Thalamic GABA may modulate cognitive control in restless legs syndrome

Neuroscience Letters 712 (2019) 134494

Contents lists available at ScienceDirect

Neuroscience Letters journal homepage: www.elsevier.com/locate/neulet

Research article

Thalamic GABA may modulate cognitive control in restless legs syndrome a

b

c

c,d

Rui Zhang , Annett Werner , Wiebke Hermann , Moritz D. Brandt , Christian Beste ⁎ Ann-Kathrin Stocka, ,1

a,1

T

,

a

Cognitive Neurophysiology, Department of Child and Adolescent Psychiatry, Technische Universität Dresden, Schubertstr. 42, 01307 Dresden, Germany Institute and Clinic for Diagnostic and Interventional Neuroradiology, Technische Universität Dresden, Fetscherstraße 74, 01307 Dresden, Germany Department of Neurology, Technische Universität Dresden, Fetscherstraße 74, 01307 Dresden, Germany d German Center for Neurodegenerative Diseases (DZNE) Dresden, Arnoldstraße 18, 01307 Dresden, Germany b c

A R T I C LE I N FO

A B S T R A C T

Keywords: Thalamic GABA Restless legs syndrome MRS Cognitive control Working memory Implicit learning

The restless legs syndrome (RLS) has repeatedly, but not exclusively, been associated with functional thalamic changes as well as changes in GABAergic neurotransmission. This has been linked to the well-known sensorymotor symptoms, but it has never been investigated whether those factors also account for potential cognitive changes in RLS, even though they are known to play an important role for cognitive control. To investigate the potential relationship between thalamic GABA concentrations and cognitive control in n = 25 RLS patients; a neuropsychological experimental paradigm was used in combination with magnetic resonance spectroscopy (MRS). Compared to n = 31 healthy controls, RLS patients displayed reduced cognitive control capacities, which were most likely based on working memory deficits. On the neurobiochemical level, (relatively) elevated thalamic GABA levels attenuated control deficits only in the RLS group, even though there were no group differences with respect to overall GABA levels. Given that RLS patients are known to display thalamic hyperactivity and associated thalamic hypoconnectivity, (relatively) higher GABA levels may have helped RLS patients to “compensate” for this pathological factor. Our findings specify the functional relevance of thalamic GABAergic neurotransmission for cognition in RLS, even though changes in GABAergic neurotransmission might not be the ultimate cause of control deficits.

1. Introduction The restless legs syndrome (RLS) is a sensory-motor disorder, which is characterized by an urge to move the legs, and sometimes the arms, during rest predominantly in the evening or at night [1]. RLS-related neuronal changes are most commonly found in cortical-basal gangliathalamic loops [2]. Typically, the involved brain regions do not show pronounced lesions or atrophy [3–5], but changes in dopaminergic, glutamatergic and GABAergic signaling [6–8]. The functional relevance of these neurotransmitter changes is underlined by the effectivity of both L-dopa and anticonvulsants in RLS treatment [6]. In addition to being implied in RLS pathophysiology [7–9], cortical-basal gangliathalamic loops and the transmitters dopamine, glutamate and GABA are known to be crucial for processes such as cognitive control [10–12]. In line with this, RLS patients have been shown to experience changes in cognitive control functions such as decision making or working memory [13,14].

The thalamus processes sensory, motor and cognitive information and substantially contributes to cognitive processes including cognitive control [15–18]. There are consistent findings of thalamic abnormalities in RLS [8,19–22]. Reduced functional thalamic connectivity with cortical areas [20], that have been suggested to be involved in cognitive control functions like working memory [23], has been reported in RLS. Furthermore, RLS patients have repeatedly been demonstrated to experience excitatory thalamic hyperactivity, as evidenced by higher functional magnetic resonance imaging blood-oxygen-level-dependent (fMRI BOLD) responses, increased grey matter density, or heightened glutamatergic activity, which may correlate with RLS motor symptoms [2,7,24]. Given that experimental electric thalamic stimulation in patients with refractory epilepsy treated with deep brain stimulation has been shown to cause more errors in executive control tasks [25], the executive control deficits observed in RLS may be linked to thalamic hyperactivity. Yet, accumulating evidence shows that the GABA system may play a

⁎

Corresponding author at: Cognitive Neurophysiology, Department of Child and Adolescent Psychiatry, Faculty of Medicine of the TU Dresden, Schubertstraße 42, D-01307 Dresden, Germany. E-mail address: [email protected] (A.-K. Stock). 1 These two authors contributed equally. https://doi.org/10.1016/j.neulet.2019.134494 Received 26 April 2019; Received in revised form 3 September 2019; Accepted 10 September 2019 Available online 11 September 2019 0304-3940/ © 2019 Elsevier B.V. All rights reserved.

Neuroscience Letters 712 (2019) 134494

R. Zhang, et al.

University hospital, TU Dresden, Germany. All patients had received a secured RLS diagnosis from an experienced neurologist based on the International Restless Legs Syndrome Study Group (IRLSSG) diagnostic criteria [1]. Furthermore, they had no signs of other neurological or psychiatric diseases (except for mild RLS-associated depressive symptoms in some of the patients), as secured by regular neurological examinations as well as an interview with an experienced neuropsychologist. Patients who reported ever having experienced RLS symptoms before noon, which indicates treatment-related augmentation or severe RLS, were not included. For comparison, we recruited n = 31 healthy control participants, who had been individually matched for sex, age ( ± 3 years), years of education ( ± 2), and further showed no signs of any neurological or psychiatric disorders (and were therefore not taking any dopaminergic or other stimulant medication). To avoid the potential artifacts induced by RLS symptoms during MRS measurement, patients and controls were scanned and subsequently tested in the morning, starting at 7 a.m., when they were free of symptoms. All participants were asked to remain awake during this procedure. Five patients had to be excluded due to data quality problems. Another patient was excluded due to poor behavioral performances (i.e. accuracy below 70% in the low demand task). The data of 25 patients (age: 60.16 ± 2.21, 18 females) and 31 controls (age: 60.25 ± 1.87, 23 females) were analyzed. All patients with dopaminergic treatment were asked to abstain from their medication prior to the appointment (please see S1 for more details). Furthermore, all participants were asked to refrain from using caffeine or any other freely available, non-pharmaceutical stimulating agents (like energy drinks) on the day of testing. All experimental protocols in the study were approved by the institutional review board of the Medical faculty of the TU Dresden in Germany (EK 27012014). All performed procedures were in accordance with the Declaration of Helsinki. Written informed consent was obtained from each participant prior to the beginning of the study. All participants received a reimbursement of 60 € for taking part in the study.

critical role in RLS. Even though no absolute thalamic GABA differences have been found in RLS patients, thalamic GABA levels seem to be correlated with the severity of RLS sensory-motor symptoms [8]. It has furthermore been proposed that thalamic GABA exerts its effect by modulating the thalamo-cortical communication [26,27]. Through multiple connections between thalamus and different parts of cortex [12], thalamic GABA does not only contribute to sensory-motor functions [8,28] but also to various cognitive functions including executive functions and working memory [27,29,30]. Hence, the thalamic GABAergic system seems to be relevant for understanding executive functions in RLS. As cognitive control is key to leading a self-serving and successful life [31,32], we set out to compare the cognitive control performances of RLS patients to that of healthy controls. For this, we investgated the potential correlation between performance and thalamic GABA levels in both groups, as assessed with the help of MR spectroscopy (MRS). We applied an experimental paradigm, which was developed by Bocenegra and Hommel [33] and measures cognitive control by comparing the behavioral performance of a high demand task and a low demand task. Due to differences in the complexity of task rules, more cognitive control and effort are required in the high demand task than in the low demand task. By comparing a participant’s performance in the two tasks, it can be investigated how well the participant coped with the increase in task complexity / control demands: The larger the performance decrease in the high demand task relative to the low demand task, the lower the participant’s cognitive control capacities. Given that RLS patients have been reported to display impairments in working memory and cognitive control [14,34–38] they should show larger task differences than healthy controls. In addition to this, the paradigm allows to investigate implicit learning and the effect of cognitive control on this process by means of an implicit feature (for details, please see methods section) [33]. This is of interest because thalamic lesions have been shown to affect both cognitive top-down processes and implicit motor learning [39], but the latter has never been investigated in RLS. In the task we used, healthy controls are usually able to benefit from the implicit feature and improve their performance when they have “free” residual cognitive processing capacities (i.e. in the low demand task), but not when their control capacities are strained (i.e. in the high demand task). Given the reported impairments in working memory, cognitive control [14,34–38], and implicit motor learning [39], RLS patients should benefit less or even not at all from the implicit feature. Since absolute thalamic GABA concentrations do not seem to differ between RLS patients and healthy controls [8], we expected that thalamic GABA may not linearly represent the potential cognitive deficits in RLS patients. But given that anticonvulsive medication, which mimics the effects of GABA, improves RLS-associated sensory-motor symptoms [6] and further given that thalamic GABA has been shown to potentially improve cognitive control performance [40], we hypothesized that higher GABA levels will lead to better cognitive control performance (i.e. reflected by smaller task differences) in both healthy controls and RLS patients. Furthermore, higher GABA levels may also facilitate the integration of implicit information. Considering that thalamo-cortical connectivity is reduced in RLS patients [20] and thalamic GABA is proposed to enhance thalamo-cortical communication [26,27] thalamic GABA levels might make larger contributions to task performances of RLS patients than healthy controls.

2.2. MRS data acquisition and processing Three tesla MRI scans were conducted at the beginning of the experiment. For optimal voxel positioning a 3D sagittal T1-weighted image was acquired first with repetition time (TR) = 2470, echo time (TE) = 4.4, inversion time (TI) = 1100 ms, flip angle = 15°, field of view = 256 × 256 mm, and 160 slices (slice thickness = 1.0 mm). To improve spectral resolution an additional Head_shim_3D shimming routine (WIP Siemens) over 64 slices (field-of-view: 384 mm, slice thickness 4 mm) was performed, followed by interactive manual local shimming. In vivo single-voxel localized 1H MR spectra were obtained, with an adapted point resolved spectroscopy (PRESS) sequence (TE = 97 ms) with two echo times of 17 ms and 80 ms according to Choi et al. [41]. PRESS acquisition parameters included TR = 2000 ms, acquisition bandwidth =2.5 kHz, and 1024 sampling points. The voxel size of MRS volumes of interest (VOIs) was 30 × 30 × 30 mm3 located on the left thalamus (see Fig. 1) and the VOI placement was checked by two experienced neuroscientists. In this context, it should be noted that a voxel of this size will always contain fractions of other brain areas that do not belong to the thalamus, but given that this volume is required for being able to properly quantify GABA within the spectrum [42], it is currently no feasible option to use a substantially smaller VOI. We further chose to only record from the left thalamus as we found no literature suggesting any asymmetry in the transmitter content of the thalamus and wanted to keep the scanning time as short as possible for the patients, who often felt uncomfortable lying in the scanner. Moreover, an unsuppressed water signal was acquired from each voxel for residual eddy current compensation using TE/TR: 30 ms/ 17 s from the same voxel. The acquired data (free induction decay/FIDs were recorded in 16 blocks) were combined off-line, using a special designed Matlab script (The MathWorks Inc., Natick, MA, USA) which was developed according to a published algorithm from Choi et al. [43].

2. Material and methods 2.1. Patients and controls N = 31 adult RLS patients with a stable medication treatment (patients with GABAergic treatment, e.g. pregabalin, gabapentin or benzodiazepine, were strictly excluded) for at least 4 weeks were recruited from the outpatient clinic of the Department of Neurology of the 2

Neuroscience Letters 712 (2019) 134494

R. Zhang, et al.

Fig. 1. Representative example of placement of the 3 × 3 × 3 cm-sized VOI in the left thalamus (on the transversal, coronal and sagittal slice, see top of the graph). Please note that while the size of the VOI does indeed exceed the borders of the thalamus, we decided not to use a smaller VOI, as this would have severely impaired the signal to noise-ratio of the obtained data (see methods section for details).

Spectral fitting was performed with LCModel software, version 6.3.1 [44], using numerically-calculated basis spectra of 20 metabolites, based on the characteristics of the sequence (echo times, pulse shapes) [41]. This adjusts for phase adjustments, frequency alignment, baseline subtraction, and eddy current correction. Before exporting the “final” raw data into LC-Model, the aforementioned MATLAB script allows for user interaction, namely to control the 17 single raw data sets (16 water suppressed / one non water suppressed) with respect to frequency shifts, signal to noise ratio and finally to eliminate single data sets in case of low quality. In this context, the linewidth (full width at half maximum / FWHM) of the unsuppressed water signal is an important criterion for shim quality. The combination of the remaining data sets yields a new. rda file, which is analyzed with LC-Model, where errors are given as Cramer Rao lower bounds. Cramer-Rao lower bounds (CRLB) as a measure of precision were returned as a percentage standard deviation (SD) by LCModel. Results with a CRLB value > 30% were excluded (Fig. 2). Metabolite concentrations from the LC-Model estimates were calculated with respect to the short-echo-time water signal using Eq. (1) [43]

( ) ( )

task. Based on the complexity of task rules, the tasks differed with respect to the required cognitive control: In the high demand task, participants had to respond to the combination of size and color: When the target was either large and red or small and green, they had to press the left button. When the target was either large and green or small and red, they had to press the right button. For the low demand task, participants had to respond to the shape only and respond on the left side for squares, and on the right side for diamonds. Each trial started with the presentation of target stimulus which remained on the screen until the first response. If no response was given within 2000 ms, stimulus presentation was stopped and the trial was coded as a “miss”. Following 700 ms after stimulus offset, a feedback sign (“+” or “-”) was presented for 500 ms. Next, a fixation cross was presented for 500 ms before the next trial presentation. All participants performed the high demand task before the low demand task to keep the task order constant across the participants (compare Bocanegra and Hommel [33]). Unbeknownst to the participants, there were also two conditions in each task: In the baseline condition (50% of trials), the target stimulus was presented in the center of the screen, while the target was slightly shifted to the top or bottom of the screen (approx. 3 cm/ 3° visual angle) in the predictive condition (50% of trials), thus constituting an implicit feature. Importantly, all stimuli requiring a left response were shifted upwards in the predictive condition, whereas stimuli requiring a right response were shifted downwards. Participants were not informed about this and did not report any explicit knowledge about the predictive feature of the stimuli at the end of the experiment. Contrasting the predictive and baseline condition allows to investigate the effect of cognitive control on the automatic exploitation of implicit stimulus feature and response bindings. Accuracy and mean response times of correct responses were collected for each combination of “task” (high vs. low demand) and “condition” (predictive vs. baseline).

TEdiff

exp T S 2 Cm = Cm ⎛ m ⎞ S ⎝ w ⎠ 1 − exp TR T ⎜

⎟

1

(1)

Sm: LC-Model estimates Sw: water signal from non suppressed SVS_se_30 file Cw: water concentration (42.3 M) Relaxation effects on metabolite signals were corrected using published metabolite T2 and T1 values (T2 = 150 ms, 230 ms and 280 ms for creatine, choline and N-acetylaspartate and 180 ms for other metabolites; T1 = 1200 ms for glutamate, glutamine and myo-inositol and 1500 ms for other metabolites) [45–47]. The observed GABA values were then integrated with the behavioral data by means of regression analyses (see statistics section for details).

2.4. Statistical analysis Independent t-tests were used for exploratory comparisons between patients and controls in several psychometric and neuropsychological scores, as well as thalamic GABA concentrations (see Table 1). Behavioral data (accuracy and hit RTs) were separately analyzed using mixed effects analyses of variance (ANOVA) comprising the within-subject factors “task” (low vs. high demand) and “condition” (predictive vs. baseline). “Group” (patients vs. controls) was used as between-subjects factor. Greenhouse–Geisser correction was applied whenever necessary. The behavioral analyses were run in order to test the hypotheses that

2.3. Experimental paradigm We applied an experimental paradigm by Bocanegra and Hommel [33] to examine differences in the degree of cognitive control exerted in challenging vs. easy tasks as well as the effect of cognitive control on automaticity. Participants were asked to respond by pressing the right and left Ctrl buttons on a regular keyboard with their right and left index fingers, respectively. The paradigm comprised two separate tasks that used the same set of target stimuli, which varied in color (red or green), size (small/∼2.5 cm in diameter or large/∼5 cm in diameter) and shape (diamond or square) (see Fig. 3). All eight possible stimulus feature combinations were used equally often in the 384 trials of each

- RLS patients should show control deficits (as reflected by larger task differences) 3

Neuroscience Letters 712 (2019) 134494

R. Zhang, et al.

(caption on next page)

4

Neuroscience Letters 712 (2019) 134494

R. Zhang, et al.

Fig. 2. MR spectroscopy. Top Graph: Representative example of a fitted spectrum (using a special basis data set within LC-Model), the metabolite estimates (in arbitrary units, right) as well as the residuals (top part of the graph). The grey lines show the measured spectrum, as well as the calculated baseline. The red line depicts the LC-Model fit (based on a specially designed basis-data set for the subsequent calculation of metabolite concentrations). The residual between the originally measured and the fitted spectrum is shown (within the upper part of the LC-Model output). The GABA peak is not apparent on the spectrum itself, as it is overlaid by other metabolites. Middle graphs: The left graph shows the LC-Model fit for selected metabolite, including the GABA signal after removal of “irrelevant”/ other metabolites (yellow line). The right graph displays the linewidth of the unsuppressed water signal (shim quality). Bottom graph: MATLAB-based data output of the exemplary dataset in the top part of this figure, including the calculated brain metabolite values (mM +/- %SD) based on the LC-Model estimates and using equation [1]. GABA = gamma aminobutyric acid; Glu = glutamate; Gln = glutamine; tCho = total choline; tNAA = total N-acetylaspartate; tCr = total creatine, Lac = lactate; Gly = glycine.

neuropsychological assessment).

- RLS patients should benefit less or even not at all from the implicit feature (as reflected by smaller condition differences)

3.2. Behavioral task performance Values are provided as mean ± standard error of the mean (SEM). MRS data was integrated with the behavioral data by means of correlation and regression analyses [40,48–50]: We correlated thalamic GABA levels

For the accuracy data (percentage of correct responses/hits), the mixed-effects ANOVA revealed a main effect of task (F(1,54) = 70.48; p < .001; η2p = .566) showing that all participants had more correct responses in the low demand task (94.79% ± 0.16) than in the high demand task (91.65% ± .39). There was also a main effect of group (F (1,54) = 12.11; p = .001; η2p = .183) showing that healthy controls had higher accuracy (94.02% ± 0.31) than RLS patients (92.41% ± .35). Interestingly, an interaction of task*group (F(1,54) = 7.10; p = .010; η2p = .116) revealed that the performance differences between the two tasks (low demand task minus high demand task) were larger in RLS patients (4.14% ± 0.65) than in healthy controls (2.15% ± .42) (Posthoc: t = 2.67; p = .010). Moreover, the worse performance of RLS patients could not be explained by shorter sleep duration or poorer sleep quality, as we found group differences (see Table 1) in sleep quality and duration, but no correlation between any of these two measures and overall accuracy performance, or the observed accuracy differences (all r < |.17|; p > .219). Of note, the interaction effect of task*group also remained significant when sleep duration and subjective sleep quality were included as covariates in the ANOVA (F (1,52) = 6.20; p = .016; η2p = .106). A main effect of condition or any other interaction effects were not found (all F < .80; p > .377). Neither general accuracy (r = .13, p = .542) nor the accuracy differences between the two tasks (r = −0.03, p = .901) were correlated to RLS

- with RLS symptom severity to test the hypothesis that thalamic GABA is linearly related to sensory-motor symptoms. - with accuracy, as we had observed group differences on the behavior level, suggesting that RLS-associated changes in thalamic GABA might linearly modulate behavioral effects. The latter was further investigated with multiple linear regression analyses were using the “enter” method to predict the behavioral differences observed between patients and healthy controls. The group effect (i.e. patients vs. controls) was also included in the regression analyses. 3. Results 3.1. Questionnaires and neuropsychological assessment Information about the clinical characteristics of the patients as well as the neuropsychological data of all subjects are shown in Table 1 (see S2 for detailed information about applied questionnaires and

Fig. 3. Illustration of the experimental paradigm. Each trial started with the target presentation which was either terminated by the first response or stopped after 2000 ms automatically (in this case, it was coded as a “miss”). It was followed by a 700 ms blank screen, a 500 ms feedback (“+” for correct and “-” for incorrect or missed responses) and a second 500 ms blank screen before next trial presentation. (b) Illustration of all employed target stimuli. Targets could vary in color, shape, and size, resulting in eight possible only four combinations of these features. For illustration purposes, only four of them are shown in this graph. The two stimuli on the left of this graph required a left hand (LH) response while the two stimuli on the right on the graph required a right hand (RH) response. (c) Illustration of the baseline vs. predictive condition. In the baseline condition, all stimuli were presented in the center of the screen while they were shifted either slightly (3 cm) upward or downward in the predictive condition. Unbeknownst to the participants, all upward stimuli required a left hand response while all downward stimuli required a right hand response.

5

Neuroscience Letters 712 (2019) 134494

R. Zhang, et al.

Table 1 Subject characteristics, neuropsychological scores for RLS patients and controls, and p-values for group comparisons.

Age Sex (Female %) High-blood pressure Diabetes Thyroid dysfunction Hyperlipidemia Rheumatism IRLS

Cigarettes per day Alcohol per day (standard units) Previous night sleep duration (h) Subjective sleep quality BDI FSMC (total) FSMC (cognitive) FSMC (motor) MEQ Stroop word (s) Stroop color (s) Stroop conflict (s) d2-R ZST MMSE Digit span forward Digit span backward

RLS (n = 25)

Control (n = 31)

Group difference (p value)

60.16 ± 2.21 72% 12 4 3 3 2 22.89 ± 1.46 (severe RLS symptoms) 1.44 ± .75 1.91 ± .42

60.25 ± 1.87 74% 8 4 5 2 0

p = .973 p > .10 p > .05 p > .10 p > .10 p > .10 p > .05

1.51 ± .83 2.08 ± .58

p = .953 p = .820

5.28 ± .33

6.45 ± .24

p = .005

.04 ± .17 8.98 ± 1.11 52.46 ± 3.53 25.66 ± 1.85 26.80 ± 1.76 59.76 ± 1.80 15.50 ± .59 21.81 ± .78 37.30 ± 1.69 134.36 ± 5.66 51.92 ± 1.60 28.80 ± .24 7.20 ± .37 5.92 ± .37

.68 ± .54 6.22 ± .95 31.61 ± 1.99 15.25 ± .95 16.35 ± 1.11 57.81 ± 1.72 14.46 ± .38 20.65 ± .53 36.82 ± 1.43 131.59 ± 6.81 52.61 ± 2.09 28.97 ± .20 6.71 ± .28 5.81 ± .32

p = .001 p = .062 p < .001 p < .001 p < .001 p = .439 p = .133 p = .212 p = .829 p = .760 p = .801 p = .588 p = .279 p = .816

Please note that t-tests were run for continuous variables like age or test scores while χ² tests were run for frequencies like sex or chronic comorbid diseases. IRLS: International RLS Rating Scale; BDI: Beck Depression Inventory; FSMC: Fatigue Scale for Motor and Cognitive functions; MEQ: MorningnessEveningness Questionaire. Lower scores represent greater eveningness and higher scores represent greater morningness; VLMT: Verbal Learning and Memory Retention Test; d2-R: Selective and sustained attention test; Stroop: Stroop color and word test determining the individual’s inhibition and cognitive flexibility; ZST: Digit symbol test (Zahlen-Symbol-Test); MMSE: Mini-Mental State Examination. Given the descriptive and exploratory nature of these analyses, we did not apply Bonferroni-correction.

Fig. 4. Scatterplots denote the correlation between thalamic GABA levels and behavioral task performance. (a) Correlation between thalamic GABA levels and general accuracy. (b) Correlation between thalamic GABA levels and the accuracy difference between the two tasks. Black dots depict RLS patients and open circles depict healthy controls. Solid lines denote slope of correlation in RLS patients and dotted lines denote the slope of correlation in healthy controls.

were used as independent variables. Sleep-related factors were not included in the regression analyses as we did not find any correlations between accuracy and sleep duration or quality. For accuracy, the regression model was significant (F(2, 39) = 4.82, p = .013, adjusted R2 = .16). Within the model, only the group factor significantly contributes to the prediction (ß=−0.450, t = -3.11, p = .004). Yet, GABA levels were not correlated with general accuracy in RLS patients or healthy controls (all r < .13; p > .568) (see Fig. 4a). For the accuracy differences between tasks, the regression model could significantly predict the behavioral performance (F(2, 39) = 5.03, p = .011, adjusted R2 = .16). Within the model, only the group factor was significant (ß = −0.393, t = 2.73, p = .010). GABA was significantly correlated with behavioral performance in the RLS group (r = −.58; p = .006), but not in the control group (r = .081; p = .726). Following this, separate regression analyses with GABA level as independent variable were conducted for the RLS and control group. The multiple linear regression analysis (F(1, 19) = 9.78, p = .006, adjusted R2 = .31) showed that thalamic GABA (ß=−0.583, t=−3.13, p = .006) significantly predicted performance differences between the high and low demand tasks in RLS patients, but not in healthy controls (ß = .081, t = .37, p = .726) (F(1, 19) = .13, p = .726, adjusted R2 = 0.05) (see Fig. 4b). Taken together, RLS patients had lower response accuracy and showed larger performance variations between high and low demand

severity, as assessed with the IRLS questionnaire. Notably, patients with medication (n = 20) and without medication (n = 5) did not differ with respect to their task performance (all t < .99; p > .331). For the RT data, the mixed-effects ANOVA revealed a main effect of task (F(1,54) = 488.60; p < .001; η2p = .900), showing that all participants responded faster in the low demand task (567 ms ± 11) than in the high demand task (875 ms ± 18). No other main or interaction effects were significant (all F < 3.145; p > .083). The RTs of RLS patients were not correlated with RLS severity, as assessed with the IRLS questionnaire (r = −0.04, p = .870).

3.3. MRS data and regression analyses Thalamic GABA levels did not significantly differ between groups (t = −0.90; p = .373; controls: .58 mM ± .05; RLS: .65 mM ± .05). Also, thalamic GABA levels were not correlated with RLS severity, as assessed with the IRLS questionnaire (r = −.26, p = .285). Against the background of the observed main and interaction effects of the group factor, a multiple linear regression was run to integrate the model-based behavioral data with MRS data. In these regression analyses, the general accuracy (total percentage of hits) and the accuracy differences between tasks (low demand task minus high demand task) were used as the dependent variables. Thalamic GABA levels and the group factor 6

Neuroscience Letters 712 (2019) 134494

R. Zhang, et al.

hyperactivity in RLS [2]. This is consistent with the beneficial therapeutic effect of anticonvulsants, which mimic the inhibitory effects of GABA via their effects of calcium channels [53–55]. As GABA is an inhibitory transmitter, thalamic GABA may hence be able to “compensate” cognitive dysfunctions resulting from aberrant thalamic activity by dampening down the thalamus to more “normal” activity levels. With respect to thalamo-cortical loops, thalamic GABA has been suggested to modulate the ongoing thalamo-cortical rhythmic oscillations [26,27] and enhance cortical activation by promoting burst firing of thalamic projection neurons [12,30]. In line with this, it has been noted that such GABAergically mediated neuronal oscillations may support task goal representations in the dorsolateral prefrontal cortex, which is cognitive control and working memory [56,57]. This could have contributed to smaller task goal representation and larger working memory deficits in RLS patients with relatively lower GABA levels. Furthermore, Ku et al. [20] found that RLS patients display reduced thalamic connectivity with the right parahippocampal gyrus, which has been presumed to contribute to the formation and maintenance of bound information (i.e. complex task rules) in working memory [23]. Due to reduced thalamo-cortical connectivity, these cognitive domains should be impaired and have a negative influence on cognitive control in RLS. Importantly, the modulation of thalamo-cortical connections by thalamic GABA has been proposed to depend on ongoing demands [58], which may explain the decreased task version differences in RLS patients with higher thalamic GABA levels, as the effect should have been larger in the high demand task. Given that healthy controls can be assumed to not suffer the same thalamic hyperactivity and/or thalamic hypoconnectivity as RLS patients, it seems logical that they did not benefit from higher thalamic GABA levels in the same way as the patients did. The lack of correlation between performance and thalamic GABA levels in healthy controls may thus be attributed to the simple fact that they did not suffer from pathological neurobiochemical or connectionistic thalamic changes that would require any kind of neurobiochemical “compensation”. Surprisingly, neither RLS patients nor healthy controls showed any sign of functionally processing the implicit task feature (i.e. no main effect of condition was found), as previously observed in younger healthy controls performing the low demand task [33,59]. While one could argue that the lack of effects in RLS patients matches our a priori hypothesis of impaired implicit learning, it does not explain why controls also showed no beneficial effects of the predictive feature in the low demand task (and also not for the high demand task, even though this was not expected). Given that the degree of implicit learning strongly relies on the degree of available residual cognitive processing capacities in the applied paradigm, a potential explanation could be provided by the observation that the capacity of cognitive control processes, including working memory, declines with age [60,61]. With an average age over 60 years, our participants were more than twice as old as the healthy controls who had previously shown an effect of implicit feature processing [33,59]. Perhaps, our sample was more cognitively challenged by both tasks due to their age-dependent decline of working memory and other cognitive control functions. As a consequence, they may have been unable to pick up the predictive feature altogether [33] due to their (comparatively) limited processing capacity, which did not provide them with enough “free” resources to process the information of the implicit feature even in the low demand task. Even though our findings are in line with most of the current literature on thalamic functioning in RLS, it is important to note that thalamic GABA levels could not explain all findings, as they did not enhance general performance (as reflected by the overall accuracy) or correlate with it. It is possible that this aspect of behavior was accounted for by other facets of the pathology underlying RLS, such as dopaminergic dysfunction [9], or abnormal glutamate/glutamine levels [7], which are also very likely to contribute to the cognitive changes

tasks than healthy controls. Within the RLS patients, higher GABA levels were correlated to smaller performance differences between tasks (but not with general increases in accuracy). However, GABA could not predict task performance differences in healthy controls. 4. Discussion We examined the potential relationship between thalamic GABA concentrations and cognitive control in RLS patients, as compared to healthy controls. The study was motivated by the fact that RLS patients have been reported to display changes in thalamic structure, connectivity, and biochemistry, especially GABAergic signaling, which are also known to contribute to cognitive control deficits [8,19,20,51]. Yet still, the role of thalamic GABA in RLS had so far only been investigated for sensory-motor symptoms, but not for the concomitant cognitive control deficits, which are key to leading a self-serving and successful life [31,32]. In line with previous reports about deficits in executive functions and cognitive control in RLS [14,36], we found lower behavioral performance (lower response accuracy) in RLS patients than in healthy controls. RLS patients furthermore displayed larger accuracy differences between the low and high demand tasks, suggesting that recruiting the additional cognitive control resources required for the high demand task puts a greater strain on RLS patients than on controls. Working memory may have played a key role for this because task demands were induced by varying working memory load: In the low demand task, relatively little cognitive control was required due to the consistent mapping between target stimulus features and responses. In contrast to this, the high demand task had a variable mapping between features and responses that required to attend to two task-relevant features (color and size) instead of just one. As working memory is therefore likely to be more exploited in the high demand task than in the low demand task, it is likely to have contributed to the observed effects on cognitive control. Importantly, the observed behavioral effects were not driven by the sleep deprivation of our patients, even though RLS patients reported an overall reduced sleep duration and lower sleep quality. Given that neither RTs nor accuracy were correlated with RLS severity and given that the RTs of patients were comparable to those of healthy controls, it seems quite unlikely that sensory-motor differences or dopamine agonist withdrawal syndrome (DAWS), for which impulsivity appears to be a major risk factor [52], caused the observed cognitive deficits. Also, no group differences were found in terms of scores of chronotype, as assessed with the Morningness-Eveningness questionnaire (see Table 1). The observed deficits can therefore also not be attributed to inter-individual differences in chronotype / circadian rhythms. Taken together, it seems more reasonable to attribute cognitive deficits in RLS to basic pathological changes in RLS rather than the consequences of sleep loss, “side effect” of RLS symptoms, DAWS, or chronotypes. In line with the findings of a study by Winkelman et al. [8], the thalamic GABA levels of RLS patients in the current study did not quantitatively differ from those of the matched healthy controls. Other than Winkelman et al. [8], we did not find a positive correlation between thalamic GABA levels and RLS severity. In this context, it should however also be mentioned that the subjects of Winkelman et al. [8] were substantially younger than those in the current study, which could have potentially influenced our results. Yet, further investigations would be required to substantiate this claim. Nonetheless, thalamic GABA levels modulated performance in the RLS group: The observed cognitive control impairments (reflected by larger task differences) were smaller in RLS patients with (relatively) higher thalamic GABA levels as compared to patients with (relatively) lower thalamic GABA levels. One of the most straightforward possibilities to explain why the RLS patients’ performance benefitted from increased thalamic GABA levels is that several lines of evidence have demonstrated thalamic 7

Neuroscience Letters 712 (2019) 134494

R. Zhang, et al.

Declaration of Competing Interest

observed in RLS and furthermore closely interact with GABA [62,63]. It could also be an interesting objective for future studies to investigate whether the observed compensatory effect relies on the GABA released by thalamic reticular nuclei, GABAergic interneurons, or GABA derived from pallido-thalamic projections. Last but not least, more studies on the effects of thalamic GABA levels on other aspects of cognitive control are needed. As working memory is fundamental for all forms of proactive control [56], thalamic GABA may attenuate the deficits of other executive functions in RLS. Lastly, it should be noted that we only assessed the left thalamus, but not the right thalamus in order to keep scanning times as short as possible for the RLS patients. While we are not aware of any studies that suggest laterality effects for thalamic glutamate or GABA, assessing both sides would of course have provided a more complete picture. It should furthermore be mentioned that even though no behavioral differences were found between participants with and without RLS medication treatment, this heterogeneity within the patient group may still have contributed to the large variance in the current study. Moreover, future studies may benefit from objective sleep measurements using actigraphy and polysomnography, which could provide objective and more detailed information about the sleep arousals, obstructive sleep apnea, and periodic limb movements occurring in most RLS patients [64–66]. These comorbidities might also provide valuable information about the heterogeneity found in our sample and further help to understand how cognitive impairments in RLS may come about.

The authors report no conflict of interest. Appendix A. Supplementary data Supplementary material related to this article can be found, in the online version, at doi:https://doi.org/10.1016/j.neulet.2019.134494. References [1] R.P. Allen, D.L. Picchietti, D. Garcia-Borreguero, W.G. Ondo, A.S. Walters, J.W. Winkelman, M. Zucconi, R. Ferri, C. Trenkwalder, H.B. Lee, International Restless Legs Syndrome Study Group, restless legs syndrome/Willis-Ekbom disease diagnostic criteria: updated International Restless Legs Syndrome Study Group (IRLSSG) consensus criteria–history, rationale, description, and significance, Sleep Med. 15 (2014) 860–873, https://doi.org/10.1016/j.sleep.2014.03.025. [2] B.B. Koo, K. Bagai, A.S. Walters, Restless legs syndrome: current concepts about disease pathophysiology, Tremor Hyperkinet. Mov. N. Y. N. 6 (2016) 401, https:// doi.org/10.7916/D83J3D2G. [3] G. Rizzo, C. Tonon, C. Testa, D. Manners, R. Vetrugno, F. Pizza, S. Marconi, E. Malucelli, F. Provini, G. Plazzi, P. Montagna, R. Lodi, Abnormal medial thalamic metabolism in patients with idiopathic restless legs syndrome, Brain J. Neurol. 135 (2012) 3712–3720, https://doi.org/10.1093/brain/aws266. [4] F. Provini, G. Chiaro, Neuroimaging in restless legs syndrome, Sleep Med. Clin. 10 (2015) 215–226, https://doi.org/10.1016/j.jsmc.2015.05.006 xi. [5] Y. Zhuo, Y. Wu, Y. Xu, L. Lu, T. Li, X. Wang, K. Li, Combined resting state functional magnetic resonance imaging and diffusion tensor imaging study in patients with idiopathic restless legs syndrome, Sleep Med. 38 (2017) 96–103, https://doi.org/ 10.1016/j.sleep.2017.06.033. [6] C. Trenkwalder, W. Paulus, Restless legs syndrome: pathophysiology, clinical presentation and management, Nat. Rev. Neurol. 6 (2010) 337–346, https://doi.org/ 10.1038/nrneurol.2010.55. [7] R.P. Allen, P.B. Barker, A. Horská, C.J. Earley, Thalamic glutamate/glutamine in restless legs syndrome: increased and related to disturbed sleep, Neurology 80 (2013) 2028–2034, https://doi.org/10.1212/WNL.0b013e318294b3f6. [8] J.W. Winkelman, L. Schoerning, S. Platt, J.E. Jensen, Restless legs syndrome and central nervous system gamma-aminobutyric acid: preliminary associations with periodic limb movements in sleep and restless leg syndrome symptom severity, Sleep Med. 15 (2014) 1225–1230, https://doi.org/10.1016/j.sleep.2014.05.019. [9] R.P. Allen, Restless leg syndrome/Willis-Ekbom disease pathophysiology, Sleep Med. Clin. 10 (2015) 207–214, https://doi.org/10.1016/j.jsmc.2015.05.022 xi. [10] I. Bar-Gad, G. Morris, H. Bergman, Information processing, dimensionality reduction and reinforcement learning in the basal ganglia, Prog. Neurobiol. 71 (2003) 439–473, https://doi.org/10.1016/j.pneurobio.2003.12.001. [11] W.X. Chmielewski, C. Beste, Action control processes in autism spectrum disorder– insights from a neurobiological and neuroanatomical perspective, Prog. Neurobiol. 124 (2015) 49–83, https://doi.org/10.1016/j.pneurobio.2014.11.002. [12] S.M. Sherman, Thalamus, Scholarpedia 1 (2006) 1583, https://doi.org/10.4249/ scholarpedia.1583. [13] S. Fulda, N. Szesny, M. Ising, A. Heck, A. Grübl, R. Lieb, S. Reppermund, Further evidence for executive dysfunction in subjects with RLS from a non-clinical sample, Sleep Med. 12 (2011) 1003–1007, https://doi.org/10.1016/j.sleep.2011.04.013. [14] S.M. Kim, J.W. Choi, C. Lee, B.U. Lee, Y.S. Koo, K.H. Kim, K.-Y. Jung, Working memory deficit in patients with restless legs syndrome: an event-related potential study, Sleep Med. 15 (2014) 808–815, https://doi.org/10.1016/j.sleep.2014.03. 010. [15] K.M. Hartikainen, L. Sun, M. Polvivaara, M. Brause, K. Lehtimäki, J. Haapasalo, T. Möttönen, K. Väyrynen, K.H. Ogawa, J. Öhman, J. Peltola, Immediate effects of deep brain stimulation of anterior thalamic nuclei on executive functions and emotion-attention interaction in humans, J. Clin. Exp. Neuropsychol. 36 (2014) 540–550, https://doi.org/10.1080/13803395.2014.913554. [16] L.M. Law, D.M. Smith, The anterior thalamus is critical for overcoming interference in a context-dependent odor discrimination task, Behav. Neurosci. 126 (2012) 710–719, https://doi.org/10.1037/a0029698. [17] D. Liebermann, C.J. Ploner, A. Kraft, U.A. Kopp, F. Ostendorf, A dysexecutive syndrome of the medial thalamus, Cortex J. Devoted Study Nerv. Syst. Behav. 49 (2013) 40–49, https://doi.org/10.1016/j.cortex.2011.11.005. [18] Y.B. Saalmann, Intralaminar and medial thalamic influence on cortical synchrony, information transmission and cognition, Front. Syst. Neurosci. 8 (2014) 83, https:// doi.org/10.3389/fnsys.2014.00083. [19] T. Etgen, B. Draganski, C. Ilg, M. Schröder, P. Geisler, G. Hajak, I. Eisensehr, D. Sander, A. May, Bilateral thalamic gray matter changes in patients with restless legs syndrome, NeuroImage 24 (2005) 1242–1247, https://doi.org/10.1016/j. neuroimage.2004.10.021. [20] J. Ku, Y.W. Cho, Y.S. Lee, H.-J. Moon, H. Chang, C.J. Earley, R.P. Allen, Functional connectivity alternation of the thalamus in restless legs syndrome patients during the asymptomatic period: a resting-state connectivity study using functional magnetic resonance imaging, Sleep Med. 15 (2014) 289–294, https://doi.org/10.1016/ j.sleep.2013.09.030. [21] J. Ku, Y.S. Lee, H. Chang, C.J. Earley, R.P. Allen, Y.W. Cho, Default mode network disturbances in restless legs syndrome/Willis–Ekbom disease, Sleep Med. 23 (2016)

5. Conclusion Compared to healthy controls, RLS patients displayed reduced cognitive control capacities, which were most likely based on working memory deficits. In the absence of absolute group differences, elevated thalamic GABA levels were correlated with reduced control deficits in the RLS group only, as they might have helped to “compensate” the RLS-related thalamic hyperactivity and associated thalamic hypoconnectivity. Taken together, our findings specify the functional relevance of thalamic GABAergic neurotransmission for cognition in RLS, even though changes in GABAergic neurotransmission might not be the ultimate cause of control deficits. Role of the funding source This work was supported by a grant from the Else Kröner-FreseniusStiftung (2014_A197) to AS. The funding source had no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript. WH has received an unrestricted research grant from TEVA and travel grants from UCB Pharma and TEVA not related to this manuscript. CB has received payment for consulting and speaker’s honoraria from GlaxoSmithKline, Novartis, Genzyme, and Teva. He has recent research grants with Novartis and Genzyme. Contributors All authors were involved in designing the study and collecting the data. RZ, AS, and AW analyzed the data. All authors were included in writing the manuscript and approved its final version. Data availability Data cannot be shared publicly due to privacy regulations / because of concerns about potential identification of study participants in case all raw data was published. All data, that does not allow to identify the participants, will be made available upon request to [email protected]. 8

Neuroscience Letters 712 (2019) 134494

R. Zhang, et al.

[43] C. Choi, S. Ganji, K. Hulsey, A. Madan, Z. Kovacs, I. Dimitrov, S. Zhang, K. Pichumani, D. Mendelsohn, B. Mickey, C. Malloy, R. Bachoo, R. DeBerardinis, E. Maher, A comparative study of short- and long-TE 1H MRS at 3 T for in vivo detection of 2-hydroxyglutarate in brain tumors, NMR Biomed. 26 (2013) 1242–1250, https://doi.org/10.1002/nbm.2943. [44] S.W. Provencher, Automatic quantitation of localized in vivo 1H spectra with LCModel, NMR Biomed. 14 (2001) 260–264. [45] S.K. Ganji, A. Banerjee, A.M. Patel, Y.D. Zhao, I.E. Dimitrov, J.D. Browning, E.S. Brown, E.A. Maher, C. Choi, T2 measurement of J-coupled metabolites in the human brain at 3T, NMR Biomed. 25 (2012) 523–529, https://doi.org/10.1002/ nbm.1767. [46] V. Mlynárik, S. Gruber, E. Moser, Proton T (1) and T (2) relaxation times of human brain metabolites at 3 Tesla, NMR Biomed. 14 (2001) 325–331. [47] F. Träber, W. Block, R. Lamerichs, J. Gieseke, H.H. Schild, 1H metabolite relaxation times at 3.0 tesla: Measurements of T1 and T2 values in normal brain and determination of regional differences in transverse relaxation, J. Magn. Reson. Imaging JMRI. 19 (2004) 537–545, https://doi.org/10.1002/jmri.20053. [48] L. Haag, C. Quetscher, S. Dharmadhikari, U. Dydak, T. Schmidt-Wilcke, C. Beste, Interrelation of resting state functional connectivity, striatal GABA levels, and cognitive control processes, Hum. Brain Mapp. 36 (2015) 4383–4393, https://doi. org/10.1002/hbm.22920. [49] C. Quetscher, A. Yildiz, S. Dharmadhikari, B. Glaubitz, T. Schmidt-Wilcke, U. Dydak, C. Beste, Striatal GABA-MRS predicts response inhibition performance and its cortical electrophysiological correlates, Brain Struct. Funct. (2014), https:// doi.org/10.1007/s00429-014-0873-y. [50] A. Yildiz, C. Quetscher, S. Dharmadhikari, W. Chmielewski, B. Glaubitz, T. SchmidtWilcke, R. Edden, U. Dydak, C. Beste, Feeling safe in the plane: neural mechanisms underlying superior action control in airplane pilot trainees—a combined EEG/MRS study, Hum. Brain Mapp. 35 (2014) 5040–5051, https://doi.org/10.1002/hbm. 22530. [51] M. Gorges, J. Rosskopf, H.-P. Müller, K. Lindemann, M. Hornyak, J. Kassubek, Patterns of increased intrinsic functional connectivity in patients with restless legs syndrome are associated with attentional control of sensory inputs, Neurosci. Lett. 617 (2016) 264–269, https://doi.org/10.1016/j.neulet.2016.02.043. [52] M.J. Nirenberg, Dopamine agonist withdrawal syndrome: implications for patient care, Drugs Aging 30 (2013) 587–592, https://doi.org/10.1007/s40266-0130090-z. [53] E. Griffin, J.N. Brown, Pregabalin for the treatment of restless legs syndrome, Ann. Pharmacother. 50 (2016) 586–591, https://doi.org/10.1177/1060028016643097. [54] E.S. Kim, E.D. Deeks, Gabapentin enacarbil: a review in restless legs syndrome, Drugs 76 (2016) 879–887, https://doi.org/10.1007/s40265-016-0584-1. [55] J.W. Winkelman, R.K. Bogan, M.H. Schmidt, J.D. Hudson, S.E. DeRossett, C.E. HillZabala, Randomized polysomnography study of gabapentin enacarbil in subjects with restless legs syndrome, Mov. Disord. Off. J. Mov. Disord. Soc. 26 (2011) 2065–2072, https://doi.org/10.1002/mds.23771. [56] D.M. Barch, A. Ceaser, Cognition in schizophrenia: core psychological and neural mechanisms, Trends Cogn. Sci. 16 (2012) 27–34, https://doi.org/10.1016/j.tics. 2011.11.015. [57] V. Prevosto, M.A. Sommer, Cognitive control of movement via the cerebellar-recipient thalamus, Front. Syst. Neurosci. 7 (2013) 56, https://doi.org/10.3389/ fnsys.2013.00056. [58] M.M. Halassa, L. Acsády, Thalamic inhibition: diverse sources, diverse scales, Trends Neurosci. 39 (2016) 680–693, https://doi.org/10.1016/j.tins.2016.08.001. [59] A.-K. Stock, L. Steenbergen, L. Colzato, C. Beste, The system neurophysiological basis of non-adaptive cognitive control: inhibition of implicit learning mediated by right prefrontal regions, Hum. Brain Mapp. 37 (2016) 4511–4522, https://doi.org/ 10.1002/hbm.23325. [60] R. Bendayan, A.M. Piccinin, S.M. Hofer, D. Cadar, B. Johansson, G. Muniz-Terrera, Decline in memory, visuospatial ability, and crystalized cognitive abilities in older adults: normative aging or terminal decline? J. Aging Res. 2017 (2017) 6210105, , https://doi.org/10.1155/2017/6210105. [61] A.X.P. Rozas, O. Juncos-Rabadán, M.S.R. González, Processing speed, inhibitory control, and working memory: three important factors to account for age-related cognitive decline, Int. J. Aging Hum. Dev. 66 (2008) 115–130, https://doi.org/10. 2190/AG.66.2.b. [62] O. Dandash, C. Pantelis, A. Fornito, Dopamine, fronto-striato-thalamic circuits and risk for psychosis, Schizophr. Res. 180 (2017) 48–57, https://doi.org/10.1016/j. schres.2016.08.020. [63] D.M. Barch, A. Ceaser, Cognition in schizophrenia: core psychological and neural mechanisms, Trends Cogn. Sci. 16 (2012) 27–34, https://doi.org/10.1016/j.tics. 2011.11.015. [64] R.P. Allen, A.S. Walters, J. Montplaisir, W. Hening, A. Myers, T.J. Bell, L. FeriniStrambi, Restless legs syndrome prevalence and impact: REST general population study, Arch. Intern. Med. 165 (2005) 1286–1292, https://doi.org/10.1001/ archinte.165.11.1286. [65] R.P. Allen, P.B. Barker, A. Horská, C.J. Earley, Thalamic glutamate/glutamine in restless legs syndrome: increased and related to disturbed sleep, Neurology 80 (2013) 2028–2034, https://doi.org/10.1212/WNL.0b013e318294b3f6. [66] D. Garcia-Borreguero, Time to REST: epidemiology and burden, Eur. J. Neurol. 13 (Suppl 3) (2006) 15–20, https://doi.org/10.1111/j.1468-1331.2006.01587.x.

6–11, https://doi.org/10.1016/j.sleep.2016.05.007. [22] J. Ku, Y.S. Lee, H.W. Chang, C.J. Earley, R.P. Allen, Y.W. Cho, Diurnal variation of default mode network in patients with restless legs syndrome, Sleep Med. 41 (2018) 1–8, https://doi.org/10.1016/j.sleep.2017.09.031. [23] D. Luck, J.-M. Danion, C. Marrer, B.-T. Pham, D. Gounot, J. Foucher, The right parahippocampal gyrus contributes to the formation and maintenance of bound information in working memory, Brain Cogn. 72 (2010) 255–263, https://doi.org/ 10.1016/j.bandc.2009.09.009. [24] P.N. Margariti, L.G. Astrakas, S.G. Tsouli, G.M. Hadjigeorgiou, S. Konitsiotis, M.I. Argyropoulou, Investigation of unmedicated early onset restless legs syndrome by voxel-based morphometry, T2 relaxometry, and functional MR imaging during the night-time hours, AJNR Am. J. Neuroradiol. 33 (2012) 667–672, https://doi. org/10.3174/ajnr.A2829. [25] J. Peräkylä, L. Sun, K. Lehtimäki, J. Peltola, J. Öhman, T. Möttönen, K.H. Ogawa, K.M. Hartikainen, Causal evidence from humans for the role of mediodorsal nucleus of the thalamus in working memory, J. Cogn. Neurosci. (2017) 1–13, https://doi. org/10.1162/jocn_a_01176. [26] R. Llinás, F.J. Urbano, E. Leznik, R.R. Ramírez, H.J.F. van Marle, Rhythmic and dysrhythmic thalamocortical dynamics: GABA systems and the edge effect, Trends Neurosci. 28 (2005) 325–333, https://doi.org/10.1016/j.tins.2005.04.006. [27] Z. Rovó, F. Mátyás, P. Barthó, A. Slézia, S. Lecci, C. Pellegrini, S. Astori, C. Dávid, B. Hangya, A. Lüthi, L. Acsády, Phasic, nonsynaptic GABA-A receptor-mediated inhibition entrains thalamocortical oscillations, J. Neurosci. Off. J. Soc. Neurosci. 34 (2014) 7137–7147, https://doi.org/10.1523/JNEUROSCI.4386-13.2014. [28] Z. Long, X.-R. Li, J. Xu, R.A.E. Edden, W.-P. Qin, L.-L. Long, J.B. Murdoch, W. Zheng, Y.-M. Jiang, U. Dydak, Thalamic GABA predicts fine motor performance in manganese-exposed smelter workers, PLoS One 9 (2014) e88220, , https://doi. org/10.1371/journal.pone.0088220. [29] C. Beste, L. Steenbergen, R. Sellaro, S. Grigoriadou, R. Zhang, W. Chmielewski, A.K. Stock, L. Colzato, Effects of concomitant stimulation of the GABAergic and norepinephrine system on inhibitory control - a study using transcutaneous vagus nerve stimulation, Brain Stimulat. 9 (2016) 811–818, https://doi.org/10.1016/j. brs.2016.07.004. [30] S.M. Sherman, Tonic and burst firing: dual modes of thalamocortical relay, Trends Neurosci. 24 (2001) 122–126. [31] J.D. Cohen, Cognitive control, in: T. Egner (Ed.), Wiley Handb. Cogn. Control, John Wiley & Sons, Ltd, 2017, pp. 1–28, , https://doi.org/10.1002/9781118920497.ch1. [32] A. Diamond, Executive functions, Annu. Rev. Psychol. 64 (2013) 135–168, https:// doi.org/10.1146/annurev-psych-113011-143750. [33] B.R. Bocanegra, B. Hommel, When cognitive control is not adaptive, Psychol. Sci. 25 (2014) 1249–1255, https://doi.org/10.1177/0956797614528522. [34] J.W. Choi, D. Ko, G.-T. Lee, K.-Y. Jung, K.H. Kim, Reduced neural synchrony in patients with restless legs syndrome during a visual oddball task, PLoS One 7 (2012) e42312, , https://doi.org/10.1371/journal.pone.0042312. [35] K.-Y. Jung, Y.-S. Koo, B.-J. Kim, D. Ko, G.-T. Lee, K.H. Kim, C.H. Im, Electrophysiologic disturbances during daytime in patients with restless legs syndrome: further evidence of cognitive dysfunction? Sleep Med. 12 (2011) 416–421, https://doi.org/10.1016/j.sleep.2010.08.018. [36] K.-Y. Jung, Cognition in restless legs syndrome, J. Sleep Med. 12 (2015) 1–6, https://doi.org/10.13078/jsm.15001. [37] K.-Y. Jung, S.M. Kim, J.-Y. Song, B.U. Lee, C. Lee, S.K. Lee, Y.S. Koo, Y.W. Cho, J.W. Choi, K.H. Kim, Sternberg working memory performance following treatment with pramipexole in patients with moderate-to-severe restless legs syndrome, Sleep Med. 16 (2015) 703–708, https://doi.org/10.1016/j.sleep.2014.10.025. [38] R. Zhang, M.D. Brandt, W. Schrempf, C. Beste, A.-K. Stock, Neurophysiological mechanisms of circadian cognitive control in RLS patients - an EEG source localization study, Neuroimage Clin. 15 (2017) 644–652, https://doi.org/10.1016/j.nicl. 2017.06.018. [39] C. Exner, G. Weniger, E. Irle, Implicit and explicit memory after focal thalamic lesions, Neurology 57 (2001) 2054–2063. [40] S. Dharmadhikari, R. Ma, C.-L. Yeh, A.-K. Stock, S. Snyder, S.E. Zauber, U. Dydak, C. Beste, Striatal and thalamic GABA level concentrations play differential roles for the modulation of response selection processes by proprioceptive information, NeuroImage 120 (2015) 36–42, https://doi.org/10.1016/j.neuroimage.2015.06. 066. [41] C. Choi, I.E. Dimitrov, D. Douglas, A. Patel, L.G. Kaiser, C.A. Amezcua, E.A. Maher, Improvement of resolution for brain coupled metabolites by optimized (1)H MRS at 7T, NMR Biomed. 23 (2010) 1044–1052, https://doi.org/10.1002/nbm.1529. [42] M. Mikkelsen, P.B. Barker, P.K. Bhattacharyya, M.K. Brix, P.F. Buur, K.M. Cecil, K.L. Chan, D.Y.-T. Chen, A.R. Craven, K. Cuypers, M. Dacko, N.W. Duncan, U. Dydak, D.A. Edmondson, G. Ende, L. Ersland, F. Gao, I. Greenhouse, A.D. Harris, N. He, S. Heba, N. Hoggard, T.-W. Hsu, J.F.A. Jansen, A. Kangarlu, T. Lange, R.M. Lebel, Y. Li, C.-Y.E. Lin, J.-K. Liou, J.-F. Lirng, F. Liu, R. Ma, C. Maes, M. Moreno-Ortega, S.O. Murray, S. Noah, R. Noeske, M.D. Noseworthy, G. Oeltzschner, J.J. Prisciandaro, N.A.J. Puts, T.P.L. Roberts, M. Sack, N. Sailasuta, M.G. Saleh, M.-P. Schallmo, N. Simard, S.P. Swinnen, M. Tegenthoff, P. Truong, G. Wang, I.D. Wilkinson, H.-J. Wittsack, H. Xu, F. Yan, C. Zhang, V. Zipunnikov, H.J. Zöllner, R.A.E. Edden, B. GABA, Edited MR spectroscopy at 24 research sites, NeuroImage 159 (2017) 32–45, https://doi.org/10.1016/j.neuroimage.2017.07. 021.

9