Dopaminergic modulation of motor network compensatory mechanisms in Parkinson's disease

Abstract

The dopaminergic system has a unique gating function in the initiation and execution of movements. When the interhemispheric imbalance of dopamine inherent to the healthy brain is disrupted, as in Parkinson's disease (PD), compensatory mechanisms act to stave off behavioral changes. It has been proposed that two such compensatory mechanisms may be (a) a decrease in motor lateralization, observed in drug-naïve PD patients and (b) reduced inhibition - increased facilitation. Seeking to investigate the differential effect of dopamine depletion and subsequent substitution on compensatory mechanisms in non-drug-naïve PD, we studied 10 PD patients and 16 healthy controls, with patients undergoing two test sessions - "ON" and "OFF" medication. Using a simple visually-cued motor response task and fMRI, we investigated cortical motor activation - in terms of laterality, contra- and ipsilateral percent BOLD signal change and effective connectivity in the parametric empirical Bayes framework. We found that decreased motor lateralization persists in non-drug-naïve PD and is concurrent with decreased contralateral activation in the cortical motor network. Normal lateralization is not reinstated by dopamine substitution. In terms of effective connectivity, disease-related changes primarily affect ipsilaterally-lateralized homotopic cortical motor connections, while medication-related changes affect contralaterally-lateralized homotopic connections. Our findings suggest that, in non-drug-naïve PD, decreased lateralization is no longer an adaptive cortical mechanism, but rather the result of maladaptive changes, related to disease progression and long-term dopamine replacement. These findings highlight the need for the development of noninvasive therapies, which would promote the adaptive mechanisms of the PD brain.

Keywords: Parkinson's disease; dynamic causal modeling; effective connectivity; fMRI; lateralization; parametric empirical Bayes.

© 2019 Wiley Periodicals, Inc.

Figures

Figure 1

Topography of the clusters containing group‐level BOLD activation maxima within the regions of interest (ROIs) used in the current study. (A) Primary motor cortex ROI peak activation clusters for the four experimental conditions (right hand movement results in LhM1 activation, left hand movement results in RhM1 activation, right foot movement results in LfM1 activation, and left foot movement results in RfM1 activation; p < .05, FWE corrected). (B) SMA and vPMC peak activation clusters for two conjunction analyses (right hand movement ∩ right foot movement results in LSMA and LvPMC activation, left hand movement ∩ left foot movement results in RSMA and RvPMC activation; MNI z = 17 in cut‐outs; p < .001, uncorrected). Only those clusters containing group‐level peaks are shown. Maxima used for VOI extraction procedure are circled and labeled [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 2

Possible contralateral‐ipsilateral activation scenarios leading to a decrease in activation laterality. “Normal” lateralization, that is, as seen in healthy participants (b), could also occur in participants with either concurrent increase (a) or concurrent decrease (c) in both contralateral and ipsilateral activation. On the other hand, a decrease in lateralization could arise from one of three possible scenarios: (d) an increase in ipsilateral activation, (e) a decrease in contralateral activation, or (f) a concurrent increase in ipsilateral and decrease in contralateral activation [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 3

Full dynamic causal model (DCM), including eight nodes — bilateral hM1, fM1, SMA, and vPMC. The arrows between the nodes of the cortical motor network represent the endogenous connectivity (DCM.A matrix), as well as the modulatory influences (DCM.B matrix) of the full model, where all possible modulatory connections are present. The four darker shaded nodes represent the targets of the driving input (DCM.C matrix), consisting of right hand, left hand, right foot, and left foot movements [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 4

The architecture of models in the DCM model space could be described by two factors: laterality and structure, which were modeled by five laterality model families (top) and 15 structure model families (bottom). The two factors were fully crossed, yielding a total of 75 models. The DCM model space was designed to test hypotheses about the architecture of the DCM.B matrix of input‐modulatory connectivity for the four movement conditions: Left hand, right hand, left foot and right foot. We tested five laterality model families: (1) symmetrical, (2) contralaterally‐lateralized, (3) ipsilaterally‐lateralized, (4) left‐lateralized, and (5) right‐lateralized, and 15 structure model families, which varied according to the presence or absence of various types of connections: interhemispheric (IE), homotopic (H), intrahemispheric (IA), and self‐inhibitory (S) [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 5

Behavioral results. The hand/foot icons at the top of the figure represent the body part (LH — left hand, RH — right hand, LF — left foot, RF — right foot) being moved in the corresponding condition. Top row: average percent correct of movements, per limb for each group. Values represent the percentage of correct movements of the total cued movements for the given limb. Bottom row: average force per limb for each group. Error bars represent 95% confidence intervals. Any significant between‐group (within‐body side, within‐limb) differences are indicated (* p < .05, ** p < .01, Bonferroni‐corrected) [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 6

Average laterality index (AveLI) for each experimental condition. The hand/foot icons at the top of the figure represent the body part (LH — left hand, RH — right hand, LF — left foot, RF — right foot) being moved in the corresponding condition. Since right hemispheric dominance is represented by negative AveLI values, for the sake of clarity of the visual representation, the sign of AveLI values for left hand and foot movement has been changed. Error bars represent 95% confidence intervals. Any significant between‐group (within‐body side, within‐limb) differences are indicated (* p < .05, ** p < .01, Bonferroni‐corrected) [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 7

Percent BOLD signal change in contralateral and ipsilateral regions of interest for each experimental condition. The hand/foot icons at the top of the figure represent the body part (LH — left hand, RH — right hand, LF — left foot, RF — right foot) being moved in the corresponding condition. Error bars represent 95% confidence intervals. Any significant between‐group (within‐body side, within‐limb, within‐ROI) differences are indicated (.p < 0.1, * p < .05, ** p < .01, Bonferroni‐corrected) [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 8

Results of Bayesian model family comparison for between‐group commonalities and differences. Left column of bar graphs: laterality model family, right column: structure model family. The y‐axis corresponds to the posterior probability, whereas the x‐axis corresponds to the model families within the given type of model family. Five laterality model families (1 — symmetrical, 2 — contralaterally‐lateralized, 3 — ipsilaterally‐lateralized, 4 — left‐lateralized, 5 — right‐lateralized) and 15 structure model families (see Figure 4) were compared

Figure 9

Results of Bayesian model comparison for between‐group commonalities and differences. The y‐axis corresponds to the posterior probability, whereas the x‐axis corresponds to the 75 models within the model space (see Figure 4)

Figure 10

Between‐group commonalities and differences in effective connectivity patterns of the motor network, as estimated through Bayesian model averaging (BMA). Top left panel: Endogenous connectivity (DCM.A matrix). Top right panel: driving input (DCM.C matrix). Bottom panel: modulatory influences (DCM.B matrix). The color of the arrow colors designates the value of the group‐level parameter estimate, i.e., the contribution of the given effect (between‐group commonality or difference) to the variance of DCM connectivity strengths. Red arrows indicate positive group‐level parameter estimates, while blue arrows indicate negative group‐level parameter estimates. Only those connections with a posterior probability greater than 0.75 are shown. Arrows in the DCM.C panel have been omitted – all connections are unidirectional, from the hand/foot input toward the cortical node [Color figure can be viewed at http://wileyonlinelibrary.com]

Figure 11

Characterization of between‐group differences in effective connectivity, as estimated through Bayesian model averaging (BMA). Left panel: Endogenous connectivity (DCM.A matrix). Right panel: driving input (DCM.C matrix). Arrows in the right panel (DCM.C) have been omitted – connections are unidirectional, from the hand/foot input toward the cortical node. Only those connections with a posterior probability – as estimated through BMA – greater than 0.75 are shown. Connections are characterized as “increased excitation” (dark orange), “reduced excitation” (light orange), “reduced inhibition” (light purple), “increased inhibition” (dark purple) [Color figure can be viewed at http://wileyonlinelibrary.com]