Interactions of Catecholamines and GABA + in Cognitive Control: Insights from EEG and 1 H-MRS

Catecholamines and amino acid transmitter systems are known to interact, the exact links and its impact on cognitive control functions is unclear. Using a multi-modal imaging approach combining EEG and Proton-magnetic resonance spectroscopy ( 1 H-MRS), we investigated the effect of different degrees of pharmacological catecholaminergic enhancement onto theta band activity (TBA) as a measure of interference control during response inhibition and execution. Central to our study was evaluating the predictive impact of in-vivo baseline GABA+ concentrations, under varying degrees of Methylphenidate (MPH) stimulation, in the striatum, the anterior cingulate cortex (ACC) and the supplemental motor area (SMA) of healthy adults. We provide evidence for a predictive interrelation of baseline GABA+ concentrations in cognitive control relevant brain areas, onto task induced TBA during response control stimulated with MPH. Baseline GABA+ concentrations in the ACC, the striatum, and the SMA had differential impact on predicting interference control-related TBA in response execution trials. GABA+ concentrations in the ACC appeared to be specifically important for TBA modulations when the cognitive effort needed for interference control was high – that is when no prior task experience exists, or in the absence of catecholaminergic enhancement with MPH. The study highlights the predictive role of baseline GABA+ concentrations in key brain areas influencing cognitive control and responsiveness to catecholaminergic enhancement, particularly in high-effort scenarios.


Introduction
The catecholaminergic system enables effective neural information processing and is central to various cognitive processes, including cognitive control mechanisms (Arnsten, 2011;Arnsten & Rubia, 2012).Adequate levels of both dopamine (DA) (Bensmann, 2020;Botvinick, 2007;Willemssen et al., 2009Willemssen et al., , 2011) ) and norepinephrine (NE) (Mückschel, Chmielewski, et al., 2017), are crucial for the processing and resolution of conflicts, which can be induced by the processing of irrelevant information.The likely mechanism underlying these observations is that catecholamines modulate sensitivity towards distinguishing relevant from irrelevant information by increasing the neural signal-to-noise ratio (SNR) (Aston-Jones & Cohen, 2005;Servan-Schreiber et al., 1990).One way to modulate the catecholaminergic system is administering the psychostimulant methylphenidate (MPH).MPH selectively inhibits DA and NE reuptake, which increases extracellular levels of DA and NE within the prefrontal cortex (Dipasquale et al., 2020;Spencer et al., 2012), and DA levels in striatal areas (Volkow et al., 2001).MPH-related temporary increases in catecholaminergic signaling appear to amplify the SNR in neural information processing, which in turn facilitates task-relevant processing and response selection (Aston-Jones & Cohen, 2005;Mückschel, Gohil, et al., 2017).Generally, catecholamines have been ascribed a biphasic nature of effects, following the well-known inverted U-shaped concentration-response function (Cools & D'Esposito, 2011;Robbins & Arnsten, 2009).That is, moderate levels of arousal/catecholamines generate optimal performance, whereas either too low or too high levels impair performance.However, currently available data on studies examining single-dose MPH effects in healthy volunteers are rather heterogeneous (Linssen et al., 2014), and indicate that there are further modulating factors affecting the size and direction of MPH effects.The catecholaminergic system is also known to be influenced by GABAergic signaling (Tritsch & Sabatini, 2012).Catecholamines modulate the sensitivity of GABAergic and glutamatergic synapses in the prefrontal cortex (PFC) and in other brain areas that play crucial roles for response selection and cognitive control (Manz et al., 2021;Plenz, 2003).In humans, MPH-induced catecholaminergic enhancement was shown to decrease glutamate levels in the PFC and striatal areas, likely due to changes in reciprocal DA/glutamatergic modulation (Carrey et al., 2002).Results from animal studies indicate that single-dose MPH administration results in the obstruction of GABAergic transmission by activation of dopamine receptor D4 (Erlij et al., 2012).Taken together, emerging evidence suggests considerable impact and interrelation of MPH-induced catecholaminergic modulations onto GABAergic transmission, and underlines its importance when investigating MPH (dose-dependent) effects.In short, catecholamines and amino acid transmitter systems are known to heavily interact, but it has never been investigated which aspects of the behavioral effects of MPH (if any) are further modulated by GABA in cognitive control-relevant areas.A better understanding of this interaction is however crucial for unraveling the previously reported, but not fully understood dose dependent effects of MPH, which had indicated further modulating factors (Linssen et al., 2014).In the current study, we used edited Proton Magnetic Resonance Spectroscopy ( 1 H-MRS) to measure total GABA+ (GABA plus macro molecules) concentrations in the striatum, anterior cingulate cortex (ACC) and supplementary motor area (SMA) of healthy adults in order to relate these baseline transmitter concentrations to the effects of MPH on response selection and inhibition during cognitive conflict/interference control.To investigate dose-dependent MPH effects, in the current study we used different degrees of catecholaminergic enhancement by administering low, medium, or high doses of MPH to healthy adults.The assessed brain regions have previously been proven essential for cognitive control processes and were associated with facilitating automatic motor activation and the suppression of pre-potent action plans (ACC and SMA; Bari & Robbins, 2013), as well as efficient response selection (ACC and striatum; Adams et al., 2017;Redgrave et al., 2011).Specifically, activity in the ACC, SMA and striatum appears to be crucial for successful execution of the experimental task and related processes employed in the current study (Beste, Mückschel, et al., 2018;Chmielewski & Beste, 2017;Mückschel, Dippel, et al., 2017).Thus, we expected GABA+ concentrations in these regions to moderate possible dose-dependent effects on both behavioral and neurophysiological level.To assess interference control, we used a combination of a Simon task and a Go/NoGo task (Chmielewski & Beste, 2017), which allows to examine interference effects during both response execution and inhibition (for task details, please refer to the methods section).Here, we expected to replicate the typical task effects, namely better performance in non-conflicting Go trials and in conflicting NoGo trials (Chmielewski & Beste, 2017;Koyun et al., 2023).With respect to the catecholaminergic modulation, we hypothesized low MPH doses to improve response control, whereas high doses of MPH were expected to push individual's catecholamine levels beyond optimum potentially leading to worse performance.Medium MPH doses were expected to tend towards improved response control, however with greater interindividual variability.As a neurophysiological correlate, we assessed theta band activity (TBA).TBA is essential for action control in the face of irrelevant competing inputs, or habitual but inappropriate stimulus-response mappings (Beste et al., 2023).Generally, theta band oscillations are associated with cognitive control, and were been shown to reflect conflict monitoring and resolution processes (Cohen, 2014;Nigbur et al., 2012), thus, likely reflecting MPH effects during context-modulated response execution and inhibition.Notably, cognitive control processes related to TBA were shown to be modulated by both the amino acid transmitter system (Quetscher et al., 2015) and the catecholaminergic system (Dockree et al., 2017).With that said, it is important to consider the close relation of TBA and GABA+ levels in the medial frontal cortex (Takei et al., 2016) and the neuroanatomically connected striatum (Quetscher et al., 2015).This becomes specifically important when investigating MPH effects, as MPH has been shown to modulate catecholaminergic activity in both of these regions (Dipasquale et al., 2020;Spencer et al., 2012)Volkow et al., 2001).Moreover, GABAergic signaling in the ACC has been shown to modulate response selection and cognitive control processes (Silveri et al., 2013).Taken together, the goal of the current study was to investigate the interrelation of baseline GABA+ levels in the ACC, striatum and SMA with regard to the modulatory effects of MPH administration on behavioral and neurophysiological measures of interference control.We hypothesized GABA+ concentrations in the ACC and striatum to be particularly important in predicting task induced modulations (as reflected in TBA), when the need for interference control and effort needed to perform well is high.In contrast to this, we predicted ACC/striatal baseline GABA+ to have comparably less impact on TBA with task practice or catecholamine enhancement.Additionally, we expected GABA+ in the SMA become more important for task induced oscillatory modulations with increasing task familiarity.

Study participants and ethical approval
A total of N =122 young and healthy participants were recruited using the following inclusion criteria: Age between 20 and 31 years, normal/corrected-to-normal vision, no current/reported history of psychiatric or neurologic disease, no developmental disorders or disorders that might interfere with normal brain functioning, no medication affecting the central nervous system, no dairy allergy, no pregnancy and MRI compatibility.Individuals who failed to meet the predetermined inclusion criteria were not eligible for participating in the study.A priori power analyses (with G*Power, Version 3.1.9.7; Faul et al., 2007) indicated that a sample size of n = 120 allows for the identification of small effect sizes down to f = 0.176 with a p (α err) <.05 and a power of 95% (assuming 6 groups and 4 conditions, as well as sphericity and a correlation of 0.5 between repeated measures).This is slightly below the smallest effect size for MPH effects found by previous work of our group, which accounted for 7% of the variance (e.g., Beste, Adelhöfer, et al., 2018).Additionally, this total sample size still allows for the evaluation of potential order effects with a minimum group size of n = 20.Participants gave written informed consent before starting the experiment and received 60€ as a reimbursement for their participation.The study was approved by the ethics committee of TU Dresden (project number: EK 420092015) and conducted in accordance with the Declaration of Helsinki and its later amendments.

Exclusion decisions
N = 21 participants were excluded from all analyses (behavioral, neurophysiological and 1 H-MRS data), due to the following reasons: n = 1 participant dropped out after the first appointment, n = 7 participants produced behavioral accuracy was below chance level in at least two task conditions (< 50%), n = 3 were excluded due to technical problems during data recording, and n = 10 participants had to be excluded due to insufficient EEG signal quality after pre-processing.This resulted in a final sample of n = 101 participants (MPH first group: n = 48; MPH second group: n = 53) that were included in the subsequent analyses of the behavioral and neurophysiological data.A sensitivity analyses in G*Power (Version 3.1.9.7;Faul et al., 2007) showed that the smallest effect size that can be detected (with a power of 95 % and p (α err) of 5%) given n = 101 (split into 2 groups, across 2 time points x 4 conditions) was f = 0.119.As for the 1 H-MRS data analyses, n = 15 participants had to be additionally excluded from the feature importance analyses due to the following reasons: n = 4 due to poor shimming quality/SNR, and/or Cramér-Rao lower bound criterion above 15% of the metabolite of interest (GABA+), and n = 11 participants had to be excluded due to incomplete data (i.e., due to movement artifacts).Thus, n = 86 participants (MPH first group: n = 41; MPH second group: n = 45) were included in the feature importance analyses.range 20-31 years *Note.Age is given as mean standard ± standard error of the mean (SEM).Low, medium and high refers to the three MPH dosage groups, respectively.Participants who had to be additionally excluded from the feature importance analyses (FI) are indicated in the brackets, where "n" refers to the last stated number.Regarding the educational background of our participants, "lower secondary/commercial school" implies 9 years of school education, whereas a "upper secondary degree" is achieved by completing 12-13 years of school education and passing a final exam.
Previous work suggested a potential influence of level of education and age on the performance in cognitive tasks as well as resting state functional connectivity measures (Montemurro et al., 2023).Consequently, we statistically compared the age and educational level among the low, medium and high MPH dosage groups included in all behavioral and neurophysiological analyses (n = 101).A one-way ANOVA revealed no significant difference in either educational background or age between the dosage groups (education: F(2,98) = 0.122; p = 0.885; age: F(2,98) = 0.407; p = 0.667).Given homogeneity of groups, neither educational background nor age were considered as covariates in the subsequent analyses.

Experimental design and methylphenidate administration
We used a double-blind MPH/placebo crossover design.The study consisted of one baseline 1 H-MRS measurement and two experimental EEG sessions spaced by seven days.Employing stratified randomization, participants were stratified by sex to establish homogenous subgroups with an evenly balanced sex ratio.Consequently, pseudorandomization, as defined by the between-subject factors of MPH dose (low, medium, and high dose) and appointment sequence (MPH on the first appointment vs. MPH on the second appointment), resulted in a balanced gender ratio within and between each subgroup.Other than that, the group assignment was random and double-blind.On one of the appointments, participants received in a double-blind fashion the respective MPH dose (low: 0.25 mg, medium: 0.50 mg, or high: 0.75 mg per kg body weight) and a lactose placebo on the other.The dosages were based on low (0.25 mg/kg) and medium (0.50 mg/kg) doses used in previous studies of our group (e.g.(Beste, Adelhöfer, et al., 2018) ), as well as an equal metric distance to obtain a high group (0.75 mg/kg) slightly below the recommended maximum dose of 0.80 mg/kg.The experiment started approximately two hours after MPH/placebo administration, as MPH plasma levels peak around 1-3 h, and maximum drug concentration occurs at about 2 h after oral administration (Challman & Lipsky, 2000).

1 H-MRS data acquisition and processing
All experimental MRI and MRS data were acquired with a Siemens 3T Prisma scanner (Siemens Healthineers, Erlangen, Germany) using a 32-channel (receive-only) head rf coil.The concentrations of GABA+ (γ-aminobutyric acid and macromolecules) in the striatum, SMA, and ACC were examined using 1 H-MRS.For exact voxel placements structural images were obtained using a high-resolution 3D T1weighted sagittal Magnetization Prepared Rapid Gradient Echo (MPRAGE) sequence (1 mm isovoxel, TE/TR/TI: 2.29 ms /2.3s /0.9 s) and reconstructed.1H-MRS was used to quantify brain metabolite (GABA+, tCr) concentrations in the striatum, ACC and SMA.Separate voxels of interest (VOIs) were individually positioned for each of these brain regions (see section 2.9 1 H-MRS voxel placement and data processing).In addition to the inbuilt shim routine, manual shimming was performed for each of the VOIs to further optimize spectral resolution.Shimming criterion was a full width at half maximum (FWHM) value below 20 Hz for the unsuppressed water signal.According to the recommendations of Peek et.al. (2023), to obtain GABA+, we ran a MEGA-PRESS (Mescher-Garwood point-resolved spectroscopy) sequence (echo time TE/ repetition time TR = 68/3000 ms, edit ON acquisitions = 128, edit OFF acquisitions = 128) developed by Edward J. Auerbach and Małgorzata Marjańska and provided by the University of Minnesota (Marjańska et al., 2013;Tremblay et al., 2014), based on a C2P license agreement with Siemens Healthineers AG Germany.
Finally, the ratio of GABA+ to tCr in the obtained spectra was estimated.After exporting the data ("edit on", "edit off", svs_se30_ in *.rda format) directly from the Spectroscopy subroutine of the scanner, difference spectra were calculated (using an inhouse python script) and loaded into LCModel software (v6.3-1H, copyright Stephen Provencher, Canada).Basis sets for MEGA-PRESS were delivered by Ulrike Dydak`s Lab at Purdue University (https://www.purdue.edu/hhs/hsci/mrslab/basis_sets.html).In the current study the "3T Siemens Difference Basis Set with Kaiser Coupling Constants", were based on updated values for chemical shifts and J-GABA coupling constants (Kaiser et al., 2008;Kreis & Bolliger, 2012;Near et al., 2013; these slightly differ from to the originally generated basis sets by Dydak et al., 2011, which used the values by (Govindaraju et al., 2000).Based on the "edit off" spectra from the same MEGA-PRESS measurement and using the corresponding "3T Siemens Edit-off Basis set", total creatine (tCr) reference values for GABA+ were estimated.For most reliable quantitation results, the spline baseline constraint of ("DKNTMN") within the LC-model routine was adapted.Generally, the DKNTMN parameter (minimum allowed spacing between spline knots) allows for flexibility in the baseline curve, which in turn could potentially account for a significant portion of the variance in GABA+ levels and result in underestimated values.Following a previously established procedure (Stock, Werner, et al., 2023), we optimized (testing within the range of 0.1 to 1.0) and finally adjusted the DKNTMN parameter to a value of 0.45, in order to minimize the measurement error of GABA+ (CRLB) without affecting the SNR.This approach was used for all three regions, therefore ensuring consistency with our previous research and minimizing potential biases in our GABA+ measurements.To warrant adequate data quality only spectra of final acceptable shim quality (FWHM of 3-7 Hz of the NAA peak) were used for the subsequent quantification in order.In the entire sample, we further assessed the GABA+ error estimate, as this measure typically has a higher error than Glx or the reference metabolite.Doing so, we obtained values below the 15% Cramér-Rao lower bound (CRLB or %SD) criterion for all three VOIs.

1 H-MRS voxel placement and data processing
Subsequent to the localizer, a high-resolution 3D T1-weighted sagittal Magnetization Prepared Rapid Gradient Echo (MPRAGE) sequence (1 mm isovoxel) was measured and multiplanar reconstructed for exact voxel placements.Next, a 30 × 30 × 30 mm voxel of interest (VOI) was placed in the right striatum, and a 20 × 30 × 40 mm VOI was placed over the midline to cover large parts of both the left and right ACC (including only relatively small fractions of neighboring brain regions).Additionally, a 20 × 30 × 40 mm VOI was positioned in a way that left and right (pre-)SMA are covered.The positioning of all three VOIs is depicted in Figure 2, and exemplary spectra of all three VOIs are provided in the supplementary material.

Task
A combined Simon and Go/NoGo paradigm (Chmielewski & Beste, 2017) was used.It allows to examine interference effects on response execution and response inhibition performance.The task design and structure are depicted in Figure 1.
The experimental paradigm consists of 6 blocks with 120 trials each, resulting in a total number of 720 trials.70% (504 trials) of the trials were Go trials, and 30% (216 trials) were NoGo trials.
During the experiment, a central white fixation cross, accompanied by two lateralized empty white frame boxes (possible stimulus positions), was continuously displayed on the screen.In each trial, a single letter target stimulus (i.e., a yellow letter "A" or "B") and a contralateral distractor stimulus (three horizontal white lines matching the letter stimulus in size) were simultaneously presented within the two white frame boxes for 200 ms.The inter-trial interval (ITI) was jittered between 1300 and 1700 ms.Importantly, the font of the letter stimulus indicated whether the current trial was a Go trial (normal font) or a NoGo trial (bold-italic font).
In response to the letter stimulus "A" in normal font, participants were instructed to respond with a left "Ctrl" button press with their left index finger.In response to the letter "B" in normal font, participants had to press the right "Ctrl" button with their right index finger.When the letter stimulus was presented in a bold-italic font (i.e., a yellow "A" or "B"), participants had to withhold all motor responses.The instructions applied irrespective of the side (left or righthand side) on which the letter stimulus was presented, creating the Simon effect.Trials in which the letter stimuli were presented in a spatial location matching the associated Go response hand (i.e., "A" or "A" presented on the left-hand side, "B" or "B" presented on the right-hand side) were coded as congruent trials.When the letter stimulus was presented at the spatially opposite location concerning the associated Go response hand (i.e., "A" or "A" presented on the right, "B" or "B" presented on the left), the trials were coded as incongruent trials.The combination of stimulus and location results in the following four conditions: (1) congruent Go trials, (2) incongruent Go trials, (3) congruent NoGo trials, and (4) incongruent NoGo trials.In both the Go and NoGo conditions, the proportion of congruent and incongruent trials was equal (50% each).The trial order within each block was randomized.Responses to Go condition trials within 1700 ms after stimulus presentation were coded as "correct" or "incorrect".In Go trials, a speed-up sign (i.e., "Schneller!" / "Faster!") was presented above the fixation cross in case no response was given within 500 ms after stimulus onset.Depending on the condition, trials in which no response was obtained post-stimulus presentation were either coded as Go trial "misses" or as NoGo trial "correct omissions".In NoGo conditions, any response obtained within 1700 ms after stimulus presentation was coded as a "false alarm".
At the first EEG session immediately after MPH/placebo administration, participants completed a practice run of 16 trials to familiarize themselves with the task.For the experiment, participants were seated approximately 60 cm away from a 24-inch LCD monitor, on which visual stimuli were presented over black background.The instruction was to respond as fast and accurately as possible so that the participants only rarely see the speed-up sign (i.e., "Faster!") appear.After each block (6 in total, 120 trials each), participants could take a selftimed break (i.e., to rest eyes), and resume via button press.The task took approximately 30 minutes to complete.

Statistical analysis of behavioral data
The behavioral data (i.e., response accuracy and Go reaction times) was analyzed using mixedeffects ANOVAs using SPSS version 29.0.0.0 (IBM Corp., Armonk, N.Y., USA).The "experimental intervention" (MPH vs. placebo), "condition" (Go vs. NoGo), and "congruency" (congruent vs. incongruent trials) were used as within-subject factors.The between-subject factors were "MPH dose group" (low vs. medium vs. high) and MPH/placebo sequence (MPH on first appointment vs. MPH on second appointment).Significant main effects or interaction effects were examined with post hoc ANOVAs and/or post-hoc t-tests.When a participant's behavioral accuracy was below chance level (< 50%) in two or more task conditions, that case was marked as an outlier, and excluded from the behavioral analysis as well as subsequent analysis of the neurophysiological and 1 H-MRS data.

EEG recording and analysis
EEG signals from 60 equidistant Ag/AgCl electrodes were recorded with a "QuickAmp" amplifier (Brain Products GmbH, Gilching, Germany) and the "BrainVision Recorder" software (Version 2.2), according to previously published methods (Wendiggensen et al., 2022 & other publications from our research group).For this, a ground electrode was positioned at the coordinates θ = 58, φ = 78, and the reference was positioned at Fpz (θ = 90, φ = 90).EEG signals were initially recorded at a sampling rate of 500 Hz, while electrode impedances were kept below 10 kΩ.The EEG preprocessing was performed using the Automagic toolbox (Pedroni et al., 2019) and the EEGLAB toolbox (Delorme & Makeig, 2004) in Matlab 2020a (The MathWorks Corp.).First, EEG data were downsampled to 256 Hz, and flat channels were removed (i.e., channels that showed activity below 5 µV for more than 5 sec).The remaining channels were then re-referenced to an average reference.Subsequently, the PREP preprocessing pipeline (Bigdely-Shamlo et al., 2015) was applied.The PREP pipeline removes line noise (for data recorded in Europe: 50 Hz) using a multi-taper algorithm, and after removing contaminations by noisy/bad channels (using high and minimum variance criterion), a robust common average reference is applied.EOG artifacts were removed using a subtraction method (EOG Regression; Parra et al., 2005).Then, the EEGLABs pop_eegfiltnew() pipeline was used to apply a high pass filter (cutoff frequency: 0.5 Hz) and low pass filter (cutoff frequency: 40 Hz), the filter order was estimated by default.To detect remaining artifactual source components in the data, the Multiple Artifact Rejection Algorithm (Winkler et al., 2011), which automatizes the process of independent component analysis (ICA), was applied.For the ICA the data was temporarily high pass filtered with 1 Hz, this option was not applied to the final pre-processed data.In the final step, removed/missing channels were interpolated using a spherical method.Considering all EEG data sets, on average 10.15 channels ± 3.31 (SD = standard deviation) were interpolated.Participants were excluded from all subsequent analysis steps (including behavioral and 1 H-MRS analysis), if the number of interpolated channels was 2 SDs above the average on both appointments (this was labelled as insufficient EEG signal quality in the exclusion decisions section).After visual inspection of each dataset, the preprocessed data was imported into Matlab (Version R2020a; The MathWorks Inc., MA, United States) for further analysis using the FieldTrip toolbox (Oostenveld et al., 2011).The EEG data was segmented into the four different trial conditions (congruent/incongruent Go and NoGo trials), and locked to the onset of the target letter stimulus.Only correct responses (Go conditions) and omissions (NoGo conditions) were segmented into epochs of 4 s (from -2000 ms before target onset to 2000 ms after).The time-frequency (TF) analyses of the Go conditions were conducted using Morlet wavelets with a width of 5 Gaussians and a Hanning taper.For the Go conditions, the average power over the theta (4-7 Hz) frequency band was calculated for each electrode and time point.Analyzing all electrodes ensures a comprehensive exploration of theta band oscillations across the scalp, reducing the risk of overlooking potentially relevant cortical interactions and/or specific spatial patterns.To investigate the interference effect in Go trials, TBA was statistically compared between congruent and incongruent Go trials.For this, channel-wise false discovery rate (FDR)-corrected t-tests (Benjamini & Hochberg, 1995) were calculated for every time point between 0 and 1000 ms relative to the stimulus onset.A total of four independent / separate comparisons were conducted for MPH first and MPH second group, as well as for the placebo and MPH appointments within each respective group.

Source estimation and beamforming analysis
The neuroanatomical sources that reflect the interference effect in Go trials (i.e., the congruent minus incongruent difference) in the theta band during response execution were estimated following a previously established procedure (Wendiggensen et al., 2022).Dynamic imaging of coherent sources (DICS) beamforming (Gross et al., 2001) was applied to identify neuroanatomical sources of substantial differences between congruent and incongruent Go conditions in the frequency domain.The source localization results were projected onto an equally spaced 0.5 cm grid created from the forward model template provided by the FieldTrip toolbox, which is based on the standard MNI (Montreal Neurological Institute) space.Power in the theta band was extracted for the period following the presentation onset of the letter stimulus in Go trials.For the theta source power difference, a contrast between congruent and incongruent Go trials was calculated and normalized on the total power of the two conditions as a ratio (Mückschel et al., 2016): Next, clusters of TBA were identified by applying the Density-Based Spatial Clustering of Applications with Noise (DBSCAN; Ester et al., 1996) algorithm as employed in Matlab, which is comparable to previous studies (Adelhöfer & Beste, 2020;Wendiggensen et al., 2022).The DICS beamforming results were restricted to negative ratios, indicating that theta power was higher in incongruent Go trials than in congruent Go trials.Thus, the negative top 1% of the power distribution in the interference/Simon effect ratio within gray matter regions in the Automatic Anatomical Labeling (Tzourio-Mazoyer et al., 2002) atlas were submitted to the DBSCAN restricting the analysis to the voxels with the largest negative differences between congruent and incongruent Go trials.The neighborhood search radius epsilon was set to 1.5 time the edge length of each voxel, to detect neighboring voxels.The resulting clusters were visually inspected based on the anatomical region and the number of voxels (cluster size).Further analyses were restricted to these ROIs.

Feature importance (𝐹𝐼) analyses
In order to evaluate the interrelation between baseline GABA+ concentrations and the neurophysiological activity of the clusters in the theta band, a machine learning approach was used.Machine learning has made it possible to identify potentially complex interrelation between a multitude of predictors (here 1 H-MRS data) and outcome (here theta band activity) variables.Especially, Artificial Neural Networks (ANN) as nonlinear regression tools attracted a great deal of attention, as they are data-driven models without requiring prior knowledge of the input data distribution, and were shown to be universal approximators (Hornik et al., 1990;Kim & Adali, 2002).For a neural network to serve as a nonlinear regressor in the present study, the network's input features are the baseline GABA+ concentrations in the striatum, SMA, and ACC.Based on the input features the neural network then attempts to estimate the theta band power of the respective neuroanatomical clusters.However, ANNs are also known to be "black box" since they are not inherently transparent about how parameters are learned in predicting outcomes.Yet, there are strategies to gain interpretability of neural networks (Nogueira et al., 2021;Pearl et al., 2021).One technique to identify the importance of input variables (predictors) in contributing to the prediction of the output variables is "one-at-a-time" feature importance analysis.In this technique the impact of changing the values of each of the chosen inputs in turn is evaluated (Chen et al., 2022;Lundberg & Lee, 2017;Winn-Nuñez et al., 2023).
Based on this method, a "Feature Importance" () score,   , is defined which quantifies the importance of i th input in predicting the j th output.It measures the amount prediction error that is changed when an i th input is removed, but the rest of the system maintains the same.
Where  |=0 is the prediction error of the j th output when i th input is removed, and   is the prediction error when all input variables are included.Here we used a Radial Basis Function (RBF) network with a single hidden layer and 7 neurons.To assess the generalizability of the model, the leave-one-out cross validation procedure has been implemented.This includes discarding one data sample, and training the network based on the remaining data.The discarded sample is considered as test data; the performance of the network in power estimation has been evaluated based on the MSE criterion.Specifically, the feature importance () score of GABA+/tCr (in VOIs) was defined for TBA (in corresponding clusters) of each group (MPH first and MPH second) and appointment (placebo and MPH) separately.
A brief overview of the applied methods and analyses Figure 3 (all details are provided in the methods section).

MPH dose group effects
The ANOVAs for both accuracy measures and reaction times did not reveal main effects (all F < .342;all p > 0.711) nor interaction effects including the between-subject factor MPH dose group (all F < 2.826; all p > 0.064).

Neurophysiology 3.2.1 Sensor level results
For both the MPH first and MPH second group, the channel-wise FDR-corrected t-tests comparing the theta band power between congruent and incongruent Go trials revealed significant differences (p <.05) for multiple electrodes.The FDR-corrected t-tests were conducted for each time point between 0 -1000 ms relative to the stimulus onset.The results are summarized in Table 3. Corresponding topographic plots for each time-window of significant differences in the theta band for each subgroup (MPH first and MPH second) and appointment (MPH and Placebo), as well as additional information on the magnitude of effects for the theta power differences for the contrast of interest are provided in the supplementary tables 1 and 2, respectively.In Figure 6, the between-condition contrast (Go congruent minus incongruent) for the theta band power are illustrated in time-frequency plots (over fronto-central electrodes) for each group (MPH first and MPH second) and appointment (MPH and Placebo) in the time window between 0 -1000 ms relative to the stimulus onset.Additionally, topographic plots for each time-window of significant differences in the theta band are provided for each subgroup and appointment in the supplementary material.

Baseline GABA+ and tCr concentrations across groups
Independent samples t-test comparing the GABA+ and tCr concentrations within the ACC, SMA and striatum revealed no significant differences in baseline levels between the MPH first and MPH second group.Mean baseline levels of GABA+ and tCr in each VOI and both groups are provided in Table 5.Overall, the obtained GABA+ and tCr levels and the resulting ratios (ACC: M=.186 ± .002;SMA: M=.223 ± .003;striatum: M=.208 ± .003) are well within the range of the previously reported concentrations in healthy adult samples (Mikkelsen et al., 2017;Quetscher et al., 2015).

Feature importance (𝐹𝐼) analyses
The results of the  analyses are summarized in Table 6, and visualized together with the respective neuroanatomical cluster in Figure 7 a and b.

Placebo
Lingual gyrus, Calcarine sulcus negative difference Superior frontal area, Supplementary Motor Area negative difference Table 6.Feature importance score ( score): predictive impact of GABA+/tCr from three VOIs on theta band power *Note: Summarized are the feature importance scores ( score) [including 95 % confidence intervals (CI)] for each group and appointment, for the predictive impact of GABA+/tCr from the three VOIs (ACC: Anterior cingulate cortex, Striatum and SMA: Supplementary motor area) onto theta band activity (TBA) modulations in the respective neuroanatomical cluster.The higher the  score, the larger is the predictive impact of GABA+ of a VOI for TBA modulations in a neuroanatomical region.

Discussion
In the current study, we investigated the predictive interrelation of amino acid transmitters (GABA+) in control-relevant regions (assessed with control-associated theta band activity (TBA), which was modulated with different degrees of catecholaminergic enhancement (i.e., low, medium and high MPH doses).To this end, we used a Simon NoGo task and specifically focused on delineating the predictive impact of GABA+ onto TBA reflecting interference control.Using EEG-beamforming techniques, we defined functional neuroanatomical networks reflecting different degrees of catecholaminergic modulation and task experience.
Here an MPH-induced increase in interference control (as compared to placebo) was specifically found in Go trials.Generally, task familiarity decreased interference effects in our sample.In particular, performance increases due to MPH modulations were largest (i.e.we obtained the overall smallest interference effects) during response execution in case of previous learning/task experience (i.e., in the placebo appointment of the MPH second group).On a neurophysiological level, higher need for interference control and /or effort during response execution increased power in theta band frequency irrespective of MPH stimulation or task familiarity.Using feature importance analyses, it is shown that the predictive impact of GABA+ concentrations in the ACC on TBA during response execution decreases as a function of both catecholaminergic stimulation and task familiarity.Importantly, when stimulating the catecholaminergic system with MPH and/or when prior task experience was present, the predictive impact of GABA+ in the ACC decreased.Taken together, baseline GABA+ levels (especially in the ACC) appeared to be more important/predictive of theta power modulations related to interference control when the performed task was still novel and there was no catecholaminergic stimulation.Contrary to this, results for the SMA seem to indicate that GABA+ concentration in that VOI become more important for task induced TBA with increasing task familiarity.This study provides noteworthy evidence for a predictive interrelation between baseline GABA+ concentrations in control relevant brain areas and task induced theta band activity during response control.Furthermore, it underlines the importance of baseline GABA+ levels, especially in the absence of (pharmacological and learning) modulation.

Effects of catecholaminergic modulation on behavioral performance
The results showed that somewhat surprisingly, different degrees of MPH-induced catecholaminergic modulation (operationalized by low, medium and high MPH doses) did not differentially effect interference and response control performance.It is likely that prior to enhancing catecholamine levels in our sample of healthy young adults, the DA system was already close to the optimal level (Beste, Adelhöfer, et al., 2018;Linssen et al., 2014).Under such conditions, MPH has been shown to most benefit individuals whose basal DA level is furthest from optimum.Given that MPH further enhances catecholamine levels, in the current study this may have positioned them close to, or at the peak of the inverted U-shaped curve, leading to ceiling effects.Consequently, further increases (operationalized by different MPH doses) did not result in additional benefits/declined performance.Furthermore, MPH-induced behavioral benefits were reported to arise from better focus due to decreasing metabolic activity in areas that are task-irrelevant (Kapur, 2020;Volkow et al., 2008).Given our healthy young adult sample, different doses of MPH might not have led to behavioral differences scaled by dosage, because baseline catecholaminergic levels likely fell within a range that was higher than what is typically observed in clinical populations (which are showing dose-dependent effects).Another point that has been raised is the fact that previously reported dose-dependent modulations differ between cognitive domains and tasks (Linssen et al., 2014).It is therefore possible that in our homogenous sample, with likely close to optimal dopamine levels, the behavioral task at hand might have been not sensitive enough to detect subtle dose-dependent differences.
The behavioral data indicate that enhancing effects of MPH onto interference control were mostly confined to the Go condition (Linssen et al., 2012).This was reflected in overall faster and more accurate responses after MPH administration, as compared to placebo.Importantly, the results show that MPH effects on cognitive control processes during response execution varied as a function of prior task experience, which corroborates previous findings (Mückschel, Eggert, et al., 2020;Mückschel, Roessner, et al., 2020).Specifically, the interference controlenhancing effects of MPH during response execution were more prominent when participants had already been familiarized with the task (as reflected in a smaller Simon effect of the group that received MPH on the second appointment).Other than previously reported (Bensmann et al., 2019), the size of the task effects in the response inhibition condition were independent of previous task experience when stimulating with MPH.This is in line with previously reported heterogenous MPH-effects (Linssen et al., 2014), and the existence of further unidentified modulating factors.However, the most relevant differences between the group that received MPH on the first and the group that received MPH on the second appointment were confined to the response execution condition, which we will discuss in the following.

Effects of catecholaminergic modulation on TBA
Irrespective of catecholaminergic stimulation and/or prior task experience, increased need for interference control (or effort) during response execution was reflected in higher TBA (in incongruent trials), as compared to trials with less interference (i.e., congruent trials) (Cavanagh & Frank, 2014;Westbrook & Braver, 2016).For all groups (MPH first and second) and appointments (MPH and placebo), neuroanatomical sources spanning medial frontal structures and parietal and occipital cortices were revealed.Theta band oscillations originating from medial frontal regions have previously been proposed as means of implementing cognitive conflict detection and control mechanisms (André et al., 2019;Cavanagh et al., 2012;Cavanagh & Frank, 2014;Cohen, 2014;Mückschel et al., 2016;Wascher et al., 2014).
When participants received a placebo on their first appointment, occipital clusters (i.e., lingual gyrus and calcarine sulcus) and a paracentral/superior frontal cluster were associated with interference-related TBA modulations.Both the lingual gyrus and calcarine sulcus have previously been implicated in attentional functions, the former due to its role in successful stimuli recognition including pro-active preparatory processes (Cloutman, 2013;Stock, Wendiggensen, et al., 2023), and the latter reflecting attentional control shifts thereby enhancing visual stimulus processing (Yamagishi et al., 2003(Yamagishi et al., , 2005)).The response conflict reflected in TBA modulations associated with the occipital clusters may therefore reflect increased attentional (selection) processes and need for adjustment of motor control.Agreeing with previous reports, TBA modulations in the paracentral/superior frontal cluster in the current study likely reflect stimulus-related conflict monitoring processes as well as behavioral strategy adjustments (Adelhöfer et al., 2019;Cavanagh & Frank, 2014;Cohen & Ridderinkhof, 2013;Giller et al., 2020).In case participants had prior task experience on the placebo appointment, interference control-related TBA modulations were located in clusters spanning the pre-, and postcentral gyrus (PCG).On a cortical network level, the PCG has long been recognized for its role as connecting hub between (attentional) control and processing networks (Bagarinao et al., 2020;Coray et al., 2023).In line with this it has been suggested that the PCG is central for sensorimotor integration processes (Chmielewski & Beste, 2016), as well as anticipatory preparation of motor responses (Confais et al., 2012).Thus, TBA modulations in the PCG likely reflect (task familiarity induced) higher allocation of conflict processing capacities for incongruent stimulus-response trials (as compared to congruent trials), and anticipatory planning for response execution.Interestingly, when stimulating with MPH (regardless of prior task familiarity) theta band modulations reflecting the response conflict during response execution were found in closely connected neuroanatomical clusters spanning the dorsal premotor areas (i.e., precuneus and (pre)SMA).In line with previous reports, modulations in the precuneus may reflect MPHinduced increases in top-down allocation of cognitive resources/suppression of precuneal cue-reactivity (Cavanna & Trimble, 2006;Jafakesh et al., 2022;Koyun et al., 2023;Stock et al., 2016), requiring more suppression control (TBA) in incongruent trials.Theta band activity in the (pre)SMA however, may reflect the MPH-induced shift/allocation of attention to conflicting response execution trials which require altered action (Pauls et al., 2012).Of note, the SMA is part of a lager network typically involved in performance monitoring and flexible adjustment of behavior in adults (Ridderinkhof et al., 2004;Rubia et al., 2011).Therefore, the MPHinduced TBA modulations associated with the (pre)SMA likely reflect stimulus-response conflict anticipation and readiness for flexible motor response adjustments (Manza et al., 2016).Of note, theta oscillations were shown to be augmented by GABA levels in both humans and nonhumans (Orzeł-Gryglewska et al., 2010;Rowland et al., 2013), indicating further modulating factors underlying the observed TBA effects.

The predictive impact of GABA+ concentrations on task induced TBA
Using feature importance analyses, we revealed interrelation of features from the macromolecule level (GABA+) with the task-induced theta band oscillations under different degrees of catecholaminergic stimulation and task familiarity.Our findings show that baseline GABA+ concentrations in control relevant regions (i.e., ACC, striatum, SMA) had differential impact on predicting interference control-related theta band activity (TBA) in response execution trials.Interestingly, the results were scaled by both catecholaminergic stimulation and task familiarity.Previously, GABAergic signaling and neural activity in the striatum, SMA and ACC was shown to play crucial roles in response control and conflict processing (Beste, Mückschel, et al., 2018;Boy et al., 2010;Chmielewski & Beste, 2017;Haag et al., 2015;Mückschel, Dippel, et al., 2017;Takacs et al., 2021).In line with this, here we show that the predictive impact of GABA+ concentrations in all control relevant VOIs, especially the ACC, varied based on cognitive demands.Specifically, the predictive value of ACC GABA+ concentrations for TBA in the respective neuroanatomical clusters was largest in the absence of catecholaminergic stimulation and when no prior task experience existed (i.e., placebo on the first appointment).Both MPH and learning have been shown to induce changes in GABAergic signaling (Frangou et al., 2018;Solleveld et al., 2017).This may be a reason why in the current study, baseline ACC GABA+ levels show overall comparably less predictive impact on TBA in a system in which GABAergic signaling has been modulated.Generally, activity in the ACC has been associated with effort, demanddependent allocation of control and (conflict) monitoring processes (Chmielewski & Beste, 2017).In the absence of task experience and/or MPH stimulation, the cognitive effort needed for interference control is usually higher than when a task was performed before/or catecholamines were enhanced (Carter et al., 2000;Westbrook et al., 2020).It therefore makes sense that GABAergic activity in the ACC is more related to TBA (in regions associated with the response conflict during response execution), when the effort needed for interference control is high (Milham et al., 2003).Task experience and MPH both modulate/decrease the (perceived) effort needed to perform well (Milham et al., 2003;Takacs et al., 2021;Westbrook et al., 2020), here mirrored in comparable decreases in the predictive impact of baseline GABA+ concentration in the ACC on TBA during response execution.Of note, when stimulating with MPH these results were irrespective of task familiarity, indicating a purely catecholamine induced increase in willingness to allocate cognitive effort (Westbrook et al., 2020).Taken together, modulating the catecholaminergic system with MPH and task-practice both lift the demand off the ACC and pass it on to areas showing comparably larger involvement in non-response (conflict) related processing and motor response control during response execution (i.e., striatum and SMA), rather than effort/interference control (i.e., ACC).In line with this, one may speculate that an opposite pattern (as in the ACC) was observed for the SMA.The results indicated a trend that with learning/task familiarity, the predictive impact of GABA+ concentrations in the SMA, regardless of MPH stimulation, increased.The SMA is an area with direct connections to the motor system, thought to be implicated in the flexible adjustment of motor behavior (Mostofsky & Simmonds, 2008;Mückschel, Dippel, et al., 2017;Rowe et al., 2010).Furthermore, GABAergic signaling in the SMA has previously been associated to have greater influence on signaling/producing the suppression of prepotent responses rather than implementing it (Boy et al., 2010;Hermans et al., 2018).Therefore, as task familiarity/practice increased (i.e., second appointment of both groups), so did automatic and prepotent response tendencies, which in turn increased the need for reactive inhibition through the SMA, especially so in conflicting response execution trials (Hermans et al., 2018). in the absence of MPH stimulation and task familiarity, GABA+ concentrations in the striatum appeared to have comparably lower to no predictive impact on TBA in the calcarine sulcus during response execution.Oscillatory activity within the calcarine sulcus has previously been associated with attentional processing and was proposed to enhance processing of visual stimuli (Yamagishi et al., 2003).Agreeing with earlier accounts, in the current study a predictive impact of GABA+ in the striatum was only found in the conflicting stimulusresponse condition, that required efficient stimulus processing to perform well.

Conclusions
In summary, we show a predictive interrelation of baseline GABA+ levels in cognitive control relevant brain regions onto task induced TBA during response control stimulated with MPH.The predictive impact of the ACC, the striatum, and the SMA varied as a function of cognitive effort, underlining differential roles of these brain regions in cognitive control related processes.GABA+ concentrations in the ACC were specifically important for TBA when the need for cognitive effort was high (in the absence of MPH stimulation or task experience), implying a major role of this area in high-effort response control.The results show that baseline GABA+ levels play pivotal predictive roles in crucial brain areas influencing cognitive control and responsiveness to catecholaminergic enhancement, particularly in high-effort situations.3 for details on the respective electrodes).

Figure 1 .
Figure 1.Illustration of the volume of interest (VOI) placements.Depicted are the VOIs (from left to right): in the right striatum, the ACC and the (pre-)SMA (please note that sides are mirrored in this scanner output).Exemplary spectra of all three VOIs are provided in the supplementary material.

Figure 2 .
Figure 2. Illustration of Simon NoGo paradigm: Depicted are the possible stimulus-response configurations for the congruent/incongruent Go condition (left) and the congruent/incongruent NoGo condition (right).Congruency indicates the stimulus-response hand mapping of Go trials.In detail, congruent trials required a Go response execution on the side where the target letter stimulus was presented.In incongruent trials, the stimulus presentation side was opposite to the side of the correct Go response hand.The NoGo condition, as indicated by bold and italic target letter stimuli, required the participants to withhold any responses.Congruency was determined in an analogue fashion to Go trials.

Figure 3 .
Figure3.Schematic overview of applied methods and analyses: Depicted is a visual summary of all methods and/or analyses applied from macromolecule level, over sensor level to source level.And how we linked features from all levels by using neural networks and applying feature importance analysis.

Figure 4 .
Figure 4. Illustrations of Go and NoGo response accuracies.Depicted are the Simon Effects (accuracy in congruent MINUS incongruent trials) for Go (upper row) and NoGo (bottom row) conditions, for both placebo (left panel) and MPH (right panel) appointments separately.Boxplots of those who received MPH on the first appointment (and placebo on the second appointment) are illustrated in blue, and boxplots of those who received MPH on the second appointment (and placebo on the first appointment) are orange.The raincloud plots illustrate the data distribution (above the boxplots) and data points (below the boxplots).The "x" and the horizontal line inside the boxplots indicate the mean and median, respectively.Asterisks indicate significant differences at p < .05,and error bars represent the 95 % confidence intervals.

Figure 5 .
Figure 5. Boxplots illustrating the Go response times.Depicted are the Simon Effects (response times in congruent MINUS incongruent trials) for Go conditions for both placebo (left) and MPH (right) appointments separately.Boxplots of those who received MPH on the first appointment (and placebo on the second) are illustrated in blue and boxplots of those who received MPH on the second appointment (and placebo on the first) are orange.The "x" and the horizontal line inside the boxplots indicate the mean and median, respectively.The raincloud plots illustrate the data distribution (above the boxplots) and data points (below the

Figure 6 .
Figure 6.Time-frequency representations of the theta band power difference in fronto-central electrodes that were significant on the sensor level (see Table3for details on the respective electrodes).

Figure 7 .
Figure 7. Main results: DBSCAN cluster and predictive impact of GABA+ on TBA.Depicted are functional neuroanatomical cluster identified by the DBSCAN algorithm during response execution.The color of the cluster scales the theta power ratio [(congruentincongruent) / (congruent + incongruent)].The neuroanatomical cluster (left) and the NE plots illustrating the predictive impact of GABA+ in the ACC, striatum and SMA (right) are displayed for (a) MPH first and (b) MPH second groups for both placebo (orange) and MPH (blue) appointments.

Table 1 .
Characteristics of participants included in the subsequent analyses.

Table 2 .
Examining Step 2. Independent samples t-tests separately examining the Simon effect (% of hits in congruent trials minus hits in incongruent trials) between MPH first and MPH second group in both conditions and appointments

Table 3 .
Theta power differences for the contrast: Go congruent minus Go incongruent.

Table 5 .
GABA+ and tCr concentrations within ACC, SMA and striatum of MPH first and