Connectivity Differences in a Heterogeneous Sample of Mice from Multiple Mouse Models of Autism

Background: Autism is often described as a disorder of aberrant connectivity. Belmonte, 2004^[1], suggests that global long range connections are weakened, an idea supported by DTI findings of altered white matter. Just, 2012^[2], contends that the diminished long range connections are localized to fronto-posterior connections rather than global ones, measured by fMRI.

Objectives: We sought to describe connectivity differences in autism-related mice. While mouse models of autism reduce phenotypic variance by controlling the underlying genetics, comparisons between mouse and human research require samples that adequately reflect this heterogeneity. We therefore asked, if a connectivity study modeled after human research was carried out in mice in which individual mice from a variety of mouse models were grouped together, do the connectivity changes reflect human patterns?

Methods: We used anatomical covariance (AC) as a measure of connectivity. In AC, correlating regional volumes between a seed region and target regions or voxels across a sample of mice provide a network of correlation coefficients across the brain for each seed. Data (currently unpublished) suggest that networks of AC correlation coefficients significantly correlate with the underlying fiber networks in mice and predict expected functional connections. Unlike DTI, AC can capture connections mediated by small fiber tracts, and unlike fMRI, AC can be done ex-vivo.

High-resolution MR images of mice from multiple autism-related mouse models were obtained and registered together in a single group, and AC networks were computed between a set of seed regions and target voxels throughout the entire brain. Ellegood, 2014^[3], found that in a study of brain region volumes across mouse models, certain structures appear to be consistently affected; we used structures from this study as seeds. We looked at which voxels consistently showed increased or diminished connectivity across all seeds. Connectivity differences were determined by taking the difference in AC correlation coefficients at every voxel between the autism group and an appropriate set of wildtype mice for every seed, and computing the t-statistic at every voxel across all seeds.

Results: We found that overall, diminished connectivity was localized to the cerebral cortex, and increased connectivity was seen in subcortical structures. Specifically, regions consistently implicated in overconnectivity (Figure 1) included parts of: striatum, nucleus accumbens, corpus callosum, thalamus, hypothalamus, hippocampus, periaqueductal grey, midbrain, superior and inferior colliculus, medulla, and cerebellar paraflocculus and vermis. Regions consistently implicated in underconnectivity (Figure 1) included parts of: frontal association cortex, somatosensory and motor cortex, lateral orbital cortex, cingulate cortex, olfactory bulbs, corticospinal tracts and the arbor vita. Preliminary results further indicate that overconnectivity is seen in regions closer to the seed, and that this decreases as a function of distance to seed.

Conclusions: Connectivity differences in mice from the autism-mouse models used are spatially ordered, with structures within the cerebral cortex showing decreased connectivity and certain subcortical structures showing increased connectivity. While these findings fit the idea that higher order cognitive and behavioural deficits arise from aberrant cortical connectivity, more needs to be done to test the two connectivity theories proposed by Belmonte and Just.