Gene transcription profiles associated with inter-modular hubs and connection distance in human functional magnetic resonance imaging networks

Abstract

Human functional magnetic resonance imaging (fMRI) brain networks have a complex topology comprising integrative components, e.g. long-distance inter-modular edges, that are theoretically associated with higher biological cost. Here, we estimated intra-modular degree, inter-modular degree and connection distance for each of 285 cortical nodes in multi-echo fMRI data from 38 healthy adults. We used the multivariate technique of partial least squares (PLS) to reduce the dimensionality of the relationships between these three nodal network parameters and prior microarray data on regional expression of 20 737 genes. The first PLS component defined a transcriptional profile associated with high intra-modular degree and short connection distance, whereas the second PLS component was associated with high inter-modular degree and long connection distance. Nodes in superior and lateral cortex with high inter-modular degree and long connection distance had local transcriptional profiles enriched for oxidative metabolism and mitochondria, and for genes specific to supragranular layers of human cortex. In contrast, primary and secondary sensory cortical nodes in posterior cortex with high intra-modular degree and short connection distance had transcriptional profiles enriched for RNA translation and nuclear components. We conclude that, as predicted, topologically integrative hubs, mediating long-distance connections between modules, are more costly in terms of mitochondrial glucose metabolism.This article is part of the themed issue 'Interpreting BOLD: a dialogue between cognitive and cellular neuroscience'.

Keywords: Allen Institute for Brain Sciences; community structure; economy; graph theory; hub; transcriptome.

Figure 1.

Complex topology of fMRI brain networks. (a) Network representation of brain functional connectivity. Colours represent eight distinct modules; the size of nodes is proportional to their degree; only the top 4% strongest connections are shown for clarity. (b) Degree distribution of the brain functional networks, pooled across subjects. (c) Normalized rich club curves of each participant's brain functional network. (d) Boxplots showing key network measures for the brain functional networks (in red) compared to randomized networks with preserved degree distribution (grey). From left to right, the metrics shown are: modularity Q, clustering C, path length L, and small-worldness σ. (e) Cortical surface map colour-coding brain regions according to fMRI modules, as in panel (a). The legend describes the approximate anatomical location of each module and defines the acronym with which each module (mod) is represented in figures 2 and 3. (f) Cortical surface map colour-coding brain regions according to von Economo & Koskinas's cytoarchitectonic classification [45]. Class 1 (purple): granular cortex, primary motor cortex. Classes 2 and 3 (blue and green): association cortex. Class 4 (orange): dysgranular cortex, primary/secondary sensory cortex. Class 5 (yellow): agranular cortex, primary sensory cortex. Class 6 (cyan): limbic regions, allocortex. Class 7 (magenta): insular cortex. The legend also defines the acronym with which each cytoarchitectonic class is represented in figures 2 and 3.

Figure 2.

Anatomical and cytoarchitectonic patterning of fMRI network hubs. (a) Binary graphs constructed at 4% connection density (for clarity) showing a sagittal view of the brain. Nodal size was scaled by five nodal metrics: from top to bottom, total degree (k), intra-modular degree (k_intra), inter-modular degree (k_inter), participation coefficient (PC), and average nodal distance (d). Nodes are coloured by module, as defined in figure 1. Nodes with high PC (>0.5) are highlighted by square markers with a magenta outline. (b) Axial view of the brain networks in panel (a). (c) Boxplots showing the distribution of nodal distance and nodal topological metrics in each of the eight modules. Modules are colour-coded and named according to the scheme shown in figure 1. (d) Boxplots showing the distribution of nodal distance and nodal topological metrics in each of the seven cytoarchitectonic classes as defined by von Economo & Koskinas's [45] classification of cortical laminar patterns. Cytoarchitectonic classes are numbered and colour-coded according to the scheme in figure 1.

Figure 3.

Gene expression profiles associated with fMRI network topology. (a,b) Binary graphs constructed at 4% connection density (for clarity) showing nodes with size and colour saturation scaled by regional scores on PLS1 (a) and PLS2 (b). Larger, darker nodes represent regions with higher PLS scores, i.e. higher expression levels of genes positively weighted on the corresponding PLS component. (c,d) Boxplots representing the distribution of regional scores on PLS1 (c) and PLS2 (d) in each of the eight network modules shown in figure 1. (e,f) Boxplots representing the distribution of regional scores on PLS1 (e) and PLS2 (f) in each of the seven cytoarchitectonic classes shown in figure 1.

Figure 4.

Enrichment analysis of PLS2 gene expression profile associated with long-distance connections and inter-modular hubs in fMRI networks. Significantly enriched GO terms are plotted in semantic space such that similar terms are represented close to one another. Markers are scaled and coloured according to the log₁₀ of the p-value for the significance of each term. Large blue circles are highly significant (P_FDR < 10⁻¹⁵), while small red circles are less so (P_FDR < 0.001).