Pharmacogenetics and mood stabilization in bipolar disorder
Abstract
Bipolar disorder is a severe psychiatric disease characterized by varying treatment response among individual patients. Effects of certain treatments, for instance, lithium, can be predicted from clinical characteristics of patients and their family histories. This led to a suggestion that a treatment response could identify subtypes of bipolar disorder particularly suited for gene-mapping studies. In this paper we review family and molecular studies of bipolar disorder responsive to lithium, as well as studies aiming to identify polymorphisms associated with the treatment response itself. While molecular genetic research and gene expression studies promise to bring new insights into the pathophysiology of the illness and the nature of treatment response, and thus provide new information for better treatment of bipolar disorder in the future, results from family studies and studies of clinical correlates of treatment response may already be utilized in the management of bipolar disorder. © 2003 Wiley-Liss, Inc.
INTRODUCTION
Bipolar disorder is a complex psychiatric illness. The etiology of bipolar disorder is not fully known, but it is recognized to include a genetic component that is well supported by the results of adoption, twin, and family studies. The lifetime prevalence is approximately 1% for the general population and rises up to 10% for first-degree relatives of people with bipolar disorder.
The etiology of bipolar disorder is not fully known, but it is recognized to include a genetic component that is well supported by the results of adoption, twin, and family studies. The lifetime prevalence is approximately 1% for the general population and rises up to 10% for first-degree relatives of people with bipolar disorder.
Bipolar disorder represents one of the major health problems due to a high rate of disability, and an increased mortality. In particular, suicides and deaths from cardiovascular causes account for the increased mortality in bipolar subjects [Osby et al., 2001]. Similarly, the socioeconomic burden of bipolar disorder is considerable. In 1991, the annual cost of bipolar disorder in the United States totaled $45 billion [Wyatt and Henter, 1995]. In the United Kingdom, the annual costs have been estimated to be £2 billion in 1999/2000 prices [Das Gupta and Guest, 2002]. Along with other psychiatric disorders—unipolar depression, alcohol use disorders, and schizophrenia—bipolar disorder ranks among major leading causes of disability. In 1990, the World Health Organization identified bipolar disorder as the sixth leading cause of disability for the age group of 15–44 years; by 2000, it had become the fifth leading cause [The World Health Report, 2001].
The management of bipolar disorder includes acute treatment of mania and depression and long-term maintenance treatment. The goal of long-term treatment is to prevent recurrences of the illness. Given the evidence from clinical trials, lithium is the most effective prophylactic treatment for bipolar disorder. In the long-term setting, lithium has been shown to reduce the risk of suicide and normalize the increased cardiovascular mortality as well [Ahrens et al., 1995]. Other medications commonly used for prophylactic treatment are anticonvulsants (e.g., valproate, carbamazepine, or lamotrigine) and atypical antipsychotics (e.g., olanzapine or clozapine). Lithium, carbamazepine, valproate, and olanzapine also exert antimanic effects. Severe mania is usually treated with an addition of antipsychotics to a mood stabilizer; depression is treated with an addition of antidepressants.
In terms of treatment efficacy, bipolar disorder appears to be a heterogeneous group of conditions. The genetic factors are likely to be responsible for part of the variability among individuals in their response to treatment. This should not be surprising, as genes encode a number of molecules such as enzymes or transporters that are involved in the mechanism of action of psychotropic drugs. The role of genetic factors in the response to a particular medication may be investigated using various approaches such as molecular genetic and family studies or studies of clinical correlates associated with response. In this paper, we will summarize the findings of studies searching for correlates of response to lithium, valproate, and carbamazepine. In the next part, we will focus on genetic research based mainly on response to lithium treatment, as there is a paucity of similar data for other drugs used in long-term management of bipolar disorder. Finally, we will discuss gene expression studies and their utilization in bipolar disorder.
CLINICAL TRAITS ASSOCIATED WITH TREATMENT OUTCOME
Lithium
Lithium, a mood stabilizer that has been used for treatment of bipolar disorder for more than five decades, still remains the first-line treatment. The response to lithium has been shown to be predictable to a considerable extent. Bipolar patients who benefit from prophylactic lithium treatment share certain characteristics with respect to their symptoms, clinical course of the illness, and family history.
Bipolar patients who benefit from prophylactic lithium treatment share certain characteristics with respect to their symptoms, clinical course of the illness, and family history.
The strongest predictors are the quality of remissions, low frequency of episodes prior to lithium treatment, nonrapid cycling course, normal profile on Minnesota Multiphasic Personality Inventory (MMPI), low psychiatric comorbidity, diagnosis of primary mood disorder, and family history [Grof, 1983; Grof et al., 1993].
Carbamazepine
Carbamazepine, an anticonvulsant drug, has been shown to be superior to placebo in the treatment of acute mania [Lerer et al., 1987; Post et al., 1987]. The group of Greil et al. [1998] carried out a randomized clinical trial—Multicenter Study of Long-term Treatment of Affective and Schizoaffective Methods (MAP) study—to investigate the efficacy of lithium vs. carbamazepine in maintenance treatment of bipolar disorder. The authors divided patients into two groups, classical (bipolar I without comorbidity and mood-incongruent delusions) and nonclassical (all other patients). A tendency in favor of carbamazepine was found for the nonclassical sample [Greil et al., 1998]. The results of subsequently published subgroup analyses of the MAP study suggest that carbamazepine is less effective in maintenance treatment of bipolar I disorder than in bipolar II disorder and bipolar disorder not otherwise specified [Greil and Kleindienst, 1999a,b].
Valproate
Valproate is another anticonvulsant that is widely used as a prophylactic treatment for bipolar disorder and has been shown to be superior to placebo in the treatment of acute mania [Bowden et al., 1994]. While the clinical features predictive of the outcome of lithium therapy are relatively well established, results of studies investigating factors associated with a response to valproate remain equivocal. Several findings suggested that valproate might be particularly effective in rapid cycling bipolar disorder and in the treatment of mixed states [Calabrese and Delucchi, 1990; Calabrese et al., 1993]. Calabrese et al. [1993] studied the outcomes of treatment with valproate in 55 bipolar rapid cyclers during a period of 7.8 months and in a follow-up study with 101 rapid cyclers for 17.2 months [Calabrese and Delucchi, 1990]. Both studies showed acute and prophylactic efficacy of valproate in mixed states and mania. The authors also suggested what might be the predictors of response to valproate in bipolar rapid cycling. They evaluated the outcome with respect to acute and prophylactic antimanic and antidepressant effects. For acute and prophylactic antimanic response the predictors were nonpsychotic mania, decreasing or stable episode frequencies, mild mania, and mixed states. For acute and prophylactic antidepressant response the predictors included nonpsychotic mania, increasingly severe mania, and the absence of borderline personality disorder [Calabrese and Woyshville, 1995]. Swann et al. [2002] studied antimanic response in 179 bipolar patients. They compared the effects of lithium, valproate, and placebo in four subtypes of mania. Valproate was superior to lithium in irritable dysphoric subtype and improved hostility and impulsivity significantly better than placebo.
In a recent study we evaluated over 120 subjects treated in a naturalistic setting. Of 69 patients treated with valproate, nine met the criteria of full response but could not be differentiated from nonresponders based on clinical profiles [Garnham et al., 2002].
These findings are important from the clinical point of view as they may assist in identifying subjects who are likely to respond to individual treatments. They may also be relevant from the research point of view. Patients who show response to a certain treatment may represent a genetically more homogeneous subgroup of bipolar disorder. Particularly, lithium responders seem to be a very consistent group of bipolar patients in terms of their course of illness. This may point to the assumption that they share a common background that could be genetic. More importantly, the support for genetic basis of lithium response has been established through family studies (see below). It is not clear to what extent genetic factors may be involved in response to other mood stabilizers. While the above data suggest that at least one subgroup of bipolar patients can preferentially benefit from valproate, there is relatively little information about their family histories. Post et al. [1987] suggested that acute antimanic response to carbamazepine was associated with family history negative for bipolar disorder. In a study comparing responders to lamotrigine or lithium, first-degree relatives of lamotrigine responders had higher risk of schizoaffective disorder, major depression, and panic attacks, while first-degree relatives of lithium responders had higher risk of bipolar disorder [Passmore et al., 2003].
Clinical observations seem to suggest that patients who benefit from one mood stabilizer may not respond to another one. To date, very little systematic work has been done to test this assumption. One such study, a prospective study evaluating 38 patients treated with lithium or divalproex in the naturalistic setting, showed that nonresponders to lithium responded well to divalproex and vice versa [Ghaemi and Goodwin, 2001].
In summary, the limited data available so far are compatible with the view of treatment response being associated with clinically distinguishable subtypes of bipolar disorder. Whether these represent truly separate entities remains to be clarified. In the case of lithium, several research groups have attempted to study treatment-responsive patients as such subgroup and/or to search for genes that explain the difference between those who do and those who do not benefit from the treatment. These studies are detailed in the next part of this review.
PHENOTYPE FOR PHARMACOGENETIC STUDIES
The critical factor in genetic research is the phenotype definition, that is, the characterization of treatment response. Treatment response should reflect a stable characteristic of an individual to ensure the validity and reproducibility of results. With this in mind, the criteria for response should allow for differentiation between the true effect of therapy and other factors that could influence the response itself as, for instance, spontaneous remissions, concomitant medication, or compliance. In order to distinguish between spontaneous and pharmacological remission, the treatment must be administered for a certain period of time, and in the case of lithium, plasma levels must be maintained within therapeutic range (which reflects compliance as well). To define lithium response, some authors proposed criteria based on high recurrence risk (judged by number and frequency of episodes prior to lithium treatment), lithium monotherapy, and no remissions on lithium therapy adequate in dose within minimum of 3 years [Grof et al., 1994]. Such strict criteria for response further help rule out the effect of natural course of bipolar disorder and identify true responders.
Several options are available to quantify treatment response. One of them is selecting cutoff points for certain categories of response. For example, the whole sample may be split into responders and nonresponders in a dichotomous manner, or investigators can select extreme groups such as unequivocal responders and nonresponders using predefined criteria. The first method has an advantage in terms of generalization of the results because it better reflects the overall bipolar population. The latter, a comparison of extreme phenotypes, has a better chance to find actual differences between groups. Another option is to evaluate the treatment outcome by means of quantitative measures. This approach includes the episode frequency on-and-off lithium treatment or Affective Morbidity Index, which, besides the frequency of episodes, also takes into account their severity [Coppen et al., 1976].
The fact that different research groups define response differently needs to be considered when interpreting sometimes contradictory results of pharmacogenetic studies.
GENETICS OF LITHIUM-RESPONSIVE BIPOLAR DISORDER
Family Studies
Studies evaluating family history of patients who benefit from lithium have been published since the 1970s. Mendlewicz et al. [1973] first reported positive association between lithium response and family history of bipolar disorder. Several subsequent papers reported findings consistent with those of Mendlewicz [Prien et al., 1974; Zvolsky et al., 1974; Svestka, 1979; Maj et al., 1984; Smeraldi et al., 1984; Abou-Saleh and Coppen, 1986; Grof et al., 1994]. Grof et al. [1994] examined the family history in 121 families. In this study, lithium responders had higher prevalence of bipolar disorder among their relatives, while nonresponders had higher rates of schizophrenia in families
Grof et al. [1994] examined the family history in 121 families. In this study, lithium responders had higher prevalence of bipolar disorder among their relatives, while nonresponders had higher rates of schizophrenia in families.
(for a more detailed review of family studies in patients on lithium, see Grof et al. [1994] and Alda [2002]). Recently, Grof et al. [2002] tested the hypothesis whether affected relatives of lithium-responsive probands have a higher likelihood of responding to the same treatment. They found that significantly more subjects from the group of relatives (67%, 16 of 24) showed clear-cut response to lithium than from the control sample (35%, 14 of 40). The above reports show clustering of bipolar disorder and the last cited study clustering of lithium response, as such in families of lithium responders. This supports the notion that bipolar disorder responsive to lithium is a distinct variant of bipolar disorder characterized by higher heritability. Alternatively, lithium responders could be viewed as more or less typical patients sharing an independent genetic factor associated with the response. However, family studies of lithium responders alone cannot separate these two possibilities. Thus, genes identified in case-control comparisons of responders and nonresponders could be relevant either for etiology or for treatment response. In the following, we report these studies together (Table I).
| Gene/marker | Design/outcome | Assessment of Li response | Sample | Results | Reference |
|---|---|---|---|---|---|
| Tyrosine hydroxylase | Li responders vs. controls | Prospective, treatment for >3 years; research criteria | 48 bipolar and 6 unipolar, 94 controls | No association with Li responsive BD |
Cavazzoni et al. [1996] |
| 6 markers on chromosome 18, including Golf | Li responders vs. controls | Prospective, treatment for >3 years; research criteria | 47 bipolar, 8 unipolar; 94 controls | No association with Li responsive BD |
Turecki et al. [1996] |
| MN blood group | BD responsive to Li vs. other diagnostic groups | Clinical, long term follow up | 174 bipolar, 176 unipolar, 98 schizophrenics, 331 controls | Lower frequency of NN in BD vs. other groups |
Alda et al. [1998] |
| PLCG1 | Li responders vs. controls | Prospective, treatment for >3 years; research criteria | 136 bipolar, 163 controls (IGSLI sample) | Positive association of Li responsive BD with PLCG1/5 (P = 0.033) |
Turecki et al. [1998] |
| INPP1 | Li treated BD vs. controls, responders vs. nonresponders | Retrospective | Norwegian sample: 23 bipolar, 20 controls; Israel sample: 54 bipolar, 50 controls | No difference between BD and controls; association with Li response in Norwegian sample |
Steen et al. [1998] |
| D3 | Li treatment outcome | Prospective, frequency of episodes | 43 bipolar and 12 major depression | No association with Li outcome |
Serretti et al. [1998] |
| MAOA | Li responders vs. controls | Prospective, treatment for >3 years; research criteria | 138 bipolar, 108 healthy controls (IGSLI sample) | No association with Li responsive BD |
Turecki et al. [1999] |
| DRD2, DRD4, GABRA1 | Li treatment outcome | Prospective, frequency of episodes | 100 bipolar, 25 major depression | No association with Li outcome |
Serretti et al. [1999a] |
| TPH | Li treatment outcome | Prospective, frequency of episodes | 90 bipolar, 18 major depression | Worse response to Li in subjects with TPH*A/A (P = 0.046) |
Serretti et al. [1999b] |
| 5-HTTLPR | Li responders, nonresponders vs. controls | Long term follow up | 67 bipolar (49 responders to Li, 18 nonresponders), 103 controls | Higher frequency of l allele in nonresponders vs. controls |
Del Zompo et al. [1999] |
| GABRA3, GABRA5, GABRB3 | Li responders vs. controls | Prospective, treatment for >3 years; research criteria | 138 bipolar, 108 healthy controls (IGSLI sample) | No association with Li responsive BD |
Duffy et al. [2000] |
| CRH, PENK | Li responders vs. controls | Prospective, treatment for >3 years; research criteria | 138 bipolar, 108 healthy controls (IGSLI sample) | No association with Li responsive BD |
Alda et al. [2000] |
| Five (CAG)n | Li responders vs. controls | Prospective, treatment for >3 years; research criteria | 138 bipolar, 108 healthy controls (IGSLI sample) | No association with Li responsive BD |
Turecki et al. [2000] |
| 5-HT2A, 5-HT2C, 5-HT1A | Li treatment outcome | Prospective, frequency of episodes | 102 bipolar, 22 major depression | 5-HT2A, 5-HT2C not associated with Li outcome; no polymorphism of 5-HT1A identified |
Serretti et al. [2000] |
| PLCG1 | Li responders vs. nonresponders and controls | Retrospective | 61 bipolar (29 Li responders, 16 nonresponders, 16 partial responders), 50 controls | PLCG1/8 more frequent in Li responders vs. controls (P = 0.032), and in BD vs. controls (P = 0.05) |
Lovlie et al. [2001] |
| 5-HTTLPR | Li treatment outcome | Prospective, frequency of episodes | 167 bipolar, 34 major depression | Subjects with ss genotype had worse response to Li than those with sl or ll |
Serretti et al. [2001] |
| COMT, MAO-A, Gβ3 | Li treatment outcome | Prospective, frequency of episodes | 160 bipolar, 41 major depression | No association with Li oucome |
Serretti et al. [2002] |
- BD, bipolar disorder.
It should be noted that several reports failed to confirm the association between family history and response to lithium [Dunner et al., 1976; Misra and Burns, 1976; Strober et al., 1988; Engstrom et al., 1997; Coryell et al., 2000]. The factors that may be responsible for such discrepant findings include methodological differences in assessing the treatment response and have been discussed elsewhere [Alda et al., 1999].
Mode of Inheritance
Several studies applying pedigree analysis on data from families of probands with defined response to lithium support the involvement of a major gene effect in lithium responders. The first evidence comes from segregation analysis by Smeraldi et al. [1984]. Alda et al. [1994] investigated the mode of inheritance in families of 71 probands, responders to long-term lithium therapy. The genetic model that fit the data best was the autosomal recessive. These findings were replicated in a follow-up study of 25 different families of lithium-responsive probands recruited in a different geographic location [Alda et al., 1997]. The results of segregation analyses need to be interpreted with caution. A finding that familial transmission is compatible with a specific genetic model does not mean that this is the true mode of transmission or the only model of inheritance possible.
A finding that familial transmission is compatible with a specific genetic model does not mean that this is the true mode of transmission or the only model of inheritance possible.
Taken together, the findings in lithium-responsive families seem to suggest that the mode of inheritance could be relatively simple, compatible with the already proposed view that this population may be well suited for gene-mapping studies.
Molecular Genetic Studies Based on Lithium Outcome
Several researches investigated responders to lithium using association and/or linkage strategies. These studies can be divided into those comparing responders to nonresponders with the goal of identifying genes involved in the response itself, and studies aiming to find susceptibility genes for bipolar disorder. These latter studies are based on the assumption that responders to lithium are a subgroup that is more homogeneous, with a stronger involvement of genetic mechanisms. However, if response to treatment indeed identifies a subtype of illness with a specific etiology, then comparisons of responders and nonresponders may lead to genes relevant for etiology as well
If response to treatment indeed identifies a subtype of illness with a specific etiology, then comparisons of responders and nonresponders may lead to genes relevant for etiology as well.
(see above). Molecular genetic studies of responders to lithium are summarized in Table I. Most results to date have been negative and/or inconclusive. The group of Steen et al. [1998] focused on the inositol polyphosphate-1-phosphatase gene (INPP1). They found an association of one polymorphism in the INPP1 gene with a lithium response in the sample from Norway but not in the sample from Israel [Steen et al., 1998]. We found an association and suggestive linkage in a subset of families to the gene for phospholipase C gamma [Turecki et al., 1998]. Lovlie et al. [2001] reported similar results with the same polymorphism in a Norwegian sample. These results are promising, but need further clarification, as others [Ftouhi-Paquin et al., 2001] were unable to identify any potentially functional polymorphism in this gene. Another set of interesting findings relates to the serotonin transporter gene. Previous studies with a functional polymorphism in the regulatory region of the gene suggested that individuals homozygous for the short (s) allele of the marker have a poorer response to serotonin reuptake inhibitors [reviewed by Alda, 2001]. Similar results for lithium have been obtained by Serretti et al. [2001]. On the other hand, Del Zompo et al. [1999] found the opposite—a higher frequency of the l allele among nonresponders in comparison with healthy controls. Mundo et al. [2001] examined the same polymorphism in patients who experienced manic or hypomanic episodes in conjunction with antidepressant treatment. These patients had a higher prevalence of the s allele in comparison with subjects with no evidence of antidepressant-induced mania.
Recently, our group performed a complete genome scan in 31 families ascertained through probands with excellent response to lithium [Turecki et al., 2001]. The analyses of 378 markers gave the best results on chromosome 15q14 (marker ACTC, lod score = 3.46, P = 1.4 × 10−5) for the phenotype of bipolar disorder and recurrent depression, using a genetic model based on results of previous segregation analysis of the studied sample. When we used the treatment response as phenotype, the best lod score was on 7q11.2 (for marker D7S1816, lod score of 1.53 with an empirical P value of 0.003).
GENE EXPRESSION STUDIES
Methods like sequential analysis of gene expression (SAGE), differential display-polymerase chain reaction (DD-PCR), and DNA arrays open a broad spectrum of possibilities in the field of genetic and pharmacogenetic research in bipolar disorder.
Similarly, as linkage and candidate gene approach, they may be used as another or complementary tool for searching for the genes involved in disease process and treatment response. In development of new treatments, studies of gene expression open up a new era, as they will facilitate the search for targets for treatment effect and side effects of drugs via tracking changes in expression of multiple genes following treatment. This approach has already been applied in several studies using DD-PCR and microarray technologies [Chen et al., 1999; Wang et al., 1999, 2002; Hua et al., 2000, 2001a,b]. Some findings, for example, changes in gene expression of GRP78 (78-kDa glucose-regulated protein) or (PEBP)2β transcription factor polyomavirus enhancer-binding protein, have been suggested to be relevant to the putative neuroprotective effect of mood stabilizers [Chen et al., 1999; Wang et al., 1999].
So far, no studies of gene expression based on treatment response have been published. Not surprisingly, methodology for such a design encounters several limitations. If we skipped the difficulties with statistical analysis and interpretation of the amount of obtained data, which is an issue inherent to large-scale expression studies in general, the first concern in our case is the source of mRNA. There is no known animal model that would be appropriate for this kind of research. Postmortem brain tissue may be an accessible source provided that response can be retrospectively assessed; however, the gain of such studies is limited by the methodological problems related to the investigation of postmortem tissue, such as variance between samples in postmortem delay and cause of death. To select a suitable control sample and collect a sample that would be big enough to allow for differentiation of the patterns unique to response may be difficult. Another option is to use lymphocytes and cell lines derived from peripheral tissues. The main concern in this case is that it is not clear to what extent these cells reflect the treatment-induced expression patterns in brain. On the other hand, the great advantage of cell lines is that they allow the tracing of changes in gene expression following in vitro treatment with various drugs in cells obtained from individuals with a defined response. It can be argued that the subject's current treatment status represents a possible confounder. However, this effect could be minimized by repeated passages; thus changes in expression patterns would reflect the true effect of experimental treatment. Moreover, sample for this method is very easy to access from peripheral blood.
It is possible that expression profiles/phenotypes associated with response to a particular drug will be determined and will become a useful tool in clinical practice for selecting an effective treatment.
CONCLUSIONS
Studies searching for clinical correlates of treatment response and family studies gained results that can be helpful in predicting benefit from treatment, particularly from lithium. Molecular genetic research yielded several interesting results, which will need to be further confirmed before they can be applied in clinical practice. Advances in gene expression technologies promise to clarify our understanding of the mechanisms of action of medication treatments and may lead to more rational treatments in the future.
Acknowledgements
Some of the studies reported in this paper have been supported by grants from Canadian Institutes of Health Research and from the National Alliance for Research in Schizophrenia and Depression to M. Alda.
