Large-scale neural network correlates of response inhibition associated with antidepressant response in major depressive disorder

Abstract

Background

A fundamental impairment of major depressive disorder (MDD) is response inhibition, which may serve as a predictor for inadequate responses to antidepressants. Nevertheless, the neurophysiological connections between treatment outcomes and response inhibition are not yet fully understood.

Methods

This study involved 149 participants, consisting of 77 healthy controls (HCs) and 72 patients with MDD. All individuals undertook a Go/No-go task while undergoing magnetoencephalography (MEG) recording. Those patients with MDD showing at least a 50% decrease in symptom severity (the short 6-item version of Hamilton Depression Rating Scale) after four weeks were classified as responders. We assessed whole-brain connectivity within the beta-band.

Results

Individuals with MDD demonstrated decreased connectivity in a right-lateralized network centred on inferior frontal gyrus. Additionally, non-responders displayed reduced functional connectivity in a left-dominant frontoparietal network centred on superior parietal gyrus and orbitofrontal cortex during response inhibition when compared to both responders and HCs. This identified dysregulation also has potential predictive value regarding the response to antidepressant treatment.

Conclusions

Hypoconnectivity in the left-lateralized frontoparietal network centred on superior parietal gyrus and orbitofrontal cortex indicate a decompensatory mechanism that may contribute to an insensitivity toward antidepressant interventions.

Background

Cognitive impairment is a well-known contributor to inadequate symptom relief from antidepressant treatments, unfavourable illness course and poor functional outcomes in major depressive disorder (MDD) [1]. One of the most common cognitive dysfunction is response inhibition [2]. Impairments in this domain are associated with negative biases in attention and memory. Response inhibition deficits prevent redirection of focus from negative content, leaving individuals stuck in negative emotional states, interfering with effective emotion regulation [3]. The inability to inhibit negative thoughts allows individuals to recur unchecked, facilitating the onset and persistence of rumination [4]. Additionally, poor baseline response inhibition might be a prognostic indicator of an increased risk of inadequate treatment response to pharmacological antidepressants [5–8]. Nevertheless, the underlying neural mechanisms connecting response inhibition with antidepressant resistance in MDD remain unclear.

The inhibitory control network plays an important role in implementing response inhibition [9, 10]. This network is comprised of various frontoparietal brain areas, such as the anterior cingulate cortex (ACC), pre-supplementary motor area (pre-SMA), inferior frontal gyrus (IFG), inferior and superior parietal gyrus (IPG, SPG) [9, 11]. Prior studies utilizing functional magnetic resonance imaging (fMRI) has found that an efficient control network could predict a good response to antidepressants [12, 13]. Therefore, the inhibitory control network may serve as the neural substrate linking response inhibition to antidepressant response. fMRI relies on blood oxygen level dependent (BOLD) signals with a temporal resolution of about 1–2 s, reflecting indirect signals of changes in blood flow, and is unable to track fast neural activity.

Studies utilizing magnetoencephalography (MEG) and electroencephalography, which offer sufficient temporal resolution to capture the dynamics of neural activity, have confirmed that regions within the inhibitory control network reach peak activity between 100 and 400 ms [14]. This pattern of neural activity includes oscillations within various frequencies, including theta (4–7 Hz), alpha (8–14 Hz), beta (15–30 Hz), and gamma (> 30 Hz) bands, each contributing to successful inhibition. Among them, the modulatory role of beta oscillation in response inhibition has been well established [15–17]. Elevated beta-band power of rIFG and pre-SMA could be observed during reaction inhibition. Our previous study identified a cognitive biotype of MDD characterized by dominant beta oscillations [18–20]. Beta band activity is involved in maintaining an ongoing cognitive state [21]. Additionally, we discovered a correlation between local beta oscillations and antidepressant efficacy [22–25]. Beyond processing local information, beta band may facilitate large-scale communication across brain networks [26]. However, the mechanisms by which areas communicate information via long-range synchrony for response inhibition remain unclear. Beta oscillations play a crucial role in top-down controlled processing [27], as they enhance communication between brain regions and contribute to the integration of information within the inhibitory control network, thereby enabling response inhibition [11]. Consequently, beta-band modulations within the inhibitory control network may serve as a bridge linking response inhibition to antidepressant response.

In this research, we used MEG to examined functional connectivity (FC) in the whole-brain range during response inhibition in MDD and its relationships with poor response to antidepressant treatment. Our focus was on the synchrony among brain regions that create networks within the beta frequency range from 100 to 400 milliseconds after stimulus presentation. According to previous studies by us and others, this time period best characterizes response inhibition-related brain activity [10, 28]. This special time period was selected as it captures the peak regional activation of response inhibition and the corresponding oscillation frequencies that facilitate large-scale communication between regions. We hypothesized that pretreatment beta-band long-range interregional communication within the inhibitory control network may serve as a general predictor of response to pharmacological antidepressants.

Methods

Study design

This investigation employed a design of cross-sectional case-control. The collected data encompassed clinical evaluations, demographic details, as well as results from MEG and MRI scans from participants with MDD and healthy controls (HCs). Clinical trial number: not applicable.

Participants

72 patients at acute depressive period were recruited from the Affiliated Brain Hospital of Nanjing Medical University in China between 2010 and 2023. Each patient met the criteria for MDD according to Fourth Edition of the Diagnostic and Statistical Manual of Mental Disorders (DSM-IV). Since this study began in 2010, we did not use the latest version (DSM-5, published in 2013).

The additional inclusion criteria for patients: (1) aged between 16 and 60 years; (2) of right-handedness; (3) with native Han Chinese ethnicity; (4) at least a junior high school education; (5) Hamilton Rating Scale for Depression 17-item (HAMD-17) ≥ 17; (6) free of antipsychotics, mood stabilizers, or antidepressants during the six weeks prior to the study.

The criteria for exclusion encompassed the following: (1) the presence of other mental or physical disorders; (2) individuals who had undergone physical therapies within the last six months; (3) any contraindications preventing them from undergoing MEG or MRI scans; and (4) current pregnancy or breastfeeding.

Additionally, 77 matched HCs without current or past mental disorder were selected from the local community. All HCs underwent the MINI by trained psychiatrists to confirm that they were free of mental illness. Additional exclusion criteria include: family history of neurological or psychiatric illness, contraindications for undergoing MEG or MRI scans, severe physical illness and recent use of psychoactive substance.

All patients received serotonin selective reuptake inhibitor (SSRI) monotherapy for at least four weeks. No systematic psychological intervention such as cognitive behavior therapy was performed. The medication doses were as follows: escitalopram (starting dose is 10 mg/day, increasing to 20 mg/day after one week depending on efficacy and tolerability). The short 6-item version of Hamilton Depression Rating Scale (HAMD-6) was employed to evaluate the response of the antidepressant treatment, owing to its superior clinimetric properties compare to the Montgomery Asberg Depression Rating Scale and HAMD-17 [29]. Following this assessment, patients with MDD were classified into two categories according to their percentage change of HAMD-6 scores following four weeks treatment: (1) the non-responder group, which exhibited reductions of less than 50%; and (2) the responder group, which demonstrated reductions of 50% or more.

This study received approval from the Affiliated Brain Hospital of Nanjing Medical University (No.2011-KY027), and it was conducted in accordance with the Declaration of Helsinki. Prior to participation, all individuals provided written informed consent.

MEG recording

Data were acquired utilizing a CTF system with 275 channels, at a sampling rate of 1200 Hz. For monitoring head movements throughout the recordings and MRI-MEG alignment, three coils were positioned: one at the nasion and two at preauricular points.

Task paradigm

While undergoing MEG scanning, individuals engaged in a Go/No-go task intended to induce brain response associated with response inhibition. Every trial began with a grey light lasting approximately 2,500 milliseconds, after which either a red or green light presented randomly for 500 milliseconds. Individuals were instructed to provide a response during the green light trials and to refrain from responding during the red light. The task consisted of 240 trials, including 60 No-go trials. Errors of omission (EO), errors of commission (EC) and response times to Go trials (RT) were recorded.

MRI recording

Structural MRI data were collected with the following parameters (Siemens Verio 3 T MRI system): TR = 1,900 ms, TE = 2.48 ms, FA = 9°, slice number = 176, slice thickness = 1 mm, voxel size = 1 × 1 × 1 mm³, FOV = 250 × 250 mm². Three fiducial markers were placed at the left, right canals and the nasion to ensure MEG and offline co-registration for further source localization.

Data processing

MEG data were analyzed in FieldTrip (version 20210720) [30]. To reduce the amount of redundant calculation, data were down-sampled to 400 Hz. To mitigate power line interference, a discrete Fourier transform filter was applied at 50 Hz. Trials and channels containing excessive variance (1 × 10⁻²⁵) were removed to ensure data cleanliness. Additionally, independent component analysis was utilized to remove components related to eye movements, and those associated with muscular and cardiac activities. MEG data needed to satisfy the criteria of maximum head motion in translation < 5 mm, and rotation < 2 degrees, while structural MRI data satisfied the criteria of maximum head motion in translation < 2 mm, and rotation < 2 degrees.

Source reconstruction

Volume conduction model of the head (head model) was construct based on an individual subject’s MRI. A regular three-dimensional grid with 6 mm spacing served as the source model, which was then aligned with the MNI152 brain template. By considering the source model, individualized head position information, and sensor locations, the leadfield was calculated for each grid point. Based on the Automated Anatomical Labelling atlas, 90 areas were defined. By the linearly constrained minimum variance (LCMV) beamformer [31], we reconstructed time series for virtual channels.

Functional connectivity network construction

The functional connectivity within beta band (15–29 Hz) during No-go trials was calculated for the time window of interest (100–400 ms) based on the virtual channel courses of 90 cortical areas. The coherence coefficient for each pair of regions was used to represent the strength of the FC, which was subsequently transformed into Fisher’s z-score to enhance normality and facilitate further analysis. A 90 × 90 functional connectivity matrix was then constructed for each participant.

Statistical analyses

A univariate ANOVA was employed to assess differences in education levels and age across groups, while chi-square tests were utilized for the gender distribution among those groups.

The functional connectivity was compared using network-based statistic (NBS) toolbox [32]. Here we implemented two independent comparisons of whole-brain FC. First, all patients were compared with HCs. Second, we compared non-responders to responders. When comparing functional network differences, two independent two-sample t tests were performed in each FC value. Connectivity that had statistical values surpassing the initial threshold (p = 0.001) was chosen to create the clusters, with the quantity of FC in every cluster designated as the observed cluster value. Data from each group were randomly permuted 5,000 times to create a reference distribution. The observed value that exceeded the 95th percentile of the reference distribution was deemed significant (p < 0.05).

Subsequently, Spearman correlation analyses were conducted to examine the association between the disrupted functional network and reduction ratios in HAMD-6 scores. Finally, logistic regression and receiver operating characteristic (ROC) curve analyses were conducted on the baseline large-scale neural oscillatory patterns during response inhibition to predict poor antidepressant response. To further evaluate the independence of neurophysiological variables from essential demographic factors, the analyses above incorporated sex, age and education as covariates.

Results

Demographics and clinical characteristics

No significant differences were found in demographic factors such as sex, age, first episode versus recurrence, duration of the current episode, years of education, or baseline disease severity between responders and non-responders (Table 1). For task performance, the three groups had no significant difference.

Table 1.

Demographic and clinical characteristics

Characteristics	Responders (n = 38)	Nonresponders (n = 34)	HC (n = 77)	P
Sex (female/male)	22/16	21/13	45/32	0.934a
Age (year)	30.26 ± 9.21	28.41 ± 9.40	28.70 ± 7.23	0.341b
Education (year)	13.66 ± 2.95	13.97 ± 2.74#	14.99 ± 2.21	0.017b
First-episode/recurrence	20/18	18/16	-	0.979a
Current period (month)	3.70 ± 3.30	4.49 ± 5.36	-	0.450c
HAMD-17, baseline	24.08 ± 5.38	23.15 ± 4.34	-	0.425c
HAMD-6, baseline	12.32 ± 3.09	11.74 ± 1.86	-	0.344c
HAMD-6, 4 weeks	3.68 ± 2.09	8.24 ± 1.86	-	< 0.001c
Errors of omission (%)	1.05 ± 0.66	0.93 ± 0.73	0.97 ± 0.89	0.804b
Errors of commission (%)	3.95 ± 2.52	4.22 ± 1.38	3.72 ± 1.66	0.432b
Response time (ms)	363.0 ± 94.2	383.1 ± 130.5	356.9 ± 92.1	0.464b

Data were presented as the mean (±SD)

^a Pearson chi-square test

^b Univariate ANOVA

^c Two-sample t test

^# Two-sample t-test between patients with response and without (p = 0.60)

Functional connectivity

Individuals with MDD exhibited reduced connectivity in a right-lateralized frontal network within the beta oscillation between 100 and 400 ms when compared to the control group (Fig. 1). Within this network, the node exhibiting the greatest degree of connectivity was the right inferior frontal gyrus, opercular part (IFGorb), which demonstrated diminished connectivity with the orbitofrontal cortex (OFC), ACC and superior frontal gyrus (SFG). Then, we averaged FC values within the significant network and compared them in three groups. The two patient groups (responders and non-responders) did not differ significantly.

Fig. 1.

Functional connectivity differences between patients and HC identified by NBS. (A) The brain map illustrates how connections are distributed, with the size of each node representing the number of connections it contains. (B) The circular plot in which a link represents a connection. (C) Bar plots showing mean connectivity obtained within the significant network. ***p < 0.001

Non-responders showed a significantly decreased left-lateralized frontoparietal network compared to responders (Fig. 2). Within this network, the node exhibiting the greatest degree of connectivity was the left SPG, which demonstrated diminished connectivity with the OFC, median cingulate cortex (MCC) and putamen. Then, we averaged FC values within the significant network and compared them among three groups. The non-responders showed significantly decreased FC relative to the HCs (p = 0.004) and responders (p < 0.001). Moreover, the responders showed significantly increased FC relative to the HCs (p < 0.001).

Fig. 2.

Functional connectivity differences between non-responder and responder identified by NBS. (A) The brain map illustrates how connections are distributed, with the size of each node representing the number of connections it contains. (B) The circular plot in which a link represents a connection. (C) Bar plots showing mean connectivity obtained within the significant network. **p < 0.01,***p < 0.001

Correlation analyses and logistic regression

The mean connectivity in left-lateralized frontoparietal network at baseline was significantly correlated with the reduction ratios of HAMD-6 scores (Fig. 3A; p < 0.001, r = −0.451). Furthermore, the mean functional couplings within the beta cloud offer insights into potential poor response, as illustrated in Fig. 3B (AUC = 0.857, p < 0.001, 95% CI: 0.772–0.942, sensitivity = 81.8%, specificity = 71.1%).

Fig. 3.

Prediction of response to antidepressant treatment based on functional couplings. (A) A significant correlation was observed between the reduction ratio in HAMD-6 and the mean connectivity in left-lateralized parietal network at baseline. (B) Performance of baseline functional couplings for discriminating the early response (AUC = 0.857, p < 0.001, 95 % CI: 0.772–0.942, sensitivity = 81.8%, specificity = 71.1%)

Discussion

This research investigated the link between neural oscillatory patterns during a Go/No-go task and the responses to antidepressants in individuals with MDD using MEG. Notably, patients who exhibited a poor early response to antidepressants showed diminished inter-regional functional connectivity within a left-lateralized frontoparietal network compared to responders. Furthermore, baseline abnormalities in beta-band long-range interregional communication patterns were effective in distinguishing non-responders from responders, thereby serving as a potential biomarker for poor antidepressant response.

Brain regions within the cognitive control network are known to be co-actived during response inhibition [33]. The right hemisphere (especially the prefrontal-basal ganglia loop) plays a dominant role in the inhibition of inappropriate behaviours or impulses. In this study, both non-responders and responders showed diminished beta FC in a right-lateralized frontal network primarily centered in IFGorb, medial SFG, OFC and ACC. MDD is characterised by dominant beta oscillations compared to other frequency oscillations [34, 35]. Anatomically, the medial SFG correspondeds with the pre-SMA. A substantial body of evidence has supported the unique roles of pre-SMA and right IFG in response inhibition [9, 36, 37]. Increased beta-band power could be observed in pre-SMA and rIFG during the process of response inhibition [38].The ACC is involved in the majority of processes related to withholding a response, including attentional control, conflict monitoring, inhibition itself, and outcome evaluation [39]. Beta oscillations play a critical role in top-down-controlled processing [27], facilitating interregional communication and the integration of information within the inhibitory control network. Notably, frontal beta has a control role in the actions, thoughts and memories [40], and they are linked to impulsive behaviour, rumination and biased memory in MDD. Consequently, the reduced beta-band connectivity suggests decreased top-down modulatory network, leading to poor response inhibition. And, no differences in this network were observed between the two patient groups, implying that this may represent a general mechanism underlying response inhibition impairment in MDD, rather than a predictor of treatment outcome.

We further found increased FC within a left-lateralized frontoparietal network centred on SPG and OFC at baseline in responders. This hyper-connectivity may indicate a compensatory mechanism that emerges in reaction to the increased challenges associated with response inhibition processes stemming from an inefficient right-lateralized frontal network. However, non-responders exhibited decreased coupling within this network compared to both responders and HCs. Previous studies have reported the role of SPG in various neurological functions, including sensory processing, cognition, visuomotor tasks, and attention [41]. MCC has been implicated in performance monitoring through conflict detection, the reallocation of attentional resources based on task-relevant information, and the formation of corresponding actions [33]. Lesions in the OFC are associated with deficits in response inhibition and fMRI studies have indicated increased OFC activity in situations requiring behaviour suppression [42]. The decompensatory pattern observed in the left-lateralized frontoparietal network centred on SPG and OFC may serve as a unique biomarker for poor antidepressant response. Furthermore, inefficient cognitive control networks in both cerebral hemispheres suggest a disruption in top-down control, thus making the implementation of response inhibition more difficult.

There is an increasing interest in categorizing patients with MDD into specific subtypes according to their neurobiological features, thus allowing for treatments to be customized based on this classification. Our findings indicate that evaluating the neurophysiological features of response inhibition could effectively differentiate between non-responders and responders, which could provide valuable insights for determining the initial treatment strategy. Patients with MDD exhibiting deficits in response inhibition might be more amenable to neuromodulation therapies. Beta frequency-specific FC could serve as a tractable biomarker for neuromodulation.

This study has several limitations. Firstly, it employed a cross-sectional design. Further longitudinal studies are needed to examine dynamic changes following intervention. Secondly, the findings of present research should be regarded as preliminary for its relatively small sample size. Thirdly, the analysis was restricted to patients undergoing SSRI monotherapy; thus, the inclusion of other pharmacological treatments and psychotherapy should be considered in future studies. Fourthly, MEG is currently a high-cost instrumentation, which may limit its widespread clinical adoption. To enhance accessibility, our findings could be extended to electroencephalography, which shares physiological origins with MEG but is far more affordable and portable.

Overall, this MEG study identified reduced connectivity within the left-lateralized frontoparietal network centred on SPG and OFC among patients exhibiting poor responses to antidepressants. These findings hold potential clinical significance for predicting antidepressant efficacy and may facilitate the identification of subsets of patients with MDD resistant to SSRI monotherapy and would benefit from alternative approaches, like psychotherapy or neuromodulation.

Abbreviations

MDD

major depressive disorder

ACC

anterior cingulate cortex

pre-SMA

pre-supplementary motor area

IFG

inferior frontal gyrus

IPG

inferior parietal gyrus

SPG

superior parietal gyrus

fMRI

functional magnetic resonance imaging

MEG

magnetoencephalography

FC

functional connectivity

HCs

Healthy controls

DSM-IV

Diagnostic and Statistical Manual of Mental Disorders Fourth Edition

HAMD-17

17-item Hamilton Rating Scale for Depression

SSRI

serotonin selective reuptake inhibitor

LCMV

Linearly constrained minimum variance

NBS

network-based statistic

ROC

receiver operating characteristic curve

OFC

orbitofrontal cortex

SFG

superior frontal gyrus

MCC

median cingulate cortex

Authors’ contributions

H. L, Y. X, X. W, Z. Y and Q. L conceived and designed this research.H. L, Y. X and X. W, performed research. H. L, Y. X and S. C analyzed data. H. L, Y. X, L. H, Z. Y and Q. L wrote the paper.All authors reviewed the manuscript.

Funding

This work was supported by the National Natural Science Foundation of China (82271568; 82151315; 82101573; 82301718); the Jiangsu Psychiatric Medical Innovation Center (CXZX202226); the Key Project of Science and Technology Innovation for Social Development in Suzhou (2022SS04); the Jiangsu Provincial Natural Science Youth Fund (BK20230154); Medical Science and Technology Development Foundation, Nanjing Commission of Health (ZKX22043).

Data availability

The datasets generated during the current study are not publicly available due to the subjects’ privacy, but are available from the corresponding author on reasonable request.

Declarations

Ethics approval and consent to participate

This study received approval from the Affiliated Brain Hospital of Nanjing Medical University (No.2011-KY027), and it was conducted in accordance with the Declaration of Helsinki. Prior to participation, all individuals provided written informed consent.

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.