Having reviewed processing accounts of cognitive impairments observed in the DS22q11.2 population, we turn now to an examination of the possible neural basis of those disabilities. Investigation of the relationship between cognitive function and neural structure is motivated by the question of whether there are patterns of unusual brain morphology in regions that lesion and functional imaging studies have shown, mostly in adults, to be critical for the processes delineated above. In other words, it may be inferred that differences such as the amount of neural tissue in a critical brain region between individuals with DS22q11.2 and typically developing controls may be related to the differential ability to carry out cognitive operations thought to depend on that region. However, in making such inferences it is important to bear in mind that some key structure/function mapping assumptions made in studies of adult populations are unlikely to be appropriate. For example, studies of the developing brain tend to focus on functional connectivity rather than strict localization, and further assume that static structure–function mappings are inappropriate for the plastic, developing brain (for an extensive review of this issue, see Johnson, Halit, Grice, & Karmiloff–Smith, 2002
). In addition, they caution against unidirectional, causal “deficit” assumptions where “damage to specific neural substrates both causes and explains the behavioral deficits observed in developmental disorders” (p. 525). Although this last constraint is an important goal for a mature theory, it is valuable to use some aspects of the deficit assumption as a means of generating candidate hypotheses for the neural basis of cognitive dysfunction, especially early in a research program like ours. Thus, to extend our limited understanding of brain/behavior interactions in this population, we next review recent findings on neuroanatomy in DS22q11.2. We pay particular attention to connections between the morphological anomalies detected and the functional impairments just described, while keeping Johnson et al.’s (2002)
exhortations in mind.
Children with DS22q11.2 have been previously found to have significant overall reductions in brain volume, compared to typically developing children; Eliez, Schmitt, White, and Reiss (2000)
and Kates, Burnette, Jabs, Rutberg, Murphy, Grados, Geraghty, Kaufmann, and Pearlson (2001)
reported whole brain volumetric reductions of 11 and 8.5%, respectively. In both studies, the reduction in white matter was slightly greater than that in gray matter. Areas of reduced volume were concentrated in the posterior and inferior regions of the brain, including the parietal, temporal, and occipital lobes, and in the cerebellum. When total brain volume was accounted for, the frontal lobes were relatively enlarged. A more diverse pattern of differences, which included posterior reductions, as well as reduced gray matter in the insula and frontal and right temporal lobes, was reported in adults with DS22q11.2 by Van Amelsvoort, Daly, Robertson, Suckling, Ng, Critchley, Owen, Henry, Murphy, and Murphy (2001)
. These differences may be due to the older age of the study population, or the fact that an IQ-matched control group was employed. Both Van Amelsvoort et al. (2001)
and Barnea–Goraly, Menon, Krasnow, Ko, Reiss, and Eliez’ (2003)
diffusion tensor imaging study reported findings consistent with increased white matter volume in the area of the splenium of the corpus callosum in the DS22q11.2 population.
We acquired MRI data that would allow us to examine the structure, volume, and white matter integrity of the brains of children with DS22q11.2 and controls (for details, see Simon, Ding, Bish, McDonald–McGinn, Zackai, & Gee, 2005
). Our findings, gathered from 18 children with DS22q11.2 and 18 controls, aged between 7 and 15 years, showed that proportion of gray and white matter and cerebrospinal fluid (CSF) that made up the total brain volume was similar in both groups and was within expected component ranges (e.g., Woods, 2004
) for overall brain tissue (approximately 55, 27, and 18% for gray matter, white matter, and CSF, respectively). However, also consistent with previous reports, we found that the total brain volumes for children with DS22q11.2 were significantly lower than those of controls (by 8.9%). This difference is accounted for by reductions in the gray matter (9.9%) and white matter (11.07%) components, but not in CSF (1.08%). It is reasonable to assume that, as tissue decreases, CSF values increase. The lack of a difference in CSF volume between children with DS22q11.2 and controls is likely due to smaller cranial volume in children with DS22q11.2 compared to controls. Although head circumference was not measured systematically by us, we do have such measures for three children in our sample taken at the time of our studies. Head circumference ranged from 0.2 to 0.67 standard deviations below the mean. Furthermore, Gerdes et al. (1999)
reported that, in a sample of 40 children with DS22q11.2 between the ages of 13 and 63 months, 20% had a head circumference below the 5th percentile, 78% had a circumference between the 5th and 50th percentile and only one child was at the 95th percentile. Thus, it is possible that the lack of significant CSF differences may be attributable to reduced cranial volume in children with DS22q11.2.
Rather than employing the region of interest approach utilized by many of the previous studies, in which volumes of entire brain regions are quantified, we used voxel-based morphometry methods (Ashburner & Friston, 2000
). This involves transforming each individual brain to fit a normalized template to reveal group differences between populations in the form of clusters of specific voxels that can be localized and identified using the standard Talairach coordinates and labels. After brain tissue is segmented into gray and white matter and CSF, the analyses are carried out. Because many of the clusters that we detected were quite large they often spanned more than a single tissue type. For example, the peak of a cluster may be in gray matter with an extent large enough to include surrounding white matter. In those cases, the complementary analyses also detected those colocalized differences, and this overlap can be observed in –. We will report our results in terms of the major regions in which differences were found, for a range of tissue types, where appropriate.
Voxel-based morphometry results depicting clusters of greater gray matter volumes (a) in control children than in those with DS22q11.2 and (b) in children with DS22q11.2 than in controls.
Voxel-based morphometry results depicting clusters of greater fractional anisotropy values (a) in control children than in those with DS22q11.2 and (b) in children with DS22q11.2 than in controls.
The biggest differences found between the brains of the children with DS22q11.2 and those of the controls encompassed midline structures in the posterior brain, as discussed below. Thus, our findings replicated most of those previously reported, which include volume reductions in occipital, parietal, temporal and cerebellar regions (e.g., Eliez, Blasey, Menon, White, Schmitt, & Reiss, 2001
; Eliez, Blasey, Schmitt, White, Hu, & Reiss, 2001
; Eliez, Schmitt, White, & Reiss, 2000
; Kates et al., 2001
). In our sample, the gray matter analysis showed that the most extensive areas of reduced volume in the brains of children with DS22q11.2 were found in regions from the anterior aspects of the medial cerebellum through the parahippocampal, fusiform, and lingual regions, up through the cuneus, precuneus, posterior corpus callosum, posterior parietal lobes, and following the cingulum from posterior to a small medial region of its anterior aspect. In cognitive processing studies of healthy adults, many of those areas have been directly implicated in precisely the kinds of visuospatial and numerical tasks for which we have reported impairments in children with DS22q11.2 (e.g., Allen, Buxton, Wong, & Courchesne, 1997
; Corbetta, 1998
; Culham et al., 2001
; Dehaene, Piazza, Pinel, & Cohen, 2003
; Mesulam et al., 2001
; Pesenti, Thioux, Seron, & De Volder, 2000
; Pinel, Le Clec’H, van de Moortele, Naccache, Le Bihan, & Dehaene, 1999
; Posner & Petersen, 1990
). Consistent with previous reports of relatively enlarged frontal lobes, we also found two regions of increased volume in the frontal regions of the brains of children with DS22q11.2. These showed more gray matter in the right middle and superior frontal gyri, and in the right insula and superior, middle, and transverse temporal gyrus ().
A complementary analysis of differences in CSF volumes detected clusters of increased CSF in children with DS22q11.2, which were colocalized with many of those of the reduced gray matter, and highlighted some other specific locations. Among the most obvious of these was an increase in all ventricular areas, especially the lateral ventricles. Ventricular dilation has been shown to have a strong relationship with impairments in visuospatial cognition in children and adults (e.g., Fletcher, Bohan, Brandt, Brookshire, Beaver, Francis, Davidson, Thompson, & Miner, 1992
; Mataro, Poca, Sahuquillo, Cuxart, Iborra, de la Calzada, & Junque, 2000
). Regionally specific increases in CSF were also detected in the fourth ventricle area, presumably resulting from the absence of what should have been tissue in the pons, medulla, and cerebellar regions. Because of increases in the lateral ventricles, regional CSF increases in the DS22q11.2 population also encompassed some of what would typically be caudate and anterior regions, corresponding to reduced volume in these regions (). Although overall white matter volumes were reduced to a larger degree than were gray matter volumes in children with DS22q11.2, only a single significant cluster emerged from our voxel-based whole brain measures. This is probably because white matter reductions are more diffuse and widespread than those involving gray matter. Therefore, they were less likely to coalesce into statistically significant localized clusters given the strict thresholds that we employed. Thus, the one significant cluster, in the right middle and superior frontal gyral regions, likely represents the most concentrated area of white matter reduction (). Three other large clusters did trend towards but not reach statistical significance (for details, see Simon et al., 2005
). These were in the same cerebellar and midbrain/brainstem regions reported for gray matter reductions and CSF increases, as well as in the inferior parietal and precuneus regions. As noted above, some of these areas are critical to aspects of visuospatial and numerical cognition.
Voxel-based morphometry results depicting clusters of greater cerebrospinal fluid volumes in children with DS22q11.2 than in controls.
Voxel-based morphometry results depicting clusters of greater white matter volumes in control children than in those with DS22q11.2.
As we shall see below, the results that we believe are of most interest from this study emerged from our diffusion tensor imaging analyses. This relatively new imaging methodology can be used to compute a measure called “fractional anisotropy” (FA), which characterizes the degree of coherence of orientation of water diffusion in the brain. As a result, it can be used as an indirect measure of the organization of white matter tracts, or myelination (Basser & Pierpaoli, 1996
; Pierpaoli & Basser, 1996
). Although some interpretive issues regarding FA measurements remain to be addressed, larger values are typically interpreted to indicate greater amounts of white matter organized in specific orientations, thus indicating the presence of neuronal fiber tracts. Our primary diffusion tensor imaging results showed that the typically developing controls had a large cluster of higher FA values, compared to children with DS22q11.2, in an area that encompassed the corpus callosum, some cingulate white matter, and thalamic white matter in the area of the pulvinar nucleus (). By contrast, the children with DS22q11.2 had a pattern of increased FA that differed in terms of its location and extent. The main cluster encompassed the midline of the cingulate gyrus, from anterior to posterior, and extended into a wedge shaped cluster that included the splenium of the corpus callosum, the precuneus and large portions of the inferior parietal lobe. One smaller cluster, in the right lateral inferior parietal area of the supra-marginal gyrus, was also detected (). This latter cluster again involves brain regions critical to visuospatial and numerical cognition that we referred to earlier although, at present, the implications of this result for cognitive function are unclear. The primary significance of these different FA patterns is that the increased lateral ventricular size of children with DS22q11.2 appears to be related to a shift in location and a change in the morphology of the corpus callosum, which may negatively impact posterior parietal connectivity. The difference can be most easily appreciated in , which depicts the major FA clusters in each group overlaid onto the clusters of increased CSF in the DS22q11.2 population. In each case, we interpret the priary clusters of increased FA to represent the corpus callosum, which carries a significant proportion of the interhemispheric connective fiber tracts. It is clearly evident from is the fact that the location of the corpus callosum in the control population falls entirely within the space taken up by the significantly dilated lateral ventricles in the DS22q11.2 population. In other words, the significant group difference in FA, which compares the degree of restricted orientation of diffusion in a given voxel, appears to represent the highly oriented, myelinated fiber tracts in this location in the brains of typically developing children, whereas this location in the DS22q11.2 brain appears devoid of any neural tissue and is instead filled with isotropically diffusing CSF.
Composite depictions of clusters of greater cerebrospinal fluid volumes in children with DS22q11.2 along with clusters of greater fractional anisotropy values (a) in control children and (b) in the same children.
shows the apparent relationship of ventricular dilation to the displacement of the DS22q11.2 corpus callosum. It is evident from this image that the interhemispheric connections of the splenium have been shifted into a far more superior and posterior location than is true for the controls (see ). The corpus callosum also appears to have been significantly altered in terms of its overall shape and size. The mechanism underlying this change is unclear, although there is no evidence that increased cranial pressure, such is the case with hydrocephalus, is responsible. We are currently working on direct morphological analyses of the callosa in these two groups. Interestingly, Shashi et al. (2004)
, using different methodology, also reported abnormal callosal morphology in children with DS22q11.2. The authors manually traced and measured the area of different sections of the corpus callosum from a midsagittal MRI image. They found the isthmus, or perisplenial region, or the corpus callosum to be larger in children with DS22q11.2 than age- and gender-matched controls.
In addition, there is substantial evidence that the executive control aspect of the ANT, specifically the monitoring mechanism that is dysfunctional in children with DS22q11.2, relies heavily in adults on the appropriate functioning of the anterior cingulate (Botvinick, Nystrom, Fissell, Carter, & Cohen, 1999
; Bush, Luu, & Posner, 2000
; Casey et al., 2000
). Our results also showed that children with DS22q11.2 have increased CSF () and abnormal FA () in an area just inferior to the area that functional neuroimaging studies with adults have implicated in conflict monitoring (Kerns, Cohen, MacDonald, Cho, Stenger, & Carter, 2004
). Thus, this anatomical difference may be related to the executive impairment we reported earlier. However, just as with other structure/function mappings found in adults, we should be cautious about assuming that the same neural substrate underlies children’s performance. Evidence for activation pattern differences between children and adults have been found by several studies of tasks requiring error monitoring, inhibition or response selection (e.g., Booth, Burman, Meyer, Lei, Trommer, Davenport, Li, Parrish, Gitelman, & Mesulam, 2003
; Bunge, Dudukovic, et al., 2002
; Durston, Thomas, Yang, Ulug, Zimmerman, & Casey, 2002
Finally, using the region of interest method in a separate study, we have also demonstrated that the same 18 children with DS22q11.2, whose whole brain volumetric data were described above, show a significant reduction in total thalamic volume compared to the 18 controls (Bish, Nguyen, Ding, Ferrante, & Simon, 2004
). The amount of the reduction of the whole thalamus was roughly equivalent to the reduction of the overall brain volume (approximately 10%). However, specific measures of the posterior thalamus, an area containing the pulvinar nucleus, revealed significantly greater volumetric reductions (22.7%) in children with DS22q11.2 relative to controls. Pulvinar reductions have previously been linked in adults to both visuospatial processing impairments (Petersen, Robinson, & Morris, 1987a
) and schizophrenia (Byne, Buchsbaum, Kemether, Hazlett, Shinwari, Mitropoulou, & Siever, 2001