The role of contralesional regions for post-stroke movements revealed by dynamic connectivity and TMS interference

Abstract

Connectivity changes after brain lesions due to stroke are tightly linked to functional outcome. Recent analyses of fMRI time series indicate that dynamic functional network connectivity (dFNC), reflecting transient states of connectivity may capture network-level disruptions distant to the lesion site. Yet, the relevance of such dynamic connectivity patterns for motor recovery remains unclear. We, therefore, combined the analysis of static and dFNC and a repetitive transcranial magnetic stimulation (rTMS) lesion approach, to test whether dFNC provides region-specific insight into motor system reorganization after stroke. We focused on the contralesional primary motor cortex (M1) and anterior intraparietal sulcus (aIPS), two regions previously shown to modulate motor performance post-stroke in a time dependent manner. In 18 individuals in the chronic phase after stroke (with either persistent or recovered deficits) and 18 healthy participants, we analyzed static and dynamic resting-state connectivity. We then applied online rTMS intereference over contralesional aIPS and M1 during hand movement tasks to assess region-specific contributions to motor behavior. Consistent with previous studies, dFNC states were associated with persisting motor deficits, whereas static connectivity was not associated with motor outcome. dFNC but not static connectivity was associated with residual motor deficits and explained TMS-induced behavioral changes, when applying rTMS over contralesional M1. For contralesional aIPS, both static and dynamic connectivity were linked to TMS effects. This indicates that dFNC - more than static connectivity - contains information on the functional relevance of brain regions for motor outcome, specifically contralesional M1. Our results highlight the added value of temporal network analysis in understanding mechanisms of stroke recovery mechanisms.

Keywords: Dynamic functional connectivity; Motor performance; Network segregation; Stroke; rTMS.

Fig. 1

Relating static and dynamic connectivity to the role of a region of interest (e.g. M1). Left (blue): dimension reduction of resting state fMRI data. Motor-related connections, which are connected to the investigated region of interest (M1 or aIPS) are transformed into an equal number of principal components for static and dynamic functional connectivity. Right (red): Effects on poststroke movements by interfering with the region of interest (M1 or aIPS) using online-rTMS. Bottom (yellow): Regression analyses test for a relationship of dynamic and static functional connectivity with the causal motor contributions of contralesional M1 and aIPS, respectively. ICA = independent component analysis, ROIs = regions of interest. FC = functional connectivity, M1 = primary motor cortex, PCA = principal component analysis. (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 2

Motor-related network. Left: Spatial maps of 20 ICA components reflecting intrinsic connectivity networks across patients and healthy participants, color coded by affiliation to larger-scale networks. BG (dark purple) = basal ganglia network, SMN (light blue) = sensorimotor network, CEN (dark green) = central executive network, SN (light green) = saliency network, CB (yellow) = cerebellar network. Right: Matrix of static connectivity values between all components of all participants (patients and healthy controls). (For interpretation of the references to color in this figure legend, the reader is referred to the web version of this article.)

Fig. 3

Different dynamic connectivity in stroke (A) Distinguishing two connectivity states reveals stroke-related differences in state 2, indicated by asterisks (ANOVA P < 0.05, FDR-corrected). (B) Reduced mean fraction and dwell times in state 1 found in patients with lower outcome, compared to patients with higher outcome and healthy controls (between-group effect P < 0.001, asterisks indicate post hoc independent t-tests ***P < 0.001 **P < 0.01; HC = healthy controls, HMO = patients with higher motor outcome, LMO = patients with lower motor outcome). Fraction times were only compared for state1 due to their symmetry with state 2 (C) After stroke, low motor performance is associated with longer fraction (Pearson r = -0.56, P = 0.014) and dwell time (Pearson r = -0.51, P = 0.030). Separate regression lines for healthy controls illustrate that this relationship is inverted at higher levels of motor performance.

Fig. 4

Connectivity linked to rTMS effects on contralesional M1. (Top) Patients’ role of contralesional M1 could be explained by dynamic but not static connectivity. Matrices indicate contributions of each connectivity (columns) to each PCA-dimension (rows). Dimension 2 was selected by the stepward regression model to explain the effects of contralesional rTMS, with a negative weight of −0.3. Note that, compared to the other dimensions, Dimension 2 receives most contributions from connections within the sensorimotor system and cerebellum. (Bottom) Connectivity reflected by Dimension 2, involving a varying connectivity of a widely connected pattern between M1 and all subnetworks in state 1, and a more restricted pattern including mainly sensorimotor and cerebellar connections in state 2. cM1 = contralesional M1, iM1 = ipsilesional M1.

Fig. 5

Connectivity linked to rTMS effects on contralesional aIPS. (Top) Patients’ role of contralesional aIPS was explained by dynamic and static. Matrices indicate contributions of each connectivity (columns) to each PCA-dimensions (rows). Dimensions selected in each model (3 and 5, respectively) were both positively weighted for rTMS effects. Compared to the other dimensions, dynamic dimension 3 and static dimension 5 are both driven by the connection between ipsilesional aIPS and the medial frontal cortex and putamen (Bottom) Anatomical visualization of connectivity patterns reflected by Dimensions 3 (dynamic connectivity) and 5 (static connectivity). caIPS = contralesional aIPS, iaIPS = ipsilesional aIPS.