Decreased thalamo-cortico connectivity during an implicit sequence motor learning task and 7 days escitalopram intake

Evidence suggests that selective serotonin reuptake inhibitors (SSRIs) reorganize neural networks via a transient window of neuroplasticity. While previous findings support an effect of SSRIs on intrinsic functional connectivity, little is known regarding the influence of SSRI-administration on connectivity during sequence motor learning. To investigate this, we administered 20 mg escitalopram or placebo for 1-week to 60 healthy female participants undergoing concurrent functional magnetic resonance imaging and sequence motor training in a double-blind randomized controlled design. We assessed task-modulated functional connectivity with a psycho-physiological interaction (PPI) analysis in the thalamus, putamen, cerebellum, dorsal premotor, primary motor, supplementary motor, and dorsolateral prefrontal cortices. Comparing an implicit sequence learning condition to a control learning condition, we observed decreased connectivity between the thalamus and bilateral motor regions after 7 days of escitalopram intake. Additionally, we observed a negative correlation between plasma escitalopram levels and PPI connectivity changes, with higher escitalopram levels being associated with greater thalamo-cortico decreases. Our results suggest that escitalopram enhances network-level processing efficiency during sequence motor learning, despite no changes in behaviour. Future studies in more diverse samples, however, with quantitative imaging of neurochemical markers of excitation and inhibition, are necessary to further assess neural responses to escitalopram.

. Results from independent sample t-tests assessing potential group differences for participant age, body mass index (BMI), Follicle stimulating hormone (FSH) levels, Luteinizing hormone (LH) levels. Mean ± Standard Deviation (M ± SD).

Demographic
Placebo (M ± SD) Escitalopram (M ± SD) t value p value www.nature.com/scientificreports/ Behavioral performance. All participants completed 5 days training on the SPFT, including a sequence learning condition (Sequence Learning) and a simpler motor learning condition (Simple Learning). A Rest condition with no behavioral input interspersed performance of the Sequence Learning and Simple Learning conditions and was retained for use in fMRI model specification. Comparison of motor performance between the Sequence Learning and Simple Learning conditions yielded a significant effect, with performance differing between conditions (see Supplemental Table 1; Supplemental Fig S3). Despite significant improvements in performance for all participants over time, we did not observe an effect of escitalopram on behavioral measures of sequence-specific motor learning. Interpretation of these findings are reported elsewhere 14 .
Group comparisons of functional connectivity changes over time. For the thalamus seed region, we observed a significant group × time interaction when comparing the PPI Learning contrast (i.e. the difference in connectivity between the Sequence and Simple Learning conditions) between steady state and baseline, with a significant decrease in connectivity within the escitalopram group, not mirrored in the placebo group. Specifically, we observed a decrease of the PPI Learning contrast between the thalamus and cortical brain regions of the motor system including bi-hemispheric primary and premotor regions ( Table 2; Fig. 1). Non-parametric permutation tests replicated this finding with a significant decrease of PPI thalamo-cortico connectivity in bilateral premotor and parietal regions (Table 3) within the escitalopram group. No significant interaction was observed when comparing single dose to baseline or single dose to steady state. Analyses of the PPI Learning contrast with the other seed-regions within the M1, SMA, cerebellum, putamen, dPMC, and dlPFC, revealed no significant group × time interaction in our statistical approach using nonparametric permutation tests including correction for multiple comparisons. As a result, no post-hoc tests were performed for these seed regions. For the PPI Motor contrast (the comparison between the combined Sequence and Simple Learning conditions to the Rest condition), no significant group × time interaction was observed for any seed-region. Thus, no post-hoc nor non-parametric analyses were performed for this contrast.
Post-hoc tests for the PPI Learning contrast with the thalamus seed region. Post-hoc paired comparisons for the escitalopram group yielded a significant change in the PPI Learning contrast from baseline to steady state between the thalamus and bilateral primary motor and parietal regions (Table 2-Post-hoc; Table 2. 2 × 2 analyses comparing group changes in the PPI Learning contrast over time show a significant decrease in connectivity from the thalamus seed region to the bilateral primary and premotor regions in response to escitalopram (Interaction). Post-hoc analyses reveal significant decreases in PPI connectivity between baseline and steady state within the thalamus seed region of the escitalopram group (Post-hoc). Results shown at a p < 0.05 Family-Wise Error corrected (FWE corr ) on the whole brain voxel level. Significant results with < 5 voxel extent not shown. L left, R right, peak MNI Montréal Neurological Institute coordinates. www.nature.com/scientificreports/ Figure 1. Orthogonal brain slices of PPI Learning contrast showing a thalamo-cortico connectivity decrease in response to escitalopram. Significant brain connectivity decrease in the PPI Learning contrast were observed after 7 days of 20 mg escitalopram administration (Difference-top row). In contrast, no task-related connectivity changes were observed within placebo (Difference-bottom row). A 2 × 2 interaction analysis revealed a significant group difference over time (Interaction). Results were obtained with p < 0.05 with FWE correction at the voxel level (yellow). Also shown is an uncorrected cluster forming threshold of p < 10 -5 with an extent of ≥ 20 voxels (red), and results from a non-parametric analysis of the interaction contrast (green) corrected for multiple comparisons (p < 0.0012). Estimates of effect size of the PPI Learning contrast. Contrast estimates for the PPI Learning contrast in the escitalopram group show a gradual decrease in intensity associated with the Sequence Learning condition over time. Specifically, results show a linear pattern of decrease in the PPI Learning contrast for the sequence motor Learning condition from baseline to steady state in each motor region resulting from our group × time interaction analysis. Contrast estimates for the placebo group indicate an increase in thalamocortico connectivity from baseline to steady state, however, these changes were not significant in post-hoc paired comparisons within the placebo group (Fig. 2).
Associations between thalamo-cortico connectivity changes and escitalopram kinetics. Consideration of plasma escitalopram levels at baseline and at steady state yielded a significant negative correlation between the PPI Learning contrast and escitalopram plasma levels, with higher plasma escitalopram levels being associated with greater decreases in connectivity ( Table 4). Extraction of the PPI Learning contrast estimates showed that higher plasma escitalopram levels are significantly associated with a lower PPI connectivity between the thalamus and the right superior frontal gyrus, the left SMA, the right M1, and the right superior parietal lobule (Fig. 3). Use of delta images (subtracting the PPI Learning contrast at steady state from baseline) yield a significant correlation between PPI connectivity change and steady state plasma escitalopram levels in bilateral cortical and subcortical regions (Supplemental Table 2; Supplemental Fig S1). Identification of high and low connectivity profiles within the escitalopram group, for each region observed in the correlation analysis, shows a significant interaction between baseline PPI connectivity and peripheral plasma escitalopram levels. This interaction shows that high baseline PPI connectivity is associated with a greater decrease at steady state (Fig. 3

, Supplemental Tables 3-6).
Associations between thalamo-cortico connectivity changes and behavioural outcome. Correlation analyses investigating a potential relationship between the PPI Learning contrast and mean sequencespecific behavioral outcomes do not yield a significant group difference when comparing escitalopram to placebo.
Motion effects inside the MR scanner. Across groups and sessions, the mean framewise displacement (FD) was consistently below 0.36 mm. Less than 0.5% of frames from the entire study indicated single head movements by more than 1 mm. We did not observe any significant group differences in any FD motion parameter.

Discussion
In this study, we employed PPI analysis to assess the effects of 1-week escitalopram-intake on functional brain connectivity during implicit sequence motor learning. By comparing a sequential Learning condition to a Simple motor learning condition (the PPI Learning contrast), our results show that, underlying a standard behavioral performance, functional connectivity from the thalamus to bilateral premotor and primary motor regions is Table 3. Non-parametric TFCE results from significant 2 × 2 group by time interaction estimated with 5000 permutations for the thalamus seed region. Results show a significant decrease in the PPI Learning contrast from the thalamus to the bilateral premotor, parietal, and occipital regions. Results shown at a p < 0.0012 family-wise error corrected (FWE corr ) on the whole brain voxel level. L left, R right, peak MNI Montréal Neurological Institute coordinates, TFCE threshold Free Cluster Enhancement. www.nature.com/scientificreports/ significantly decreased in the Sequence Learning condition after 1-week of drug intake, compared to baseline. Additionally, we show that this decrease correlates with increases in escitalopram plasma levels between baseline and steady state, suggesting a parallel development between the degree of task-modulated connectivity decrease and the establishment of steady state escitalopram plasma levels. We did not observe any significant effect of escitalopram intake on PPI connectivity in any other seed region, relative to placebo. Moreover, we did not find any significant effect of escitalopram on any seed region in the PPI Motor contrast (i.e., comparing a combination of the Sequence and Simple Learning conditions to the Rest condition). Finally, we did not observe any significant change within the placebo group, with either the PPI Learning or PPI Motor contrasts. Our main result indicates an escitalopram-induced decrease in thalamo-cortico PPI connectivity during implicit sequence motor learning. However, rather than a decrease of functional connectivity with the Sequence Learning condition, an alternative interpretation is an increase in connectivity associated with the Simple Learning condition. This seems less likely, however, given previous findings regarding escitalopram (and its racemic compound citalopram) on functional architecture, with observations of decreased connectivity in the presence of multiple doses 17,35,36 . Moreover, functional activity responses to sequence motor learning, particularly in premotor and parietal regions 37 , are hypothesized to constitute a decrease rather than an increase following training, Figure 2. Contrast estimates from motor regions showing a gradual decreases in the thalamo-cortico PPI Learning contrast from baseline to steady state. Contrast estimates extracted from the escitalopram group motor regions at each time point show a gradual decrease in the PPI Learning contrast over the course of the administration week. Orthogonal brain slices show regions significant in a 2 × 2 flexible factorial analysis comparing groups from baseline to steady state. Results are shown at both a p < 10 -5 cluster corrected threshold (red) and p < 0.05 with FWE correction at the voxel level (yellow). Table 4. Brain regions where decreases connectivity of the PPI Learning contrast from the thalamus correlates with increased plasma escitalopram levels. Results shown at p < 10 -5 cluster forming threshold corrected with family-wise error (FWE) on the cluster level (≥ 20 voxels). L left, R right, peak MNI Montréal Neurological Institute coordinates, FEW family-wise error. www.nature.com/scientificreports/ a pattern that may extend to underlying connectivity. Considering our statistical model specification, we argue that our results represent a decrease in the Sequence Learning condition relative to the Simple Learning condition with respect to thalamo-cortico connectivity in response to escitalopram. This decrease may reflect heightened automation of performance 38 , reduced neural energy consumption 39 , or increased independence between regions 40 possibly via alterations in inhibitory tone, which has been proposed to occur in response to SSRIs 41 . Support for this explanation comes from findings of an inverse relationship between GABA and functional brain responses, both during motor learning 42 and at rest 43,44 , with decreased fMRI response corresponding to increased GABA concentrations. An alternative explanation is the modulatory effect of escitalopram intake on inhibitory serotonin 1A receptor (5HT 1AR ) binding 45 , which itself is not only thought to be a mediator of subcortical neurogenesis 46,47 but to have reciprocal interactions with action at the serotonin 2A receptor (5HT 2AR ) subtype 48 . Enhanced signaling of 5HT 2AR is hypothesized to lead to increased neural adaptability and plasticity, with acute pharmacological stimulation of these receptors leading to widespread decreases in functional network connectivity 49 , similar to the direction of change observed in this study. SSRI-induced modulation of receptor binding and action on GABAergic signaling could shift excitatory and inhibitory balance, thus allowing for improvements in neural automation of task performance, despite no difference in the behavioral performance between groups.

Region Cluster p(FWE corr ) Voxels t-value z-value
Complimentary to this interpretation is our observation of a negative correlation between connectivity change and peripheral plasma escitalopram levels. This finding indicates that higher plasma escitalopram levels at steady state are associated with greater decreases in thalamo-cortico PPI connectivity, particularly in participants with a higher baseline connectivity profile. While this variance could be attributable to trait differences in 5HT 1AR receptor density, which may determine baseline serotonin availability 50 , this finding may also be due to inherent differences in baseline connectivity. For example, baseline sensorimotor connectivity has been associated with variance in motor learning 51 while whole-brain connectivity has been observed as a predictor of responses to SSRI intake 52 . This suggests that initial connectivity strength from the thalamus seed region may reflect baseline heterogeneity in individual network responsivity, thereby influencing the degree to which escitalopram exerts its neuromodulatory effects, possibly via a floor effect in participants with lower baseline connectivity. Future studies with larger sample sizes specifically designed to test this hypothesis are needed to clarify further.
Against our initial hypotheses, however, we did not observe any significant changes in PPI connectivity when using the M1, SMA, dlPFC, putamen, cerebellar or dPMC as seed regions. This may be due to a higher density of serotonergic binding sites in thalamic regions relative to the cortex 53 and particularly, the cerebellum 54 . While transporter occupancy in the putamen and thalamus following SSRI-intake is comparable 55 , thalamic nuclei uniquely exhibit behaviorally relevant and reciprocal interactions with cortical motor regions, in which sustained bidirectional thalamo-cortico signaling modulates preparatory motor activity 56,57 . Moreover, these thalamo-cortico neural circuits operate on a cell-specific level, with excitatory loops, in particular, mediating Inclusion of plasma escitalopram levels at both baseline and steady state as a covariate of interest shows a significant negative correlation between task dependent differences in functional thalamo-cortico connectivity in multiple motor regions (yellow). Results indicate a greater decrease in task-based thalamic connectivity with greater levels of escitalopram at steady state. Overlaid in red are the clusters from the significant group by time interaction (Fig. 1 -Interaction). All results are presented at a p < 10 -5 cluster corrected threshold with a minimum cluster extent of 20 voxels. BOLD blood-oxygen-level-dependent. www.nature.com/scientificreports/ this interaction 58,59 . This distinctive role of the thalamus as a central hub of motor-related neural circuitry 60 may contribute to the specificity of our findings to this region. Additionally, we did not observe any significant changes in connectivity when comparing single dose with either baseline or steady state for any seed region. As prolonged SSRI-intake is required to downregulate auto-receptor feedback and to reduce serotonin reuptake 61 , early variance is 5HT 1AR responses to escitalopram may mitigate SSRI-induced alterations in PPI connectivity at single dose. Future studies with larger sample sizes may be needed to detect more subtle changes in PPI motor connectivity at early phases of SSRI-intake. We also did not observe a significant group difference in a correlation between thalamo-cortico connectivity and the mean behavioral performance, an observation that is consistent with our behavioral findings which also did not show any difference in performance between groups. While this outcome could be a result of statistical power, or possibly a non-linear relationship between brain connectivity and behavioral outcomes, the use of an implicit sequence motor task may also be a contributor. Specifically, findings in unilateral stroke patients have shown that sequence learning undertaken with prior knowledge of sequences leads to improved performance when compared to extended practice of implicit learning alone 62 . This approach differs to the current study, in which participants implicitly learned the required sequence of pinch contractions and received no feedback regarding their performance. It is therefore possible that escitalopram may not modulate performance of implicitly-learned sequences, leading to a comparable learning curve between groups. In this case, the observed changes in thalamo-cortico connectivity may not necessarily correlate with this standard learning curve. Interestingly, we also did not observe any effect for the PPI Motor contrast. One possible explanation for this absence of evidence may be a possible loss of sensitivity resulting from the combination of the Sequence and Simple Learning conditions, which reflect different forms of motor performance. Moreover, the Rest condition was interleaved between motor conditions, possibly masking a "true" baseline, given the comparatively short duration of the condition. Specifically, this interleaved fashion could lead to a variable signal in Rest, possibly diminishing detectable effects of escitalopram when compared to this generalized motor condition. Studies with tasks with more explicit sequence learning and a more distinct Rest condition could provide more clarity on this question.
There are study limitations, however, that should be considered. First, we cannot rule out that our findings may be influenced by a difference in complexity between task conditions. Given that both the Sequence and Simple Learning conditions reflect motor learning, it is possible that the heightened difficulty associated with the Sequence Learning condition may draw on other domains not related to motor learning, such as attention, heightened visual perception, and general cognitive strategizing. The SPFT, however, was designed to exhibit a minimal learning effect for the Simple Learning condition, and recent preliminary findings have shown differentiated functional connectivity patterns when comparing these conditions at rest 63 , suggesting a distinction in the neural processing of each condition. Secondly, we also cannot rule out generalized SSRI effects that may be unrelated to the task. Even so, we employed an undirected modelling of task conditions 22 and observed brain regions known to be involved in task processing (specifically, the bilateral premotor cortices). Additionally, taskindependent thalamic connectivity has also been shown to be locally increased in response to escitalopram 18 , which differs to our observations during motor learning. Secondly, our sample consisted only of healthy female participants on oral contraceptives, making these findings difficult to generalize to male, naturally cycling female, healthy older, and stroke participants. However, this was a deliberate restriction to control for age, sex, pathology, and sex-hormone fluctuations on both escitalopram responsivity and motor learning. Third, we cannot comment on prolonged SSRI effects as intake took place across only 1-week. Future studies over several weeks are required to investigate long-term serotonergic modulation of PPI connectivity.
Fourth, we acknowledge that we have acquired a peripheral measure of escitalopram concentrations. As such, our findings of a relationship between PPI connectivity and plasma escitalopram levels should be interpreted with some caution. Future studies could address this with direct measures of the neural pharmacokinetics of SSRIs. Fifth, we acknowledge the relatively short training regimen and the early plateau in behavioral performance, which may limit the interpretability of these findings. This choice of 5 days, however, was made based on previous studies on sequence motor learning in health 26,29 and in patients 20 and to allow for escitalopram levels to reach a steady state plasma level, which is thought to occur after approximately 1-week of intake 64 . Finally, fMRI cannot assess the cellular effects of escitalopram on brain connectivity. Quantitative neuroimaging methodologies such as 1 H MR spectroscopy estimates of glutamate and GABA, which have successfully imaged pharmacological interventions in thalamic regions and motor responses in cortical regions 65,66 respectively, are needed to further discuss these interpretations.
In conclusion, we present results from the first study on the effects of 1-week SSRI-administration on functional connectivity during implicit sequence motor learning in health. While our results do not indicate an effect on behavioral performance, they do show that 1-week administration of escitalopram decreases thalamo-cortico connectivity when comparing two different levels of difficulty on a sequential motor learning task, and that these decreases are associated with peripheral levels of plasma escitalopram. As recent clinical trials in stroke patients assessed only behavioral responses to SSRIs, these findings highlight the importance of considering changes in fundamental brain architecture in response to combined sequence motor training and SSRI-intake. Given that motor learning is a viable model for post-stroke motor recovery 67 , our findings could further assist in advancing neural models of patient rehabilitation.

Methods
Sample. Participants were right-handed, 18-35 years old, with a body mass index (BMI) between 18.5 and 25 kg/m 2 , and female on oral contraceptives for ≥ 3 months (to control for sex differences in SSRI responsivity 68 , motor learning 13,69 , and sex-hormone effects on serotonin transporter density and serotonergic transmission 70 ). Exclusion criteria were medication use, tobacco or alcohol intake, positive drug or pregnancy tests, professional www.nature.com/scientificreports/ musicianship or athleticism, an average of > 2 h video gaming per week, contraindications for MRI or SSRIintake, and history of neurological or psychiatric illnesses. A physical examination, electrocardiogram recording, and hormonal analyses were performed prior to enrolment. Seventy-one participants enrolled, 60 of whom were included in analyses (escitalopram n = 29, placebo n = 31). During the experiment, 6 (escitalopram n = 4) participants voluntarily discontinued while data quality assessment excluded 2 participants (escitalopram = 1) due to head movement. Two further participants (escitalopram n = 1) were excluded due to MRI-related quality concerns. Finally, 1 placebo participant was excluded due to a pre-analytical error in plasma sample acquisition (Supplemental Fig S2). All participants provided written informed consent and received compensation.
Design. The Ethics Committee of the Faculty of Medicine at Leipzig University approved all experimental procedures (approval number 390/16-ek), as governed by the Declaration of Helsinki, 2013. Seed regions were defined as a secondary outcome measure to a previously published study 14 . Following a baseline fMRI and motor learning measurement, participants were randomly assigned to orally receive either 20 mg escitalopram or placebo (mannitol/aerosol) from indistinguishable sequentially numbered containers, at fixed times, for seven consecutive days in a double-blind design. Task-fMRI was initially measured at baseline, prior to escitalopram or placebo intake. Following a baseline measurement, participants were randomized to either the escitalopram or placebo condition using an independent block randomization with a 1:1 allocation ratio, conducted by the Pharmacy of the University Clinic at Leipzig University. Task and fMRI measurements were subsequently performed after single dose escitalopram or placebo intake (single dose), and after 7 days of administration (steady state) (Fig. 4A). All measurements took place 3-4 h after intake, to allow for escitalopram to reach peak serum concentrations 64 . Both the experimenter and all participants were blind to group allocation.
Sequence motor learning paradigm. We administered the sequential pinch force task (SPFT) during baseline, single dose, and steady state fMRI measurements with 2 additional behavioral assessments on days five and six of the experiment. Prior to entering the scanner, participants were verbally instructed on task requirements, which involved controlling the height of a yellow bar via the right-hand thumb and index finger, with stronger contractions leading to a greater increase in the bar's height. Participants were asked to match their pinch strength to that of a computer controlled blue bar, the height of which rose and fell independently. No further verbal descriptions of either the Sequence Learning or Simple Learning conditions were given. Thus, participants were unaware of the specific requirements of differing task conditions and their interchanging onsets.
Prior to the commencement of the measurements, participants performed an initial pinch, while positioned in the scanner. This served two purposes; to ensure participants understood task requirements and to act as an indicator of individual pinch strength. Task parameters were subsequently attenuated to account for this individual strength (measured in kg) to avoid variance associated with task difficulty. Motor learning was assessed with 2 conditions: a "Simple Learning" condition in which the blue bar moves sinusoidally, and a "Sequence Learning" condition in which the blue bar moves in a sequential pattern (stable across training sessions). Both conditions were prompted with on-screen text. A "Rest" condition, with no behavioral input in which the blue and yellow bars remained stationary, interspersed motor performance to avoid fatigue (Fig. 4B). Each task condition was repeated 5 times and participants received no feedback regarding performance.
Analysis of screening variables. We used independent samples t tests using the R statistical programming language 71 to assess potential group differences in age, BMI, and downregulated hormonal profiles. We quantified plasma escitalopram levels via a liquid chromatographic method 72 using blood samples obtained at single dose and steady state from each participant.
Analysis of sequential pinch force task performance. We used linear mixed effects modelling to analyze behavioral performance on the SPFT for the full sample (n = 60). To determine sequence-specific performance with regard to the Sequence Learning condition, we employed a linear model using the 'lmer' in R with the Lag (temporal deviation between the height of the computer-controlled and participant-controlled bars) as the outcome measure and time and condition (Sequence Learning/Simple Learning) as factors. We also applied a between-group linear modelling analysis, again with Lag as the outcome measure, and group and time as factors, to assess potential group differences in sequence motor learning behavior. A full description of these analyses and their results are reported elsewhere 14  www.nature.com/scientificreports/ Within the first level analysis, PPI contrasts were generated for each participant with the psycho-physiologic interaction module in SPM12. We used a general linear model (GLM) to specify two PPI contrasts: (1) the PPI Learning contrast comparing the difference in connectivity between the Sequence and the Simple Learning conditions whereby Sequence Learning connectivity patterns are greater than Simple Learning connectivity patterns. The Simple Learning condition, representing a simplistic contraction and release of the pinch device, was thus used to define a baseline of PPI connectivity to which sequence motor conditions could be compared. (2) The PPI Motor contrast, which combined the Sequence and the Simple Learning conditions into a single motor condition to compare general motor execution to the Rest condition.
To specify the physiological component for each contrast, we selected the left hemisphere thalamus, primary motor cortex (M1), supplementary motor area (SMA), putamen, cerebellum, dorsal premotor cortex, and the right hemisphere dorsolateral prefrontal cortex (dlPFC) as 3 mm spherical seed regions of interest (Fig. 5). Using a t-contrast, we modelled the interaction between task conditions and the physiological neural time series by setting the psychological and physiological variables to zero and the interaction term to one. We approximated this interaction term with a deconvolution of the representative hemodynamic response function in order to normalize both psychological and physiological epochs to a uniform temporal resolution. All resulting contrast images for baseline, single dose and steady state were subsequently processed in second-level analyses.
Second level analyses. Following PPI contrast specification, we employed a 2 × 2 flexible factorial model in SPM12 comparing group (escitalopram, placebo) changes in the PPI Learning contrast between: (1) baseline to single dose, (2) baseline to steady state, and (3) single dose to steady state, for each seed region. Within each model, we specified a 'subject' factor for pairwise analysis, a 'time' factor (time 1, time 2), and a 'group' factor (escitalopram, placebo). We specified t-contrasts to test for an interaction between the factors group and time, using both a positive and negative contrast to test for both directions of the interaction. Significant results were obtained using p < 0.05 family-wise error (FWE) corrected at the whole-brain voxel-level. In order to detect potential type II errors, we additionally employed an uncorrected threshold of p < 10 -5 with a minimal extent threshold of 20 voxels. In addition, all interaction analyses that yielded a significant result were then validated with a non-parametric permutation analysis (5000 permutations) using the Threshold-free Cluster Enhancement (TFCE) toolbox in SPM12. All non-parametric tests were considered significant at a Bonferroni-corrected FWE α-threshold of p < 0.0012 (0.05/42-for the total number of bi-directional interaction contrasts employed across all seed seven regions and timepoints) to account for multiple testing. Only significant interactions replicated with non-parametric permutation tests were retained for post-hoc analyses.  showing a significant group × time interaction, we extracted beta values obtained by the parameter estimation from each significant region within the escitalopram group (n = 29). Additionally, we extracted betas from matching coordinates of the placebo group, for each of baseline, single dose, and steady state, in order to show effect size changes over the course of escitalopram administration, compared to placebo.
Thalamo-cortico correlational analyses. We also assessed whether the observed changes in the thalamo-cortico PPI Learning contrast were associated with (1) plasma escitalopram levels within the escitalopram group and (2) sequence motor learning performance in the full sample. Here, we specified two separate flexible factorial designs, one for each analysis. In the first, plasma escitalopram levels were entered as a variable of interest 75 , while mean behavioral performance for each participant in the full sample were entered in the second. We assessed both directions for a potential correlation with thalamo-cortico connectivity. Analyses yielding a significant result were validated with a second approach by correlating baseline-steady state delta images (computed using the 'fslmaths' subtraction command) and entering these images in a third design to assess the correlation between the rate of change in thalamo-cortico connectivity and steady state plasma escitalopram levels.
Interaction between baseline connectivity and plasma escitalopram levels. To assess individual variation in baseline connectivity on the association with plasma escitalopram levels, we performed an exploratory median split within the escitalopram group (n = 29) for each significant region observed in correlation analyses. Here, we specified two groups: high and low baseline connectivity, with one median split for each significant region. We specified an interaction term in SPM12 with the factors of profile (high, low) and the covariate plasma escitalopram levels, at baseline and steady state. With a t-contrast, we modelled the interaction between baseline connectivity profile and plasma escitalopram levels.
Analysis of head motion inside the MR scanner. We assessed potential differences in head motion between groups and scans by assessing the framewise displacement (FD). Here, we used translational and rota- www.nature.com/scientificreports/ tional motion parameters obtained by SPM motion correction. All FD time courses were characterized by the mean and maximum FD, and the maximum FD after eliminating the largest 5% of the FD values, and the number of FD values exceeding 1 mm 76 . Consent to publish. The authors affirm that human research participants provided informed consent for publication.