Conservative and disruptive modes of adolescent change in brain functional connectivity

Abstract

Adolescent changes in human brain function are not entirely understood. Here we used multi-echo functional magnetic resonance imaging (fMRI) to measure developmental change in functional connectivity (FC) of resting-state oscillations between pairs of 330 cortical regions and 16 subcortical regions in N=298 healthy adolescents. Participants were aged 14-26 years and were scanned on two or more occasions at least 6 months apart. We found two distinct modes of age-related change in FC: "conservative" and "disruptive". Conservative development was characteristic of primary cortex, which was strongly connected at 14 years and became even more connected in the period 14-26 years. Disruptive development was characteristic of association cortex, hippocampus and amygdala, which were not strongly connected at 14 years but became more strongly connected during adolescence. We defined the maturational index (MI) as the signed coefficient of the linear relationship between baseline FC (at 14 years, F C 14 ) and adolescent change in FC (∆F C 14−26 ). Disruptive systems (with negative MI) were functionally specialised for social cognition and autobiographical memory and were significantly co-located with prior maps of aerobic glycolysis (AG), AG-related gene expression, post-natal expansion of cortical surface area, and adolescent shrinkage of cortical depth. We conclude that human brain organization is disrupted during adolescence by the emergence of strong functional connectivity of subcortical nuclei and association cortical areas, representing metabolically expensive re-modelling of synaptic connectivity between brain regions that were not strongly connected in childhood. We suggest that this re-modelling process may support emergence of social skills and self-awareness during healthy human adolescence.

Dynamical consequences of regional heterogeneity in the brain’s transcriptional landscape

“…Finally, to further ensure that any results obtained for the E:I model cannot be explained by a generic spatial gradient, we also evaluated model performance after permuting the assignment of E:I values to regions. A total of 10,000 permutations were performed by spatially rotating the expression values using the approach described in ( 72 ). Critically, this approach preserves the spatial autocorrelation of the expression values, which is essential for ensuring that any resulting effects are not driven by low-order spatial gradients (see also ( 31 )).…”

Dynamical consequences of regional heterogeneity in the brain’s transcriptional landscape

“…Last, to further ensure that any results obtained for the E:I model cannot be explained by a generic spatial gradient, we also evaluated model performance after permuting the assignment of E:I values to regions. A total of 10,000 permutations were performed by spatially rotating the expression values using the approach described in (87). Critically, this approach preserves the spatial autocorrelation of the expression values, which is essential for ensuring that any resulting effects are not driven by low-order spatial gradients [see also (32)].…”

Functional MRI connectivity accurately distinguishes cases with psychotic disorders from healthy controls, based on cortical features associated with neurodevelopment

“…The glycolytic index (GI) is a measure of aerobic glycolysis, a metabolic pathway that is utilised specifically by neurodevelopmental processes (20,46). Maturational index (MI) is a novel measure of adolescent change in functional connectivity measured in a longitudinal fMRI study of healthy young people: M I < 0 indi-cates a region that has "disruptively" increased connectivity during adolescence and early adulthood (14-26 years) from a low baseline level at age 14 (21). Both maps were correlated with the machine diagnostic feature map for psychosis, which is compatible with the theory that psychotic disorders result from aberrant brain network development (16)(17)(18).…”

“…The map of glycolytic index (GI) was negatively correlated with the map of maturational index (MI) (Spearman's ρ = −0.54, P < 0.001 (21)). The GI map was positively correlated with the fMRI diagnostic feature maps from the Maastricht, Dublin and Cobre datasets, with (r = 0.19, df = 246, P = 0.0025), (r = 0.24, df = 288, P =< 0.001) and (r = 0.34, df = 291, P < 0.001), respectively; see Figure 3.…”

“…We compared ML diagnostic feature maps to maps of the group mean differences between cases and controls. In view of evidence that psychotic disorders are related to abnormal brain network development (16)(17)(18), we tested the hypothesis that the diagnostic feature map for fMRI connectivity would be co-located with a prior map of aerobic glycolysis, a metabolic pathway that is utilised specifically by neurodevelopmental processes (20), and with a prior map of change in functional connectivity during healthy human adolescence and early adulthood (21). Finally, since psychosis is heritable (22), we tested the hypothesis that fMRI data on non-psychotic siblings of cases would be assigned intermediate probabilities of psychosis by the ML algorithm trained on case and control data.…”