Antidepressant medications reduce subcortical–cortical resting-state functional connectivity in healthy volunteers

1Alterations in resting-state connectivity have been observed in depression across multiple networks. For example parts of the cognitive control network (anterior cingulate, prefrontal cortex) which are involved in decision making, attention and resolving conflicts have been found to be altered in depression as has the default mode network a network involved in self-referential activity and emotional regulation (e.g. precuneus, hippocampus, medial prefrontal cortex) and parts of the affective network (orbitofrontal cortex, striatum, amygdala) known to be involved in the processing of emotional and rewarding information.

2Recently Sheline and colleagues reported increased dorsal medial prefrontal cortex connectivity (dmPFC) in depression and suggested that differences in default mode processing may contribute to depression and thus may be a target for antidepressant drug treatment. Consistent with this idea, recent research examining emotional processing in depressed patients has also shown that dysfunctional activity in these regions is attenuated after pharmacological treatment and that this in turn correlates with treatment response. Furthermore a recent study in depressed patients examining orbitofrontal cortex (OFC) connectivity before and after antidepressant treatment found increased OFC connectivity prior to treatment differentiated responders from non-responders.

3Some preliminary examination of our resting-state dataset revealed decreased connectivity between the dmPFC and hippocampus following citalopram treatment compared to placebo. However, it is unknown whether this is a general effect of all antidepressants or specific to those with serotonergic action. Thus we decided to investigate the dmPFC connectivity in the reboxetine group and hypothesised that unlike citalopram, reboxetine would have its effects specifically in areas where catecholaminergic pathways have a role, such as the striatum and prefrontal cortex. We also decided to examine other seed regions within the affective network that we have shown in our previous studies to be modulated by reboxetine and citalopram during task-based fMRI such as the nucleus accumbens, the subgenual cingulate and the amygdala. We investigated the effects in healthy volunteers with no history of depression or use of psychiatric medications. We hypothesised that antidepressant medications would decrease connectivity within the affective network and that this would be consistent with the recent hypotheses proposed by and.

4The time course of the current study (7 days) was limited compared to the duration of clinical treatment with antidepressants and it would therefore also be relevant to study the effect of medications on the neural basis of resting-state after longer treatment periods. However, a meta-analysis of clinical trials has shown that, relative to placebo, SSRIs significantly decrease depressive symptomatology after 1 week of treatment and indeed the effect size of active treatment relative to placebo is numerically greater during the first week of therapy than in subsequent weeks. Further, it has been demonstrated in healthy volunteers that 7 days treatment with reboxetine produces positive biases in measures of emotional perception and memory, suggesting that therapeutically relevant changes in neuropsychological function are indeed apparent during the first week of antidepressant administration.

5Forty healthy volunteers were randomised to receive 7 days oral treatment with citalopram (20 mg/day, n = 12), reboxetine (4 mg/b.i.d., n = 13) or placebo (n = 15), in a double blind between groups design. Medication was taken twice a day, once in the morning and once in the evening, to maintain blinding. Ethical approval was provided by the Oxford Research Ethics Committee B and written informed consent was obtained from all participants before screening and after a complete description of the study was given. Exclusion criteria for all subjects were current or past Axis-1 disorder on the Structured Clinical Interview for DSM-IV and any contraindications to MRI e.g. pacemaker, mechanical heart valve, hip replacement, metal implants.

6None of the participants took current medication apart from the contraceptive pill. Before drug administration and to ensure group matching, baseline information was collected using a neuropsychological battery that included: Beck Depression Inventory (BDI), State-Trait Anxiety Inventory (STAI), Fawcett–Clarke Pleasure Scale (FCPS), and Snaith–Hamilton Pleasure Scale (SHAPS). Body mass index (BMI) was also calculated for each volunteer. To assess the effects of the treatment the following questionnaires were taken before and after the treatment: Visual Analogue Scales (VAS) of happiness, sadness, anger, disgust, alertness and anxiety; and the State Anxiety Inventory.

7MRI-derived measures of brain function, based on blood-oxygenation-level-dependent (BOLD) contrast, were used to compare brain responses at rest across the three drug groups and on the 7th day approximately 5 h after the last treatment. The resting-state data were acquired after the volunteers had completed both a reward/aversion task and a structural scan. Subjects were instructed to lie in dimmed light with their eyes open, think of nothing in particular, and not to fall asleep.

8Images were acquired with a 3.0-T VARIAN/SIEMENS whole-body scanner at the Centre for Functional Magnetic Resonance Imaging at Oxford (FMRIB), where T2* weighted EPI slices were acquired every 2 seconds (TR = 2). Imaging parameters were selected to minimise susceptibility and distortion artefact in the orbitofrontal cortex. Coronal slices (25) with in-plane resolution of 3 × 3 mm and between plane spacing of 4 mm were obtained. The matrix size was 64 × 64 and the field of view was 192 × 192 mm. Acquisition was carried out during the resting scan yielding 150 volumes in total. An anatomical T1-weighted sequence with coronal plane slice, thickness 3 mm and in-plane resolution of 1.0 × 1.0 mm was also acquired to improve the registration process.

9Data analysis was carried out using FSL tools (www.fmrib.ox.ac.uk/fsl). After discarding the first four volumes to allow for T1 equilibration effects, fMRI data analysis was carried out using FEAT (FMRI Expert Analysis Tool) Version 5.98, part of FSL (FMRIB's Software Library, www.fmrib.ox.ac.uk/fsl). The pre-processing included the following steps: motion correction non-brain removal; spatial smoothing using a Gaussian kernel of full-width at half maximum (FWHM) of 6.0 mm; grand-mean intensity normalisation of the entire 4D dataset by a single multiplicative factor and high pass temporal filtering (Gaussian-weighted least-squares straight line fitting, with sigma = 75.0 s). fMRI volumes were registered to the individual's structural scan and standard space images using FMRIB's Linear Image Registration Tool (FLIRT).

10Based on a-priori hypothesis, we created seed ROIs for the right dmPFC (18 34 29) and the left dmPFC (− 24 35 28). These areas have been shown to have increased connectivity within the default mode network and the affective network in the recent resting-state functional connectivity study in depressed patients by. We used Wake Forest University Pickatlas tool in SPM5 to create the dorsal nexus spheres of 10-mm radius. We also created seed ROIs for the right nucleus accumbens (10 10 − 8), the subgenual cingulate (2 22 − 18) and the right amygdala (24 − 4 − 18) as these areas are part of the affective network and have been shown to be modulated by the action of antidepressant drugs in our previous studies. Further these areas are specifically implicated in fMRI studies in depression. For the subgenual cingulate cortex we used Wake Forest University Pickatlas tool in SPM5 to create the sphere of 10-mm radius. The nucleus accumbens and the amygdala spheres were created with the Harvard-Oxford subcortical structural atlas probabilistic atlas so as to maximise the exact coverage of these smaller differentially shaped structures, furthermore the nucleus accumbens and the amygdala masks were thresholded by 20%.

11An FSL tool (Featquery) was used to extract the time series within each of the five seed regions for each individual subject after having pre-processed the raw fMRI data. Time series were averaged across all voxels in each seed ROI. Using the extracted time series we performed multiple regression analysis using FMRI data processing FEAT. These analyses produced separate individual participant-level correlation maps of all voxels that were positively or negatively correlated with each of the five seeds. Afterwards, higher level (group level) analysis was carried out using FMRIB's Local Analysis of Mixed Effects (FLAME). The general linear model (GLM) was applied to test for group averages and differences among the three groups (citalopram, reboxetine and placebo). The Z statistic images were thresholded using clusters determined by Z > 2.3, and a whole brain family-wise error-corrected cluster significance threshold of p < 0.05 was applied to the superthreshold clusters. Graphs of % BOLD signal change (Figs. 1–3) were created with the FSl tool, Featquery (www.fmrib.ox.ac.uk/fsl).

12In order to account for potential indeterminate noise, nine covariates of no interest (nuisance) were identified for inclusion in our analyses. In detail, global signal, white matter (WM), cerebrospinal fluid (CSF), and the 6 motion parameters for each individual were added. As the global signal is thought to reflect a combination of physiological processes (such as cardiac and respiratory fluctuations) and scanner drift, it was included as a nuisance signal to minimise the influence of such factors. In order to extract the nuisance covariate time series for WM and CSF, we firstly segmented each individual's high-resolution structural image, using FSL's FAST segmentation program. The resulting segmented WM and CSF images were then thresholded to ensure 80% tissue type probability. These thresholded masks were then applied to each individual's time series, and a mean time series was calculated by averaging across all voxels within the mask.

13There were no significant differences among the three groups as determined by one-way analyses of variance (ANOVAs) for age, gender or body mass index, p > 0.06 (Table 1). There were no significant differences between the three groups as determined by one-way ANOVAs for measures of anhedonia (Snaith–Hamilton Pleasure Scale, Fawcett–Clarke Pleasure Scale) trait anxiety (TRAIT) or mood (Beck Depression Inventory), p > 0.1 (Table 1). Treatment with reboxetine or citalopram also failed to affect subjective state and mood measured over the 7-day experimental period on visual analogue scales (alertness, disgust, drowsiness, anxiety, happiness, nausea, and sadness) as determined by repeated-measures ANOVAs, p > 0.1 (Table S1 in Supplement 1).

14For each of the five seed regions previously associated with depression and the effects of antidepressant medications on neural processing, seed-based correlation analyses were employed to characterise their associated functional systems during rest. The functional connectivity for the five seed regions is described in Supplementary Table 2 for the placebo group alone (baseline). Overall, the patterns of connectivity associated with each of the seed regions are consistent with other neuroimaging studies examining the correlations between these seed regions and other brain areas in resting-state and functional connectivity experiments in healthy controls and depressed patients.

15Compared to the placebo group there was reduced functional connectivity between the amygdala and the ventral medial prefrontal cortex (vmPFC) in the citalopram group compared to placebo (Fig. 1). There was also reduced functional connectivity between the amygdala and the OFC in the reboxetine group compared to placebo (Fig. 2) (Table 2).

16Compared to the placebo group there was reduced functional connectivity between the nucleus accumbens and the mid OFC in the reboxetine group (Fig. 3). There were no differences between the placebo and the citalopram group for this seed region (Table 2).

17There were no differences between the placebo and the reboxetine group for this seed region. Compared to the placebo group there was reduced functional connectivity between the dmPFC seed region and the hippocampus in the citalopram group (previously published ).

18There were no differences between the placebo group and either of the drug groups citalopram or reboxetine in subgenual functional connectivity.

19Our study reveals that 7 day administration with the antidepressants citalopram and reboxetine modulates resting-state functional connectivity in a double-blind placebo controlled design in healthy volunteers. Specifically, resting-state functional connectivity was reduced between the prefrontal cortex and the amygdala for both the reboxetine and the citalopram groups. Further there was reduced connectivity between the OFC and the striatum in the reboxetine group compared to the placebo group. Consistently previous studies using task related functional MRI have also shown these limbic areas to be modulated by citalopram and reboxetine and previous studies of resting-state functional connectivity in depressed patients reveal dysfunction in these regions. Unlike our previous finding of reduced dmPFC connectivity with the hippocampus in the citalopram group compared to placebo we found no effects on connectivity with the dmPFC for the reboxetine group. We also found no effects of the antidepressants on connectivity with the subgenual cingulate cortex.

20In our study both antidepressants reduced connectivity between the amygdala and parts of the prefrontal cortex. Such effects are potentially relevant to the treatment of depression as a number of task related studies have revealed reduced amygdala activity after antidepressant drug treatment in healthy and depressed volunteers. Furthermore a recent study by revealed that depressed and bipolar patients were associated with increased amygdala-orbitofrontal cortex connectivity compared to healthy controls. Also the patients currently treated with antidepressants had reduced functional connectivity in these areas and the authors concluded that the increased connectivity may serve as a state marker for depression. Our results indicate that using a randomised placebo controlled design we were able to extend these findings to resting-state functional connectivity in healthy volunteers with antidepressant treatment and we suggest that the normalisation of connectivity could be the direct result of the pharmacological treatment.

21We also found that the connectivity between the OFC and the striatum was reduced with reboxetine treatment during our resting-state scan. The striatum has been implicated previously in studies on resting-state functional connectivity in depression and a recent study in depressed patients found that there was increased functional connectivity in the striatum, the amygdala and the hippocampus during negative emotional processing. The authors concluded that depressed individuals over-recruit neural networks involved more generally in enhancing memory for affective stimuli and that the degree to which they over-recruit this system is related to the severity of clinical symptomatology. Our results are the first to show that these areas which are known to be involved in the affective salience of stimuli, the representation of reward value and self-referential processing can be modulated by currently used antidepressant medications outside of a task situation and in unmedicated healthy control volunteers.

22The OFC/vmPFC has been described as the principal mediator in a reciprocal relationship between the subcortical–cortical regions in depression and a recent study has revealed aberrant mutual excitation between the medial prefrontal cortex and part of the ventral anterior cingulate in depressed patients which also positively correlated with increased scores on a depressive rumination scale. This region is very close to the region of vmPFC that we have found to have reduced connectivity with the amygdala in the citalopram group (Table 2). As the medial PFC has been reported to underlie the evaluation of the self it has been hypothesised that over activity in this region might be responsible for the maladaptive rumination seen in depression. Furthermore a recent study by discusses how the integrative function of the OFC with its convergence of connections from the subcortical limbic regions is a key player in the pathophysiology of major depression. Disrupted connectivity between the vmPFC/OFC and regions outlined here such as the amygdala and the striatum may thus underlay the aberrant negative and positive emotional processing witnessed in depression. Specifically related to the recent study on resting-state functional connectivity in depressed patients we have previously examined the dmPFC (dorsal nexus) as a seed region and found that the antidepressant citalopram reduced the connectivity between the dmPFC and a part of the default mode network the hippocampus in our healthy volunteers, we concluded that antidepressant medications may work by reducing the aberrant increased connectivity seen in depression. Interestingly though in this study we did not find any effects of reboxetine on dmPFC connectivity. This may simply be due to power and perhaps with increased numbers of volunteers differences may have been revealed. Or it could also be due to the relatively less efficacious nature of reboxetine as a treatment for depression. Although we ran our resting-state approximately 10–15 min after a task it is possible that examining resting-state after other tasks could effect the results and thus we intend for future experiments to run resting-state scans before any other tasks.

23As hypothesised reboxetine decreased mesolimbic striatal–OFC connectivity (indicating catecholaminergic pathway involvement) whereas citalopram decreased amygdala-vmPFC connectivity ((and in our previous study hippocampal–dmPFC connectivity) indicating serotonergic pathway involvement). These results are consistent with recent studies in animals using pharmacological challenges to unpick rat brain connectivity. Schwarz and colleagues examined pharmacological challenges such as amphetamine (catecholamine) and fluoxetine (SSRI) on rat brain functional connectivity. Similarly they report that catecholaminergic drugs modulate striatal and prefrontal connections whereas serotonergic drugs modulate hippocampal, amygdala and thalamus connections. Yet this is the first examination, to the best of our knowledge, of the effects of SSRI and SNRI antidepressant drug action in the resting human brain. Similar to Schwarz et al. we also found that the regions affected by the drugs closely reflect the known pathways in the neurotransmitter systems targeted by these drugs. Further work on receptor mapping would of course also aid our understanding of the brains architecture and would be especially relevant for pharmacological studies such as this.

24There was no effect of either of the antidepressant medications on connectivity with the subgenual cingulate cortex seed, despite reports of increased subgenual connectivity in depressed patients resting-state data. The subgenual cingulate cortex is an important area targeted in deep brain stimulation for treatment-resistant depression perhaps then antidepressant improvement may lie in the development of drugs that can better target regions such as the subgenual cingulate cortex as well as the dmPFC. The subgenual cingulate is also a region that is hard to image and thus it cannot be ruled out that our results might in some way reflect signal drop-out in this area.

25In conclusion, the results from this study show that both serotonergic and noradrenergic antidepressants decrease resting-state functional connectivity in areas of the brain that correspond with the central activity of these drugs. The results also support our hypothesis that antidepressant medications may work by rebalancing cortical control in depression. It will be of great interest to examine resting-state functional connectivity in unmedicated depressed patients before and after antidepressant treatment as the next step in unravelling antidepressant drug action at the neural level.