We found a high correlation between FA and bimanual co-ordination scores in the body of the corpus callosum (). The voxels on the mid-sagittal plane formed part of a cluster of supra-threshold voxels spanning the corpus callosum (43 voxels, mean t-stat=3.4, max t-stat=4.8).
Region of corpus callosum showing correlation between FA and bimanual co-ordination skill
When this cluster was used as a seed mask for probabilistic tractography, we found paths connecting the seed with the supplementary motor area and the caudal cingulate motor area (). The same approach did not reveal paths connecting the seed callosal area with the (lateral) primary motor cortex, except for in one individual subject ().
Paths from callosal region where FA correlates with bimanual co-ordination performance
Effects of thresholding on pathways
This suggests that inter-individual variations in bimanual co-ordination reflect variability in white matter integrity in callosal pathways - most strongly in paths connecting the supplementary motor areas. The high spatial specificity of the region of high correlation suggests that variation between individuals differs across functional systems. Although there are some suggestions that inter-individual variation in brain structure can be captured by a general IQ marker (Colom et al., 2006
), there is also abundant evidence for regional variations in brain structure and function reflecting specific patterns of skill and experience (Bengtsson et al., 2005
; Gaser and Schlaug, 2003
; Maguire et al., 2000
; Sluming et al., 2002
Previous studies have suggested that normal variation in white matter fractional anisotropy (FA) in specific brain pathways is related to varying performance in related visuospatial and cognitive tasks (Deutsch et al., 2005
; Madden et al., 2004
; Schulte et al., 2005
; Tuch et al., 2005
; Wolbers et al., 2006
). For example, increased FA in white matter underlying the inferior parietal cortex is associated with higher efficiency in mental rotation (Wolbers et al., 2006
); FA along the optic radiations correlates with response time in a choice reaction-time task but in this case the correlation is positive: increasing (i.e. slower) response times correlate with higher FA (Tuch et al., 2005
). This relationship is less intuitive, as a simple-minded interpretation of FA suggests that higher FA should indicate higher ‘integrity’, for example reflecting increased packing density or myelination. FA is, however, a complex measure that will be influenced not only by myelination, axon size, and axon density (Beaulieu, 2002
), but also by path geometry and the presence of crossing fibre pathways. In the current study, we tested the relationship between FA and behaviour in a strongly coherent, parallel fibre system, namely the corpus callosum, and therefore expected to find increasing FA associated with improving behavioural performance.
Inter-individual variation in callosal size has previously been shown to correlate with fMRI activation in medial motor areas during performance of a bimanual motor task (Stancak et al., 2003
). In vivo
measures of callosal size have been shown to relate to the number of myelinated and non-myelinated fibres (Aboitiz et al., 1992
) both in the healthy brain and in disease states such as multiple sclerosis (MS), where axonal degeneration plays a role in accumulating disability (Evangelou et al., 2000
). Further, in MS, reduced callosal size correlates with reduced FA within the normal appearing white matter of the callosum, suggesting that measures of FA within the callosum reflect local axonal density (Cader et al, unpublished observations). Previous studies have shown that visuomotor inter-hemispheric transfer correlates with FA in the splenium and genu of the corpus callosum in healthy control subjects, suggesting that structural variation has functional consequences (Schulte et al., 2005
). Here, we tested the relationship between callosal FA and performance on a purely motor inter-hemispheric task and were particularly interested in the spatial specificity of correlations with FA within the callosum. By using sites of local FA correlation as seeds for probabilistic tractography we were able not only to identify the sites of local FA change, but also to implicate specific cortico-cortical connections in task performance.
We found that the callosal region where FA correlated with bimanual co-ordination performance gave rise to pathways connecting the supplementary motor areas and caudal cingulate areas in the two hemispheres. Tracer studies in macaque demonstrate that the hand area of SMA has dense homotopic transcallosal connections, in contrast to the modest transcallosal connections between primary motor hand areas (Rouiller et al., 1994
). SMA also has heterotopic transcallosal connections that are dense with cingulate motor areas, moderate with lateral premotor cortex, and sparse with M1 in the opposite hemisphere (Rouiller et al., 1994
). The cingulate motor areas also have dense homotopic transcallosal projections (R.J. Morecraft, personal communication). The relative strength of transcallosal connectivity between SMAs and CMAs has not yet been systematically studied using anatomical tracers. Damage to these medial motor areas impairs bimanual co-ordination (Geffen et al., 1994
; Larson et al., 2002
; Serrien et al., 2001
; Zaidel and Sperry, 1977
). The correlation between callosal size and movement-related fMRI activity is much stronger for medial wall motor areas than for lateral motor areas (Stancak et al., 2003
). Our finding is therefore consistent with previous work suggesting that callosal connections between medial wall motor areas are most critical in bimanual co-ordination. It is also the case, however, that diffusion tractography tends to be most sensitive to medial fibres when seeds are placed in the corpus callosum, as the presence of longitudinal fibres, such as the superior longitudinal fasciculus make tracking to lateral areas less likely. In an attempt to reduce this problem we used a tractography approach that can model multiple fibre directions at each voxel (Behrens et al., in press
). Even with this model, connections to lateral motor areas were rarely seen from the region of CC showing high correlation with bimanual performance (Figures ,). To test whether lateral connections could be traced from other regions in the CC we seeded all voxels within the CC on the mid-sagittal plane. In addition to strong connections with medial wall areas, we also found lateral connections in prefrontal, premotor and parietal cortices in some subjects but a striking absence of lateral connections to the primary sensorimotor cortices (). The absence of lateral connections was found not only for the levels of the motor cortex corresponding to the hand representation, but also for more ventral portions of primary motor cortex where the face representation would be found. By contrast, medial parts of M1, including lower limb representations, showed strong inter-hemipsheric connectivity using tractography ().
Connections from the whole callosum
It is tempting to interpret this as human evidence for the anatomical finding in macaque monkey of sparse connectivity between lateral motor areas (Rouiller et al., 1994
), but it is also possible that this reflects limitations of the tractography technique. Callosal connections to lateral motor areas would have to cross not only the longitudinally oriented SLF, but also the superior-inferiorly oriented cortico-spinal tract. Therefore, it is likely that there are voxels along this route that contain at least three major fibre directions. We have previously shown although diffusion data such as the type acquired here (60 diffusion encoding directions, b-value of 1000smm−2
) provides sufficient evidence to allow for fitting of two fibre directions at many white matter voxels, such data does not provide evidence for three or more directions at any white matter voxels (Behrens et al., in press
). It is likely that diffusion data with greater angular resolution and higher b-values, which could enable modelling of three or more directions when they are present, may help to establish the relative strength of transcallosal connections to lateral motor cortices.
We have demonstrated a specific association between integrity of white matter pathways and a functional skill in a healthy adult population. This suggests that inter-individual variability in brain structure has functional consequences. The current study provides us with a snapshot that makes it impossible to establish the direction of causality between structure and function: it is possible that innate variation in brain structure influences subsequent skill levels. It has recently been demonstrated, for example, that genotype not only influences brain structure (Bueller et al., 2006
), but also influences the degree of functional plasticity in the motor system (Kleim et al., 2006
). It is likely that, in addition to any innate influences on brain structure, differences in experience, manifest as variable activity along specific brain pathways, result in plastic changes that could be reflected in local measures of fractional anisotropy. Piano training, for example, results in different patterns of FA change depending on the age at which training is given (Bengtsson et al., 2005
), suggesting that environmental influences at specific developmental stages trigger long lasting structural changes that are reflected in FA measures along those pathways. FA reflects numerous structural properties of the white matter (Beaulieu, 2002
), some of which may be subject to experience-dependent changes. Exposure to an enriched environment in rats, for example, results in increases in the size or number of axons in the corpus callosum (Juraska and Kopcik, 1988
). Future longitudinal studies using diffusion MRI should test whether alterations in experience, such as intensive skill training, result in observable changes in white matter structure in the adult human brain.