme that may metabolize d-serine, and its activator is DAOA, encoded by G72. However, the role of d-serine in bipolar disorder has not been well studied as yet.
The association between bipolar disorder and G72 was first reported by Hattori et al.22 They found significant over-transmission of the most common haplotype in two sets of samples: Clinical Neurogenetics (CNG) pedigrees and National Institute of Mental Health (NIMH) pedigrees. Another study subsequently confirmed this association in a small sample set of 139 patients and 113 controls.23 However, the associated haplotype and single nucleotide polymorphism (SNP) were different from the original report, and the association with one SNP (rs1935062) was in the opposite direction. In a third report (a case–control study using 300 patients and the same number of controls), the global P for the association with the G72 haplotype did not reach significance. Although one SNP marker, M23, had significant association with bipolar disorder, this marker was not associated in the original report by Hattori et al. and no association was found for rs1935062, which had been found to be associated in previous studies.24 A recent report shows association at the haplotype level, in which one associated SNP (rs1341402) was common to the original report by Hattori et al. but in the opposite direction.25
Several studies have suggested an association between G72 and persecutory delusion20 or psychosis,76 rather than with bipolar disorder itself. However, a recent study did not find an association with psychosis.25
Association of bipolar disorder with G72 was regarded as very robust at first, but recent findings do not consistently support this idea.74 In addition, no base substitution with a functional consequence has been identified yet in G72. Gershon’s group, who reported the association of bipolar disorder with G72, is still searching for other nearby gene(s) at 13q34 responsible for the linkage with bipolar disorder.77 It is evident that G72 is one of the most promising candidate genes, but it is still premature to say that G72 is a genetic risk factor for bipolar disorder.
DISC1 (disrupted in schizophrenia 1), the only gene that has been established as a causative gene for a mental disorder, was first cloned from a translocation break point (1q42.1) of a pedigree in which schizophrenia and mood disorders are linked with a balanced chromosomal translocation.78 Many researchers have recently conducted studies of its function, and they have found that DISC1 is widely distributed throughout neurons, nuclei, mitochondria, and neurites. It interacts with many proteins such as NUDEL, Kendrin, PDE4B, FEZ1, Citron, and ATF4/5, and it is related to many functions such as neurite extension, neuronal migration, dendrite plasticity, and neurotransmitter signaling.78 In contrast with an early report that DISC1 function is tightly linked to neural development,79 recent studies show multiple faces of this molecule, and this has obscured the neurobiological mechanism of its relationship with mental disorders.
Recently, a new pedigree with a loss-of-function mutation of DISC1 was reported.80 In this pedigree, a 4-bp deletion causing a frame shift was found in the proband and his two siblings with schizophrenia or schizoaffective disorder. However, their father, who also carries the mutation, is healthy, and two patients with major depression, which was regarded as one of phenotypes in the first pedigree, did not have the mutation. Thus, the linkage of this frame shift mutation with mental disorders remains inconclusive. Indeed, this frame shift mutation was not found in 655 patients with schizophrenia but was found in two of 694 healthy subjects.81
Although SNP analysis showed association between DISC1 and bipolar disorder,6 this should be discussed separately from the linkage with loss-of-function mutations in the two pedigrees.
Positional candidate approach
Previous linkage studies and meta-analyses showed linkage of bipolar disorder with many loci: 2p, 4p, 4q, 6q, 8q, 9p, 10q, 11p, 12q, 13q, 14q, 16p, 16q, 18p, 18q, 21q, 22q and Xq.1,82 The most straightforward strategy in molecular genetics is to increase the number of markers for dense mapping, and to search for a causative mutation in the candidate gene in that locus. Such dense mapping showed progress in several linkage loci such as 3q29, 5q31-33 and 18q22-23.83,84 After the linkage locus is well defined, all candidate genes in that region should be examined. Shink et al. examined 32 candidate genes at 12q24.3185 and found nominal association with polymorphisms in three genes: KIAA1595, FLJ22471, and HM74. They also found a CAG repeat in a candidate gene in this locus, SMRT/N-CoR2, but it was not associated with bipolar disorder.86
Blair et al. examined 17 brain-expressed genes on 4p3587,88 and found that the haplotype covering the FAT gene, encoding a cadherin family protein, is associated with bipolar disorder in four independent sample sets.11 The direction of association was not always consistent between the sample sets. They also found that lithium treatment in mice downregulates the FAT gene, and concluded that this gene confers a genetic risk for bipolar disorder.
In addition to these two studies, association with many other genes located at linkage loci have also been reported in the past few years (Table 1): TRPM2 (21q22.3),33GPR50 (Xq28),37Citron (12q24),21CHMP1.5 (18p11.2),30GCHI (14q22-24),26MLC1 (22q13),35GABRA5 (15q11-q13),27BCR (22q11),34CUX2, FLJ32356 (12q23-q24),19 and NAPG (18p11).29
Several other reports showed no association of bipolar disorder with PIP5K2A (10p12.2.,54CHRNA2 (8p21-22),53Nogo (2p14-p13),40SYNJ1 (21q22),65 or NOS1 (12q24).61
Syndromal bipolar disorder
Most researchers have performed molecular genetic studies on bipolar disorder following the common disease common variant (CDCV) model. In this hypothesis, only relatively common variants are considered, and rare haplotypes are not analyzed. However, this strategy might overlook rare variants that are causative for a subgroup of those afflicted. Therefore, some researchers are searching for rare mutations causing marked functional disturbance. In this context, it may be a promising strategy to study the causative gene in a pedigree of bipolar disorder accompanied by a specific somatic syndrome.
We have been focusing on the relationship between mitochondrial diseases and bipolar disorder.89 Recent studies showed that mood disorder is statistically more frequently seen in maternal relatives of children with mitochondrial diseases, compared with their paternal relatives, or compared with maternal relatives of the children with other metabolic diseases.90,91 We sequenced whole mitochondrial DNA (mtDNA) from six patients with bipolar disorder who developed some somatic symptoms characteristic of mitochondrial disease, such as ptosis, muscle weakness, cardiomyopathy, and diabetes mellitus. We found that the mtDNA 3644 mutation reducing mitochondrial membrane potential was associated with bipolar disorder.38
Among the mitochondrial diseases, autosomal dominantly inherited chronic progressive ophthalmoplegia (adCPEO) is sometimes comorbid with mood disorders.92 We recently developed transgenic mice carrying mutant POLG encoding mtDNA polymerase only in neurons. These mice have bipolar disorder-like behavioral and pharmacological characteristics. This finding suggests that mood disorder seen in patients with adCPEO is not a reaction to the somatic disease but the consequence of accumulation of mtDNA mutations in the brain.92
Linkage with an autosomal dominantly inherited dermatological disease, Darier’s disease, has been well studied. However, it is still controversial whether this linkage is caused by a pleiotropic effect of the ATP2A2 mutation that causes Darier’s disease or whether it resulted from tight linkage of the two causative genes, one for Darier’s disease (ATP2A2, encoding endoplasmic reticulum Ca2+ pump) and the other for bipolar disorder. Current research is focusing on the latter possibility.93
A linkage between autosomal-dominant medullary cystic kidney disease and bipolar disorder has also been reported.94
From gene expression analysis to genetic analysis
Comprehensive gene expression analyses in the brain or lymphocytes have also indicated candidate genes for bipolar disorder.10,15,71,95
Among the genes differentially expressed in the post-mortem brains of bipolar disorder patients, PDLIM5 had robust changes in lymphoblastoid cells96,97 and in a second sample set of post-mortem brains.10PDLIM5 may also be a state-dependent marker of major depression.98PDLIM5 (LIM) encodes an adaptor protein linking PKCɛ and N-type calcium channels. Genetic analysis showed association of bipolar disorder with promoter SNP in two sample sets.10 One of these SNP may alter the binding to transcription factors.99 These SNP were not associated with major depression.98
Nakatani et al. performed gene expression analysis with DNA microarray in post-mortem brains and found that nine genes including SST (somatostatin) and NDUFV2 were altered.71 By analyzing 43 SNP in these genes, they found an association between bipolar disorder and a haplotype of SST. NDUFV2 is one of two genes differentially expressed both in frontal cortex and hippocampus in depression-model rats using the learned helplessness paradigm.8 In addition, we analyzed NDUFV2 as a positional candidate gene based on the mitochondrial dysfunction hypothesis and reported association of promoter SNP of NDUFV2 with bipolar disorder in Japanese and parents–proband trios from NIMH Genetics Initiative.100 However, the NDUFV2 haplotype was not associated with bipolar disorder in an extended trio sample from NIMH.71
We analyzed the DNA microarray data focusing on mitochondria-related genes and found that LARS2 is upregulated in the post-mortem brains of patients with bipolar disorder.39LARS2 encodes an enzyme that catalyzes aminoacylation of mitochondrial tRNALeu. The mtDNA 3243 mutation in tRNALeuUUR, the most established mtDNA mutation causing mitochondrial myopathy encephalopathy, lactic acidosis with stroke-like episodes, is known to impair aminoacylation of this tRNA. We found that LARS2 is upregulated in cybrids carrying the 3243 mutation, possibly by a compensatory mechanism. By applying a highly sensitive assay using peptide nucleic acid clamped polymerase chain reaction (PCR) to detect the small amount of mutation, we found that three patients, two with bipolar disorder and one with schizophrenia, had a small amount of the 3243 mutation.
We also performed gene expression analysis in the lymphoblastoid cell lines derived from two pairs of monozygotic twins discordant for bipolar disorder and found that two endoplasmic reticulum stress-related genes, XBP1 and HSPA5, were downregulated in the patients.101 Although we initially reported that a promoter polymorphism of XBP1 was associated with bipolar disorder, this was not replicated in a larger association study in Caucasians67 and a smaller study in a Chinese population.68 We also performed an association study of HSPA5, and found nominal association of functional promoter polymorphisms of HSPA5 with bipolar disorder in a Japanese population.15 However, no association was found in NIMH trios.
WFS1, a causative gene for Wolfram disease, an autosomal recessive disease that is frequently comorbid with bipolar disorder, was recently identified as the gene regulated by XBP1.102 Although a hypothesis that many patients with mood disorder are carriers of Wolfram syndrome mutations was not supported by previous studies,103 the first study showing association of a WFS1 haplotype with bipolar disorder may provide a new approach for the study of the role of the WFS1 gene in bipolar disorder.9
Candidate gene approach
The most active areas in molecular genetic research of bipolar disorder are replication studies of previously reported candidate genes and analysis of related genes. Among the previously reported genes, the following genes have recently been examined: DRD2, DRD3, DRD4,42HTT,63DRD1,12,13MAOA,69DAT1,48TPH1,58HTR2A,62 and IMPA2.72,73 Most replication studies do not support the initial positive findings. In fact, there has been a failure to confirm the association with apparently promising candidate genes such as HTTLPR63 or the Val66Met polymorphism of BDNF.55–57
A previous study showed association of bipolar disorder with the DRD1 haplotype but not with the −48A/G polymorphism.104 In contrast, two recent studies showed association with the −48A/G polymorphism of DRD1.12,13 There are positive studies for DAT148 and IMPA2.73
Several new candidate genes have been tested for possible involvement in the previously implicated molecular cascades: TPH2, HTR3B,18ADRA1C, ADRA2A, and ADRA2C47 in the monoamine pathway; PIK3C331 in the intracellular signaling system; GAD1,41GABAAβ2, GABAAγ249 in the GABAergic system; PCDH11Y,70 a proto-cadherin; and GSK3β,43–45 a target molecule of lithium. These studies showed mostly negative results, but positive association was reported for HTR3B18 and PIK3C3.31
DePaulo’s group has been continuously studying the role of triplet repeat expansion in the pathophysiology of bipolar disorder, and they have reported an association with a newly identified CTG/CAG repeat.32,105 A replication study of their previous finding of association with triplet repeat expansion in ERDA1 and CTG18.1 did not support the association.64
Several molecular cascades that have not been well studied via molecular genetics were also examined. Studies of circadian rhythm-related genes16,106 indicated an association of bipolar disorder with BmaL1 (ARNTL) from two independent groups.3,16 Association with TIMELESS16 and PERIOD33 was also reported.
Cytokine genes in the IL-1 cluster are also reportedly associated with bipolar disorder.7
New approaches in linkage studies
Traditional linkage studies have revealed many candidate loci such as 2p, 4p, 4q, 6q, 8q, 11p, 12q, 13q, 16p, 16q, 18p, 18q, 21q, 22q, and Xq.82 An NIMH Genetics Initiative showed significant linkage signals at 10q25, 10p12, 16q24, 16p13, and 16p12.107 In an isolated Swedish population, linkage with 9q31.1-q34, 6q23-q24, and 2q33-q36 was reported.108 A linkage locus, 12q24, which was reported in an isolated population in Quebec109 was extensively investigated, but still no causative gene has been identified.85,86 Extensive analysis in European pedigrees confirmed previously reported loci, 4q31 and 6q24, and also showed a new linkage locus, 1p35-p36.110
In addition to such straightforward approaches, researchers have been trying to improve their methods of selecting pedigrees appropriate for linkage analysis. Some of these are reanalysis based on large-scale collaboration,111,112 focusing on single large families to avoid genetic heterogeneity,113 analysis taking geographic origin into account,114 and analysis of isolated populations.108,115
In psychiatry genetics, one of the most difficult problems is phenotype definition. As seen in the case of DISC1 pedigrees, linkage results largely depend on how the phenotype is defined. Thus, researchers are searching for phenotypes directly related to the genotype that may differ from the traditional disease definition. Psychotic symptoms,107,116,117 age at onset,118–120 combination with schizophrenia,121,122 comorbidity with panic disorder,107 suicide,107 and a phenotype of wellness rather than disease123 have been used as phenotypes for linkage analysis. Although pediatric bipolar disorder is a matter of debate, association with the BDNF polymorphism in those cases has drawn attention.124 Several associations were found between clinical features and candidate genes: antidepressant-induced manic switch was associated with HTTLPR,125 temperament was associated with monoamine-related genes,126 and sleep disturbance was associated with CLOCK.127
Interaction of two loci or the parent-of-origin effect was found for previously reported loci, 6q16.3-22.128
Endophenotype or intermediate phenotype may also be used for genetic studies. Neuroimaging findings such as decreased volume of anterior cingulated cortex129 or cognitive function130 may also be used as phenotypes for molecular genetic studies.
Compared with bipolar disorder, major depression is much more heterogeneous and the contribution of environmental factors is larger than that of genetic factors. Thus, in the molecular genetic studies of major depression, environmental factors should be taken into account, and a larger sample size may be necessary. Recently, molecular genetic studies of major depression examining such gene–environmental interactions in a large sample have begun to be published. Most studies have focused on well-studied functional polymorphisms such as HTTLPR and BDNF V66M. Association of COMT with early onset depression has also been reported.131
The number of reports of large-scale linkage studies of depression is also increasing. In 497 pairs of affected siblings of recurrent depression, linkage with 1p36,12q23.3-q24.11, and 13q31.1-q31.3 was found.132 In 297 families with 415 affected sibling pairs, linkage with 15q25.3-26.2 was reported.133 Linkage of recurrent early onset depression with 3p12.3-q12.3 and 18q21.33-q22.2, both of which are linked to bipolar disorder, was found in 87 families including 718 patients.134
Recent DNA microarray studies showed a gene expression change of SSAT in three brain regions in 24 suicide victims and 12 controls. A functional SNP in this gene (342A/C) was associated with suicide in 181 suicide completers and 80 male controls.135
Interesting gene–environmental interaction in the onset of depression has attracted much attention. Caspi et al. reported that subjects carrying the S allele of HTTLPR were more susceptible to depression triggered by environmental stress or childhood maltreatment in a cohort study of 847 subjects.136 Several studies using smaller numbers of subjects supported this finding, but only partially.137–139 In addition to HTT and stress, a three-way interaction with BDNF was also reported.140 A similar interaction was also found in female monkeys (n = 190).141 However, recent large-scale studies have failed to replicate this gene–environment interaction.142,143 Surtees et al. examined the presence of major depressive episode in the previous year and adverse experiences during the previous 5 years in 4175 subjects from a cohort study of cancer and nutrition. Among the participants, 298 subjects had major depression, and marked association of depression with adverse experiences was found, but no significant effect of HTT was found.144 Gillespie et al. studied 1091 twins but could not detect any gene–environment interaction in HTT and stressful life events.143
Replication by small-scale studies and lack of replication in large-scale studies could arise as a result of publication bias.
Although a functional polymorphism V66M of BDNF, which affects activity-dependent secretion and cognitive functions,145 was reported to be associated with bipolar disorder, this was not supported by subsequent studies.55–57
Schumacher et al. examined 2376 subjects, including 465 with major depression, but did not find significant association of V66M with major depression. However, three marker haplotypes had a significant association.57 Surtees et al. also did not find significant association of major depression with V66M in 7389 subjects, including 1214 (16.4%) with a history of major depression.142 Association with BDNF at the haplotype level was not supported in another smaller study.146 Together, these results do not support a role of BDNF in the onset of depression.
An enzyme synthesizing serotonin has drawn attention in the context of the serotonin hypothesis of depression. Recently, TPH2 was identified as a homolog of tryptophan hydroxylase (TPH), and found to be a brain-specific isoform. A rare mutation of TPH2 causing a substitution of a well-conserved amino acid (G1463A, R441H) is a loss-of-function mutation. Zhang et al. reported that this mutation was not seen in 219 healthy controls, but it was found in nine of 87 patients with major depression.147 Although this discovery caused excitement, no subsequent studies could identify the same mutation in a large number of patients with major depression148–152 and in treatment-resistant depression.151 The cause of this inconsistency has not been identified yet, but one possibility is a genotyping error.
No association between the TPH2 promoter haplotype and bipolar disorder or suicide was reported.153,154
Previous studies showed association of the S/S genotype of HTTLPR with poor lithium response in bipolar disorder.155 However, a recent report did not support this association,156 while another study did.157
One study suggested association of positive lithium response with the Met allele of the BDNF V66M polymorphism,158 but this association was not supported by two other studies.159,160 Association of XBP1-116G with poor lithium response was reported separately in two independent sample sets.161,162
The previous finding of a poor antidepressant response in depressive patients with the HTTLPR S/S genotype was tested and a similar trend was reported in recent studies.156,163 However, recently the dichotomy of HTTLPR into the S and L alleles has been questioned (as discussed later). An SNP causing functional difference between L alleles was reported to be associated with antidepressant response.164 Association of antidepressant response with GNB3 (beta3 subunit of G protein)165 and DAT1163 was also reported.
Binder et al. examined 57 SNP in eight candidate genes related to the hypothalamic–pituitary–adrenal (HPA) axis in 294 patients with depression, and they searched for the SNP associated with antidepressant response. A functional polymorphism of FKBP5, a co-chaperone of Hsp90 regulating glucocorticoid receptor function, was associated with rapid response to antidepressant. Because these patients had a higher incidence of relapse, this SNP may be related to a subgroup of depression, rather than antidepressant response.166 More recently, they reported that functional polymorphisms of glucocorticoid receptor were associated with depression.167
In a large-scale study entitled STAR*D (sequenced treatment alternatives to relieve depression), 768 SNP in 68 candidate genes were examined in 1953 patients with major depression, and only one SNP marker in HTR2A (rs7997012) was significantly associated with treatment response.168 While 79.9% of AA carriers were responders, 62.4% of GG carriers were responders.
The locus 12q24 has been identified as being responsible for antipsychotics-induced obesity. PMCH (pro-melanin-concentrating hormone) is a candidate gene in this locus.169 Gastrointestinal side-effects by fluvoxamine was associated with HTR2A and CYP2D6, with interaction of these two genes.170
Severe side-effects caused by carbamazepine (CBZ), named Stevens–Johnson syndrome (SJS), have been associated with Human Leukocyte Antigen in a Taiwanese population. While all 44 patients with SJS had Human Leukocyte Antigen–B*1502, this allele was found in only 8.6% of 93 controls and 3% of 101 patients who did not develop SJS during CBZ treatment.171 Although this may be a promising marker for SJS, such an association was not seen in Caucasian subjects.172
New technology for comprehensive genomic analysis
The SNP chip (Affymetrix, Santa Clara, CA, USA) and a beads array (Illumina, San Diego, CA, USA) enable high-throughput analysis of SNP markers, and they have already been used for linkage analysis.173 Ewald et al. examined 1494 SNP using the SNP chip for homozygosity mapping in 24 patients with bipolar disorder in 22 inbred families to identify the linkage loci. They found a linkage with 17q24-q25, which had been implicated in a previous study.174
It was recently reported that copy number polymorphisms (CNP) are frequently found in the human genome. Comprehensive analysis of CNP is now available using bacterial artificial chromosome (BAC) arrays, oligonucleotide tiling arrays, or the SNP chip. Using BAC arrays, Wilson et al. comprehensively searched CNP associated with bipolar disorder using DNA obtained from post-mortem brain samples.175 Only a few of the loci detected by BAC array could be confirmed by real-time PCR, and ultimately they found a robust copy number alteration at four loci. Three of them (1p34.3, 14q23.3, 22q12.3) were replicated in a different sample set. By scrutinizing brain-expressed genes in these loci, they found an increased copy number of GLUR7 (GRIK3), which encodes a kinate-type glutamate receptor, in seven patients (four with bipolar disorder, two with major depression, and one with schizophrenia), but in none of the controls. Two of four bipolar patients had an intermediate copy number. This may be explained by a mixture of cells having different copy number. Otherwise, wild-type allele may contain multiple copies of this gene. Increased copies of AKAP5, encoding A-kinase anchor protein, was found in three patients (one with bipolar disorder, one with major depression, and one with schizophrenia). One patient with schizophrenia had a deletion of CACNG, which encodes a subunit of voltage-gated calcium channels.
Genotyping of complex polymorphisms
There are two instances of new research showing that a previously reported polymorphism is not as simple as originally thought.
Okada et al. closely examined the dinucleotide repeat polymorphism in the 1 kb upstream of the BDNF gene. They found that this is not a simple dinucleotide repeat polymorphism, but a complex polymorphism having 23 different alleles arising from the combination of three polymorphisms: dinucleotide repeat, insertion/deletion, and base substitution (BDNF-linked complex polymorphic region, BDNF-LCPR).176 They genotyped this polymorphism using Pyrosequencer (Biotage, Uppsala, Sweden), and found that a major variant reducing promoter activity was associated with bipolar disorder. Similarly, Nakamura et al. found that two HTTLPR alleles in this polymorphic region actually consisted of at least 14 alleles,177 having functional differences.178 Among these alleles, recent studies focused on a base substitution, A to G, in the L-type allele creating an AP-2 binding site.164 Promoter activity of the LG allele is lower than that of LA and more similar to that of the S allele.179 By dividing the L allele into two subtypes, the LA allele was found to be significantly associated with obsessive compulsive disorders179 and better antidepressant response in depression.164 Thus, hundreds of papers studying the association of HTTLPR with various conditions should be reconsidered, and these papers should not be cited without a comment on this situation.
CURRENT STATUS AND FUTURE DIRECTIONS
The availability of rich resources such as the SNP consortium (http://snp.cshl.org) and International HapMap Project (http://www.hapmap.org) now stimulates the genetic association studies of complex disorders. By genotyping representative tag SNP using high-throughput genotyping methods, large-scale association analysis has become easier.
In spite of extensive studies, however, we have not yet obtained conclusive evidence from genetic association studies. As discussed here, most of the initial positive findings in molecular genetic studies of mood disorders were not replicated in subsequent studies. Because positive findings tend to be published more easily, meta-analysis is not free from false-positive results.
Among 166 putative associations that were studied at least three times each, only six associations were replicated in more than 75% of studies.180 Such frequent false positives can be explained by small number of samples, population stratification, phenotype definition, genetic heterogeneity, low relative risk, multiple testing, genotyping error, selection bias especially for control group, and many other factors.180 Recent studies of G72 and DISC1 have reminded us of the difficulty in phenotype definition in the molecular genetic studies of mental disorders.
How to manage the uncertainties in association studies?
How can we manage such a situation in which most of the published papers are false positives and in which there are few clinically useful association findings?
Some journals have already faced this situation and have established criteria for publication of association studies, such as replication in two or more sets of samples, suitably large number of samples, suitably small P, and so on.181,182
The Human Genome Epidemiology Network (HuGENet) is also planning to develop guidelines for genetic epidemiology, for investigators, peer reviewers, and editors, to improve the quality of association studies.183 They also plan to build a database or online journal for negative results that currently tend not to be published, in order to avoid the effect of publication bias. Similar guidelines are also being established for microarray studies (MIAME, Minimum Information About a Microarray Experiment) (http://www.mged.org/Workgroups/MIAME), and some journals demand that the authors of submitted work conform the guidelines and submit the raw data to public database.
Although the study of copy number variations (CNV) in mental disorders has just begun and still suffers from technical problems, analysis of CNV at the genome-wide level would be the most important area of research in the next decade. CNV might contribute to the interindividual genetic variability even more than SNP. As discussed here, few studies succeeded in identifying the causative mutations at the linkage loci. Further studies may identify the causative copy number variations in such loci.
Recent advances in molecular genetic studies of mood disorders are reviewed here. Some of the most promising and striking results, such as the gene–environment interaction found between HTTLPR and stress or the loss of function mutation of TPH2 in depression, have not yet been replicated in subsequent studies. Researchers are continuously searching for better, more robust strategies for identifying genetic risk factors or causative genes for mood disorders.
Due to limited space, recent advances in other areas such as gene expression analysis and epigenetics could not be addressed in this review. These topics are reviewed elsewhere.184,185 Although the current status in the molecular genetics of mood disorders is that of a ping-pong game of positive and negative findings, some clues about the genetic determinants of mood disorders are beginning to emerge from the strategies mentioned here.