Phenotypic effects of a bipolar liability gene among individuals with major depressive disorder†
How to Cite this Article: Casamassima F, Huang J, Fava M, Sachs GS, Smoller JW, Cassano GB, Lattanzi L, Fagerness J, Stange JP, Perlis RH. 2010. Phenotypic Effects of a Bipolar Liability Gene Among Individuals With Major Depressive Disorder. Am J Med Genet Part B 153B:303–309.
Abstract
Variations in voltage-dependent calcium channel L-type, alpha 1C subunit (CACNA1C) gene have been associated with bipolar disorder in a recent meta-analysis of genome-wide association studies [Ferreira et al., 2008]. The impact of these variations on other psychiatric disorders has not been yet investigated. Caucasian non-Hispanic participants in the STAR*D study of treatment for depression for whom DNA was available (N = 1213) were genotyped at two single-nucleotide polymorphisms (SNPs) (rs10848635 and rs1006737) in the CACNA1C gene. We examined putative phenotypic indicators of bipolarity among patients with major depression and elements of longitudinal course suggestive of latent bipolarity. We also considered remission and depression severity following citalopram treatment. The rs10848635 risk allele was significantly associated with lower levels of baseline agitation (P = 0.03; β = −0.09). The rs1006737 risk allele was significantly associated with lesser baseline depression severity (P = 0.04; β = −0.4) and decreased likelihood of insomnia (P = 0.047; β = −0.22). Both markers were associated with an increased risk of citalopram-emergent suicidality (rs10848635: OR = 1.29, P = 0.04; rs1006737: OR = 1.34, P = 0.02). In this exploratory analysis, treatment-emergent suicidality was associated with two risk alleles in a putative bipolar liability gene. © 2009 Wiley-Liss, Inc.
INTRODUCTION
The bipolar spectrum concept, derived from Kraepelin's unitary view of manic-depressive illness, was first revived by Akiskal et al. 1977, and was subsequently endorsed by other researchers trying to address the limitations of categorical classification [Klerman, 1981; Cassano et al., 1999, 2002, 2004; Ghaemi et al., 2002; Sato et al., 2003; Angst and Cassano, 2005]. This dimensional concept suggests a continuum between major depressive and bipolar I disorder, with bipolar II representing an intermediate form [Angst, 1978, 2007].
In clinical studies describing differences between unipolar and bipolar depression as well as in long-term follow-up studies aimed at identifying the characteristics of “converters” from unipolar depression to bipolar I or II disorder, it is generally agreed that patients with bipolar spectrum illness have an earlier age of onset [Andreasen et al., 1988; Weissman et al., 1996; Perlis et al., 2006; Solomon et al., 2006], more depressive recurrences [Roy-Byrne et al., 1985; Andreasen et al., 1988; Perlis et al., 2006; Solomon et al., 2006] and a greater likelihood of having a family history of bipolar disorder [Akiskal et al., 1983; Perlis et al., 2006]. Differences in clinical presentation between unipolar and bipolar depression have also been suggested, such as a greater prevalence in bipolar subjects of psychomotor disturbance/retardation [Dunner et al., 1976; Akiskal et al., 1983; Parker et al., 2000], psychotic symptoms [Guze et al., 1975; Strober and Carlson, 1982; Akiskal et al., 1983; Coryell et al., 1995; Goldberg et al., 2001; Mitchell et al., 2001; Baethge et al., 2005; Goes et al., 2007], intraepisode irritability, anger, agitation and restlessness [Abrams and Taylor, 1980; Deckersbach et al., 2004; Mammen et al., 2004; Perlis et al., 2004; Benazzi and Akiskal, 2006], mood lability [Brockington et al., 1982; Akiskal et al., 1995], suicidality [Olfson et al., 2005; Perlis et al., 2006], and atypical features [Abrams and Taylor, 1980; Perugi et al., 1999; Benazzi and Rihmer, 2000; Mitchell et al., 2001; Serretti et al., 2002; Benazzi, 2003].
Despite the great interest in the topic, the bipolar spectrum concept is difficult to test empirically. Genetic studies may provide novel means of illuminating disease models such as the bipolar spectrum [Angst, 2007]. In particular, Sklar et al. 2008 recently described an association between single-nucleotide polymorphisms (SNPs) within the voltage-dependent calcium channel, L-type, alpha 1C subunit (CACNA1C) gene and bipolar disorder. We hypothesized that individuals with CACNA1C risk alleles diagnosed with major depressive disorder (MDD) might nonetheless display some clinical features suggestive of bipolar disorder.
MATERIALS AND METHODS
Clinical Methods
The STAR*D multisite trial was intended to determine prospectively which of several treatments are more effective in treating subjects diagnosed with non-psychotic MDD. All patients received citalopram as first-line treatment at doses up to 60 mg for up to 14 weeks.
The rationale, design, and methods of STAR*D are illustrated thoroughly elsewhere [Fava et al., 2003; Rush et al., 2004]. A summary of the clinical methods is presented in supplemental text.
Only subjects who spontaneously sought treatment at clinical sites were enrolled in the STAR*D study and advertising was proscribed to best mirror the reality of common clinical practice.
Criteria for eligibility were deliberately broad: male and female outpatients, age 18–75 years, with a DSM-IV diagnosis of non-psychotic MDD and a baseline score ≥14 on the 17-item Hamilton Rating Scale for Depression [Hamilton, 1960] as rated by the Clinical Research Coordinator.
Phenotype Definition
Two of the authors (F.C., R.P.) reviewed the literature on bipolar spectrum indicators and identified by consensus clinical features suggestive of bipolar illness which could be assessed in STAR*D. These included at baseline a self-reported family history of bipolar disorder, a history of suicide attempts, an early age at onset of the disease (≤18 years old), and a recurrent course of depression (i.e., more than one episode including the current one). We also examined baseline symptoms which could be associated with bipolar spectrum illness, including atypical symptoms, agitation, suicidality, and irritability. For longitudinal course, we considered treatment-related adverse effects emerging within 4 weeks from the start of antidepressant therapy. Among these symptoms we again focused on those suggested in previous reports to arise from a bipolar predisposition, namely, suicidal ideation, agitation, and signs of unfavorable activation/mixed features or prodromes of a switch process (restlessness, insomnia, irritability, worsening of concentration), defined as a 1-point or greater increase in Quick Inventory of Depressive Symptomatology-self-report version (QIDS-SR) item [Rush et al., 1996, 2003; Trivedi et al., 2004]. As three items on the QIDS-SR concern insomnia, we summed the three measures of sleep disturbance (initial insomnia, middle insomnia, and terminal insomnia) and required a 1-point increase. Emergence of anxiety, insomnia, poor concentration, and restlessness were also identified using the FIBSER [Wisniewski et al., 2006], a self-report measure administered at all post-baseline visits. For these analyses, an adverse effect was considered to be present if it was reported on at least one visit. Finally, we evaluated treatment response, both as a continuous measure and as protocol-defined remission of symptomatology (exit QIDS-SR score ≤5), because of the proposed role of a bipolar diathesis in the pathogenesis of treatment resistant depression.
Genotyping Methods
We selected two SNPs (rs10848635, rs1006737; r2 = 0.65) identified in a recent meta-analysis of genome-wide association studies of bipolar disorder [Ferreira et al., 2008], including rs1006737, the SNP showing the strongest association signal in that report, genotyped as previously described [Sklar et al., 2008].
Of 1,213 Caucasian non-Hispanic individuals genotyped in this fashion, 1,206 (99.4%) were successfully genotyped for at least one locus, and 1,167 (96.2%) for both. Neither SNP deviated from Hardy–Weinberg equilibrium (HWE) (P > 0.05 for both).
Statistical Methods
All analyses utilized chi-square or regression-based tests, as appropriate, implemented in PLINK v1.03 [Purcell et al., 2007, http://pngu.mgh.harvard.edu/purcell/plink/]. Linear regression and logistic regression were adopted for quantitative and categorical traits respectively.
For the observed minor allele frequencies (0.35 and 0.37 for rs1006737 and rs10848635 SNPs respectively) power is 99.0% to detect a quantitative trait locus accounting for 1% of variance, assuming alpha value at 0.05, with a sample size of ∼1,200 subjects [Purcell et al., 2003]. As this was an exploratory analysis, no correction for multiple comparisons was performed.
RESULTS
Table I displays the association of baseline bipolar features and the two SNPs in the CACNA1C gene. Data indicating the association of family history of bipolar depression, QIDS-SR total score at study outset and a measure of sleep disturbance (including initial, middle, and terminal insomnia) are also presented in Table I.
| Baseline feature | N (%) | G827_rs10848635 | G826_rs1006737 | ||
|---|---|---|---|---|---|
| P | OR [95% CI] | P | OR [95% CI] | ||
| Family history of depression | 721 (60.1) | 0.89 | 0.99 [0.83–1.17] | 0.68 | 1.04 [0.87–1.24] |
| Family history of bipolar disorder | 126 (10.5) | 0.46 | 1.11 [0.85–1.44] | 0.06 | 1.29 [0.98–1.69] |
| History of suicide attempts | 171 (14.2) | 0.48 | 1.09 [0.86–1.38] | 0.74 | 1.04 [0.81–1.34] |
| Recurrent course | 894 (79.1) | 0.32 | 1.11 [0.90–1.37] | 0.41 | 1.1 [0.88–1.37] |
| Early onset | 504 (41.8) | 0.62 | 1.04 [0.88–1.23] | 0.81 | 1.02 [0.85–1.2] |
| Atypical symptoms | 205 (17.7) | 0.72 | 0.96 [0.77–1.20] | 0.42 | 0.91 [0.72–1.15] |
| Baseline feature | Mean (SD) | G827_rs10848635 | G826_rs1006737 | ||
|---|---|---|---|---|---|
| P | β [95%CI] | P | β [95%CI] | ||
| Depression severity | 15.0 (4.51) | 0.20 | −0.23 [−0.58/0.12] | 0.037 | −0.40 [−0.77/−0.02] |
| Agitation | 1.05 (1.12) | 0.03 | −0.09 [−0.16/−0.007] | 0.06 | −0.07 [−0.16/0.005] |
| Suicidality | 0.66 (0.93) | 0.59 | 0.02 [−0.04/0.08] | 0.31 | −0.03 [−0.1/−0.03] |
| Insomnia | 4.79 (2.49) | 0.12 | −0.16 [−0.37/0.04] | 0.047 | −0.22 [−0.43/−0.003] |
| Poor concentration | 1.55 (0.98) | 0.18 | −0.04 [−0.11/0.02] | 0.13 | −0.05 [−0.12/0.015] |
| Irritability scorea | 32.3 (14.69) | 0.91 | −0.04 [−0.75/0.67] | 0.54 | 0.23 [−0.52/0.99] |
- All OR's and β refer to the minor (A) allele for both SNPs.
- a Adjusted for baseline QIDS-SR score.
rs10848635 was nominally associated with baseline agitation. rs1006737 SNP produced nominally significant evidence of association with QIDS-SR total score and overall baseline presence of sleep disturbance. However, in all cases risk allele was associated with lesser risk of these symptoms at baseline.
Variables representing symptomatic worsening during antidepressant treatment and outcome measures are presented in Table II.
| N (%) | G827_rs10848635 | G826_rs1006737 | |||
|---|---|---|---|---|---|
| P | OR [95% CI] | P | OR [95% CI] | ||
| Symptoms worsening | |||||
| Agitationa | 287 (23.7) | 0.03 | 0.80 [0.66–0.98] | 0.21 | 0.87 [0.71–1.08] |
| Insomniab | 368 (30.3) | 0.4 | 0.93 [0.77–1.12] | 0.77 | 1.03 [0.85–1.25] |
| Suicidalityc | 151 (12.5) | 0.047 | 1.28 [1.00–1.63] | 0.03 | 1.31 [1.02–1.69] |
| Concentrationd | 195 (16.1) | 0.096 | 1.21 [0.97–1.50] | 0.05 | 1.26 [1.00–1.59] |
| Side effects | |||||
| Anxiety | 617 (53.1) | 0.46 | 1.06 [0.90–1.26] | 0.36 | 1.09 [0.91–1.30] |
| Sleep disturbance | 377 (32.6) | 0.07 | 0.85 [0.71–1.01] | 0.09 | 0.85 [0.70–1.03] |
| Poor concentration | 602 (51.8) | 0.06 | 0.85 [0.72–1.01] | 0.39 | 0.93 [0.78–1.10] |
| Restlessness | 527 (45.4) | 0.88 | 0.99 [0.83–1.17] | 0.78 | 0.97 [0.82–1.16] |
| Outcomes | |||||
| Remissione | 511 (42.3) | 1.00 A | 1.00 [0.84–1.19] | 0.33 | 1.09 [0.91–1.31] |
| Mean (SD) | G827_rs10848635 | G826_rs1006737 | |||
|---|---|---|---|---|---|
| P | β [95% CI] | P | β [95% CI] | ||
| Depression severity at level 1 exite | 7.66 (5.94) | 0.26 | −0.24 [−0.67/0.18] | 0.19 | −0.30 [−0.75/0.15] |
- All OR's and β refer to the minor (A) allele for both SNPs.
- a Adjusted for baseline agitation score on QIDS-SR.
- b Adjusted for baseline insomnia scores on QIDS-SR.
- c Adjusted for baseline suicidality score on QIDS-SR.
- d Adjusted for baseline concentration score on QIDS-SR.
- e Adjusted for baseline QIDS-SR score.
A significant association of both risk alleles with an increased risk of treatment-emergent suicidality was observed. rs10848635 was also significantly associated with worsening of agitation while rs1006737 was associated with worsening of concentration under treatment.
We conducted the same sequence of analyses post hoc (Supplemental Table I), on patients ≤25 years old (N = 151), hypothesizing that this younger group might be enriched for “bipolar spectrum” subjects. In this subsample, both risk alleles were significantly associated with a recurrent course (rs10848635: P = 0.029, OR = 1.98; rs1006737: P = 0.033, OR = 2.09), early illness onset (rs10848635: P = 0.037, OR = 1.80; rs1006737: P = 0.05, OR = 1.79) and with the side effect of anxiety (rs10848635: P = 0.017, OR = 1.80; rs1006737: P = 0.001, OR = 2.44). The rs1006737 marker was also nominally associated with family history of bipolar disorder (P = 0.04, OR = 2.05) and with poor concentration at baseline (P = 0.017, β = 0.20).
DISCUSSION
To our knowledge, this report is one of the first to attempt to validate the clinical heuristic of “bipolar spectrum disorders” using genetic association. As an exploratory analysis, we did not correct for multiple comparisons, recognizing that risk of type I error is high.
We found that among the considered baseline features the two SNPs in CACNA1C gene had a protective effect on agitation whereas only the rs1006737 SNP was associated with baseline depression severity and insomnia, again yielding a protective effect.
Both risk alleles were associated with an increased risk of treatment-emergent suicidality. However, the two selected markers were not associated with the other indicators of bipolar susceptibility, including history of suicide attempts, early onset of depression, recurrent course, atypical symptoms, baseline suicidality and irritability, anxiety and restlessness under treatment, nor were either of the two markers associated with outcome measures (e.g., remission and depression severity at level 1 exit).
The most striking finding of the current study is the association of the polymorphisms investigated with treatment-emergent worsening of suicidal ideation. This finding was also confirmed after adjustment for baseline suicidal ideation. Many concerns have arisen about antidepressant-induced suicidal ideas and “behavioral toxicity,” especially in children and adolescents [Damluji and Ferguson, 1988; Teicher et al., 1990; King et al., 1991; Masand et al., 1991; Jick et al., 2004; Olfson and Marcus, 2008], inducing the FDA to implement a “black box” warning about the risk of suicidality with antidepressant use in this subpopulation of patients [US FDA, 2004].
On the other hand, meta-analyses of placebo-controlled antidepressant trials in MDD and other reports underscore the neutral or protective effect of antidepressants on suicide risk [Storosum et al., 2001; Khan et al., 2003; Simon et al., 2006; Mulder et al., 2008]. More likely, there may be a subgroup of subjects experiencing both exacerbation of suicidal ideation and de novo suicidality under treatment, along with worsening of other psychopathologic dimensions [Horwitz et al., 1996; Jick et al., 2004]. Furthermore, the appearance of suicidal ideas and behaviors under antidepressant treatment could reflect an underlying bipolar liability [Shi et al., 2004].
Albeit more modest in its statistical significance, this is the third report from the STAR*D study suggesting the association between genetic markers and treatment-emergent suicidality. Perlis et al. 2007 found an interaction of 2 SNPs in CREB1 with treatment emergent-suicidality only in males, while Laje et al. 2007 identified a strong association between two markers within the glutamate signaling (GRIK2 and GRIA3 genes) and the same phenotype. Twin studies have demonstrated that suicidal behavior heritability estimates range from 30% to 50% [Statham et al., 1998; Fu et al., 2002] and even though antidepressant-related worsening of suicidal ideation is an infrequent sub-phenotype [US FDA, 2004], elucidating its biological basis would be of manifest clinical relevance.
While preliminary, our post hoc analyses also suggest bipolar spectrum effects may be more apparent in younger individuals, often misdiagnosed with unipolar depression before receiving a correct diagnosis of bipolar disorder [Ghaemi et al., 2000; Hirschfeld et al., 2003].
Limitations
We note several caveats in interpreting our results. First, STAR*D study criteria excluded subjects with psychotic symptoms and non-response to antidepressants, both of which have been suggested to be related to bipolar disorder [Guze et al., 1975; Strober and Carlson, 1982; Akiskal et al., 1983; Akiskal and Mallya, 1987; Coryell et al., 1995; Goldberg et al., 2001; Mitchell et al., 2001; Fagiolini and Kupfer, 2003; Baethge et al., 2005; Sharma et al., 2005; Goes et al., 2007; Forty et al., 2008]. These exclusions might “de-enrich” this cohort for bipolar spectrum patients and limit our ability to detect associations.
A second limitation arises from the fact that the choice of phenotypes pertaining to the bipolar spectrum concept is subjective and depends on accumulated evidence gathered from the literature. The design and phenotypic data of our study, mainly aimed at exploring external validation and treatment-related features of soft bipolarity, also did not permit us to explore other documented characteristics of bipolar spectrum, such as postpartum onset, anxiety comorbidity and personality profiles of patients [Perugi et al., 1999; Akiskal et al., 2003, 2006; Hantouche et al., 2003].
The specificity of our association also merits further study as CACNA1C variations have also been associated with other neuropsychiatric [Gargus, 2006] and non-psychiatric diseases or phenotypes [Bremer et al., 2006].
Future Directions and Conclusions
Further study of genetic variation may provide an opportunity to address the dilemma raised by Perris 1966 more than 40 years ago, when he divided the classical Kraepelinian diagnosis of manic-depressive illness into bipolar and unipolar subtypes. Our results illustrate one approach that may help to elucidate these debates, which were present in the earliest modern conceptions of psychiatric illness, and continue during the current debates about refinement of the DSM and the management of psychiatric disorders [Fava et al., 2007]. Finally, additional study may be warranted examining CACNA1C relative to other candidate genes previously associated with suicidality, such as SCN8A gene encoding a neuronal voltage gated sodium channel [Wasserman et al., 2005; Wang et al., 2008] and with treatment-emergent suicidality [Laje et al., 2007; Perlis et al., 2007].
