A Success at the End of an Era, and a Glimpse of Things to Come

The history of genetic studies of bipolar disorder can be divided into eras distinguished by the methodology that drove the big advances. The first era started roughly with Kraepelin in the 1890s; it rested on descriptive phenomenology and established the definition of what geneticists call the bipolar phenotype. The second era started with the early twin studies in the 1930s and relied on genetic epidemiology to show that bipolar disorder was largely a genetic disease. The third era, which we can call the linkage era, probably began with the first genetic linkage study of bipolar disorder in the late 1960s but did not really get under way until molecular markers became available for genetic mapping in the late 1980s.

The linkage era is now passing and may soon be overshadowed by the findings that are emerging from genetic association studies, especially the genome-wide studies that will come to dominate the field in the coming years. Linkage surveys the genomic landscape and reveals only the general topography; linkage alone cannot pinpoint the underlying disease-related gene. A more powerful, more sensitive tool has long been needed. Unlike linkage studies, which depend on the cosegregation of genetic markers with disease within families, association studies depend on the detection of markers whose alleles are more or less common among individuals with a disease compared with a group without that disease. (Association studies can also be carried out with family data, but the fundamental test still depends on comparing allele frequencies between families.) Linkage studies work best when a single gene plays the major role in disease risk. Association studies can detect genes with smaller effects but can still fail when there is no common genetic variant that contributes to disease in a substantial proportion of cases. In the worst-case scenario, where many different, individually rare variants could be involved, large-scale resequencing may be required (1) .

In the field of bipolar disorder genetics, the linkage era will not be widely mourned. The many linkage studies conducted over a period of more than 20 years have not produced signals that are robust and localized enough to allow the efficient identification of risk alleles. Even when risk alleles have been successfully identified within linkage regions, such as in the region on chromosome 13 harboring the gene DAOA (2) , the alleles have not been consistently replicated across samples (3) . Still, many of the linkage findings to date will probably prove to harbor one or more genes involved in bipolar disorder susceptibility. The problem is one of sorting the wheat from the chaff. Genome-wide association studies may provide some answers in the near future by revealing which of the linkage regions actually contain susceptibility genes. In this way, the modest findings of the linkage era can still inform our choices as we prioritize the many signals that will likely emerge from genome-wide association results.

There is a special case where linkage can produce strong signals, even in the face of substantial genetic heterogeneity. When the phenotype can be redefined in such a way that the revised definition of a case encompasses a narrower and more genetically homogeneous group, linkage studies may be able to detect a robust signal. Imagine that the standard case definition actually encompasses a set of diseases, each with their own set of genetic risk factors. Under this scenario, a narrower case definition might actually encompass a smaller set of genetic risk factors. This approach has been used with some success with forms of bipolar disorder that differ by clinical features, such as bipolar II (4) , lithium response (5) , bipolar disorder with comorbid panic (6) , psychotic bipolar disorder (7) , and bipolar disorder with mania at onset (8) . In each of these examples, weak linkage signals became substantially stronger when key clinical features were considered.

Now, in this issue, Jones et al. have shown that yet another clinical feature, in this case postpartum or puerperal onset, substantially increases linkage evidence to regions of chromosomes 16 and 8 that have been implicated in other studies (8 – 10) . Episodes of bipolar disorder have long been known to occur with increased frequency during the immediate postpartum period. As Jones et al. point out, these are typically severe episodes characterized by psychosis and, occasionally, violent or suicidal behavior. Since the postpartum period is accompanied by a host of psychological and physiological stressors, it is perhaps not surprising that vulnerable people might suffer onset or relapse of bipolar disorder during this time. Nevertheless, the fact that this clinical feature strongly implicates one or two specific chromosomal regions implies that bipolar disorder with postpartum onset may indeed be a genetically more homogeneous illness than bipolar disorder more generally defined.

These encouraging results are not without limitations. The postpartum timing of episodes was established retrospectively and could be subject to recall bias, although childbirth is often a strong memory signpost for women. Of course, most of the cases of bipolar disorder in the families studied by Jones et al. did not have a postpartum onset. Thus, we cannot conclude that bipolar disorder with postpartum onset is a distinct disease entity. Previously, Kassem et al. (8) showed that bipolar disorder with mania at onset was also linked to the same region on chromosome 16 implicated by Jones et al. Since bipolar disorder with postpartum onset often takes the form of a severe mania, it is possible that both studies have detected the same susceptibility locus whose alleles especially predispose to mania. This hypothesis can now be tested in future studies that focus on genes in the region.

Also in this issue, Matsuzawa et al. provide a taste of the power of genetic association studies to detect alleles that contribute to psychiatric disease, in this case, methamphetamine psychosis. Previous studies have shown that methamphetamine primarily targets the dopamine transporter, whose trafficking is controlled in part by protein kinase C alpha binding protein, encoded by the gene PICK1. Interestingly, PICK1 resides in a region on chromosome 22 that has been widely linked to schizophrenia, the psychotic features of which resemble those seen in methamphetamine psychosis. Matsuzawa et al. hypothesized that genetic variation in PICK1 may contribute to the risk of methamphetamine abuse and its associated psychotic complications. After genotyping markers in and around the gene in 209 methamphetamine abusers and 218 healthy comparison subjects, Matsuzawa et al. find that several markers and haplotypes are associated with methamphetamine abuse, psychosis, and spontaneous psychotic relapse. They also show that the associated alleles at some of these markers increase PICK1 transcription in a gene reporter assay. These findings must be viewed with caution, especially in light of the small sample size, and it remains to be seen whether they will be replicated in other studies. But this study suggests that the same kind of precision phenotyping that has been helpful in the study of bipolar disorder can also be useful with other psychiatric phenotypes.

The candidate gene approach may follow linkage approaches into obsolescence as denser sets of markers for genome-wide association studies become available, but there will probably always be room for good hypothesis-driven candidate gene studies, especially for phenotypes that might not easily provide the large sample sizes needed for genome-wide studies.