Altered network organization with shift to local processing during adaptive behaviour in multiple sclerosis

To adjust to changing environmental demands, the brain needs to organize and (re)configure mental resources in so-called task sets. Following a network perspective of cognitive functions, these processes are considered to rely on dynamic exchange of information between distant brain regions, possibly synchronized by slow-wave theta oscillations. However, few attempts have been made to describe the directional transfer of information in theta networks during behavioural adaptation. In the current study, we examined directed neural information transfer in 57 healthy individuals and 57 patients with multiple sclerosis with a focus on brain-behaviour associations during task switching. For this, we calculated behavioural switch costs and further applied an artificial neural network-based approach on neurophysiological EEG data to estimate linear and non-linear effective connectivity measures within cortical networks during task switching. Source reconstruction and connectivity analyses revealed distinct neural network architectures and communication strategies during task switching in healthy controls compared to patients with multiple sclerosis. In healthy controls, enhanced modulations of long-range fronto-parietal connectivities were associated with higher reaction time switch costs, possibly reflecting increased cognitive demands rather than adaptive processes. In contrast, patients with multiple sclerosis relied more strongly on local intra-regional modulations, which is consistent with the idea of a dysconnectivity syndrome in multiple sclerosis. Strikingly, stronger modulations in self-connectivities were associated with better accuracy measures in patients with multiple sclerosis, suggesting a compensatory mechanism to support performance. Notably, only non-linear directed connectivity measures were related to behavioural switch costs, emphasizing the importance of non-linear models for adaptive cognitive processes.