Gene transcription profiles associated with inter-modular hubs and connection distance in human fMRI networks

Graph theoretical methods have been widely used to investigate the topology of large-scale human brain networks constructed from resting state functional magnetic resonance imaging (fMRI). It has been demonstrated that such human functional connectomes have a complex topology comprising integrative components, such as hubs and inter-modular edges, that are associated with proxy markers of greater biological cost. In the absence of secure knowledge of the neurovascular mechanisms linking ensemble oscillations of neuronal populations to low frequency coupling or functional connectivity between regional fMRI time series, it has been challenging to validate fMRI network properties reductionistically. Supportive evidence to date has been mostly provided by analogous results on the relationships between integrative topology and biological cost in other nervous systems. Here, we use microarray data on brain regional expression of 20,737 genes to explore the relationships between fMRI network topology and transcription of genes annotated for biological processes and cellular components. We show that intra-modular degree and inter-modular degree are differently patterned in anatomical space, are differently associated with cytoarchitectonic classes of cortex, and are associated with distinct and statistically independent gene expression profiles. Genes strongly associated with nodes mediating many long-distance and inter-modular connections are significantly enriched for oxidative metabolism and mitochondria as well as for a subset of genes specifically enriched in human supragranular cortical layers. These results are directly supportive of the concept of high cost / high value network hubs in fMRI networks and point to the nascent opportunity to resolve the molecular and cellular substrates of human brain graphs