Molecular genetics of bipolar disorder


E. P. Hayden, Institute of Psychiatric Research, Indiana University School of Medicine, 791 Union Drive, Indianapolis, IN 46202-4887, USA. E-mail:


Bipolar disorder (BPD) is an often devastating illness characterized by extreme mood dysregulation. Although family, twin and adoption studies consistently indicate a strong genetic component, specific genes that contribute to the illness remain unclear. This study gives an overview of linkage studies of BPD, concluding that the regions with the best evidence for linkage include areas on chromosomes 2p, 4p, 4q, 6q, 8q, 11p, 12q, 13q, 16p, 16q, 18p, 18q, 21q, 22q and Xq. Association studies are summarized, which support a possible role for numerous candidate genes in BPD including COMT, DAT, HTR4, DRD4, DRD2, HTR2A, 5-HTT, the G72/G30 complex, DISC1, P2RX7, MAOA and BDNF. Animal models related to bipolar illness are also reviewed, with special attention paid to those with clear genetic implications. We conclude with suggestions for strategies that may help clarify the genetic bases of this complex illness.

Current systems of classifying mental illness (American Psychiatric Association 1994) describe the bipolar disorders (BPDs) as a group of affective illnesses characterized by excessive shifts in mood. Patients with BPD experience episodes of either mania (bipolar I disorder) or hypomania (bipolar II disorder). Manic episodes are persistent periods of abnormally elevated or irritable mood accompanied by several associated symptoms such as grandiosity, decreased sleep, excessive talkativeness, racing thoughts, distractibility, and increases in goal-directed and pleasurable activities. Hypomanic states are similar in terms of symptomatology but may be of shorter duration and have less associated impairment. Most patients with BPD also experience major depressive episodes, which are periods of either sad mood or loss of interest accompanied by symptoms such as changes in weight, appetite, sleep and activity, along with fatigue, guilt, impaired concentration and thoughts of death. Bipolar disorder is relatively common, with bipolar I illness affecting 0.5–1% of the population.

Twin studies show a markedly elevated concordance rate of BPD in monozygotic twins compared to dizygotic twins (Bertelsen et al. 1977; Cardno et al. 1999), and BPD is more common among the biological parents than the adoptive parents of BP adoptees (Mendelwicz & Rainer 1977). Family studies have established that severe forms of affective illness, including BPD, run in families and appear to be highly heritable (Nurnberger et al. 1994). Thus, although twin, family and adoption studies do not identify specific vulnerability genes, such designs consistently indicate a strong genetic component to BPD susceptibility.

Strategies for elucidating specific genetic bases for BPD include linkage and association methods. Linkage methods test the location of vulnerability genes by studying chromosomal fragments that are inherited together with an illness. Such analyses often test LOD scores (the logarithm of the odds that loci are linked), with higher LODs reflecting greater probability of linkage. Although recommended cutoffs for LODs vary slightly depending upon the type of analysis, a LOD of 1.9 is the minimum score suggestive of linkage, while LODs of 3.3–3.6 reflect significant linkage in genome-wide surveys of complex disorders (Lander & Kruglyak 1995). Parametric linkage analyses specify a mode of gene inheritance (e.g. dominant and recessive), while non-parametric methods simply measure sharing of gene variants or alleles without indicating a mode of inheritance. Association methods examine whether a given gene variant is associated with illness. Because it is unclear which phenotype best captures the underlying genetic mechanisms of the disorder, affected status is often defined in multiple ways in studies. For example, studies may use narrow, intermediate and broad disease models in analyses, meaning that disease status can refer to narrowly defined BPD only or can include a relatively broad spectrum of affective illnesses.

The present paper reviews current research from linkage, association and animal studies on the molecular bases for BPD, with an emphasis on papers published since 1999. In determining which linkage studies to include, we generally follow Lander and Kruglyak's (1995) cutoffs for LOD scores, except in instances where multiple studies implicate the same region. In such cases, we may also include studies reporting LOD scores that approach these cutoffs. While some studies report P-values for significance of LOD scores, we did not use this information in determining which linkage studies to include as these values can be misleading. While LOD scores give odds of probability of linkage, these values for probability are often misinterpreted as P-values. As genome-wide studies include multiple tests, a LOD score of 3 (indicating an odds of 1000:1 for linkage) corresponds to a P-value of 0.05, not 0.001. With respect to association studies, we included those studies which we felt had the strongest designs, based on factors such as sample size and statistical methods. In the rare instance where a finding can be considered confirmed, we have noted this. We reviewed animal models related to bipolar illness that we felt had the clearest implications for genetic research.

Linkage and association studies

Chromosome 1

Detera-Wadleigh and colleagues (1999) reported a suggestive linkage to 1q31-32 in a genome-wide scan of 22 pedigrees. Linkage of multiple psychiatric diagnoses, including BPD, to 1q42 was found in a family with a translocation (Millar et al. 2004). Millar et al. also report an association between the disrupted-in-schizophrenia 1 gene (DISC1; a gene at 1q42 coding for a neuronal structural protein) and BPD in the Scottish population.

Chromosome 2

Liu and colleagues (2003) examined an Israeli and American sample of 57 extended families (1508 Caucasian individuals) with BPD, reporting a two-point parametric LOD score of 3.20 for the region 2p13-16 using an intermediate disease phenotype and a dominant model of transmission.

Chromosome 4

Significant linkage to chromosome 4p was initially reported by Blackwood et al. (1996) in a Scottish pedigree, and Detera-Wadleigh et al. (1999) also reported linkage to 4p16-p14. Suggestive linkage was reported to chromosome 4q35 by Adams et al. (1998). Badenhop et al. (2003) examined a 55-pedigree sample comprised of 674 individuals, conducting two-point parametric LOD score analyses on chromosome 4q35. Several markers in this region showed evidence for linkage, including D4S3051 (LOD = 2.32), D4S426 (LOD = 2.49) and D4S1652 (LOD = 3.19), all under a broad disease model. Liu et al. (2003) report a suggestive two-point LOD score of 3.16 at D4S1625 (on 4q31) under a dominant model and broad disease phenotype.

Chromosome 5

Using non-parametric multipoint linkage analysis, Dick et al. (2002) analyzed chromosomes 5, 15, 16, 17 and 22 in a replication sample of 56 multiplex families from the National Institute of Mental Health (NIMH) Genetics Initiative for BPD. Sibling-pair analysis revealed a suggestive LOD score of 2.8 for a broad disease model at marker D5S207. However, the LOD score for this marker decreased to 2.0 when the replication and original sample were combined for an analysis restricted to sibling pairs with genotyped parents.

Greenwood et al. (2001) reported differential transmission of a haplotype (a group of closely linked alleles inherited together) of five single-nucleotide polymorphisms (SNPs; common variants in the genome sequence) within the dopamine (DA) transporter (DAT) gene. DAT, which has been mapped to 5p15.3, mediates reuptake of DA. Ohtsuki and colleagues (2002) examined polymorphisms of the serotonin 4 receptor (HTR4) gene on 5q32 in a case–control sample of 48 patients with mood disorder, finding that four polymorphisms at or in close proximity to exon d showed an association with BPD with odds ratios of 1.5–2. HTR4 encodes the serotonin 4 receptor gene and influences DA secretion.

Chromosome 6

Ginns et al. (1996) reported suggestive linkage at marker D6S7 (on proximal 6p) in an Amish pedigree. Dick et al. (2003) conducted genome-wide linkage analyses on 1152 individuals from 250 families in the NIMH Genetics Initiative Bipolar Survey, reporting that chromosome 6 yielded a suggestive multipoint maximum LOD score of 2.2 (near marker D6S1021), under a broad disease model. Combined analysis of 399 NIMH pedigrees (including those in Dick et al. 2003) yielded a significant LOD of 3.8 at 113 cM on 6q (A. Hinrichs et al. unpublished data).

Chromosome 7

Using affected sib-pair (ASP) analyses, Liu et al. (2003) report a suggestive multipoint LOD score of 2.78 at 7q34 using an intermediate disease phenotype. Additional suggestive evidence for linkage to 7q has been reported by Detera-Wadleigh and colleagues (1997, 1999).

Chromosome 8

Segurado et al. (2003) applied meta-analytic techniques to 18 BP genome scans (see Levinson et al. 2003 for a review of the methods). Chromosome 8q (8q24.21-qter) appeared linked to BPD under narrow and broad disease models, suggesting that loci with small effects on BPD may be located in this region. Dick et al. (2003) also report evidence for linkage to 8q in this same region, finding a suggestive LOD score of 2.46 under a narrowly defined disease phenotype, near the marker D8S256. Also near this marker, McInnis and colleagues (2003a) report a suggestive non-parametric LOD of 2.1 on 8q24 using an intermediate model of disease.

Chromosome 9

The Segurado et al.‘s (2003) study described above produced modest evidence that a region near the centromere on chromosome 9 contains loci that influence BPD. Additionally, a number of candidate genes for BPD are located on chromosome 9. As lithium and valproate may produce some of their effects by action on N-methyl-d-aspartate receptors (NMDAR), genes that code for the subunits of NMDAR are candidates for BPD. GRIN1, on chromosome 9q34.3, codes for the zeta-1 subunit of NMDA receptors. Mundo et al. (2003) examined three polymorphisms of this gene for linkage disequilibrium in BPD in 288 probands with narrowly defined BPD and related disorders and their parents. A preferential transmission of the G allele was found for the 1001G/C and 6608G/C variants of the GRIN1 in affected individuals.

Chromosome 10

McInnis et al. (2003b) conducted a genome-wide scan of 153 pedigrees as part of the NIMH Genetics Initiative, finding a suggestive non-parametric LOD of 2.2 for the marker D10S1423 on chromosome 10p12 under an intermediate disease model. This linkage was also reported by Foroud et al. (2000) in a subset of the aforementioned sample, and the region has been implicated in linkage studies of schizophrenia as well (Faraone et al. 1998; Schwab et al. 1998). Older reports from Ewald et al. (1999) and Cichon et al. (2001) reported linkage to 10q. More recently, Liu et al. (2003) obtained a suggestive LOD of 2.33 at 10q24 under a dominant, narrowly defined phenotype. Under a narrowly defined model, Segurado et al. (2003) found evidence that the region 10q11.21-q22.1 may influence BPD.

Chromosome 11

Since the first indication of linkage in Old Order Amish kindred (Egeland et al. 1987), chromosome 11 has been of interest to investigators. Zandi et al. (2003) scanned chromosomes 2, 11, 13, 14 and X in 56 families of 354 individuals who were part of the NIMH Genetics Initiative on BPD. Parametric analysis revealed a heterogeneity LOD score of 2.0 near the marker D11S1923 under a dominant, intermediate disease model.

The DA D4 receptor (DRD4) and tyrosine hydroxylase (TH) genes were examined in 145 Canadian patients and their biological parents (Muglia et al. 2002). Both DRD4, which encodes the D4 subtype of the DA receptor, and TH, a rate-limiting enzyme in the synthesis of catecholamines, are located on 11p. Biases in TH allele transmission were not found, consistent with several older studies finding no association with BPD and TH (Turecki et al. 1997; Souery et al. 2001; although see also Meloni et al. 1995). However, excess transmission of the DRD4 four-repeat alleles was detected (associated with an increased odds of BPD of 1.7) while the two-repeat allele was transmitted at reduced rates (non-transmission associated with an increased odds of BPD of 6.88), leading the authors to propose that this allele may provide protection from risk for BPD.

Sklar et al. (2002) genotyped SNPs in 136 patient–parent triads from the same sample as McInnis et al. (2003a), finding an association between BPD and SNPs in the brain-derived neurotrophic factor (BDNF) gene on chromosome 11p13-15. Sklar et al. confirmed this association in two independent samples of BP patients (although multiple family members seem to have been treated as independent cases). Further evidence for the role of BDNF in BPD was reported by Neves-Pereira et al. (2002), although negative results have also been reported (Nakata et al. 2003). BDNF, which has also been implicated in unipolar depression, encodes a nerve growth factor protein and its transcription is highly susceptible to modulation by antidepressants (Ivy et al. 2003).

In 469 BP patients and 524 matched controls, Massat et al. (2002a) investigated the DA D2 receptor (DRD2). Mapped to 11q22.2-22.3, DRD2 encodes the D2 subtype of the DA receptor. An increased odds ratio for BPD of 1.9 was found for allele 5 (especially the 5-5 genotype). Nearby on 11q23.1, polymorphisms of the neural cell-adhesion molecule 1 (NCAM1) gene were found to be nominally associated with BPD in a Japanese sample of 151 patients compared to 357 controls (Arai et al. 2004). NCAM1 is involved in neuronal growth and pathway formation.

Chromosome 12

Craddock et al. (1999) reported linkage to chromosome 12 in a family in which major affective disorder cosegregated with Darier's disease. Other studies suggest that while regions near the Darier's disease gene may confer vulnerability to BP, it may not be the Darier's disease gene itself that does so (Jacobsen et al. 2001; Jones et al. 2002). Morissette et al. (1999) have shown evidence for linkage to 12q24 in large French Canadian families. Also in this region, Barden et al. (2004) recently reported evidence from linkage and association studies indicating that the purinergic receptor P2X, ligand-gated ion channel, 7 (P2RX7) gene is a susceptibility gene for BPD and major depression. P2RX7 influences neurotransmitter release and neurogenesis.

Chromosome 13

Stine and colleagues (1997) reported modest evidence of linkage to chromosome 13q32 in the NIMH Genetics Initiative pedigrees, and further support for this finding was reported by Detera-Wadleigh et al. (1999) in the neurogenetics sample. Liu et al. (1999) and Kelsoe et al. (2001) have also reported suggestive findings of linkage on chromosome 13q. More recently, Potash et al. (2003) examined four chromosomal regions thought to confer susceptibility to both schizophrenia and BPD in a sample of 65 BP probands and 237 relatives affected with a major mood disorder. Subsets of families were created based upon the number of members with psychotic mood disorder (mood disturbances accompanied by hallucinations and/or delusions). Families with multiple cases of psychotic mood disorders showed evidence of non-parametric linkage to 13q31, with a LOD score of 2.52; these regions showed little evidence of linkage when the sample was examined in its entirety. Badner and Gershon (2002) conducted a meta-analysis of published whole-genome scans of BPD and schizophrenia, reporting that 13q shows significant evidence for linkage to both disorders. Finally, Liu et al. (2003) report a suggestive multipoint ASP LOD score of 2.2 under an intermediate diagnostic model for the D13S779 marker on 13q32.

With respect to association studies, Hattori et al. (2003) examined the relationship between the G72/G30 gene locus on 13q33 and BPD in two pedigrees, one from the Clinical Neurogenetics branch of NIMH (Berrettini et al. 1991) and another from the NIMH Genetics Initiative. The investigators performed transmission/disequilibrium testing and haplotype analysis. A similar haplotype was overtransmitted in both samples, which suggests that a susceptibility variant for BPD exists in this region. DePaulo (2004) recently suggested that the G72/G30 complex be considered the first confirmed gene related to BPD, as several independent groups have reported findings consistent with Hattori et al. (Chen et al. 2004; Schumacher et al. 2004). The G72 gene appears to influence activity of NMDA receptors, while the function of the G30 gene is unclear.

Ranade et al. (2003) examined linkage and association between BPD and serotonin 2A receptor (HTR2A) gene polymorphisms in a sample of 93 patients and their parents. Comparing the BP patients to 92 controls revealed an association of BPD with SNPs on exons 2 and 3, consistent with haplotype differences. Examining patients and their parents suggested significant linkage and association with 1354C/T and haplotypes containing this SNP. HTR2A, located on 13q14-q21, may mediate the effects of some types of antidepressants.

Chromosome 16

After finding suggestive linkage signals at marker D16S2619 in both the original and the replication samples from the NIMH Genetics Initiative, Dick et al. (2002) examined the combined samples for additional evidence of linkage. Using non-parametric affected relative pair analysis, they identified a region containing four markers that all yielded LOD scores greater than 2.0, with the highest LOD occurring at D16S749 (LOD = 2.8).

Within the NIMH Genetics Initiative sample, Itokawa et al. (2003) examined a functional polymorphism in the promoter region of the GRIN2A gene (which codes for an NMDA receptor subunit), located on 16p13.3. In this sample, association analysis of a panel of 96 multiplex BP pedigrees indicated a statistically significant bias in the transmission of longer alleles. Results suggested that longer than average alleles resulted in decreased glutamatergic neurotransmission, which in turn contributes to BP susceptibility. Also on 16p13, the adenylate cyclase type 9 (ADCY9) gene is a candidate gene for BPD. Adenylate cyclases influence neuronal signaling and may be targets of antidepressants. However, findings from association studies have been inconsistent (Toyota et al. 2002a, 2002b).

Chromosome 17

Examining affected relative pairs, Dick et al. (2003) found suggestive evidence for linkage on chromosome 17q where they obtained a multipoint, non-parametric LOD score of 2.4 under an intermediate disease model. Liu et al. (2003) reported a suggestive two-point parametric LOD score of 2.68 at D17S921 under a dominant model of narrowly defined disease.

Collier et al. (1996) found an association of the short allele of the serotonin transporter (5-HTT) gene, which maps to 17q11.1-12, with BPD and major depression. Rotondo et al. (2002) examined the 5-HTT gene in a sample of Italian BP patients with (n = 49) and without (n = 62) a co-occurring diagnosis of panic disorder (an anxiety disorder) and 127 healthy subjects. Relative to the healthy subjects, BP patients who did not also have panic disorder had significantly higher frequencies of the short allele of the 5-HTT gene-linked polymorphic region; 58% of the non-comorbid patients had a short allele compared to 43% of the controls. In addition, a recent meta-analysis of 5-HTT indicated a positive association with BPD (Anguelova et al. 2003). A repeat length polymorphism in the promoter of this gene affects the rate of serotonin uptake.

Chromosome 18

Berrettini and colleagues (1994) and Detera-Wadleigh et al. (1999) reported suggestive and significant linkage to the pericentromeric region of chromosome 18. Additionally, Costa Rican pedigrees supported linkage to the tip of 18p and 18q22-23 (Garner et al. 2001). Segurado et al. (2003) identified several regions on chromosome 18 as potentially encompassing susceptibility loci for BPD, including 18pter-p11 and 18p11-q12.3. Genes in this region potentially associated with BPD include CHMP1.5 (or C18-ORF2, unknown function) (Berrettini 2003) and G-olf (or GNAL; guanine nucleotide-binding protein, alpha-stimulating, olfactory type), although negative findings have been published regarding the role of G-olf in BPD (Turecki et al. 1996; Zill et al. 2003).

Chromosome 20

Willour et al. (2003) analyzed 56 multiplex bipolar pedigrees from the wave 2 sample of the NIMH Genetics Initiative for BPD, examining chromosomes 4, 7, 9, 18, 19, 20 and 21. While evidence for linkage was modest in the wave 2 sample alone, analysis of the combined samples from waves 1 and 2 detected a suggestive non-parametric LOD score of 2.38 at D20S162, under a broad disease model.

Chromosome 21

Straub et al. (1994) reported significant linkage to chromosome 21q22, and additional evidence for linkage to this region was observed by Detera-Wadleigh et al. (1996) in two independent samples. In a recent extension of the Straub et al.‘s study, Liu et al. (2001) reported additional evidence of linkage in 56 families to chromosome 21.

Chromosome 22

Several groups have reported linkage to chromosome 22 in bipolar samples, including the NIMH Genetics Initiative samples, the NIMH Neurogenetics pedigrees and Kelsoe et al. (2001). Badner and Gershon's (2002) meta-analysis provided strong evidence that 22q harbors a common susceptibility locus for both BPD and schizophrenia. Potash et al. (2003) found evidence of linkage in families with psychotic mood disorders to 22q12, reporting a non-parametric LOD score of 3.06.

Lachman et al. (1996) found a relationship between an allele for a variant of the catechol O-methyltransferase (COMT) gene and rapid cycling in BPD. Rotondo et al. (2002) examined the frequency of the polymorphisms for the COMT gene in affected and unaffected groups. Relative to the healthy subjects, BP patients without an additional diagnosis of panic disorder had significantly higher frequencies of the COMT Met158 allele (56% vs. 39%). COMT catalyzes the neurotransmitters DA, epinephrine and norepinephrine. Recently, Barrett et al. (2003) examined the role of the G-protein receptor kinase 3 (GRK3) gene in two independent sets of families with BP probands. An SNP was found to be associated with disease in the families of Northern European descent in this sample. GRK3 appears to regulate the brain's response to DA.

Chromosome X

McInnis et al. (1999) reported linkage to the X chromosome on Xp22.1 with a heterogeneity LOD of 2.3 in their analyses of the NIMH Genetics Initiative pedigrees (waves 1 and 2, 153 families). Ekholm et al. (2002) examined the relationship between the X chromosome and BPD in a sample of 341 BP Finnish individuals from 41 families. Using a dominant model of inheritance, a suggestive maximum two-point LOD score of 2.78 was found at marker DXS1047 under a narrow disease model. Previous research from this group has also supported linkage between BPD and markers on Xq24-q27.1 (Pekkarinen et al. 1995). Zandi et al. (2003) report a suggestive parametric heterogeneity LOD of 2.25 at marker GATA144D04 at Xp11.3 under a narrow, recessive model.

A dysfunction in gamma amino butyric acid (GABA) system activity has been hypothesized to play a role in BP vulnerability. Massat et al. (2002b) examined the GABA receptor (GABRA3) dinucleotide polymorphism, which maps to Xq28, in a matched European sample of 185 BP patients and 370 controls, and found that BP patients were much more likely to have the 1-1 genotype than control subjects, with increased odds of 2.5 of having BPD with this genotype. In addition, a meta-analysis of MAOA, which breaks down an array of monoamines, indicated a significant association (Preisig et al. 2000). MAOA is located on the short arm of chromosome X.

Animal models of BPD

One line of research in animal models of mood disorders examines the relationship between genetic characteristics of animals and behavior during laboratory situations that replicate environments thought to cause depression in humans (e.g. early maternal separation and chronic stress). In the frequently used learned helplessness paradigm, animals are administered an inescapable aversive stimulus, often an electric shock. Following the inescapable portion of the paradigm, animals are again administered the aversive stimulus, but are capable of acting to escape or terminate the stimulus. Longer latencies to escape in this portion of the task are seen as indicative of a ‘learned helpless’, depressotypic response to the task. That escape latencies are shortened in response to antidepressant dosing is taken as evidence for the validity of this task as a depression model (Nestler et al. 2002). The Porsolt swim test (Porsolt et al. 1977), another animal model of depression, entails placing an animal in water in an enclosed space. The period of time spent floating motionlessly, as opposed to the amount of time spent actively seeking escape, is seen as an index of depressive behavior. Although this model has somewhat limited face validity, behavior tapped during the task does show strong responsivity to antidepressants (Porsolt et al. 1991).

Lira et al. (2003) examined serotonin transporter-deficient mice in this task, finding that such mice had increased latency to escape and also showed depressive-like behavior in a variety of other laboratory tasks. Additionally, Kohen et al. (2003) found that congenitally helpless rats had abnormalities in signal transduction and regulation of apoptosis. These rats had reduced expression of cAMP-response element-binding protein (CREB) messenger ribonucleic acid in the hippocampus and increased levels of the antiapoptotic protein bcl-2 mRNA in prefrontal cortex, among other changes. Effects of the non-receptor tyrosine kinase Pyk2 on behavior during a learned helplessness task have also been investigated by Sheehan et al. (2003), who reported that enhancing levels of Pyk2 in the lateral septum increased escape behavior. Enhancing expression of CREB in the rat hippocampus appears to produce similar effects on behavior during a learned helplessness task (Chen et al. 2001).

Synaptotagmin IV knockout mice were examined during a modified version of the Porsolt swim test and were found to be highly sensitive to the effects of the antidepressant imipramine (Ferguson et al. 2004). Synaptotagmin IV is one of the families of proteins that regulate vesicle trafficking in neurons and appears to be downregulated in some forms of psychiatric illnesses. Male DRD5 null mutant mice were found to show less immobility during the Porsolt swim test (Holmes et al. 2001), although these mice did not appear different on other behavioral tasks tapping depressive-like behavior.

Currently, more animal models of depressive illness exist than of mania or of the cyclical nature of BPD (Nestler et al. 2002). Lithium response is viewed as a validating characteristic for such models (Nestler et al. 2002). Based on this criterion, probably the most valid of these models is animal activity level (Nestler et al. 2002). BDNF heterozygous mice have abnormalities in general activity level (Kernie et al. 2000), and extracellular signal-related kinases also appear to influence motor behavior in mice (Selcher et al. 2001). The genetic bases for other animal behaviors potentially relevant to BPD have been also studied, including eating (Kernie et al. 2000) and aggression (Lyons et al. 1999).

Future directions

As can be seen, multiple regions are putative susceptibility loci in BPD, with 2p, 4p, 4q, 6q, 8q, 11p, 12q, 13q, 16p, 16q, 18p, 18q, 21q, 22q and Xq arguably showing the most support. Studies have produced positive findings for a number of candidate genes, including COMT, DAT, HTR4, DRD4, DRD2, HTR2A, 5-HTT, the G72/G30 complex, DISC1, P2RX7, MAOA and BDNF, among others (Table 1). To date, the G72/G30 complex is the single finding we would consider fully replicated.

Table 1.  Overview of recent evidence for candidate genes for bipolar disorder (BPD)
GeneLocationFunctional significanceSupportive evidence
  • DISC1, disrupted-in-schizophrenia 1; DA, dopamine; DRD, dopamine D; DAT, dopamine transporter; NMDAR, N-methyl-d-aspartate receptors; BDNF, brain-derived neurotrophic factor; NCAM1, neural cell-adhesion molecule 1; P2RX7, purinergic receptor P2X, ligand-gated ion channel, 7; ADCY9, adenylate cyclase type 9; COMT, catechol O-methyltransferase; GRK3, G-protein receptor kinase 3; GABRA3, gamma amino butyric acid receptor 3; GABA, gamma amino butyric acid.

  • *

    Animal study.

DISC11q42Neuronal structural proteinMillar et al. (2004)
DRD54p16.1DA system regulates emotion and motivationHolmes et al. (2001)*
DAT5pMediates reuptake of DAGreenwood et al. (2001)
HTR45qEncodes the 5-HT4 receptor, which influences DA secretionOhtsuki et al. (2002)
GRIN19qCodes for a critical NMDA receptor subunit; lithium may act
Mundo et al. (2003)
DRD411pDA system regulates emotion and motivationMuglia et al. (2002)
BDNF11pNeuronal growth factor involved in stress and antidepressant responseSklar et al. (2002), Neves-Pereira et al. (2002), Kernie et al. (2000)*, Dluzen et al. (2001)*, Lyons et al. (1999)*
DRD211qDA system regulates emotion and motivationMassat et al. (2002a)
NCAM111qInvolved in neuronal growth and pathway formationArai et al. (2004)
P2RX712qCalcium-stimulated ATPase; influences neurotransmitter release and neurogenesisBarden et al. (2004)
G72/G3013qG72 interacts with d-amino acid oxidase; G30 unknownHattori et al. (2003), Chen et al. (2004), Schumacher et al. (2004)
HTR2A13qMay mediate effects of serotonin reuptake inhibitorsRanade et al. (2003)
GRIN2A16pGlutamate receptor subunitItokawa et al. (2003)
ADCY916pSecond messenger in neuronal signaling; may be
antidepressant target
Toyota et al. (2002b)
5-HTT17qPromoter alleles affect transcriptional efficiency of 5-HTTCollier et al. (1996), Rotondo et al. (2002)
CHMP1.518pAffects G-protein signalingBerrettini (2003)
COMT22qCOMT alleles affect enzymatic activityLachman et al. (1996), Rotondo et al. (2002)
GRK322qRegulates homeostatic brain response to DABarrett et al. (2003), Niculescu et al. (2000)*
GABRA3XqBPD may stem in part from GABA deficitMassat et al. (2002b)
MAOAXpDegrades DA, serotonin, norepinephrinePreisig et al. (2000)

A combination of linkage, association and other approaches will probably be necessary to clarify the genetic mechanisms of BPD. Findings in BP genetics may prove more robust through increased use of designs that incorporate the complex nature of BPD. For example, while genetic risk for BPD likely results from the effects of multiple genes that interact with one another, few studies have examined interaction effects. Methods that treat individual genetic effects as independent of one another are likely to be incomplete. Future association studies may be able to study multiple candidate genes known to play a role in mood regulation and their interactive effects.

Although the influence of genes in BPD is clearly critical, evidence suggests that environmental factors play a significant role in the course of the illness. For example, several studies have provided evidence that life events may precipitate episo des (Amberlas 1979; Johnson et al. 2000; Mortensen et al. 2003). Gene-environment models have recently been successfully applied to major depression (Caspi et al. 2003), and sufficient research on psychosocial influences on BPD (e.g. stressful or goal-attaining life events) currently exists to suggest some useful starting points for multivariate studies.

Future studies of the genetics of BPD may benefit from the investigation of phenotypes of increased specificity and increased breadth. For example, several studies have used the strategy of examining subgroups of BP patients based on factors such as early age of onset (Faraone et al. 2004), psychotic symptomatology (Potash et al. 2003), treatment response (Turecki et al. 2001) and comorbid anxiety disorders (Rotondo et al. 2002). Other investigations have yielded intriguing results by examining whether BPD and schizophrenia have a common underlying genetic diathesis (Badner & Gershon 2002). Further exploration is needed to clarify whether such phenotypes yield greater consistency of linkage and association findings.

Genetic research on BPD may also benefit from the increased use of endophenotypes (i.e. traits associated with the disease that are heritable and possibly precede disease onset and are present in unaffected relatives). As such traits are often quantitative, they may more accurately reflect the underlying genetic phenomena and may lend greater power to statistical analyses (Baron 2002). In schizophrenia research, endophenotypes have received much attention but are less widely used in BPD research. Candidate endophenotypes include circadian rhythm disruption, response to sleep deprivation, psychostimulants, tryptophan depletion and white matter hyperintensities (see Lenox et al. 2002 for a review of these), as well as temperament (e.g. high behavioral activation, hyperthymic temperament; Johnson et al. 2000; Kwapil et al. 2000; Lozano & Johnson 2001), and melatonin levels (Nurnberger et al. 2000). Further research is needed to establish whether these markers validly reflect underlying genetic vulnerability.

Animal studies of endophenotypes may also provide clues about the genetics of BPD. For example, studies have examined the genetic bases of circadian rhythms in animals (Hofstetter et al. 2003). As disruption of circadian processes may contribute to BPD, and as animal data suggest susceptibility to circadian rhythm disruption is heritable (Mayeda & Nurnberger 1998), such approaches may prove fruitful. In addition, research into the genetic loci regulating prepulse inhibition in rats (Palmer et al. 2003) may be informative as prepulse inhibition appears to be abnormal in manic BP patients (Perry et al. 2001; although studies of prepulse inhibition in euthymic patients have yet to be carried out). Studies have also examined activity level in response to stimulants, finding that BDNF heterozygous mice show significant increases in locomotor behavior when administered an amphetamine challenge (Dluzen et al. 2001); Niculescu et al. (2000) found that a variety of candidate genes for BPD, including GRK3, were upregulated in the prefrontal cortex of rats following a methylamphetamine injection. Future animal studies of endophenotypes may wish to emphasize underexplored areas such as response to sleep deprivation. Rats demonstrate a variety of manic-like behaviors in response to sleep deprivation, including hyperactivity and irritability (Gessa et al. 1995); this may provide a substrate for genetic studies targeting inbred strains or examining genetically modified animals. Examination of the genetic underpinnings of reward sensitivity and motivation (Depue et al. 1987), is also probably warranted. Convergent functional genomics, which integrates such animal models with human linkage data to identify high-probability candidate genes (Niculescu et al. 2000), may also prove to be a productive strategy.


Portions of this work were supported by AA07462, MH059545, and a grant to the clinical laboratories at the Institute of Psychiatric Research from the Indiana Division of Mental Health and Addiction.