To our knowledge, this is the first study to use computational mesh-based mapping methods to investigate callosal structure in twins discordant for bipolar disorder. Four key findings emerged from this study: 1) Localized regions of callosal thinning, most pronounced in the genu and splenium, were observed in bipolar probands, relative to both healthy control twins and to their non-bipolar co-twins. Corresponding reductions in callosal area were identified in bipolar probands using traditional morphometric methods; 2) Significantly altered callosal curvature was found in bipolar I probands, relative to both healthy control twins and their non-bipolar co-twins. 3) Within bipolar patients and their co-twins, callosal measures were significantly correlated with cognitive performance on measures of verbal processing speed and response inhibition, indicating that quantitative variation in CC size has functional significance. And finally, 4) no differences could be detected between co-twins and controls on measures of callosal structure, suggesting that callosal alterations in bipolar patients are disease related.
Callosal aberrations in bipolar probands were most pronounced in the genu and splenium and also observed within the callosal midbody but were not present within the isthmus, a callosal region containing fibers mainly projecting to parietal regions (Hofer and Frahm 2006
). Our findings are highly consistent with those of a recent study using similar methods to map the CC in elderly patients with major depressive disorder, in which significant callosal thinning was restricted to the genu in early-onset depressed patients but evident in both the genu and the splenium in patients with late-onset depression (Ballmaier et al. 2008
). Similarly, in patients with established bipolar illness, Walterfang, Malhi, et al. (2009a)
found evidence for disproportionately reduced callosal thickness in the splenium. Notably, this pattern is also consistent with that seen in patients with Alzheimer disease, who tend to show the greatest callosal atrophy in the rostrum and splenium, with relative sparing of the callosal body, a pattern which has been proposed to reflect specific loss of large pyramidal neurons in cortical layers III and V of frontal and parietooccipital association areas (Hampel et al. 1998).
Although no previous studies have examined callosal structure in twins discordant for bipolar I disorder, a prior study—using a different methodology—examined callosal size and shape in BPI patients and their first-degree relatives (both siblings and adult offspring of bipolar patients). Consistent with our findings, Walterfang, Wood, et al. (2009b)
found significant global and regional reductions in callosal thickness in BD patients, but first-degree relatives did not differ in callosal size or shape from controls. This pattern of findings provides further support for our conclusion that CC abnormalities may be linked to disease expression in bipolar disorder and thus do not appear to represent a marker of familial predisposition. These investigators also found that BD patients on lithium treatment showed a thicker anterior midbody than those on other psychotropics. While we did not find a medication effect in our sample, the majority of BPI patients in our sample were on lithium, thus limiting our power to compare the effects of lithium to those of other medications. However, it is important to note that the effect of lithium detected in the study by Walterfang and colleagues was in the opposite direction of the effect of diagnosis; this is consistent with prior studies showing volumetric increases associated with lithium treatment (Moore et al. 2000
; Bearden et al. 2007
). Thus, effects of lithium in our patient sample would be likely to result—if anything—in attenuation of the observed callosal reductions.
A variety of methods can be used to normalize raw neuroimaging data, and the results from these methodologies are not always interchangeable (Bermudez and Zatorre 2001
). Due to concerns that the relationship between callosal structure and overall brain size is not necessarily linear and also because the relationships between callosal size and brain size may differ across biological risk groups, we chose—for our primary analysis— not to use a global scaling transform but rather to adjust for brain size statistically in our analyses. As shown in Supplementary Figure 2
, the group differences observed for our “unscaled” callosal thickness maps remained significant when a scaling parameter is used.
Decreased callosal size may be associated with a reduction in fibers or decreased myelination of fibers connecting the prefrontal and parietal cortices. This might interfere with interhemispheric communication required for sustained attention and response inhibition, thus contributing to symptoms of distractibility, impulsivity, and inattention, frequently observed in patients with bipolar disorder. Supporting this view, we found significant associations of genu and splenium midsagittal areas with verbal processing speed and response inhibition, and a similar trend for a measure of auditory attention. While these tasks cannot be considered specific to interhemispheric connectivity, they are nonetheless measures requiring integrated, efficient information processing and thus are likely to be disrupted by abnormalities of major white matter fiber tracts (Karlsgodt et al. 2008
; Madsen et al. 2010
At present the cellular changes leading to the observed callosal alterations in bipolar patients are not known. Although postmortem neuropathology studies have reported decreased neuronal and glial density in the dorsolateral prefrontal cortex in bipolar patients (Rajkowska et al. 2001
). Beasley et al. (2002)
found no differences in the spatial pattern distribution or density of interstitial white matter neurons in frontal cortex. A follow-up study (Beasley et al. 2005
) demonstrated that postmortem brain cholesterol (a major component of myelin and of cell membranes) was reduced by 10% in individuals with bipolar disorder compared with controls, suggesting that reduced brain cholesterol levels and/or a reduction in synapses may underlie findings of white matter volume reduction in bipolar disorder. Loss of NRG1-erbB signaling in oligodendrocytes in the adult brain has been shown to lead to reduced myelin thickness and increased levels of dopamine (D1-like and D2-like) receptors and transporter (dopamine active transporter), and behavioral alterations that are consistent with mental disorders, for example, heightened anxiety-like behavior and increased amphetamine sensitization. This signaling pathway may also contribute to mental disorders through nongenetic mechanisms, as both NRG1 and erbB4 expression can be altered by environmental stressors such as forced locomotion (Roy et al. 2007). Further studies are needed to determine whether the observed regional callosal thinning reflects neurodevelopmental alterations and/or whether bipolar disorder is characterized by progressive callosal atrophy.
Our findings of abnormal callosal structure in bipolar I probands are also supported by recent DTI studies of white matter microstructure (Bellani et al. 2009
). Wang et al. (2008)
recently observed reduced fractional anisotropy (FA) in the anterior and middle CC in bipolar patients relative to healthy controls, suggesting reduced myelination and/or axonal damage in these regions. Only one study, to our knowledge, has applied DTI methods to first-degree relatives of bipolar patients (Chaddock et al. 2009
); these investigators found lower FA in bipolar patients compared with controls in the genu of the CC and in major intrahemispheric white matter tracts (i.e., right inferior longitudinal fasciculus and left superior longitudinal fasciculus). Unaffected relatives did not show significant FA differences relative to controls in a straightforward group-level comparison, but they did show intermediate FA values in the same clusters that showed reduced FA in bipolar patients. Increasing genetic liability for bipolar disorder was significantly associated with lower FA across distributed white matter regions.
Similar to what has been reported in schizophrenia (e.g., Casanova et al. 1990
), we did find evidence for altered morphology of the CC in bipolar patients. These disorders may therefore involve common processes such as deficient myelination and abnormal white matter integrity (Casanova et al. 1990
; Narr et al. 2000
). Nevertheless, the pattern of shape alteration observed in bipolar probands was not identical to that typically seen in schizophrenia patients, involving an upward bowing of the callosum (which may reflect ventricular enlargement). Here, we observed differences in posterior CC curvature in bipolar probands, a finding which is unlikely to result from enlarged lateral ventricles.
In contrast to findings in relatives of schizophrenia patients (Casanova et al. 1990
; Narr, Cannon, et al. 2002
), we did not find evidence that these alterations are a result of genetic liability to bipolar disorder, as non-bipolar co-twins of bipolar probands did not differ from controls on any callosal measures. Overall callosal area in co-twins was intermediate between bipolar probands and control twins, but this difference was nonsignificant and the effect size was small (eta-squared = 0.005). Genetic effects on callosal structure—if present—may be very subtle, and thus a much larger twin sample would be needed to detect such effects. Thus, while global decreases in white matter volume may be related to genetic risk for bipolar disorder (Kieseppa et al. 2003
; van der Schot et al. 2009
), to our knowledge, no studies have yet identified genetic contributions to regionally specific white matter variation in bipolar disorder.
Certain limitations of this study should be noted. In particular, the limited number of MZ twin pairs in our sample precluded comparison of MZ with dizygotic twin pairs. Nevertheless, groups were very well matched on demographic variables, and the correlative nature of the data from twin pairs was taken into account in the statistical analyses. Second, in order to maximize our sample, we included some BP1 probands without a matching co-twin. Nevertheless, excluding these 2 subjects from the analysis did not substantively alter our results. Also, despite the field strength of 1.0 T, we were able to get good gray–white contrast for the images (see Supplementary Fig. 1
). While this field strength could potentially be a limitation for imaging smaller structures (e.g., amygdala), the CC was clearly visible in our images, and our intrarater reliability was excellent (r.m.s. error < 1 mm). Finally, our study was limited to structural MR methodology; future studies employing multimodel neuroimaging techniques will be important to determine the direct relationship between regional callosal thickness and white matter integrity.
In summary, this was the first MRI study to examine the integrity of the CC and its functional correlates in twins with bipolar I disorder, their co-twins, and appropriate control twin subjects. We found evidence for structural alteration in the CC associated with bipolar illness, which appeared to be a disease-specific, rather than genetically mediated, effect. Callosal reduction in the genu and splenium was significantly associated with measures of cognitive processing speed (verbal fluency) and response inhibition in bipolar probands and co-twins. Our findings further implicate disrupted interhemispheric connectivity in bipolar disorder. Further studies are needed to determine the etiology of callosal alterations and when in the course of illness they develop.