Heterogeneity and the genetics of bipolar disorder
INTRODUCTION
As reviewed elsewhere in this issue, family, twin, and adoption studies clearly show bipolar disorder to be a highly heritable condition [Tsuang and Faraone, 1990; Cardno et al., 1999]. Yet, although some regions of the genome have been implicated, the genes that increase susceptibility to bipolar disorder remain to be discovered [Tsuang and Faraone, 2000; Berrettini, 2001]. Even when data from several genome scans were jointly examined by meta-analysis, the best evidence for linkage at particular loci would only be considered suggestive according to traditional criteria of genome-wide statistical significance, despite the increased power afforded by a pooled sample of over 1,000 affected individuals [Badner and Gershon, 2002; Segurado et al., 2002].
This difficulty in finding genes for bipolar disorder may be due, at least in part, to the etiological heterogeneity of the disorder. To confirm etiological heterogeneity requires that bipolar patients could be separated into at least two classes having different known etiologies and, perhaps, different pathophysiological signatures. In contrast, the homogeneity hypothesis asserts that there is a single necessary and sufficient cause or configuration of causes of bipolar disorder.
Our definition of heterogeneity has two consequences. First, because the assumption of one etiological class (i.e., homogeneity) is more parsimonious than the assumption of two or more, Occam's razor requires us to accept homogeneity as our null hypothesis in the absence of disconfirming data. Second, to assert either etiological heterogeneity or homogeneity implies that we can fully describe one or more mechanism of etiology. Although there are many clues to the etiological puzzle, they do not allow a full description because the etiological mechanisms for bipolar disorder have not been worked out in sufficient detail. Thus, we are tempted to examine descriptive, phenotypic data and infer the presence of one or more etiologies.
Unfortunately, the correspondence between etiology and phenotype is not isomorphic. Differing etiologies may lead to similar phenotypes, For example, the clinical syndrome known as Alzheimers disease can arise from either genetic or nongenetic factors. Moreover, linkage studies of Alzheimers disease have replicated linkages to chromosomes 21 and 14 in early-onset families [Tsuang et al., 1993a]. On chromosome 21, mutations in the amyloid precursor protein gene have been identified. Moreover, one study reported linkage to chromosome 19 in late-onset families [Pericak-Vance et al., 1991]. Additional studies have implicated the apolipoprotein E gene on this chromosome as a risk factor for Alzheimers disease [Corder et al., 1993; Saunders et al., 1993]. On the other hand, although the observation of dissimilar phenotypes (clinical heterogeneity) may suggest etiological heterogeneity, such inferences are also limited because a single disease entity, with a homogeneous etiology, can have variable expression due to moderating factors [Tsuang et al., 1993b]. Pleiotropy is not unusual for human genetic diseases [Vogel and Motulsky, 1986].
Thus, the link between phenotypic heterogeneity and etiological heterogeneity is tenuous, so attempts to use purely phenotypic data to infer etiologic heterogeneity must be viewed cautiously. For example, although it is tempting to use unimodal distributions of phenotypes to assert etiological homogeneity and multimodal ones to assert heterogeneity, the utility of phenotypic distributions to inform nosological theory is limited by the need for large samples to detect modality and even larger samples to detect low prevalence subtypes [Everitt, 1981; Ghosh and Sen, 1985; Hartigan, 1985; Schork and Schork, 1988]. With these cautions in mind, in this paper we consider two dimensions of heterogeneity for bipolar disorder that may have implications for future molecular genetic studies.
GENETIC IMPLICATIONS OF PSYCHOSIS IN BIPOLAR DISORDER
The idea that psychotic bipolar disorder is a distinct disorder has received some attention in the literature. Clinical studies show that some patients with bipolar disorder experience delusions and hallucinations that are similar to the psychotic symptoms of schizophrenia. In one study, over 50% of bipolar patients reported a psychotic episode at least once in their lifetime [Dunayevich and Keck, 2000], with psychosis occurring in either the manic or depressive phases of the illness. This study found grandiose delusions to be the most common symptom, but hallucinations, thought disorder, and mood-incongruent psychotic symptoms have also been reported [Dunayevich and Keck, 2000; Potash et al., 2001].
Winokur et al. [1985] compared rates of psychosis among relatives of 30 bipolar patients classified as either psychotic or not. They did not find the relatives of psychotic bipolar patients to have a higher risk of psychosis than the relatives of nonpsychotic bipolar patients. In a larger study, Potash et al. [2001] reported that 64% of families of psychotic bipolar patients had at least one relative with a psychotic mood disorder. This was significantly greater than the 28% figure found for relatives of nonpsychotic bipolar patients. These researchers also found that 34% of the affectively ill relatives of psychotic patients had psychotic symptoms, versus 11% of the affectively ill relatives of nonpsychotic patients. Although other studies are not definitive on this issue, there is a trend in this small literature to support the idea that psychotic symptoms tend to cluster in families with mood disorders [Leckman et al., 1984; Kendler et al., 1993; Coryell, 1997; Goldstein et al., 1998].
Given the cofamilial transmission of bipolar disorder and depression, studies of psychotic depression also provide some insights into the familial transmission of psychosis in mood disorders. Several studies compared the levels of risk for mood disorders among families of psychotic versus nonpsychotic depressed patients. Prusoff et al. [1984] found that first-degree relatives of delusional depressives were 1.5 times more likely to have major depression than were the relatives of nondelusional depressed patients, and 3.5 times more likely to have major depression than were relatives of normal controls. In contrast, Coryell et al. [1984] did not detect a difference in family risk for depression between patients with psychotic depression versus patients with nonpsychotic depression; however, patients with psychotic depression were more likely to have a family history of schizophrenia.
Coryell et al. [1982] contrasted the family histories of patients with mood-congruent psychotic depression and mood-incongruent psychotic depression. Relatives of the mood-incongruent group were found to be at greater risk for mood disorder and schizophrenia than were relatives of individuals with nonpsychotic depression, but they did not differ in levels from relatives of patients with schizophrenia. In contrast, the risk to relatives for mood disorders and schizophrenia among depressed patients with mood-congruent psychoses was similar to that of nonpsychotic depressed patients. Those family history data were regarded to suggest that depression with mood-incongruent psychotic features is more heterogeneous, but that those patients might be somewhat more likely to be genetically susceptible to schizophrenia than those with mood-congruent symptoms. In another study, Coryell et al. [1985] found that relatives of mood-incongruent patients had a higher rate of schizophrenia (as well as a lower rate of bipolar disorder).
Clearly, more work is needed to understand psychosis as a potential cleavage point for understanding the heterogeneity of bipolar disorder. To date, only two studies have specifically studied the familial transmission of psychosis in bipolar families. More studies have examined psychotic depression, but it still remains unclear whether the relatives of psychotic depressed probands are more at risk for either bipolar disorder or major depression than relatives of nonpsychotic depressed probands, or if psychosis aggregates preferentially in the families of probands with psychotic depression. Linkage studies have not yet made the distinction between mood disorders with and without psychotic symptoms. Nor have they systematically examined psychosis as a quantitative trait in bipolar disorder families.
THE AGE AT ONSET OF BIPOLAR DISORDER
Studies of Age at Onset as a Quantitative Trait
The age at onset of bipolar disorder has been recognized for some time as a potentially important marker for disease severity and, possibly, a more substantial genetic basis. Initial studies clearly supported this possibility, especially in the degree to which early onset was found to predict the amount of familial aggregation of the disease. For example, Rice et al. [1997] found a strong relationship between a patient's age at onset of bipolar disorder (range = 13–52) and the risk of the disease to his first-degree relatives. Subsequently, in a study of the Old Order Amish, Pauls et al. [1992] found increased lifetime rates of bipolar disorder among the first-degree relatives of patients who experienced their first affective symptoms before age 12, compared to relatives of those with onset at age 13 or later. A similar elevation in bipolar disorder prevalence among relatives was seen for probands who first met Research Diagnostic Criteria [Spitzer et al., 1978] for bipolar disorder before age 21 vs. those who had met disease criteria after age 25. These findings are representative of this now well-documented phenomenon and have been replicated by others. Tsuang and Faraone [1990] reviewed 16 such family studies assessing the age at onset of mood disorders. They found an elevated risk for mood disorders among relatives of probands who experienced an early onset of their mood disorder vs. relatives of those with later onset. This was true for both bipolar disorder and unipolar depression. But because these studies defined early and late onset by dichotomizing the age at onset in either early or mid-adulthood, their relevance to juvenile-onset bipolar disorder was not clear. It is becoming clearer that the dividing line between early- and later-onset cases is not as late as once thought, nor, in fact, is one dividing line sufficient to describe the age-at-onset distribution for the disorder. Other work suggests three distinct onset peaks for bipolar disorder, at 16.9, 26.9, and 46.2 years [Bellivier et al., 2001], and also confirms previous findings of a greater frequency of psychotic symptoms and a higher prevalence of familial cases in the group with the earliest onset.
Because age at onset is both familial [Omahony et al., 2002] and heritable [Visscher et al., 2001], is it reasonable to use linkage analysis to detect genes that regulate this trait in bipolar families? To address this issue, Faraone et al. (submitted for publication) examined the heritability of the age at onset of mania in bipolar disorder pedigrees collected through the NIMH genetics initiative program [Detera-Wadleigh et al., 1997; Edenberg et al., 1997; Rice et al., 1997; Stine et al., 1997]. Their analyses suggested that three regions of the genome (chromosomes 12p, 14p, and 15p) may contain genes that influence the age at onset of mania in bipolar disorder. These chromosomal regions had not been implicated as high-risk loci for bipolar disorder in either of the previously published meta-analyses of bipolar disorder [Badner and Gershon, 2002; Segurado et al., 2002], or in previous linkage analyses of the data set focused on bipolar disorder [Detera-Wadleigh et al., 1997; Edenberg et al., 1997; Rice et al., 1997; Stine et al., 1997]. Although preliminary and in need of replication, these results suggest that genes influencing the age at onset of mania in bipolar disorder may be distinct from those influencing the liability of developing the disorder.
Studies of Pediatric-Onset Bipolar Disorder
Although the diagnosis of mania in children has generated much debate [Biederman, 1998; Klein et al., 1998], there is growing recognition that a group of seriously disturbed children with severe affective dysregulation and high levels of agitation, aggression, and dyscontrol show clinical features consistent with the diagnosis of bipolar disorder. Notably, several reviews of the literature suggest that the diagnosis of bipolar disorder is warranted for at least some of these youth [Faedda et al., 1995; Weller et al., 1995; Geller and Luby, 1997; Biederman et al., 2000b].
Much of the controversy about apparently bipolar children derives from their clinical picture, which is atypical by adult standards. Bipolar children have a mixed presentation, a chronic course, poor response to mood stabilizers and high levels of comorbidity with attention deficit hyperactivity disorder (ADHD) [Wozniak et al., 1995a; Faraone et al., 1997c; Biederman et al., 1998, 2000b; Carlson et al., 2000; Schurhoff et al., 2000]. Yet, because a similar atypical picture is seen in about one-third of adults with bipolar disorder [McElroy et al., 1992], we cannot rule out the diagnosis in children based on atypical signs.
Subsequent work has addressed this issue using samples strictly ascertained for juvenile-onset bipolar disorder. Strober et al. [1988] found elevated rates of both bipolar disorder and major depression in first-degree relatives of all bipolar patients, but also found a higher prevalence of bipolarity in relatives of pediatric bipolar cases (29.4%) than in older cases (7.4%). Neuman et al. [1997] observed a similar pattern, in which relatives of pediatric-onset bipolar patients were more than twice as likely to have bipolar disorder than were relatives of later-onset patients. Such observations led these authors and others to posit a greater genetic role in early-onset cases, and to regard them as more genetically homogeneous. These ideas have precedent in other diseases, such as breast cancer and Alzheimers disease, where the earlier-onset forms appear to be caused by fewer genes with higher penetrance than the late-onset forms [Bishop, 1999; St. George-Hyslop, 2000]. Interestingly, the familial risk for both breast cancer and Alzheimers disease is higher for relatives of patients with these early-onset forms, much like the increased risk experienced by relatives of patients with early-onset bipolar disorder. Furthermore, early-onset breast cancer is considered a more severe form of the disease (due to an increased incidence of bilateral disease [Hall et al., 1990]) with a poorer prognosis, mirroring the increased severity and worse long-term outcome of early-onset bipolar disorder as evidenced by its chronicity, resistance to mood stabilizers, and higher rate of psychotic symptoms [McGlashan, 1988; Wozniak et al., 1995a; Faraone et al., 1997c; Biederman et al., 1998, 2000b; Carlson et al., 2000; Schurhoff et al., 2000]. Since twin and adoption studies of pediatric bipolar disorder have yet to surface, the heritability of this specific bipolar disorder subtype is unknown; however, segregation analyses initially excluded purely environmental transmission [Todd et al., 1993] and, subsequently, supported non-Mendelian major gene inheritance with a polygenic component [Grigoroiu-Serbanescu et al., 2001].
Of the molecular genetic studies of pediatric bipolar disorder that have been published to date, most have investigated allelic associations with the disorder, sometimes with contradictory results. For example, Geller and Cook [1999] found no evidence of association between prepubertal-onset bipolar disorder and the short/long polymorphism in the promoter region of the gene coding for the serotonin transporter (HTT), whereas Ospina-Duque et al. [2000] observed significantly more short HTT alleles in younger bipolar patients. Unfortunately, this discrepancy is not easily reconciled, as these are the only two studies of this association in pediatric bipolar disorder patients to date. However, for the more common adult-onset form of bipolar disorder, the influence of this polymorphism on disease risk has been examined many times; in fact, Craddock et al. [2001] identified eight separate studies of 11 different samples that provided data on this association and performed a meta-analysis of their findings. In this pooled analysis, there was little evidence to support a role for this polymorphism in the susceptibility to common bipolar disorder. If pediatric-onset forms of the disease represent a more severe, but related, form of bipolar disorder caused by a simple enrichment of genetic risk factors, then all of the eventually elucidated risk genes for adult-onset bipolar disorder should be found to confer risk for early-onset forms. The findings of Geller and Cook [1999] are consistent with this concept. However, if the findings of Ospina-Duque et al. [2000] are supported by future studies, the early-onset form of the disease may become recognized as a separate entity masquerading as bipolar disorder (i.e., a phenocopy).
Genes, such as HTT, that are expressed in monoamine neurotransmitter pathways are a priori good candidates for association with bipolar disorder for a number of reasons. Foremost among these is the fact that these genes are abundantly expressed in limbic brain regions that are critical for emotional regulation. In addition, the serotonin transporter is the principal binding site of antidepressants such as fluoxetine, which can actually induce mania in depressed patients [Hon and Preskorn, 1989]. However, some evidence suggests that monoaminergic genes may only contribute to specific features of the disorder. For example, in the absence of an overall association with pediatric bipolar disorder, Serretti and Smeraldi [1999] found excess short-repeat alleles of an intronic polymorphism of DRD2, the dopamine D2 receptor gene, in patients with disorganized symptoms. Another candidate gene, COMT, which codes for catechol-o-methyltransferase, a dopamine-metabolizing enzyme, had no detectable association with early-onset bipolar disorder or ultradian rapid cycling [Geller and Cook, 2000], which contrasts with the small but reliable association noted for adult-onset bipolar disorder [Craddock et al., 2001]. The failure to document this association in the early-onset sample may simply reflect inefficient power of a relatively small sample to detect a significant association of such small magnitude, or it may indicate a difference in the genetic etiologies of early- and later-onset bipolar disorders. Of the candidate genes examined to date, none seems particularly promising for a major role in early-onset forms of the illness, but contributions to specific symptoms or subtypes of the disorder cannot be excluded.
Based on numerous reports of phenotypic anticipation in bipolar disorder [McInnis et al., 1993], much effort has been directed at identifying the genetic basis for the phenomenon. In other disorders that show patterns of anticipation, such as Huntingtons disease [Morell, 1993], fragile-X syndrome [Kremer et al., 1991], spinal and bulbar muscular atrophy [La Spada et al., 1992], and myotonic dystrophy [Tsilfidis et al., 1992], anticipation is known to involve expansion of segments of the genome consisting of trinucleotide repeats. Thus, it is reasonable that such a mechanism may account for the earlier onset and increased severity of bipolar disorder in subsequent generations of families through which the disorder is segregating. Several investigations have sought expansions of genomic CAG/CTG repeats (coding for polyglutamine tracts) in younger individuals affected with bipolar disorder. Schurhoff et al. [2000] found no evidence for such a process in a limited sample of individuals with adolescent-onset bipolar disorder; however, this conclusion was based on the absence of protein products containing expanded polyglutamine tracts, rather than the absence of expanded nucleotide sequence repeats.
Vincent et al. [1999] obtained some evidence to support a repeat-expansion mechanism in pediatric bipolar disorder, but the patient sample in which a higher prevalence of long-repeat alleles was observed was comprised of individuals with schizophrenia, bipolar disorder, juvenile-onset depression, or borderline personality disorder. Although specific information was not provided for each patient group, larger alleles were reported as most prevalent in schizophrenic patients, and no other diagnostic group, including bipolar disorder, differed significantly from the control group in the prevalence of longer-repeat alleles. Furthermore, examination of a specific CAG/CTG repeat polymorphism within the Expanded Repeat Domain 1 on chromosome 17 actually revealed a higher prevalence of low-repeat alleles among patients with pediatric onset of bipolar disorder [Verheyen et al., 1999], but the significance of this difference did not withstand corrections for multiple testing. Thus, the role of polyglutamine tract expansion in the etiology of early-onset bipolar disorder remains unresolved, but an appreciable contribution to risk seems unlikely.
An elevated rate of early-onset bipolar disorder has been observed among patients with velo-cardio-facial syndrome, a genetic disorder that produces numerous physical and psychological anomalies [Papolos et al., 1996]. Because the cause of velo-cardio-facial syndrome was identified as an interstitial microdeletion of chromosome 22 [Scambler et al., 1992], it is possible that a pediatric bipolar disorder locus is located in this region. Patients with 22q deletion syndrome also exhibit a higher rate of schizophrenia [Murphy et al., 1999; Usiskin et al., 1999]. This may point to a shared susceptibility gene for bipolar disorder and schizophrenia, as has been postulated [Berrettini, 2000].
Psychiatric Comorbidity and Early Age at Onset
There is robust evidence for comorbidity between ADHD, conduct disorder (CD), and bipolar disorder. Epidemiological studies of child psychiatric disorders suggest that comorbidity is more the rule than the exception [Caron and Rutter, 1991; Angold et al., 1999]. For ADHD, comorbidity with CD [Biederman et al., 1991] and with bipolar has been well documented [Borchardt and Bernstein, 1995; West et al., 1995; Biederman et al., 1996], as has comorbidity between conduct and bipolar disorders [Kutcher et al., 1989; Geller et al., 1994; Kovacs and Pollock, 1995; Wozniak et al., 1995a; Biederman et al., 1996; Faraone et al., 1998b]. For example, Kovacs and Pollock [1995] reported a 69% rate of CD in a referred sample of youth with mania. Another study found that 42% of hospitalized youth with mania had comorbid CD [Kutcher et al., 1989].
Analyses of two independent data sets [Faraone et al., 1991; Biederman et al., 1992] found elevated rates of CD and antisocial personality disorder among relatives of ADHD + CD boys, but not among relatives of ADHD boys without CD. This suggested a qualitative familial difference between ADHD probands with and without CD. Moreover, only the relatives of ADHD + CD probands were at increased risk for ADHD + CD, and rates of ADHD + CD among relatives of other ADHD probands were no different from those observed among relatives of controls. This pattern of results was maintained over a 4-year follow-up period [Faraone et al., 1997a] and was replicated in a sample of ADHD girls [Faraone et al., 2000]. Taken together, these results supported the hypothesis that ADHD + CD is familially distinct from other cases of ADHD. If ADHD and CD had been independently transmitted, there should be no comorbidity between ADHD and CD among the relatives of CD + ADHD probands. Additionally, the low risk for CD among relatives of ADHD probands without CD suggests that these probands are qualitatively distinct from the probands with comorbid antisocial disorders instead of quantitatively different.
To address these findings further, Faraone et al. [1995] proposed a family-based stratification that defined a family as antisocial if either the proband had CD or a parent had antisocial personality disorder. This was an extension of the work of August and Stewart [1983], who subtyped families solely on the presence of parental antisocial disorders. Even if ADHD + CD is a familial disorder, we would not expect all ADHD children with the underlying predisposition to exhibit CD. For example, CD may not be expressed if the child had been assessed prior to the onset of CD or if CD does not always result from the familial predisposition. However, in cases where CD was not expressed, the presence of antisocial personality in a parent would signal the presence of the familial predisposition. Compared with parents from nonantisocial families, parents from antisocial families had significantly higher rates of ADHD, major depression, anxiety disorders, and substance use disorders. Compared with siblings from nonantisocial families, those from antisocial families had significantly higher rates of ADHD, CD, drug abuse or dependence, and major depression [Faraone et al., 1995]. The differences between antisocial and nonantisocial families held up over a 4-year prospective follow-up [Faraone et al., 1998a]. These differences between families provided further support for our hypothesis that the combination of ADHD and antisocial disorders indicated a distinct subtype of ADHD.
Data from other groups support the distinction of ADHD + CD as a familial subtype. Notably, Szatmari et al. [1993] confirmed the familial coaggregation of ADHD and CD in a population-based, epidemiologic family study, as did Silberg et al. [1996] in a population-based twin study. The twin study found that the genes influencing variation in hyperactivity scores were also responsible for variation in conduct problems. Between 76% and 88% of the correlation between hyperactivity and conduct scores were attributed to genes. The authors concluded that the results were consistent with the existence of a biologically based group of children who manifest both hyperactivity and conduct disturbances.
In a pilot study, Wozniak et al. [1995b] examined children with bipolar disorder and their relatives with structured psychiatric interviews. Among the first-degree relatives of children with bipolar disorder, a significantly greater number met criteria for mania with associated impairment than did the relatives of the ADHD only and normal control groups. The relatives of ADHD and control probands did not differ in their risk for bipolar disorder. Both the bipolar disorder and ADHD groups had significantly higher rates of ADHD than normal controls. Among the ADHD first-degree relatives of bipolar disorder children who met criteria for ADHD, 42% also met criteria for bipolar disorder. In contrast, only 3% of the non-ADHD relatives met criteria for bipolar disorder (P < .01). Thus, there was significant comorbidity between bipolar disorder and ADHD among the relatives of children with bipolar disorder (i.e., the two disorders cosegregated in these families).
The above findings indicate that when bipolar disorder and ADHD exist comorbidly, both diagnoses are present. Furthermore, the finding of cosegregation between mania and ADHD is consistent with the hypothesis that bipolar disorder plus ADHD may be a distinct familial syndrome [Pauls et al., 1986a, 1986b].
Faraone et al. [1997b] replicated the above findings in an entirely different sample. Relatives of ADHD/bipolar disorder children had significantly higher rates of both ADHD and bipolar disorder than controls. However, compared with relatives of ADHD without bipolar disorder children, the relatives of the ADHD/bipolar disorder children had significantly higher rates of bipolar disorder, but not of ADHD. Compared with controls, the relatives of nonbipolar ADHD probands had significantly higher rates of ADHD, but not bipolar disorder. Among relatives, rates of ADHD/bipolar disorder were significantly higher in relatives of ADHD/bipolar disorder probands. Among the ADHD relatives of the ADHD/bipolar disorder probands, 55% had bipolar disorder. In contrast, only 5% of the non-ADHD relatives of the bipolar, ADHD probands had bipolar disorder. Thus, like ADHD and CD, ADHD and bipolar disorder cosegregated among the relatives of the ADHD/bipolar disorder probands. Faraone et al. [2001] extended the above findings in boys to a sample of girls.
Faraone et al. [1998b] sought to clarify if antisocial-ADHD and bipolar-ADHD were the same or different subtypes of ADHD. They reported significant associations between family history of ADHD and family history of antisocial disorder, family history of ADHD and family history of bipolar disorder, and family history of bipolar disorder and family history of antisocial disorder. Each of these associations had a similar interpretation: having a family history of one disorder made it more likely that the proband had a family history of the other disorder.
The log linear analyses also found a significant three-way association between the three family history variables. This three-way association indicates that the pairwise associations between any two family history variables depend on whether the third family history variable is positive or negative. For example, for probands that did not have a family history of antisocial disorders, the odds ratio for the association between ADHD family history and bipolar family history was 11.9, indicating a robust association. For other probands, the odds ratio was 1.1, indicating no association. Moreover, when families were stratified into bipolar, antisocial, and other types, few differences emerged between the bipolar and antisocial families in terms of psychopathology, measures of delinquency and aggressive behavior, psychosocial functioning, and cognitive functioning. These results provide fairly consistent support for the hypothesis that the antisocial- and bipolar-ADHD subtypes are different manifestations of the same familial condition.
To confirm these findings, Biederman et al. [2000a] examined competing hypotheses regarding the relationship between bipolar disorder and CD using family study methodology. They compared relatives of four proband groups defined by the presence or absence of CD and bipolar disorder in the proband: 1) CD + bipolar disorder; 2) bipolar disorder without CD (bipolar disorder); 3) CD without bipolar disorder (CD); and 4) controls without bipolar disorder or CD. Results showed that relatives of bipolar disorder probands had an increased risk for bipolar disorder but not CD. Relatives of CD probands had an increased risk for CD but not for bipolar disorder. And relatives of CD + bipolar disorder probands had an elevated risk for both disorders. Among relatives in this latter group, bipolar disorder and antisocial disorders showed significant cosegregation. That is, CD + bipolar disorder was significantly elevated among relatives of CD + bipolar disorder probands but was rare among the relatives of the other proband groups, suggesting that this configuration of disorders represents a distinct familial subtype. These family study results also support and extend previous findings in probands that documented that children satisfying diagnostic criteria for both CD and bipolar disorder have prototypical symptoms of both disorders irrespective of the comorbidity with the other disorder, thus indicating that both syndromes were present in the affected children.
SUMMARY AND CONCLUSIONS
The familial aggregation and genetic transmission of bipolar disorder has been consistently demonstrated through family, twin, and adoption studies. These studies show that early onset of the disorder confers a greater familial risk to relatives, but relatively little is known about genetic transmission in families having juvenile-onset cases.
Although there are few genetic epidemiologic studies of juvenile-onset bipolar disorder, the evidence from extant studies is clear: this form of bipolar disorder aggregates in families and appears to be highly heritable. It is not yet known if this aggregation is due to genetic or environmental sources as twin studies of juvenile-onset bipolar disorder are lacking; however, twin and adoption studies of broad-spectrum bipolar disorder clearly demonstrate the high heritability of the disorder, with a minority of disease risk attributable to shared and unique environmental experiences. As a potentially more severe and more organic form of bipolar disorder, it is reasonable to suspect that pediatric-onset bipolar disorder is at least as heritable as bipolar disorder ascertained without regard to age of onset, although this remains to be established.
Molecular genetic studies of early-onset bipolarity have failed to reveal reliably significant contributors to disease risk, but such efforts have only recently begun. As suggested by Todd et al. [1993], because it may be a more genetically influenced condition, early-onset bipolar disorder may be the most useful bipolar subtype in identifying genes that influence risk for all cases of the illness. The study of early-onset disease forms has proven extremely useful in identifying genetic contributors to various diseases, including Alzheimers disease [Levy-Lahad et al., 1995a,b] and diabetes mellitus [Davies et al., 1994].
Presumably, individuals with an early onset of disease will produce stronger genetic signals in linkage and association studies due to less genetic heterogeneity, greater penetrance of risk alleles, or both. Furthermore, such cases may have a higher genetic loading of risk alleles (i.e., they possess a greater number of the critical alleles of risk genes), thus facilitating the detection of any one or more risk genes out of the many that may exist. Although early-onset bipolar disorder may be a product of an accumulation or convergence of numerous environmental and, especially, genetic risk factors in individual members of multiply-affected families, the early-onset form of the disorder may actually be due to altogether different genes than the more typical adult-onset form (i.e., early-onset bipolar disorder may be a phenocopy of the adult-onset form, or vice versa).
The extent studies of pediatric bipolar disorder should be viewed in light of the fact that the phenotype is complex and comorbid with many other disorders, including CD, anxiety disorders, substance use disorders, and ADHD [Carlson, 1990; Carlson and Weintraub, 1993; Wozniak et al., 1995a, 2001, 2002; Geller and Luby, 1997; Geller et al., 2000, 2002; Carlson et al., 2002]. Thus, future genetic studies should address the breadth of the phenotype, its discrimination from other disorders, and its continuity with bipolar disorder in adulthood. As reviewed above, some evidence suggests that the early-onset bipolar disorder phenotype is associated with CD and ADHD. Thus the co-occurrence of these disorders may define a homogeneous subgroup of patients that may be useful for molecular genetic research.
Acknowledgements
This work was supported in part by National Institutes of Health grants R01MH57934, R01HD37694, and R13MH59126 (to Dr. Faraone) and grants RO1MH43518, R01MH59624, and R01MH60485 (to Dr. Tsuang), and by grants (principal investigator J. Biederman) from the Stanley Foundation and Johnson and Johnson.
