Functional segregation and integration within fronto-parietal networks

Abstract

Experimental data on monkeys and functional studies in humans support the existence of a complex fronto-parietal system activating for cognitive and motor tasks, which may be anatomically supported by the superior longitudinal fasciculus (SLF). Advanced tractography methods have recently allowed the separation of the three branches of the SLF but are not suitable for their functional investigation. In order to gather comprehensive information about the functional organisation of these fronto-parietal connections, we used an innovative method, which combined tractography of the SLF in the largest dataset so far (129 participants) with 14 meta-analyses of functional magnetic resonance imaging (fMRI) studies. We found that frontal and parietal functions can be clustered into a dorsal spatial/motor network associated with the SLF I, and a ventral non-spatial/motor network associated with the SLF III. Further, all the investigated functions activated a middle network mostly associated with the SLF II. Our findings suggest that dorsal and ventral fronto-parietal networks are segregated but also share regions of activation, which may support flexible response properties or conscious processing. In sum, our novel combined approach provided novel findings on the functional organisation of fronto-parietal networks, and may be successfully applied to other brain connections.

Keywords: Diffusion tractography; Frontal parietal; Functional magnetic resonance imaging (fMRI); Meta-analysis; Superior longitudinal fasciculus.

Fig. 1

Mapping of the Superior Longitudinal Fasciculus (SLF). The top panel displays the average reconstruction of the SLF I (light blue), SLF II (navy blue) and SLF III (purple). The lower panel displays the axial sections of the three branches of the SLF. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

Meta-analytic maps. The maps of the 14 investigated functions are shown projected onto 3D-renderings of the brain (lateral and medial surfaces). A description of these maps can be found in Supplementary Results.

Fig. 3

Cross-correlation. This panel displays the cross-correlations between the 14 meta-analytic maps. Two main clusters can be observed, one including spatial/motor functions and one including non-spatial/motor functions.

Fig. 4

Principal component analysis. Panel ‘a’ shows the graph of the principal components (x) according to their eigenvalue sizes (y). Component 1 (pink) and component 2 (light blue) accounted for 70% of the total variance of the fronto-parietal activations. Panel ‘b’ and ‘c’ respectively show dorsolateral and medial tridimensional views and axial views of the two main components identified with the principal component analysis. Note that the intersection between the two components is displayed in dark blue. The raw weights for the different functions on the first two components are reported in Table 1. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 5

Areas of shared activation. Panel ‘a’ displays the map of fronto-parietal regions that are more probably activated by the 14 investigated functions (lateral and medial surfaces). BA: Brodmann area. Panel ‘b’ shows that the areas of shared activation are mostly associated with the SLF II. Average Z values of the functional maps at the location of the projections of the three branches of the SLF (with a 50% threshold) are reported. Error bars indicate confidence intervals (p<0.001).

Fig. 6

Functional roles of the Superior Longitudinal Fasciculus (SLF). Panel ‘a’ displays the cortical projections of the three branches of the SLF (lateral and medial view). Panel ‘b’ shows their functional correlates. We quantified the contribution of the SLFs to the spatial/motor and non-spatial/motor fronto-parietal meta-analytic maps. The SLF I appears to be primarily involved in spatial/motor functions, whereas the SLF III in non-spatial/motor functions. The SLF II was associated with both functions (see also Fig. 5b). Average Z values of the functional maps at the location of the projections of the three branches of the SLF (with a 50% threshold) are reported. Error bars indicate confidence intervals (p<0.001).