In this paper we have presented evidence indicative of opposing patterns of neural connectivity in the frontal and parietal lobes of children with 22q11.2DS and TD controls. These data take the form of commonly located clusters within major longitudinal fasciculi in which complementary measures that can be computed from the diffusion tensor produced patterns of differences that were essentially the opposite of one another in the two groups. Specifically, children with 22q11.2DS had significantly higher fractional anisotropy values in bilateral parietal and bilateral frontal clusters compared to TD control children. By contrast, TD children had significantly higher radial diffusion values in the same locations compared to children with 22q11.2DS. The high FA values in the 22q11.2DS group clusters suggest a pattern of connectivity that is primarily parallel to the major fiber tracts in which they are located. These are the superior longitudinal fasciculus for the parietal clusters and the frontal-occipital fasciculus for the frontal clusters (see Table and Figure for details). Conversely, the high values of RD in the TD control group clusters clearly indicate a pattern of connectivity that is greater in the perpendicular plane to the major fiber tracts in which they are located. We interpret this as evidence of greater connectivity with contiguous parietal and frontal cortical regions. It is also worth noting that in Figure most of clusters with greater mean diffusivity in the TD than the 22q11.2DS group overlap their clusters of higher radial diffusion. Since mean diffusivity consists of both axial and radial diffusion, this finding is consistent with evidence of a more widespread pattern of greater branching from major white matter tracts into adjoining cortex in this group, or higher complexity in their connective patterns derived from the RD component of this measure. Based on our statistical analyses we made the straightforward inference that higher radial values probably indicated greater connectivity to contiguous parietal and frontal cortex in TD children than those with 22q11.2DS, and that the higher FA values in the 22q11.2DS group represented a reduction of this connectivity matrix.
We further reported significant, and similarly complementary, patterns of correlations of visuospatial attention scores to diffusivity values in focal sub-regions of those clusters. These results generate some important hypotheses about neural connectivity and their effect on cognitive function that we intend to explore in future studies. The apparently counter-intuitive relationship between higher factional anisotropy in the right parietal cluster and worse performance on the attention task in the DS22q11 group is explained to a degree by their correlations of axial diffusion and cognitive performance scores. Both parietal clusters showed extraordinarily high positive correlations indicating that higher axial diffusion was related to worse performance. At the same time, while higher FA had correlated with better performance for the TD group (albeit in the right frontal cluster), there was again a positive correlation of AD with invalid cue cost size in that group in the left parietal cluster. These results suggest that increasing amounts of axial diffusion in parietal clusters relate to poorer performance on visuospatial attention tasks. This is presumably because axial diffusion is the complement of radial diffusion and so may reflect reduced complexity in or degree of connectivity to critical surrounding cortex. The fact that children with 22q11.2DS had much higher FA values in their parietal (and frontal) clusters than typical controls was taken to indicate that their connectivity perpendicular to the major fasciculus into adjoining cortex was reduced. The correlations with AD, a purer measure of orientation along a single axis than FA, appear to confirm that impression. At the same time, even though radial diffusion, and thus inferred cortical connectivity, was significantly higher in TD controls for all clusters, there is still necessarily a component of total (mean) diffusivity that is axial in nature. Apparently, the more axial diffusion there is on these clusters the greater the negative impact it has on performance. Correlations were also carried out with performance data from higher-level numerical cognition tasks such as speeded enumeration of random dot patterns and analog and numerical magnitude comparison. While some results were significant and broadly consistent with those presented above, no strong pattern like the one just described was evident. Our interpretation is that, because tasks such as these depend on acquired knowledge and strategy choice to a much greater degree than is the case for simple spatial cueing of attention, the relationship between basic neurobiological measures and cognitive function is bound to be much weaker. However, Barnea-Goraly and colleagues [25
] recently reported a similar correlation between FA and scores on a standardized Math test that suggest just such a relationship does exist in a slightly older group of participants with 22q11.2DS.
One question that is important to ask when interpreting neuroimaging findings in atypically developing populations is whether finding is more likely to indicate the delayed progression of typical development or that which arises from a completely different developmental trajectory [26
]. It appears that the findings we report here are more likely to be an instance of the latter than the former. This is because a recent study [27
] reported that the superior longitudinal fasciculus, where our most significant findings occurred, undergoes a prolonged maturation in typically developing individuals. In infancy and early childhood FA values are low and angle ± values are high, denoting reduced organization of fibers in the tract. That pattern reverses into the adult profile by 5 years of age. By contrast our developmentally delayed population of 7–14-year-old children with DS22q11 had higher FA values than our typically developing cohort in the cluster that we described, thus indicating that a completely different pattern of development had taken place. Whether this is likely to normalize over time can only be determined by future cross sectional or longitudinal studies involving wider age ranges. Interestingly, Hoeft et al. [28
], reported very similar findings to ours in children with Williams syndrome. That is another disorder characterized in part by significant impairments in visuospatial ability and their findings of increased FA in the right superior longitudinal fasciiculus correlated with poorer scores on a standardized measure of visuospatial ability. Together, these findings do indeed appear to indicate that atypical connectivity in the longitudinal fasciculus is a biomarker of and associated with visuospatial cognitive impairments in at least two neurogenetic disorders.
Despite the apparent evidence in favor of our interpretation of the data presented, it must still be kept in mind that the precise neuroanatomical implications of diffusion tensor MRI data remain exceedingly difficult to evaluate. For example, Pierpaoli et al. [29
] discuss in detail their conclusion that no clear relationship has been defined between measures of anisotropy or other diffusion measures and packing density of fibers, myelin density or distribution. In fact, Pierpaoli et al. conclude that they were "unable to identify a single microstructural factor or a combination of them to account for the observed differences in diffusion anisotropy in all regions of white matter in the normal human brain" (p. 646). However, some of the more direct measures of the directionality of the main axes of diffusion may have a sounder interpretive basis. For example, Song et al. [30
] suggest that one interpretation of increased radial diffusion measures in some populations, such as those with multiple sclerosis, may actually be a signal of dismyelination. Only further analyses that include higher resolution data combined with cognitive performance data and other imaging measures will be able to converge upon a more definitive account of the patterns in and functional implications of neural connectivity in clinical populations such as the one described here. Our ongoing research program has just such goals and will resolve two of the major limitations of our current results by providing complete brain coverage and by acquiring data in 12 directions at higher resolution on a 3T scanner. These advances will allow us to examine connectivity patterns in our data in more detail, to generate reasonable visualization of differences that are detected, and also to relate them to a wider range of cognitive processes that are impaired in children with 22q11.2DS. In doing so we expect to further advance our understanding of the neural foundations of cognitive impairments in 22q11.2DS and to use that knowledge in the design of effective interventions.