DOPAMINE EFFECTS ON FRONTAL CORTICAL BLOOD FLOW AND MOTOR INHIBITION IN PARKINSON’S DISEASE

Abstract

Parkinson’s disease (PD) is characterized by dysfunction in frontal cortical and striatal networks that regulate action control. We investigated the pharmacological effect of dopamine agonist replacement therapy on frontal cortical activity and motor inhibition. Using Arterial Spin Labeling MRI, we examined 26 PD patients in the off- and on-dopamine agonist medication states to assess the effect of dopamine agonists on frontal cortical regional cerebral blood flow. Motor inhibition was measured by the Simon task in both medication states. We applied the dual process activation suppression model to dissociate fast response impulses from motor inhibition of incorrect responses. General linear regression model analyses determined the medication effect on regional cerebral blood flow and motor inhibition, and the relationship between regional cerebral blood flow and motor inhibitory proficiency. We show that dopamine agonist administration increases frontal cerebral blood flow, particularly in the pre-supplementary motor area (pre-SMA) and the dorsolateral prefrontal cortex (DLPFC). Higher regional blood flow in the pre-SMA, DLPFC and motor cortex was associated with better inhibitory control, suggesting that treatments which improve frontal cortical activity could ameliorate motor inhibition deficiency in PD patients.

1. INTRODUCTION

The ability to execute goal-relevant actions and to suppress irrelevant prepotent action impulses is crucial for navigating dynamic action-oriented environments. Cortico-striatal circuits linking the prefrontal cortices to the direct and indirect basal ganglia pathways coordinate, respectively, action selection and suppression of irrelevant action impulses (Alexander et al., 1986; Mink, 1996; Redgrave et al., 1999; Ridderinkhof et al., 2004). Stopping prepotent response tendencies is an important aspect of inhibitory control that is modulated by key frontal cortical areas including the pre-supplementary motor area (pre-SMA), inferior frontal cortex (IFC), and dorsolateral prefrontal cortex (DLPFC) (Aron et al., 2003, 2007; Frank, 2006; Alexander et al., 2007; Forstmann et al., 2008a; Duann et al., 2009; van Gaal et al., 2011; Jahfari et al., 2011; Ridderinkhof et al., 2011; van Belle et al., 2014; Jahanshahi et al., 2015). These cortical-basal ganglia circuits are responsive to dopamine, and thus, diseases that disrupt these networks, like Parkinson’s disease (PD), can be characterized by deficits in inhibitory action control.

Dopamine therapy generally remediates PD-related inhibitory control deficiencies (Wylie et al., 2012; van Wouwe et al., 2016). However, the effect of dopamine therapy on inhibitory control in PD also depends on individual differences in disease duration (Manza et al., 2017), motor severity (Tolleson et al., 2017), and behavioral symptoms that can emerge with dopamine therapy (Wylie et al., 2012; Claassen et al., 2015). While neuroimaging studies comparing PD patients with healthy controls have emphasized functional deficits in regions involved in motor inhibition, including the IFC, pre-SMA, striatum, and thalamus (Fernández-Seara et al., 2012; Herz et al., 2014; Ye et al., 2014; Vriend et al., 2015; Criaud et al., 2016), the effect of dopamine medication on regional neuronal activity in PD patients remains unclear.

This study describes the pharmacological effect of dopamine agonist (DAA) therapy on frontal cerebral blood flow (CBF) in patients with PD, and the relationship between frontal cortical regional CBF (rCBF) and individual differences in inhibitory action control. DAA therapy improves motor symptoms in PD, and can alter the proficiency of inhibiting action impulses (Wylie et al., 2012; Claassen et al., 2015; Yang et al., 2016, 2018). Using non-invasive arterial spin labeling (ASL) MRI (Claassen et al., 2017), we examined a cohort of PD patients in the off- and on-DAA medication states to assess the effect of DAA on rCBF in frontal regions involved in motor inhibition, and the relationship between rCBF and action control performance as measured by a well-established cognitive test, the Simon task (Simon, 1969). Given our previous findings of improved action control in response to DAA therapy, we predicted that DAA administration would improve the suppression of action impulses, and we hypothesized that individual differences in rCBF in frontal regions are associated with proficiency of motor inhibition.

2. METHODS

2.1. Participants

Patients with idiopathic PD meeting UK Brain Bank criteria (Gibb and Lees, 1988) treated with dopaminergic therapy were recruited from the Movement Disorders Clinic at Vanderbilt University Medical Center. The study has been carried out in accordance with The Declaration of Helsinki, and all subjects provided written, informed consent before participating in the study in compliance with the standards of ethical conduct in human investigation regulated by the Vanderbilt Institutional Review Board.

The extent of PD severity was assessed by a board-certified neurologist (DOC). The Movement Disorders Society-United Parkinson’s Disease Rating Scale (MDS-UPDRS) parts II and III were used to assess self-reported quality of life and motor symptom severity, respectively (Goetz et al., 2007). Cognitive screening was performed using the Montreal Cognitive Assessment (MoCA) to rule out patients with frank dementia (Nasreddine et al., 2005), requiring a score of at least 22. Premorbid intelligence was screened using the American version of the National Adult Reading Test (AMNART), and depression symptoms were screened using the Center for Epidemiologic Studies Depression Scale Revised (CESD-R) (Radloff, 1977). During the clinical assessment, participants also completed the Questionnaire for Impulsive-Compulsive Disorders in Parkinson’s Disease-Rating Scale (QUIP-RS). Patients’ current prescribed dosages of dopaminergic medication, including Levodopa and DAA, were converted to levodopa equivalent daily dose (LEDD) using the conversion factors and formulae reported in (Tomlinson et al., 2010). The DAA medications and doses taken by each subject are listed in the Supplementary material 1.

All participants had normal or corrected-to-normal vision. Patients were excluded if they had (i) history of neurological diseases other than PD, (ii) clinical symptoms of dementia, depression, cerebrovascular, or cardiovascular disease, (iii) an implanted deep brain stimulator, or (iv) implanted hardware that was contraindicated for 3T MRI.

All patients enrolled in the study completed two experimental sessions, each including MRI and Simon task. One session was performed following withdrawal from their dopaminergic medication (Off-DAA), and the other when patients were in their optimal state of dopamine agonist therapy (On-DAA). In the Off-DAA condition, patients refrained from all dopaminergic medications for a total time of 5-half lives of DAA. Practically, this was at least 36 hours for DAA, and 16 hours for Levodopa due to differences in pharmacokinetic properties (Fabbrini et al., 1987; Tompson and Oliver-Willwong, 2009). This period was deemed sufficient to eliminate DAA effects, while minimizing potential patient discomfort. In the On-DAA state, patients were evaluated after taking their prescribed DAA medication, having withheld Levodopa for at least 16 hours. Extended release DAA compounds (taken by 8 patients) were administered 6 hours before MR scanning, whereas non-extended release DAA (taken by 16 patients) were administered 2 hours before scanning. No changes in medication dosages or addition or discontinuation of either drug for clinical purposes were made at any time during study participation. Demographic and clinical features for patients meeting the inclusion criteria are presented in Table 1.

Table 1.

Demographic and clinical data

Sex (male/female)	18/8
Age (years)	61.2 ± 9.6
Disease Duration (years)	5.4 ± 3.4
Montreal Cognitive Assessment (MoCA) [min=0, max=30]	25.6 ± 2.5
American version of the National Adult Reading Test (AMNART)	117.5 ± 9.3
Center for Epidemiologic Studies Depression Scale (CES-D) [min=0, max=60]	17.0 ± 9.7
Questionnaire for Impulsive-Compulsive Disorders in PD-Rating Scale (QUIP-RS) [min=0, max=112]	28.4 ± 13.8
Movement Disorders Society-United Parkinson’s Disease Rating Scale (MDS-UPDRS)
Part II [min=0, max=52]	23.0 ± 9.1
Part III (Off) [min=0, max=56]	30.1 ± 12.2
Part III (On) [min=0, max=56]	21.2 ± 11.1
Dopamine Replacement Therapy
Total Levodopa Equivalent Daily Dose (mg/day)	565.0 ± 360
Agonist Single Dose Equivalent (mg/day)	113.2 ± 64.8

Data are shown as mean ± standard deviation

2.2. Simon task

The Simon task was administrated on a PC as previously described in the literature (see Wylie et al., 2012; van Wouwe et al., 2016 for details on stimuli and description of individual trials). In brief, participants were instructed to manually respond to the color of a circle that would appear left or right of a fixation point in the center of the screen (e.g., green circle=right thumb press; blue circle=left thumb press). To elicit the Simon effect, two trial types were presented to the patients: 1) Corresponding trials (Cs) in which the circle appeared to the side of fixation that matched the response side signaled by the color of the stimulus (e.g., a green circle calling for a right-hand response appeared to the right side of fixation), and 2) Non-corresponding (Nc) trials, in which the circle appeared on the side of fixation opposite the side of the response signaled by the circle’s color (e.g., a green circle calling for a right-hand response appeared on the left side of fixation). Cs and Nc trial types were presented randomly, but with equal probability, within each block of trials. Reaction times (RT) and accuracy were measured for each trial, and the Simon effect was estimated as Nc trials minus Cs trials for RT and error rates.

2.3. Behavioral analysis

RT latencies for Cs and Nc trials faster than 150 ms (anticipatory reactions) and slower than three standard deviations of the mean within each condition were excluded, but accounted for fewer than 1% of trials across participants (Wylie et al., 2010a). For each level of correspondence (i.e. Cs and Nc), mean RT and log transformed accuracy rates were calculated to analyze mean Simon effect on RT and on accuracy (i.e. the difference in RT or accuracy between Cs and Nc trials).

The dual process activation suppression (DPAS) model provided the conceptual and analytical framework to dissociate fast response impulses and inhibition of incorrect responses by means of distributional analyses (Ridderinkhof, 2002; Wylie et al., 2010a, 2010b). Distributional analyses provide more insight into dynamic control processes that are often masked in global RT or interference measures. Distributional analyses were implemented by rank-ordering single-trial RTs from fastest to slowest and by dividing rank-ordered RTs into 6 equal sized bins (each bin contained the same number of trials).

According to the DPAS model, impulse capture can be measured by plotting the accuracy rates against the RT for each level of correspondence (Conditional Accuracy Function, CAF). The proportion of fast errors on non-corresponding trials (Nc) (i.e. within fastest RT bin) reflects the strength of the incorrect response capture (van den Wildenberg et al., 2010).

The DPAS also allows measurement of the proficiency of inhibitory control by means of delta plots, which shows the size of the Simon effect on RT as a function of the RT. This yields a pattern of increasing interference across fast to intermediate response latencies that is followed by a significant reduction (c.f., Luce, 1986) in interference toward the slow end of the RT distribution (Proctor et al., 2011). According to the DPAS model, when an incorrect response has been triggered by non-corresponding stimulus information, it takes time for inhibitory control to build up. Therefore, inhibitory control is most clearly reflected at the slow end of the RT distribution; the slope of the interference reduction between the final two bins of the delta plot (delta-slope) provides the most sensitive metric of the proficiency of inhibitory control over conflicting motor impulses. This has been supported empirically across several studies using both non-clinical and clinical populations (Burle et al., 2002; Ridderinkhof et al., 2005; Wijnen and Ridderinkhof, 2007; Wylie et al., 2007, 2009, 2010b; for a review, see Ridderinkhof et al., 2011).

2.4. MRI acquisition

Patients were scanned using a 3T MRI scanner (Philips Healthcare, Best, The Netherlands). Anatomical MRI scans, including (i) T₁-weighted (MPRAGE; spatial resolution=1×1×1 mm³; TR/TE=8.9/4.6 ms), and (ii) T₂-weighted FLAIR (spatial resolution=1×1×1 mm³; TR/TE=4000/120ms; TI=2800 ms), were obtained to exclude coexisting central nervous system disorders. Pseudo-continuous ASL (pCASL) data were acquired using 2D single-shot echo-planar-imaging (field-of-view = 220×220×119 mm³, slices=20; spatial resolution=3.5×3.5×5 mm³; gap = 1 mm, TR/TE=4000/12 ms with post-labeling delay and labeling pulse train length both set to 1500 ms). The field-of-view for the ASL scan was placed perpendicular to the feeding arteries and covered an area between the inferior border of the pons and the superior sagittal sinus.

2.5. Image analysis

All the analyses were performed on each subject’s native space. Image pre-processing was performed using the FMRIB software library (FSL v5.0.2.1, FMRIB, Oxford, UK). This included brain extraction using BET, and affine time-course motion correction using MCFLIRT (Jenkinson et al., 2002). CBF quantification was performed in Matlab (Mathworks, Natick, MA, USA). Motion-corrected images were pair-wise subtracted using the surround subtraction approach (Lu et al., 2006), and the mean across measurements was computed to obtain a mean difference magnetization (ΔM). The difference magnetization was then normalized by the equilibrium magnetization (M₀), which was calculated by converting the control image magnetization to equilibrium magnetization by dividing by the term [1-exp(−TR/T_1t)], where TR is the repetition time (TR=4000 ms), and the T_1t is the 3T T₁ of gray matter tissue (T_1t=1200 ms). Next, the ΔM/M₀ image was converted to absolute CBF (ml/100g/min) using the simplified kinetic model recommended by the ISMRM perfusion study group (Alsop et al., 2015) and accounting for slice-time correction in the 2D EPI readout (duration of EPI readout per slice = 23 ms).

The regions of interest (ROIs) including the medial pre-frontal cortex (MPFC), dorsolateral pre-frontal cortex (DLPFC), ventrolateral pre-frontal cortex (VLPFC), supplementary motor area proper (SMA-p), pre- supplementary motor area (pre-SMA), primary motor cortex (M1), and anterior cingulate cortex (ACC) were selected due to previous evidence of their importance in motor inhibition (Aron et al., 2003, 2007; Frank, 2006; Alexander et al., 2007; Forstmann et al., 2008a; Duann et al., 2009; Jahfari et al., 2011; van Belle et al., 2014), and segmented on each subject’s T₁-weighted image using FreeSurfer (Version 5.3.0, http://surfer.nmr.mgh.harvard.edu). The cortical parcellation was performed using the Desikan-Killiany atlas. The cortical labels obtained from the automatic cortical parcellation in surface-space (aparc) was mapped to the automatic segmentation volume (aseg) using the function mri_aparc2aseg from FreeSurfer, which finds each aseg voxel labeled as cortex and assign it the label of the closest cortical vertex.

The Brodmann areas atlas was used to parcellate Brodmann area 6 (Fig. 1A), which was used to define the SMA-p and the pre-SMA. We defined the verticofrontal (VCA) line perpendicular to the anterior commissure - posterior commissure (AC-PC) line at the level of the anterior commissure to divide the pre-SMA anteriorly from the SMA-p (Picard and Strick, 1996, 2003; Vergani et al., 2014) (Fig. 1B).

Fig. 1.

Illustration of the location of ROIs on a T1-weighted template. A) Broadman area 6 was obtained using the Brodmann areas atlas, and the VCA line was used to separate the pre-SMA anteriorly from the SMA-p. B) Sagittal, coronal and axial representations showing all ROIs. C) Axial slices at different levels.

The rest of the ROIs were defined using the cortical parcellations obtained with the Desikan-Killiany atlas. Specifically, M1 corresponded to the precentral gyrus; MPFC was created by merging the superior frontal and frontal pole (excluding the pre-SMA and SMA-p); DLPFC by merging the rostral middle frontal and caudal middle frontal (excluding the pre-SMA and SMA-p), VLPFC by merging the pars opercularis, pars triangularis, and pars orbitalis; and ACC by merging the rostral anterior cingulate and caudal anterior cingulate (Fig. 1C). Finally, the CBF maps were co-registered to the T₁-weighted images using FSL’s FLIRT with 6 degrees of freedom, and the median rCBF value (instead of mean) was recorded for each ROI to alleviate the effect of outlying voxels.

Finally, to account for the possible effect of global (whole-brain) CBF, we calculated the CBF for total gray matter for each patient in both of the medication conditions. The total gray matter mask was obtained by merging all the cortical and subcortical structures from FreeSurfer.

2.6. Experimental design and statistical analysis

To assess the effect of DAA on action control, we performed general linear regression model (GLM) analyses using each of the cognitive measures (RT, Simon effect on RT, Simon effect on accuracy, response capture, and delta-slope) as dependent variables, and medication status as independent variable. To ensure that changes in the cognitive performance were not a result of potential confounding factors, we considered age, UPDRS-II, UPDRS-III (Off) and MoCA, and disease duration as possible covariates, i.e., inter-subject variation in those confounding factors. To avoid overfitting, we performed a principal component analysis (PCA) including age, UPDRS-II, UPDRS-III (Off), MoCA, and disease duration, and used the first principal component (PC1), which explained 40% of the variance of the data, as covariate in the GLMs (model 1).

To test the hypothesis that DAA medication has an effect on rCBF in frontal areas, we performed GLM analyses specifying within-ROI rCBF as the dependent variable, and medication condition as the independent variable. To ensure that differences in rCBF were not a result of potential confounding factors, we also included PC1 (as described above) as covariate in the GLMs (model 2). A separate GLM analysis was performed for each of the seven ROIs.

To evaluate if rCBF is related to action control, we performed GLM analyses with each of the Simon task measures as dependent variables, ROI rCBF as independent variable, and PC1 as covariate. These analyses were performed separately for every ROI in each medication condition (model 3).

All analyses were performed using R version 3.5.1 (R Foundation for Statistical Computing, Vienna, 2016). The results were considered significant at the level of false discovery rate (FDR) of 0.10 in concordance with the threshold recommended in the first description of the method (Benjamini and Hochberg, 1995), which has been used in previous neuroimaging studies (Petersen et al., 2018; Stark et al., 2018a, 2018b). The FDR corrections were applied separately to the results from model 1 to account for the multiple cognitive measures, and to the results from models 2 and 3 to account for the multiple ROIs evaluated in the study. In each GLM analysis, we also reported 95% confidence interval (CI) associated with each regression coefficient of interest, using the t-statistic of the coefficient.

3. RESULTS

3.1. Action control performance

Table 2 summarizes the results for the effect of medication of the cognitive measures. The RT did not differ between medication states. The Simon effect on RT but not on accuracy rates varied by medication state. In the Off-DAA condition, patients showed a Simon effect that was 17 ms larger than in the On-DAA state (P=0.005). The analysis on the response capture rates (accuracy rates in the fastest RT bin of the non-corresponding trials) confirmed that fast impulsive errors on Nc trials were similar across medication states (Fig. 2A). The proficiency of inhibitory control (i.e. delta-slope between the last two bins of the RT distribution) showed a trend for improvement in the On-DAA condition (i.e. more negative slope On- versus Off-DAA), but this finding was not statistically significant (P=0.19) (Fig. 2B, Table 2).

Table 2.

DAA medication effect on action control (Model 1: cognitive measures ~ medication status + PC1)

Cognitive Measure	t-statistic (DF=25)	95 % CI	coefficient	P-valuea
Overall RT across trials	1.103	[−16.180, 53.513]	18.666	0.280
RT for Cs trials	1.454	[−11.339, 65.797]	27.228	0.158
RT for Nc trials	0.655	[−21.641, 41.844]	10.101	0.518
Simon Effect (RT)	−3.056	[−28.670, −5.585]	−17.127	0.005b
Simon Effect (accuracy)	−0.614	[−0.018, 0.010]	−0.004	0.544
Response capture	−0.124	[−0.591, 0.524]	−0.034	0.902
Delta-slope	−1.354	[−0.271, 0.056]	−0.108	0.188

^aUncorrected p-value

^bSignificant p-value at FDR = 0.1

DF = Degrees of freedom

Negative coefficients represent better performance on action control in the On-DAA condition.

Fig. 2.

A) Conditional Accuracy Function (CAF) for corresponding (Cs) and non-corresponding (Nc) trials. B) Reaction times (RT) delta plots.

3.2. CBF and DAA medication status

Fig. 3 shows the mean CBF maps for the Off- and On-DAA conditions. When examining the effect of DAA medication status on total gray matter CBF, we did not find significant differences between the Off-DAA (42.53 ± 9.27) and On-DAA (44.68 ± 6.51) states (P=0.12). We observed increases in rCBF in the On-DAA condition. In particular, we found significantly increased rCBF values in the DLPFC (P=0.011) in response to DAA administration. The rCBF in the pre-SMA also showed an increase in the On-DAA condition but did not survive the FDR correction (P=0.041). Table 3 summarizes these results.

Fig. 3.

A) Orthogonal representation of a T₁-weighted anatomical atlas, along with mean quantitative CBF maps (ml/100 g/min) across subjects in the (B) Off-DAA and (C) On-DAA states. Increases in rCBF in several regions, including the DLPFC (indicated by the cursor) and pre-SMA, were observed in the On-DAA condition.

Table 3.

DAA medication effect on rCBF (Model 2: rCBF ~ medication status + PC1)

ROI	t-statistic (DF=25)	95 % CI	coefficient	P-valuea
Total gray matter	1.627	[−0.572, 4.872]	2.150	0.116
ACC	0.516	[−3.772, 6.291]	1.260	0.611
MPFC	1.714	[−0.539, 5.879]	2.670	0.099
DLPFC	2.742	[1.007, 7.087]	4.047	0.011b
VLPFC	2.017	[−0.061, 5.806]	2.872	0.055
SMA-p	1.217	[−2.018, 7.842]	2.912	0.235
Pre-SMA	2.155	[0.165, 7.252]	3.709	0.041
M1	1.155	[−1.498, 5.320]	1.911	0.259

^aUncorrected P-value

^bSignificant P-value at FDR = 0.1

DF = Degrees of freedom

Positive coefficients represent higher rCBF in the On-DAA condition.

3.3. CBF and action control

In the On-DAA condition, the delta-slope showed a significant association with the rCBF in the DLPFC (P=0.038), pre-SMA (P=0.006) and M1 (P=0.015) (Table 4), with the strongest relationship in the pre-SMA (Fig. 4). Conversely, no significant relationship was observed between the rCBF and the Simon effect on RT, Simon effect on accuracy, or the response capture in any of the ROIs, regardless of medication state (Table 4).

Table 4.

Relationship between rCBF and action control (Model 3: action control ~ rCBF + PC1)

Off-DAA				On-DAA
	t(DF=23) [95 % CI]	Coeff.	P-valuea	t(DF=23) [95 % CI]	Coeff.	P-valuea
Simon Effect on RT
ACC	0.954 [−0.551, 1.496]	0.472	0.3497	−0.441 [−1.322, 0.858]	−0.232	0.663
MPFC	0.235 [−1.185, 1.489]	0.152	0.816	−0.221 [−1.341, 1.082]	−0.129	0.827
DLPFC	−0.209 [−1.209, 0.987]	−0.111	0.837	0.300 [−0.910, 1.219]	0.154	0.767
VLPFC	0.208 [−1.176, 1.438]	0.131	0.837	0.135 [−1.248, 1.422]	0.087	0.894
SMA-p	−0.721 [−1.384, 0.668]	−0.358	0.478	−0.239 [−1.220, 0.917]	−0.151	0.772
Pre-SMA	−0.869 [−1.495, 0.611]	−0.442	0.394	−0.604 [−1.502, 0.823]	−0.339	0.552
M1	−0.585 [−1.432, 0.800]	−0.316	0.564	−0.100 [−1.142, 1.037]	−0.052	0.922
Simon Effect on accuracy
ACC	−0.458 [−0.002, 0.001]	−4e-04	0.652	1.054 [−0.001, 0.003]	8e-04	0.303
MPFC	−0.488 [−0.003, 0.002]	−5e-04	0.630	1.149 [−0.001, 0.003]	1e-03	0.262
DLPFC	−0.937 [−0.002, 0.001]	−8e-04	0.359	−0.020 [−0.002, 0.002]	1e-08	0.984
VLPFC	−0.550 [−0.003, 0.002]	−5e-04	0.588	0.542 [−0.001, 0.003]	5e-04	0.593
SMA-p	−0.775 [−0.002, 0.001]	−6e-04	0.446	0.266 [−0.001, 0.003]	2e-04	0.792
Pre-SMA	0.675 [−0.001, 0.002]	5e-04	0.506	1.165 [−0.001, 0.003]	1e-03	0.256
M1	−0.622 [−0.002, 0.001]	−5e-04	0.540	1.161 [−0.001, 0.003]	9e-04	0.258
Response capture
ACC	−0.885 [−0.080, 0.032]	−0.024	0.385	1.542 [−0.019, 0.127]	0.054	0.137
MPFC	−1.080 [−0.108, 0.034]	−0.037	0.291	1.583 [−0.019, 0.142]	0.062	0.127
DLPFC	−0.885 [−0.084, 0.034]	−0.025	0.385	0.775 [−0.046, 0.101]	0.028	0.446
VLPFC	−1.357 [−0.113, 0.024]	−0.045	0.188	0.906 [−0.052, 0.132]	0.040	0.374
SMA-p	0.156 [−0.052, 0.061]	0.004	0.877	1.251 [−0.029, 0.116]	0.044	0.224
Pre-SMA	0.175 [−0.053, 0.063]	0.005	0.863	1.868 [0.007, 0.146]	0.069	0.075
M1	−0.355 [−0.071, 0.050]	−0.010	0.726	1.217 [−0.030, 0.117]	0.044	0.236
Delta-slope
ACC	−1.473 [−0.021, 0.004]	−0.009	0.154	−1.400 [−0.024, 0.005]	−0.010	0.175
MPFC	−1.470 [−0.027, 0.005]	−0.011	0.155	−1.835 [−0.029, 0.002]	−0.014	0.080
DLPFC	−1.961 [−0.024, 0.001]	−0.012	0.062	−2.158 [−0.027 −0.001]	−0.014	0.042b
VLPFC	−1.362 [−0.026, 0.005]	−0.010	0.187	−1.881 [−0.032, 0.002]	−0.015	0.073
SMA-p	−1.652 [−0.022, 0.003]	−0.009	0.112	−1.794 [−0.025, 0.002]	−0.012	0.086
Pre-SMA	−1.925 [−0.024, 0.001]	−0.011	0.067	−2.980 [−0.033, −0.006]	−0.019	0.007b
M1	−2.064 [−0.025, 0.000]	−0.013	0.050	−2.678 [−0.029, −0.004]	−0.017	0.013b

^aUncorrected p-value

^bSignificant p-value at FDR = 0.1

t(DF) = t-statistic(degrees-of-freedom); Coeff=Coefficient

Fig. 4.

Scatter plot and linear regression showing the relationship between the delta-slope and the rCBF in the pre-SMA in the On-DAA state.

4. DISCUSSION

We evaluated the effect of DAA medication on rCBF in frontal cortical areas and examined how action control in PD is informed by rCBF. We found that single-dose DAA administration increases rCBF in the DLPFC and pre-SMA, supporting the hypothesis that DAA administration results in regional brain activity changes to cortical regions involved in motor inhibition. Our results emphasize that greater rCBF in the pre-SMA, DLPFC and M1 reflect better proficiency of inhibitory control. These findings extend our understanding of the cortical localization of dopamine effects on motor inhibition and the source of individual differences in action control.

4.1. Effects of DAA medication on action control

In this study, we find similar motor speed between medication states. However, based on the DPAS model and previous findings of medication on inhibitory control in PD (van Wouwe et al., 2016), we did not expect that medication effects would be purely additive across the RT distribution. Instead, the DPAS model predicts that suppression of interfering information slowly builds up, and becomes most evident at the slow end of the RT distribution where the Simon effect reduces. Based on our previous work that showed that dopamine therapy significantly improves reactive inhibitory control processes engaged to suppress interference from the spontaneously activated impulses (Wylie et al., 2012; van Wouwe et al., 2016), we predicted that this suppression would be more efficient On- compared to Off-DAA, and would be reflected in a sharper decline of the delta slope function in the On-DAA condition (Fig. 3b). Here, DAA significantly improved the Simon effect, and while the GLM analysis did not show significant differences in delta-slope between Off- and On-DAA, the direction of the delta slope is more negative for On-DAA, which is consistent with previous work, and indicates that DAA therapy improves reactive inhibitory control.

4.2. Effects of DAA medication on rCBF

We find that a single-dose of DAA results in rCBF increases in the DLPFC and pre-SMA. Under the common suppression that changes in perfusion are coupled to changes in glucose metabolism and associated neuronal activity, we interpret these results to indicate that DAA administration induces increased neural activity in key cortical regions that subserve motor inhibition and action control. This network effect is likely mediated via reciprocal connections to basal ganglia and thalamic networks. In healthy neurovascular coupling, CBF measured using ASL is a surrogate marker of glucose metabolism and associated neuronal activity (Buxton, 2005; Wolk and Detre, 2012). Our interpretation, that DAA modifies regional CBF, depends on the coupling between rCBF, metabolism, and neuronal activity. Previous animal studies indicate that this coupling is preserved in the presence of non-ergot DAA (McCulloch and Edvisson, 1980; McCulloch et al., 1982). As a supplementary analysis, we evaluated whether global CBF (rather than regional), varied with medication status. In this case, we did not observe a significant DAA effect on global CBF, suggesting that the DAA induced changes in rCBF reflect regional effects, rather than a global hemodynamic change. This supplementary analysis also supports the observation that CBF findings are not being driven by technical variations due to image artifacts, for instance due to variations in labeling efficiency in the pCASL sequence from scan to scan that would manifest as a globally altered CBF measurement.

Previous PET and MRI studies investigating the effects of levodopa, the most commonly used dopamine precursor therapy for PD, indicate that treatment alters rCBF in several regions: the striatum and thalamus (Kobari et al., 1995), nigrostriatal pathway, occipital cortex, and inferior parietal areas (Chen et al., 2015). While both levodopa and DAA augment dopaminergic tone in PD and both appear to increase rCBF in areas associated with PD symptoms, despite the differing mechanisms of action. Pharmacological neuroimaging studies in non-human primates with levodopa (Hershey et al., 2000) and non-ergot DAAs (Black et al., 1997, 2002, 2011) indicate that different dopaminergic agents have unique rCBF responses, suggesting that rCBF changes are mediated by the action as a dopamine precursor (levodopa) or receptor agonist (DAA). Moreover, studies in healthy volunteers investigating the relationship between CBF and activity changes in specific neurotransmitter systems (Donahue et al., 2014; Dukart et al., 2018) showed that CBF reflects specific metabolic demands from diverse underlying neurotransmitter systems, supporting the notion that different pharmacological agents provide unique patters of CBF changes associated to receptor availability, affinity and function. In particular, the CBF changes induced by dopaminergic medications showed a significant positive association with the underlying D₁ and D₂ receptor densities. These findings are consistent with previous research reporting increased CBF after administration of both dopamine agonist and antagonist (Mu et al., 2007; Fernández-Seara et al., 2011; Handley et al., 2013; Schouw et al., 2013).

In this study, patients were given non-ergoline DAAs (ropinirole, pramipexole, rotigotine) which exert their effects by acting directly on dopamine receptors. These DAAs primarily target D₂-like receptors, particularly D₃ (Mierau et al., 1995; Piercey, 1998; Millan et al., 2002; Newman-Tancredi et al., 2002; Scheller et al., 2009; Wood et al., 2015), which are localized in greater density to the mesolimbic and mesocortical regions (Sokoloff et al., 1990, 1992; Schwartz et al., 1993; Murray et al., 1994). Previous studies suggest that D₃ receptors can influence cognitive function via their inhibitory effect on mesocortical dopaminergic activity (Cole et al., 2012; Gross et al., 2013; Nakajima et al., 2013). We find that DAA produces rCBF increases in the DLPFC and pre-SMA which are important for top-down control of subcortical brain regions that implement selective stopping.

4.3. rCBF and motor inhibition

In the On-DAA condition, higher rCBF in several frontal cortical areas, including the pre-SMA, DLPFC and M1, was associated with better inhibitory control over conflicting action impulses (as measured by the negative slope of the delta plot). These cortical areas play a critical role in selecting task-relevant information during conflict situations (Ullsperger and von Cramon, 2001; Peterson et al., 2002; Mars et al., 2009; Neubert and Klein, 2010; Duque et al., 2012; Soutschek et al., 2013). Previous studies of neural activation patterns associated with response interference trials on conflict tasks indicate the involvement of fronto-parietal and fronto-striatal networks (Ridderinkhof et al., 2004; Nee et al., 2007). The resolution of conflict situations involving two competitive actions, an incorrect prepotent action versus a desired action, may come from diverse sources, including inhibition of motor areas and increased activation of brain areas involved in action selection. When stopping an already-started action, the pre-SMA works together with the right IFC to send a stop command to intercept prepotent response tendencies via the basal-ganglia networks, resulting in the suppression of basal-ganglia output with global inhibitory effects on M1 (Aron and Poldrack, 2006; Badry et al., 2009); when control is needed, the pre-SMA may act in concert with the right IFC to implement stopping.

This is supported by neuroimaging studies that have shown that the network composed of the right IFC, pre-SMA, and STN (i.e. hyperdirect pathway) plays a crucial role in reactive stopping (Aron and Poldrack, 2006; Aron et al., 2007), that individual differences in inhibition are accompanied by differences in brain function in those areas (Forstmann et al., 2008b), and that stimulation of the pre-SMA influences frontal rCBF and has a significant impact on response inhibition (Obeso et al., 2013). The strong connections between the pre-SMA, right IFC and STN (Johansen-Berg et al., 2004) lead to global inhibitory effects of the motor system (Aron et al., 2007; Stinear et al., 2009; Swann et al., 2009). Additionally, this network for reactive stopping can also be prepared in advance, and thus, action control can be proactive (Jahfari et al., 2010).

In this study, we showed a significant association between rCBF in the pre-SMA, DLPFC, and M1 and the ability to suppress irrelevant action impulses, suggesting that higher rCBF in these areas could reflect individual differences in the ability to implement control in conflict situations. Our results are consistent with previous studies (Chikazoe et al., 2009; Jahfari et al., 2010), and indicate that if the stopping networks are preactivated, stopping is more proficient when required. Interestingly, we did not observe an association between rCBF and the mean Simon effect nor with response capture (as reflected by accuracy rates in the fast bin of the RT distribution), which could indicate that individual differences in rCBF may be specifically related to selective response inhibition. The process of selective stopping could be prepared in advance via a top-down influence of the DLPFC over the striatum (Vink et al., 2005; Chikazoe et al., 2009; Zandbelt and Vink, 2010), which takes time to build-up and allows for more selective motor control by targeting particular representations in M1 (Aron and Verbruggen, 2008; Claffey et al., 2010). DLPFC activity seems to be associated with goal-directed activation, resulting in advance preparation of task-relevant areas with subsequent performance benefits (Wylie et al., 2006; Yeung et al., 2006).

Overall, the present results corroborate and extend earlier findings: the role of frontal cortical areas, particularly the pre-SMA and DLPFC, in the implementation of selective response inhibition. We found that individual differences in rCBF in the pre-SMA and DLPFC significantly correlate with performance on the proficiency of selective inhibitory control over conflicting motor impulses. When multiple competing actions are possible, the demands on action control are highest, and selecting the correct action may require stronger activation of the pre-SMA and DLPFC (Ridderinkhof et al., 2011; Carbonnell et al., 2013). Finally, we also observed a trend for a correlation between rCBF in the pre-SMA, DLPFC and M1 and delta-slope for the Off-DAA condition, although it was weaker compared to the On-DAA state. This could mean that, even in the absence of a DAA effect, higher rCBF in the pre-SMA is associated with better predisposition to implement action control when needed. However, we hypothesize that DAA administration likely restores a deficient dopamine tone, where increases in rCBF in the pre-SMA and DLPFC reflect stronger frontal-cortical function and more proficient action control (more negative slope).

4.4. Limitations and future directions

In the On-DAA condition we examined patients in their optimal DAA dosage; however, they were still withdrawn from levodopa, and thus, they were not in their optimal ‘On’ state. This design was purposeful as we wished to study agonist-only effects on rCBF. Moreover, in the On-DAA condition, we examined patients after the administration of their routine prescribed DAA, and we did not standardize the acute dose across patients. This was a convenient method, which allowed for us to assess a range of medication doses on behavioral and imaging endpoints. Future studies should take this into consideration in the study design to reduce variability and replicate these results.

Although all imaging and behavioral assessments were performed after dopaminergic washout, we did not address chronic effects of dopamine medication to rCBF and performance in inhibitory control. Future pharmacological neuroimaging and cognitive studies may benefit from study designs that evaluate neuronal function and cognitive performance under the selective influence of dopamine-modifying medications and different time courses (acute vs chronic), and ideally, DAA naive PD patients could be imaged at baseline, and with acute DAA treatment and then followed longitudinally. Additional studies assessing acute DAA response in healthy controls may be useful. Future studies should also investigate rCBF changes during task performance, which would provide a better understanding on how rCBF varies with demands in control.

5. CONCLUSIONS

Our results suggest that DAA medication in PD patients induces rCBF changes in key areas involved in motor inhibition. Increased rCBF in the pre-SMA and DLPFC is linked to improved proficiency of selective inhibitory control over conflicting motor impulses, suggesting that targeted therapies designed to improve frontal cortical function may promote action control.

ABBREVIATIONS

ACC

Anterior cingulate cortex

AMNART

American version of the National Adult Reading Test (AMNART)

ASL

Arterial spin labeling

rCBF

Regional Cerebral blood flow

CESD-R

Center for Epidemiologic Studies Depression Scale Revised

DAA

Dopamine agonist

DLPFC

Dorsolateral prefrontal cortex

IFC

Inferior frontal cortex

M1

Primary motor cortex

MDS-UPDRS

Movement Disorders Society-United Parkinson’s Disease Rating Scale

MoCA

Montreal Cognitive Assessment (MoCA)

MPFC

Medial prefrontal cortex

MRI

Magnetic resonance imaging

PD

Parkinson’s disease

PET

Positron emission tomography

Pre-SMA

Pre-supplementary motor area

QUIP-RS

Questionnaire for Impulsive-Compulsive Disorders in Parkinson’s Disease-Rating Scale

RT

Reaction times

ROI

Region of interest

SMA-p

Supplementary motor area proper

STN

Subthalamic nucleus

VLPFC

Ventrolateral prefrontal cortex