Brain Function in Carriers of a Genome-wide Supported Bipolar Disorder Variant

Context  The neural abnormalities underlying genetic risk for bipolar disorder, a severe, common, and highly heritable psychiatric condition, are largely unknown. An opportunity to define these mechanisms is provided by the recent discovery, through genome-wide association, of a single-nucleotide polymorphism (rs1006737) strongly associated with bipolar disorder within the CACNA1C gene, encoding the α subunit of the L-type voltage-dependent calcium channel Ca_v1.2.

Objective  To determine whether the genetic risk associated with rs1006737 is mediated through hippocampal function.

Design  Functional magnetic resonance imaging study.

Setting  University hospital.

Participants  A total of 110 healthy volunteers of both sexes and of German descent in the Hardy-Weinberg equilibrium for rs1006737.

Main Outcome Measures  Blood oxygen level–dependent signal during an episodic memory task and behavioral and psychopathological measures.

Results  Using an intermediate phenotype approach, we show that healthy carriers of the CACNA1C risk variant exhibit a pronounced reduction of bilateral hippocampal activation during episodic memory recall and diminished functional coupling between left and right hippocampal regions. Furthermore, risk allele carriers exhibit activation deficits of the subgenual anterior cingulate cortex, a region repeatedly associated with affective disorders and the mediation of adaptive stress-related responses. The relevance of these findings for affective disorders is supported by significantly higher psychopathology scores for depression, anxiety, obsessive-compulsive thoughts, interpersonal sensitivity, and neuroticism in risk allele carriers, correlating negatively with the observed regional brain activation.

Conclusions  Our data demonstrate that rs1006737 or genetic variants in linkage disequilibrium with it are functional in the human brain and provide a neurogenetic risk mechanism for bipolar disorder backed by genome-wide evidence.

Bipolar disorder, an often chronic and devastating disease affecting approximately 1% to 3% of the population, has one of the highest heritability rates (estimated at 0.85) of all known psychiatric conditions.¹ Recently, a meta-analysis of 3 genome-wide association studies reported strong evidence for an association between bipolar disorder and polymorphism rs1006737 located within the CACNA1C gene (OMIM *114205) on chromosome 12p13 (combined P = 7.0 × 10⁻⁸, odds ratio, 1.21).^2-5 In addition, a recent study⁶ indicates this variant also confers risk with similar effect sizes for schizophrenia and major depression. CACNA1C encodes the α subunit of the L-type voltage-dependent calcium channel Ca_v1.2. The neuronal L-type channel mediates synaptic plasticity in an N-methyl-D-aspartate receptor–independent way by activating the protein kinase extracellular signal-regulated kinase and cyclic adenosine monophosphate response element-binding–dependent transcription.^7,8 Calcium influx through Ca_v1.2 channels triggers processes that underlie hippocampus-dependent memory.^7-10 In particular, mice with a selective inactivation of the CACNA1C gene, lacking Ca_v1.2 in the hippocampus, show a defect in N-methyl-D-aspartate receptor–independent long-term potentiation in the CA1 region of the hippocampus, paralleled by a severe deficit in spatial memory.⁷This function is closely linked to episodic memory, which is critically dependent on hippocampal function. Taken together, these data clearly show a link between the CACNA1C-mediated L-type voltage-dependent calcium channel Ca_v1.2 and hippocampal function, including episodic memory.

Notably, impairment of declarative memory is the most consistent cognitive dysfunction found in patients with bipolar disorder.^11-13 It also has been observed in the unaffected relatives of these patients, suggesting that declarative memory impairment is related to genetic risk for bipolar disorder.^12,13 Neuroimaging studies have provided a substrate for these neuropsychological findings because hippocampal abnormalities have been repeatedly observed in patients with bipolar disorder. Although analyses of structural changes in hippocampal volume have shown divergent results,¹¹ functional neuroimaging studies^14,15 have consistently indicated that a dysfunction of the hippocampus and closely related regions underpins abnormal affective responses and dysfunctional emotion regulation in bipolar disorder. Moreover, studies using proton magnetic resonance spectroscopy demonstrated significantly lower concentrations of N-acetyl-aspartate in the bilateral hippocampus in patients^16,17 early in the course of illness,¹⁶ suggesting abnormal neuronal viability and impaired energy metabolism. Results of postmortem studies provide further evidence for the hypothesis that hippocampal abnormalities are associated with altered synaptic plasticity and diminished resilience in bipolar disorder.¹¹

Despite this compelling and convergent evidence, a direct demonstration of the role of CACNA1C for human hippocampal function in general and the relevance of these neural systems for heritable risk for bipolar disorder in particular has not been provided. Toward these goals, we examined hippocampal function in healthy carriers of the recently documented CACNA1C risk variant. We used brain activation as an intermediate phenotype in an imaging genetics approach. We use the term intermediatephenotype rather than the more commonly used endophenotype because there is nothing “endon” (Greek for “hidden”) about these imaging phenotypes,¹⁸ which are used under the assumption that genes will show higher penetrance for brain activation phenotypes closer to gene biology than behavioral or clinical measures.^18-20 We investigated a sample of 110 healthy volunteers of both sexes and of German descent genotyped for rs1006737 (Table 1). We used functional magnetic resonance imaging (fMRI) and an episodic memory task optimized for imaging genetics that allows probing of different cognitive subprocesses (ie, encoding, recall, and recognition of associative episodic information) and yields stable and reliable activations in core regions of episodic memory processing. We studied regional brain activation during task performance and functional coupling between brain regions using functional connectivity, a correlative measure that, although not directly indicative of structural or causal connections, has proved useful in parsing the functional anatomy of neural systems mediating genetic risk for psychiatric disorders.^18,19

Fifty healthy German volunteers with parents and grandparents of European origin (as determined by self-report) were recruited in Bonn and 60 in Mannheim as part of an ongoing study on neurogenetic mechanisms of mood disorder and schizophrenia, as reported previously.¹⁹ All participants gave prior written informed consent. No participant reported a lifetime or family history of schizophrenia or affective disorder. Of our sample (n = 110), 60 participants were rs1006737 GG homozygotes, 41 were GA heterozygotes, and 9 were AA homozygotes. The allele frequencies were in Hardy-Weinberg equilibrium (χ² = 0.10, P = .75). Genotype distributions did not differ between sites (χ² = 0.78, P = .72). Sex, age, handedness, and level of education did not differ significantly among the genotype groups (Table 1). The study was approved by the local ethics committees of the universities of Heidelberg and Bonn.

Genomic DNA was extracted from ethylenediaminetetraacedic acid anticoagulated venous blood by using a Chemagic Magnetic Separation Module I (Chemagen AG; Baesweiler, Germany) according to the manufacturer's recommendations. We genotyped rs1006737 using a TaqMan 5′ nuclease assay (Life Technologies Corporation, Carlsbad, California). Accuracy was assessed by duplicating 15.0% of the sample; reproducibility was 100%.

During fMRI scanning, participants completed 3 consecutive memory tasks (ie, encoding, recall, and recognition of face-profession pairs eFigure, with an overall duration of 13 minutes, based on a paradigm previously used for imaging genetics.²¹ This task was part of a functional imaging genetics battery. The encoding task consisted of 16 face-profession pairs and 24 head contours as the control condition, with 4 blocks of 4 face-profession pairs and 4 blocks of 6 head contours each, respectively. Face-profession pairs were presented for 6 seconds each and head contours for 4 seconds each. Thus, each block lasted 24 seconds. Participants were instructed to imagine the depicted person behaving in a scenario typical of the given profession; participants were then asked to indicate whether the profession suited the person's face in order to induce deep encoding. During the control condition, participants had to indicate which ear of the depicted head contour was larger. The alternating sequence of 4 face-profession association blocks and 4 control blocks was presented twice to ensure successful encoding. During recall, faces were presented again, and participants were asked whether the depicted person had to complete apprenticeship or academic studies to qualify for the respective profession introduced during the encoding stage. Participants were then asked to indicate, by pressing a button, which qualification was correct. The stimulus duration and control condition were similar to that from the encoding stage; blocks were presented only once. For recognition testing, faces were presented together with 2 written professions, and participants were asked to indicate which profession was correct. The stimulus duration for recognition was 3 seconds. The control condition consisted of 4 blocks of 4 head contours each (for 3 seconds each). Thus, each recognition block lasted 12 seconds. Similar to the recall stage, blocks were only presented once.

On the second day, participants underwent neuropsychological assessment and testing for verbal intelligence (Mehrfachwahl-Wortschatz-Intelligenztest, version B) and memory (Verbal Learning and Memory Test, a translated verbal version of the Rey Auditory Learning Task) and completed versions of the Beck Depression Inventory (BDI), the State Trait Anxiety Inventory, trait version (STAI-T), the Symptom Checklist–90-Revised (SCL-90-R), and the Neuroticism, Extraversion, and Openness to experience Five-Factor Inventory (NEO-FFI).

Blood oxygen level–dependent fMRI was performed using 2 scanners (Siemens Trio 3T; Siemens Medical Solutions, Erlangen, Germany) at the Central Institute of Mental Health Mannheim and the University of Bonn. At these sites identical sequences and scanner protocols were used (parameters: 33 slices, axially tilted [−30°]; slice thickness, 2.4 + 0.6-mm gap; field of view, 192 mm; repetition time, 1.96 seconds; echo time, 30.00 milliseconds; flip angle, 80°). High-resolution 3-dimensional T1-weighted images were acquired with 160 contiguous sagittal slices of 1-mm thickness (field of view, 256 mm; repetition time, 1.57 seconds; echo time, 3.42 milliseconds; inversion time, 800.0 milliseconds; flip angle, 15°).

Multicenter fMRI studies have the advantage of increasing sample size, and thus enhanced statistical power, applied to a specific research question. One important challenge, however, is the consistency of the measurement from one scanner to another.²² We applied a multicenter quality assurance protocol that has been documented to improve the quality of fMRI data.²³ Quality assurance measures were conducted on every measurement day at both sites according to protocol²³; stable signals over time and similar quality between sites were revealed. In addition, stringently accounting for any remaining differences in the signal to noise ratio across sites, site was used as a confounding covariate for all statistical analyses. Furthermore, in post hoc testing, we used analysis of variance to examine whether scanner-site-by-genotype interactions were found in our regions of interest (ROIs). No significant interactions were observed. Moreover, we tested whether the observed genotype effects could be verified for each site independently (see the “Effect of CACNA1C Genetic Variation on Brain Function” subsection of the “Results” section).

Image processing and statistical analyses were conducted using statistical parametric mapping methods as implemented in Statistical Parametric Mapping 5 (SPM5) statistical software (http://www.fil.ion.ucl.ac.uk/spm/software/spm5/) and were similar for all tasks. Briefly, images were realigned to a mean image (movement parameters were confined to <3-mm translation and <3° rotation between volumes), slice-time corrected, spatially normalized to a standard stereotactic space (a brain template created by the Montreal Neurological Institute) with volume units (voxels) of 2 × 2 × 2 mm, and smoothed with an 8-mm full-width-at-half-maximum gaussian filter; ratio was normalized to the whole-brain global mean. A first-level fixed-effects model was computed for each participant. Regressors were created from the time course of the 2 experimental conditions (memory and control) and convolved with a canonical hemodynamic response function. Movement parameters were included in the first-level model as regressors of no interest. For each participant, statistical contrast images of memory vs control were obtained. To test for genetic association, these images were analyzed using the general linear model in a second-level random-effects analysis (2-sample t test; group 1: GG; group 2: AA + GA; because of the small number of AA individuals, we pooled individuals with either 1 or 2 copies of the risk allele [GA or AA] for subsequent analyses), with the scanner site as the nuisance covariate, to identify genotype effects on activation or connectivity.

Analyses to measure functional connectivity used a seed region approach.¹⁹ For each participant, time series were extracted from the left and right hippocampal regions using first eigenvariates from all voxels within the respective hippocampal mask. Following previous practice,¹⁹ to derive a robust summary measure of activity in the respective hippocampal ROI, we excluded white matter by restricting the averaging to voxels related to task at a P <.50 level (note that this level was not used for statistical inference). Using the SPM5 software, seed time series were high pass filtered (128 seconds) and task-related variance was removed to avoid measuring coactivation solely owing to temporal correlation with the experimental paradigm. To account for unspecific noise, first eigenvariates from masks covering cerebrospinal fluid and white matter were extracted for each individual. These eigenvariates were entered, together with movement covariates and task regressors, into whole-brain multiple regression analyses where the respective seed region time series (ie, the time series from the right or left hippocampal region) was the covariate of interest. Thus, we identified voxels whose activity shows significant covariation with the left or right hippocampal region. Here, brain areas are called functionally connected if their blood oxygen level–dependent signal time series show covariance over time. The resulting maps of partial correlation with the left or right hippocampal seed region were then each subjected to a random-effects analysis in SPM5 using a 2-sample t test as described herein, with the scanning site as the covariate of no interest.

For all imaging methods, the significance threshold was set to P < .05, corrected for multiple comparisons across the whole brain or within our single a priori–defined anatomical ROI, the bilateral hippocampus. For all analyses, we used conservative analysis statistics by using the false discovery rate (FDR), a widely used frequentist method previously shown to exert strong control of type I error over multiple comparisons in imaging genetics.²⁴ The hippocampal ROI for seed voxel extraction and ROI analysis was defined a priori and created using anatomical labels provided by the Wake Forest University PickAtlas.²⁵

First, we investigated the effect of the CACNA1C genotype on task-related functional brain activation. As detailed in the “Functional Imaging Task” subsection of the “Methods” section, the task included encoding, recall, and recognition of face-profession associations in 3 consecutive sessions based on a paradigm previously used for imaging genetics,²¹ showing highly significant activations in core regions of episodic memory processing (for detailed results, see eTables 1-4. Whereas no significant association of the rs1006737 variant with brain activation was observed during encoding and recognition, carriers of the CACNA1C risk allele showed a highly significant reduction of bilateral hippocampal activation during recall (Figure 1 A, Table 2, and eTable 5). Moreover, in the same recall condition, risk allele carriers also exhibited significantly reduced activation in the subgenual anterior cingulate cortex (sgACC; Figure 1B), dorsal anterior cingulate cortex, ventral striatum, and superior frontal and temporal cortices (for all results, P < .05 FDR corrected for multiple comparisons across the whole brain; Table 3).

Activation of the bilateral hippocampus and sgACC was also significantly reduced in risk allele carriers for each site separately (eTable 6). In addition, functional connectivity of the hippocampus (as a measure of correlation of brain activity over time) revealed significantly diminished coupling between the left hippocampal seed region with a cluster within the right hippocampal region in risk allele carriers (24/−12/−14, Z = 3.68, P < .05 FDR corrected for multiple comparisons within the hippocampal ROI; Figure 2; for results of within-group connectivity, see eTable 7). No region outside the right hippocampus revealed altered connectivity with the left hippocampal seed region, and no altered connectivity was found for the right hippocampal seed region. Altered functional coupling was exclusively observed during memory recall. Again, diminished coupling between the left and right hippocampal regions in risk allele carriers was observed during recall for each site separately, although at a lower statistical level (eTable 8.

The brain regions found to have reduced activation during the recall task show overlap (although not completely) with the default mode network. Thus, an alternative explanation of our results would be that the risk allele confers impaired modulation of large-scale networks in general, such as the default mode network, rather than hippocampal dysfunction specifically. To investigate this possibility, we performed an independent component analysis on a resting-state data set in the same sample (for methodological details, see eAppendix 1). We found no influence of the CACNA1C risk allele on components including regions of the default mode network (eg, posterior cingulate cortex/precuneus, superior temporal sulcus, and medial prefrontal cortex), even when lowering the threshold to P < .001 uncorrected. This finding supports the conclusion that the risk allele confers vulnerability to hippocampal dysfunction specifically rather than to large-scale default mode networks.

To test whether partial volume effects in the hippocampus may have contributed to our findings of reduced hippocampal activation in risk allele carriers, we performed an additional analysis on T1-weighted structural scans using a sensitive volumetric approach (ie, Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra²⁶; for methodological details, see eAppendix 2) able to detect even subtle volume differences between groups.²⁷ There was no effect of the CACNA1C variant on the hippocampal volume or any other brain region, even when lowering the threshold to P < .001 uncorrected. Thus, our volumetric analysis provided no evidence for a partial-volume effect that might have contributed to our findings.

Furthermore, to demonstrate the specificity of our connectivity results, we conducted a control analysis of functional connectivity in which we used an “irrelevant” ROI, the occipital gyrus BA17 (visual cortex), a region strongly active during the task but not specifically involved in memory or implicated in CACNA1C action. No significant connectivity differences between the 2 groups were found, even when lowering the threshold to P < .001 uncorrected (for methodological details, see eAppendix 3).

We found a significant effect of the CACNA1C genetic variation on the BDI; STAI-T; depression, anxiety, obsessive-compulsive, and interpersonal sensitivity subscales of the SCL-90-R; and neuroticism scores of the NEO-FFI, with risk allele carriers exhibiting significantly higher scores on all measures (Table 2). No association of the CACNA1C variant was found with task performance during scanning or with neuropsychological tests of episodic memory (ie, early and late recall and recognition in the Verbal Learning and Memory Test). Moreover, no effect of the CACNA1C variant on verbal intelligence, as measured by the Mehrfachwahl-Wortschatz-Intelligenztest, version B, was observed.

To test for a link between genotype-dependent differences in brain activation and psychopathological measures, we extracted individual beta weights from the peak activated voxel in regions showing a significant group effect (ie, the bilateral hippocampus and sgACC). The beta weights were externally correlated (SPSS statistical software, version 17; SPSS Inc, Chicago, Illinois) with individual scores on the BDI; STAI-T; depression, anxiety, obsessive-compulsive, and interpersonal sensitivity subscales of the SCL-90-R; and the neuroticism subscale of the NEO-FFI. We observed a significant negative correlation between regional activation in the left and right hippocampus and depression (results marked with an asterisk are significant after FDR correction for multiple testing; left: r = −0.19, P = .03; right: r = −0.23, P = .008*), anxiety (left: r = −0.25, P = .004*; right: r = −0.27, P = .002*), interpersonal sensitivity (left: r = −0.18, P = .03; right: r = −0.17, P = .04), and obsessive-compulsive (left: r = −0.23, P = .009*; right: r = −2, P = .02) subscales of the SCL-90-R. Moreover, we found a negative correlation between regional activation in the left hippocampal region and BDI scores (r = −0.24, P = .006*) and a trend for a negative correlation of left hippocampal region activation with neuroticism (r = −0.15, P = .055). Furthermore, we observed a negative correlation between regional brain activation in sgACC and the anxiety (r = −0.16, P = .045) and obsessive-compulsive (r = −0.2, P = .02) subscales of the SCL-90-R.

Converging evidence in animal behavior shows that CACNA1C, the gene encoding the L-type voltage-dependent calcium channel α subunit 1C, plays a crucial role in N-methyl-D-aspartate–independent hippocampal memory function.^7,10 Recent genome-wide association studies have identified a single-nucleotide polymorphism (rs1006737) within the CACNA1C gene as being strongly associated with the risk of bipolar disorder.^2-5CACNA1C appears to be a biologically plausible candidate for bipolar disorder: it seems to occur in pathways affected by lithium carbonate, the criterion standard for treatment of this disease, which has been shown to downregulate subunits of the calcium channel in the mouse brain.²⁸

Although it is evident there will be no single gene of major effect to explain most cases of bipolar disorder, the understanding of the neural mechanisms by which genetic variation influences susceptibility to psychiatric disorder is of crucial relevance. More biologically based phenotypes that are closer to the genetic substrate reduce phenotypic heterogeneity and enhance the penetrance of gene effects.¹⁸ Intermediate phenotypes in psychiatry can thus not only complement and underline existent genetic evidence but also, in particular, elucidate the biological and functional relevance of this evidence.

We investigated the neurogenetic mechanisms for bipolar disorder associated with rs1006737 using an imaging genetics approach with brain activation during episodic memory as an intermediate phenotype.^18,20 On the basis of preclinical^7,8,10 and clinical¹¹ evidence, we expected that CACNA1C also would be functional in the human brain and that genetic risk for bipolar disorder associated with rs1006737 would be mediated through hippocampal function.

Consistent with our hypothesis, in risk allele carriers, regional brain activation revealed a highly significant reduction of bilateral hippocampal activation. This effect appears cognitively specific in that it was observed only during recall but not during encoding or recognition (Figure 1A and Table 3). In fact, memory recall has been proposed to rely most critically on the hippocampus.^29,30 Thus, our data provide clear evidence that CACNA1C is also functional in the human hippocampus. These pronounced differences in neural activity in healthy risk allele carriers were found despite unimpaired cognitive performance: groups did not differ with respect to recall and recognition performance, scores in an episodic memory test, and verbal intelligence (Table 2).

The observed strongly reduced hippocampal activity in healthy human carriers of the CACNA1C risk allele indicates a possible link between the role of CACNA1C in genetic risk for bipolar disorder and its functional relevance for hippocampal function. The variant rs1006737 is located in the third intron of CACNA1C; no direct functional effect, such as conferring differences in gene expression or splicing, has been attributed to it. At present it is unclear whether the observed effects are owing to rs1006737 or another genetic variant in linkage disequilibrium. The possibility remains that variation at a different gene locus than CACNA1C, which is in linkage disequilibrium with rs1006737, is responsible for the observed effect. However, the most straightforward interpretation of our data suggests that CACNA1C is the gene causally implicated in the observed changes. In particular, recall-specific attenuation of hippocampal activity provides a clear human analogue to the findings in knockout mice and in rats injected with pharmacologic blockers of Ca_v1.2, showing preserved acquisition of memory but a marked difference of performance in a retention test.^7,10 Diminished coupling between the left and right hippocampal regions in risk allele carriers (Figure 2) identifies a further mechanism of altered hippocampal function consistent with preclinical evidence: recently, it has been shown in rodents that local strengthening of synaptic weights in the hippocampus modulates its functional connectivity in a way that potentiates bilateral hippocampal communication, improving memory recall.³¹

We did not expect, based on the concept of intermediate phenotypes,^18,20 to observe an effect of rs1006737 on behavioral readouts of hippocampal function, such as memory tests: biologically more proximate measures (ie, brain function) should exhibit higher penetrance than behavioral measures further removed from the genetic substrate. Because we observed pronounced functional abnormalities of the hippocampal function in the absence of a cognitive behavioral effect, our data are compatible with the assumption that it is not impaired episodic memory per se that increases risk for bipolar disorder but hippocampal functional impairment, which is indexed by behavioral and imaging methods at different levels of power and sensitivity.

However, the CACNA1C genotype showed a pronounced effect on measures of psychosocial functioning. Carriers of the rs1006737 risk allele exhibited significantly higher scores on tests for depression, anxiety, interpersonal sensitivity, obsessive-compulsive thoughts, and neuroticism (Table 2). Speculatively, this finding may point to increased susceptibility to psychological stress. Moreover, the respective psychopathology scores correlated negatively with hippocampal activation, suggesting a link between psychopathology scores and reduced hippocampal functioning in risk allele carriers. High trait anxiety and depressive temperament have been reported in patients with remitted bipolar disorder and their unaffected relatives, providing evidence of heritability and genetic influence.³² Neuroticism has been related to major depression and bipolar disorder, particularly to the severity of depressive symptoms among persons with bipolar disorder or persons with undiagnosed bipolar symptoms.³³ Thus, the observed link between increased scores on psychopathological scales and reduced hippocampal function in healthy risk allele carriers may provide preliminary evidence for subclinical instability and anxiety and depression proneness associated with the CACNA1C genotype. Future work is needed to further characterize and replicate this relation.

In addition to the hypothesized diminished hippocampal recruitment found during recall, we observed a highly significant reduction of sgACC activation (Figure 1B and Table 1). The sgACC is crucially involved in the regulation of emotional behavior and stress responses.^14,15,34 Because emotion dysregulation is one of the most apparent characteristics of bipolar disorder¹⁴ and functional abnormalities in the neural systems underlying emotion regulation might predispose individuals to the development of this disease, the sgACC has been proposed to play a key role in its pathophysiology. This is especially true in depressive episodes,^14,15,34 which are present in most patients with bipolar disorder. Neuroimaging findings in patients include reduced gray matter volume and altered glucose metabolism in sgACC.^14,34 Moreover, volume reductions within the sgACC have been linked to genetic risk for bipolar disorder because these structural abnormalities have not only been found in patients but also in their unaffected relatives.³⁵ In rodents, the homologue of this region participates in a visceromotor network that modulates autonomic responses to fear and stress.³⁶

Besides marked reductions in hippocampal and sgACC activation, we also observed decreased activation in the dorsal anterior cingulate and ventral striatum in carriers of the CACNA1C risk allele during recall.³⁷ A recent study,³⁵ using resting 18-fluorodeoxyglucose positron emission tomography in patients with depression and bipolar disorder, reported reduced regional metabolism in a number of regions, including the sgACC, anterior cingulate, and ventral striatum. Unaffected relatives of individuals with bipolar disorder have been reported to exhibit local gray matter reductions in the ventral striatum and more dorsal parts of the anterior cingulate cortex, both of which are brain-circuitry components for emotion processing and salience detection.

Speculatively, our data implicating the hippocampus and sgACC in genetic risk for bipolar disorder also suggest a mechanism for gene-environment interaction. Exposure to repeated stress results in dendritic atrophy and glia cell reductions in both structures.^38,39 Moderate to high levels of stress, which may result from an impairment of automatic or voluntary emotion regulation,¹⁴ and administration of exogenous corticosteroids can impair memory processing in animals and humans.⁴⁰ Glucocorticoid and mineralocorticoid receptors that are highly expressed in the animal and human hippocampus mediate corticosteroid-induced alterations of gene expression, leading to adaptive stress responses.⁴¹ Because a dysregulation of the hypothalamic-pituitary-adrenocortical axis and a genetically determined corticoid receptor imbalance have been associated with the pathophysiology of affective disorders,^41,42 it is conceivable that maladaptive regulatory mechanisms can lead to exaggerated levels of stress hormones and, finally, hippocampal damage. Most notably in the context of the present study, stress or corticosteroid-modulated synaptic plasticity in the hippocampus crucially involves voltage-dependent L-type calcium channels.⁴³ Because the CACNA1C gene specifically contributes to the encoding of L-type calcium channels, one can further speculate that the variant, significantly associated with risk for bipolar disorder,^2,4 might not only result in hippocampal dysfunction and memory impairment, as shown in rodents,^7-9 but also contribute to heightened vulnerability to stress, which may affect hippocampal integrity. Heightened scores in psychopathology measures observed in our sample of healthy risk allele carriers that might indicate psychological stress may strengthen this proposed mechanism. Further work should measure current and long-term levels of stress hormones and relate them to the neural circuitry outlined in the present genetic study.

Notably, variation in rs1006737, showing maximum association with susceptibility to bipolar disorder, has been found also to be associated with unipolar depression and schizophrenia; this finding points to a possible influence of CACNA1C variants in susceptibility across the mood-psychosis spectrum and less clear-cut boundaries between diagnostic entities.⁶ In this context, our findings provide a potential pathophysiologic mechanism conferred by the CACNA1C risk variant centered on the hippocampus that mediates risk for symptom dimensions shared among bipolar disorder, major depression, and schizophrenia.

Taken together, our findings show that the recently identified risk variant in the CACNA1C gene is associated with functional effects in the human brain even in the absence of overt disease. Moreover, these findings identify a system of altered functional activation in the hippocampus and sgACC that provides the first neural circuit for risk for bipolar disorder supported by genome-wide evidence. We have speculated that this result suggests a mechanism for gene-environment interactions through a dysfunctional adaptation to stress. Finally, Ca_v1.2 is an established drug target because it is the binding site for calcium channel blockers such as verapamil, which has shown some evidence of efficacy in mood stabilization in bipolar disorder.⁴⁴ Therefore, it is hopeful that the present systems-level findings may also contribute to the identification of drug targets for this common and commonly disabling disease.

Correspondence: Susanne Erk, MD, PhD, Department of Psychiatry, Division of Mind and Brain Research, Charité Medical University Berlin, Charité Campus Mitte, Charitéplatz 1, 11017 Berlin, Germany (susanne.erk@charite.de).

Submitted for Publication: September 4, 2009; final revision received February 3, 2010; accepted February 6, 2010.

Author Contributions: Drs Erk, Meyer-Lindenberg, Schnell, Esslinger, Kirsch, Grimm, Witt, Cichon, Nöthen, Rietschel, and Walter and Mss Arnold and Haddad had full access to all the data in the study and take responsibility for the integrity of the data and the accuracy of the data analysis.

Financial Disclosure: None reported.

Funding/Support: Funding for this study was provided by the German Ministry for Education and Research (BMBF) grant NGFNplus MooDS and by the German Research Foundation grant SFB 636-B7.

Additional Information: Drs Erk, Meyer-Lindenberg, and Schnell contributed equally to this work.

Additional Contributions: Beate Newport, Dagmar Gass, Daniela Mier, DiplPsych, Carina Sauer, and Kyeon Raab provided help with data acquisition. Helene Dukal provided expert technical assistance. Jan-Peter Lamke provided assistance in structural analysis.