Temporal dynamic changes in functional connectivity of reward network in depressed adolescents and young adults with and without suicidal attempts

The prevalence of depression has sharply increased among adolescents and young adults over the past decade. Depression in adolescence and young adulthood has raised growing concern because it is associated with an increased risk of suicide, thus heightening the likelihood of future suicidal behavior and completed suicide. However, the neurobiological underpinnings of suicidal behavior in depressed adolescents and young adults remain poorly understood. In this study, we investigated the dynamic functional connectivity (dFC) between brain regions in reward network and identified alterations in the dynamics of functional connectivity patterns within this network in adolescents and young adults with depression, both with and without a history of suicide attempts, using dFC analysis. Our findings revealed significant differences in dynamic functional connectivity between several key brain regions, including the left ventral striatum and putamen, the left anterior insula and the left medial superior frontal gyrus, the right anterior insula and the right anterior central and left medial superior frontal gyrus, as well as the anterior auxiliary motor area and the inferior frontal gyrus, among the sMDDgroup, nMDD group and HC group. These results demonstrate that the temporal dynamic changes in functional connectivity of reward network were associated with suicidality, suggesting that such changes may serve as an important biomarker for adolescent suicidality.

Clinical trial number

Not applicable.

Introduction

Major depressive disorder (MDD) is a prevalent psychiatric disorder worldwide, predicted to become one of the top causes of disease burden by 2030. The prevalence of depression has sharply increased among adolescents and young adults over the past decade [1, 2]. This rise might be attributable to heightened social demands and stressors, hormonal changes, and brain development. Depression during adolescence and young adulthood has raised growing concern because this life stage encompasses rapid social, emotional, and cognitive development, as well as key life transitions, all critically important for adolescent brain development [3]. Depression in adolescents and young adults may lead to adverse outcomes, including poor educational performance, social dysfunction, and an increased risk of substance abuse and suicide attempts. The presence of depression during adolescence and young adulthood is associated with an increased risk of suicide, thus heightening the likelihood of future suicidal behavior and completed suicide. Previous studies indicate that earlier onset of depression correlates with higher suicidal intent and more pronounced suicidal ideation [4]. However, accurately identifying and preventing depression and suicide in adolescents and young adults remains challenging. Therefore, greater attention should be directed toward investigating the neurobiological features of the suicide attempts in depressed adolescents and young adults. This critical period of brain plasticity encompasses the development governing the growth of cognitive and social competencies as well as puberty-related changes, elucidating the neural correlates of suicide risk in this vulnerable population holds significant implications for creating adolescent-specific interventions aimed at reducing suicidality during this high-risk period.

The reward network, which involves dopaminergic neurons in the limbic system, is composed of the nucleus accumbens(NAcc), ventral tegmental area (VTA), ventral striatum(VS), thalamus, cingulate gyrus, and other deep brain nuclei, as well as the medial prefrontal cortex(mPFC). Based on the meta-analysis by Bartra et al. [5], 11 regions of interest (ROIs), including the bilateral VS, bilateral anterior insula, bilateral thalamus, anterior and posterior cingulate, ventromedial prefrontal cortex, ventral temporal area, and presupplementary motor area have been identified as the core area of the reward network. This network plays a critical role in adolescent development. A longitudinal multi-site study found that nucleus accumbens response to reward increases exponentially with pubertal stage, while dorsolateral prefrontal cortex༈dlPFC༉activation remains flat [6]. Dopaminergic projections to the ventral striatum and medial prefrontal cortex undergo rapid remodeling between the ages of 12 and 18, contributing to the well-documented “reward hypersensitivity” phenotype [7]. During adolescence, the rapid maturation and heightened sensitivity of the reward circuit amplify approach motivation toward novel, social, and rewarding stimuli, thereby facilitating learning, identity exploration, skill acquisition, and environmental adaptation. However, excessive activation can bias decision-making or promote risk-taking behaviors. Elevated activity in the reward circuit is associated with adolescents’ impulsive behavior and vulnerability to addiction. Augmented dopamine release can intensify the pursuit of immediate rewards, whereas the slower-developing prefrontal cortex is not yet fully capable of regulating these impulses. This period of heightened reward reactivity coincides with the most pronounced increase in suicidal ideation and attempts, highlighting the developmental significance of this neural circuit. Previous studies have revealed that the reward circuit is of interest regarding the mechanism of suicide, as it plays a crucial role in regulating emotions and behaviors [8]. Evidence indicates that dysregulation within the reward system is linked to an elevated risk of suicidal thoughts and behaviors. Neuroimaging studies in humans reveal that adolescents with suicidal ideation exhibit a blunted ventral striatal response during reward anticipation. Furthermore, a meta-analysis has identified deficits in reward learning as the most robust distinguishing feature between suicidal and non-suicidal youth [9]. Therefore, investigating the local and network functional changes in the reward circuit of depressed patients and those with suicide attempts dating back to adolescence and early adulthood may help in the discovery of biomarkers of depression and depression-related suicide during critical brain development. Simultaneously, it might aid in exploring the neurobiological mechanisms of suicidal behavior, and provide key clues for the search for important brain intervention targets, which is of great significance for the early identification, intervention and prognosis of depression and depression-related suicide.

Resting-state functional magnetic resonance imaging (R-fMRI) is an emerging neuroimaging technique used to study the functional connectivity in the brain and holds great potential in exploring the neurobiological mechanisms of suicide. It has the benefits of being noninvasive and easy to perform and it offers high temporal and spatial resolution. As a result, it has played a significant role in the understanding of the mechanisms underlying risk for suicide attempt, and is a superior technique to search potential neuroimaging markers for identifying suicide risk in adolescents and young adults [10–13]. Since spontaneous brain function connectivity(FC)may dynamically change over time, a number of methodological developments have tried to track these temporal fluctuations. The dynamic functional connectivity (dFC) analysis offers a powerful method to explore changes in the brain’s intrinsic functional organization [14]. It is proposed based on the assumption that the brain changes dynamically, emphasizing that the connectivity between different brain regions changes over time, allowing the study of rs-fMRI time series at a finer scale (e.g., at a specific time point or within a predefined time window). The analysis of dFC allows for the detection of changes in complex brain activity patterns, providing insights into common and unique processes underlying neuropsychiatric disorders [14, 15]. It has the following unique advantages compared to static functional connectivity (sFC). On one hand, considering the “non-stationarity” of the resting brain, dFC allows investigating the R-fMRI time series on a much finer scale (e.g., at specific time points or within predefined time windows), which helps to observe the details averaged out in sFC. Because sFC is calculated across the entire scan, it cannot reveal moment-to-moment fluctuations in inter-regional coupling. By estimating FC at each time-point (or within short sliding windows), dFC uncovers transient interactions that are averaged out in the sFC measure. Capturing these rapid dynamics may therefore offer additional insight into the mechanisms that sustain and modulate functional communication in the brain, which may provide greater insight into the fundamental mechanisms of FC. On the other hand, dFC is able to capture spontaneously recurring FC patterns (i.e., FC states), which is crucial for understanding temporal variability in the brain’s internal organization. Additionally, the dynamic change is more in line with the actual activity of the brain, and previous studies have also shown static functional connectivity changes dynamically over time. Based on these advantages, many researchers consider dFC to be a potentially sensitive biomarker for neuropsychiatric disorders (such as schizophrenia [16], autism [17], Alzheimer’s disease [15, 18] and Parkinson’s disease [15, 18]. Thus, dynamic connectivity analysis could be conducted to calculate the variability of functional connectivity over time by considering temporal fluctuations within varying windows [19]. However, few studies have investigated the variability of dFC in adolescent and young MDD patients, especially in the aforementioned reward network.

In conclusion, to elucidate markers of risk for suicidal behavior in adolescents and young adults and to help identify neurobiological underpinnings of pathophysiologic mechanisms. We measured dynamic functional connectivity (dFC) between brain regions within the reward network using dFC analysis, and detected significant alterations in the dynamic functional connectivity of reward network in adolescent and young depressed patients with and without suicide attempts compared to normal controls based on R-fMRI. Additionally, we identified distinct patterns of brain network connectivity between adolescent and young depressed patients with and without suicide attempts. We hope this study clarifies the neural features of adolescent and young individuals with MDD who have attempted suicide through the application of dynamic functional connectivity analyses, and can provide a profound insight into the role of resting state functional connectivity of the reward network in distinguishing depressed adolescents and young adults with and without suicide attempts. This approach could be used in the future to improve the prediction of suicide risk in MDD and may serve as targets for future neurobiological interventions.

Participants and methods

Participants

Adolescents and young adults with MDD were recruited from the outpatient and inpatient departments of the First Affiliated Hospital of Chongqing Medical University. The Structured Clinical Interview of the Diagnostic and Statistical Manual of Mental Disorders, 4th edition (DSM-IV), Patients version (SCID-I/P) was administered for the screening diagnostic tool. The patients were screened and diagnosed by two qualified psychiatrists who used the Structured Clinical Interview for DSM-IV Axis I disorders (SCID-I). All participants were Han Chinese, right-handed (determined by patient interview and parental confirmation), age between 15 and 24 years and had normal to corrected vision. There were no contraindications to MRI for any study subjects. Exclusion criteria included other DSM-IV Axis I comorbidities, a history of neurological diseases or seizures, significant head injury or concussion, history of substance/alcohol abuse or dependence, history of psychiatric disorders or suicide among first-degree relatives, and other clinically relevant abnormalities based on medical history or laboratory examination. The MDD patients were categorized into the sMDD group who had attempted suicide within the last six months and the nMDD group who had never attempted suicide. Suicide attempt was defined as self-injurious behavior with some degree of intent to die. The suicide history interview was assessed with the Columbia Suicide History Form, and was confirmed according to their medical records or a history of visits to the hospital emergency department following a suicide attempt.

In addition, 52 right-handed healthy controls (HC) matched for gender, age, and education, were recruited from the local community through poster advertisements and assessed using the Structured Clinical Interview for DSM-IV, non-patient edition (SCID-NP). Individuals with a past or current DSM-IV axis-I psychiatric diagnosis, neurological illness, a history of head trauma with loss of consciousness, or a family history of psychiatric disorders were excluded.

The present study was approved through the Medical Ethics Committee of the First Affiliated Hospital of Chongqing Medical University, P.R. China (NO.2020-97-2), and the protocols were performed according to the standards of the Declaration of Helsinki. All the participants recruited into the study were fully informed of the details of the study and agreed to participate in this experiment. For individuals under the age of 16, their parents signed the informed consent after the study procedures were explained in detail.Written informed consent was obtained from each participant or parents (for under 16 years old).

Clinical assessment

The 17-item Hamilton Depression Scale (HAMD) was administered to evaluate the severity of depression. The Colombia Suicide Severity Rating Scale (C-SSRS) [20] was used to estimate the severity of suicidal ideation, suicide attempts, and fatal outcomes associated with suicide attempts. Additionally, the Scale for Suicide Ideation (SSI) was utilized to measure the severity of suicidal ideation [21].

MRI data acquisition

All participants were scanned on a 3T GE Signa HDxt MRI scanner (General Electric Healthcare, Chicago, Illinois, USA) with a standard eight-channel head coil array. Foam pads were used to minimize head motion and machine noise, and the participants were instructed to relax in a comfortable position with their eyes closed, remain still and stay awake. R-fMRI data were acquired using an echo-planar imaging (EPI) pulse sequence and the following parameters: repetition time (TR) = 2000 ms, echo time (TE) = 40 ms, flip angle(FA) = 90^◦, number of slices = 33, slice thickness = 4.0 mm, field of view(FOV) = 240 mm×240 mm, and in-plane resolution = 64 × 64. A total of 240 time points were axially recorded. In addition, conventional T1-weighted images were acquired using fast spin echo (FSE) with the following parameters: TR/TE = 500 ms /14 ms, thickness/gap = 5.0/0 mm, FA = 45°, NEX = 1, FOV = 240 mm×240 mm, and matrix = 256 × 126. A high-resolution 3D T1-weighted MRI image was obtained with a fast gradient echo (FGRE) sequence using the following scan parameters: TR = 24 ms, TE = 9 ms, FA = 90◦, thickness/gap = 1.0/0 mm, FOV = 240 mm×240 mm, and matrix = 256 × 256. The MR images were reviewed by an experienced neuroradiologist to check for incidental findings and no obvious brain abnormalities on conventional MRI were observed, and none of the subjects felt discomfort during or after the procedure.

Preprocessing

The R-fMRI preprocessing was conducted using the Data Processing and Analysis for (Resting-State) Brain Imaging (DPABI) software available at http://www.restfmri.net, running in MATLAB (Mathworks, Natick, MA, USA). For each participant, the preprocessing steps included discarding the first ten volumes, correcting slice timing, realigning for head motion, normalizing to the Montreal Neurological Institute (MNI) template, spatial smoothing, resampling to 3 × 3 × 3 mm³, regressing out the nuisance signal, removing linear trend and filtering (0.01–0.08 Hz). The initial ten EPI time points were excluded to allow the signal to reach equilibrium and to enable participants to acclimate to the scanning noise. Subsequently, slice timing correction, and realignment for head motion were performed across the time series incorporating nuisance covariate regression using the Friston 24-parameter model [22]. Signals from segmentations of the white matter (WM) and cerebrospinal fluid (CSF) compartments in the 3D T1-weighted image as regressors to mitigate respiratory and cardiac effects. Additionally, linear trend removal and filtering (0.01–0.08 Hz) was included as a regressor to account for potential drifts in the BOLD signal. The resulting aligned image time series for each subject were each co-registered with the corresponding 3D T1-weighted image and The Diffeomorphic Anatomical Registration Through Exponentiated Lie Algebra (DARTEL) tool [23] was utilized to normalize the data for all subjects to Montreal Neurological Institute (MNI) space (resampling with 3 × 3 × 3 mm³ resolution), and spatially smoothed with a Gaussian kernel of 4 mm full-width half-maximum (FWHM).

Definition of the regions of interest (ROIs) of the reward system

In reference to Bartra et al.’s meta-analysis [5], 11 coordinates within the MNI space were selected as regions of interest (ROIs). The 11 sets of ROI coordinates (in MNI space) were considered the core areas of the reward system with a diameter of 5 mm around each core region corresponding to these coordinates. The 11 ROIs of the reward network were defined by referring to the method of Satterthwaite et al. [24], which was also adopted in the recent study by Pan et al. [25] and proven to be effective. The identified 11 ROIs include: ROI1: left ventral striatum (l-VS) (-12, 12, -7), ROI2: right ventral striatum (r-VS) (12, 10, -6), ROI3: ventromedial prefrontal cortex (vmPFC) (2, 46, -8), ROI4: left anterior insula (l-AI) (-30, 22, -6), ROI5:right anterior insula (r-AI) (32, 20, -6), ROI6: posterior cingulate cortex (PCC) (-4, -30, 36), ROI7: ventral tegmental area (VTA) (-2,-22 ,12), ROI8: anterior cingulate cortex (ACC) (-2 ,28 ,28), ROI9: presupplementary motor area (Pre-SMA) (-2 ,28 ,46), ROI10: left thalamus(l-TA) (-6 ,-8 ,6), and ROI11: right thalamus (r-TA) (6 ,-8 ,6). Their spatial locations are illustrated in Table 1 and Fig. 1.

Table 1.

Meta-analytic MNI coordinates used as ROI in reward network

ROI	Brain area	Abbreviation	MNI coordinates
			x y z
ROI1	Left Ventral Striatum	l-VS	-12	12	-7
ROI2	Right Ventral Striatum	r-VS	12	10	-6
ROI3	Ventromedial Prefrontal Cortex	vmPFC	2	46	-8
ROI4	Left Anterior Insula	l-AI	-30	22	-6
ROI5	Right Anterior Insula	r-AI	32	20	-6
ROI6	Posterior Cingulate Cortex	PCC	-4	-30	36
ROI7	Ventral Tegmental Area	VTA	-2	-22	12
ROI8	Anterior Cingulate Cortex	ACC	-2	28	28
ROI9	Presupplementary Motor Area	Pre-SMA	-2	28	46
ROI10	Left Thalamus	l-TA	-6	-8	6
ROI11	Right Thalamus	r-TA	6	-8	6

Fig. 1.

The spatial locations of 11 ROIs within the reward network

Dynamic functional connectivity (dFC) analysis

Dynamic functional connectivity (dFC) analysis was carried out using Temporal Dynamic Analysis package in DPABI software. Previous studies have shown that dFC analysis based on sliding time windows can more accurately reflect the time-varying patterns of individual brain neural activity, which has important clinical significance for understanding the potential mechanism of brain functional network changes in normal people or patients with diseases. In sliding window analysis, window width and step size are critical parameters. Previous studies have shown that the length of the window needs to take into account two factors: (1) It needs to obtain relatively stable brain connectivity. This requires relatively long windows to capture features of rapidly changing functional states [26]. However, this may not be sensitive enough to temporal changes in functional connections, preventing the detection of effective connection states; (2) To better detect the instantaneous changes in the functional state, it is necessary to consider selecting a smaller window width should be considered. However, this may result in too few time points in each window to generate stable connections and may lead to misjudging noise-induced false changes as alterations in functional connectivity.

Considering these two points, the window length in this study was set to 10 TRs (i.e., 30 s) based on previous studies, with the time window advanced by a step of one TR (i.e., 3 s). This means that each sliding time window contains all 33 time points in the time series, and the starting point of each sliding time window is exactly 3 TRs later than the starting point of the previous window.

After MRI image preprocessing, the time series of 11 ROIs in all participants were extracted using the ROI Signal Extrctor in DPABI. Subsequently, dFC was estimated using the sliding-window method through the Temporal Dynamic Analysis toolkit. Specifically, the whole time series was re-divided into several overlapping time series (sub-time series) windows, and the dynamic functional connectivity network was constructed with 11 ROI window time series. We computed Pearson correlation coefficients between 11 ROIs using the time course segment within a time window, and generated a connectivity correlation matrix of size 11 × 11 for each time window. Finally, the standard deviation of the correlation coefficient was calculated for each pair of ROIs over all sub-time series. The 11 × 11 standard deviation matrix is the dynamic functional connectivity matrix, which reflects the time variability of the time series.

Statistical analysis

For group-level analyses, one-sample t tests were performed with group-level correlation maps for the sMDD, nMDD, and HC groups at a relatively liberal threshold of P < 0.05 (uncorrected) for both positive and negative correlations for each ROI, as the following analyses were performed within this mask. The normality of the data distribution was determined. Subsequently, one-way analysis of variance (one-way ANOVA) was performed for dFC variability on each voxel using the Statistical Analysis module in DPABI to compare the differences in dFC variability patterns among the three groups, with FDR correction (P < 0.05), and Bonferroni multiple comparisons were conducted when discrepancies were observed. To investigate whether FC dynamics were related to the clinical performance, we controlled for age, sex, and education level, and employed Pearson’s correlation analysis to analyze the relationships between changed dynamic functional connectivity and SSI scores, HAMD scores, considering two-tailed p-values less than 0.05 as statistically significant.

Results

Demographic and clinical characteristics

In this study, a total of 41 patients with MDD who had attempted suicide were included in the sMDD group, comprising 12 males and 29 females, with an average age of 18.78 ± 1.98 years. The average number of previous suicide attempts was 2.97, and the most recent suicide attempt occurred 3.7 months prior to this MRI scan.

There were no significant differences in gender (P = 0.623), age (P = 0.508) or education level (P = 0.721) among sMDD, nMDD and HC group. However, the scores of HAMD and SSI in sMDD group and nMDD group were significantly higher than those in HC group(P < 0.001). Post-hoc analysis revealed that there was no significant difference in the total scores of HAMD between the sMDD group and nMDD group, but the SSI score in the sMDD group was significantly higher than that in the nMDD group (P = 0.036) as shown in Table 2.

Table 2.

Participant demographic and clinical characteristics

	sMDD(n = 41)	nMDD(n = 44)	HC(n = 52)	P-value
Gender (male/female)	12/29	13/32	17/35	0.623a
Age (years)	18.78 ± 1.98	18.60 ± 2.58	18.78 ± 1.98	0.508b
Education level (years)	12.63 ± 1.88	12.0 ± 2.35	13.73 ± 1.61	0.721b
HAMD	33.23 ± 4.63	31.50 ± 6.92	3.67 ± 2.45	< 0.001b
SSI	21.06 ± 4.35	15.22 ± 3.45	2.65 ± 2.71	< 0.001b

^a Chi-square test; ^b one-way analysis of variance

Differences in dynamic functional connectivity of the Reward Network

In the dynamic functional connectivity of the 11 ROIs of the reward network with the whole brain, significant differences were observed in the left ventral striatum (l-VS) (ROI1), left anterior insula (l-AI) (ROI4), right anterior insula (r-AI) (ROI5), ventral tegmental area (VTA) (ROI7), and presupplementary motor area (Pre-SMA) (ROI9) among the three groups. Specifically, there were significant differences in the dynamic FC of the left ventral striatum with the putamen between sMDD group and nMDD group and between nMDD group and HC group (all P < 0.001). Compared to the HC group, the dynamic FC of the left anterior insula with the left medial superior frontal gyrus was decreased in both the sMDD group and nMDD group (P < 0.001 and P = 0.001, respectively). Conversely increased dynamic functional connectivity of the right anterior insula with right precuneus in the sMDD group and nMDD group compared with that in the HC group (P < 0.001, P = 0.002, respectively). Additionally, the dynamic FC of the right anterior insula with the left medial superior frontal gyrus was higher in the sMDD group compared to the nMDD group. The dynamic FC between the ventral tegmental area and the right occipital cortex was decreased in sMDD group and nMDD compared to HC group (P = 0.049, p < 0.001). The decreased dynamic FC of the pre-SMA with the inferior frontal gyrus was also found in the sMDD group compared with that in the HC group (P = 0.022) (see Fig. 2).

Fig. 2.

Group differences in dynamic functional connectivity (dFC) of the Reward Network among among sMDD, nMDD and HC groups. (A) Statistical analysis of dynamic functional connections between the Left Ventral Striatum and the putamen_L(***means p < 0.001). (B) Statistical analysis of dynamic functional connections between the Left Anterior Insula and the Frontal_Sup_Medial_L (**means 0.001 < p ≤ 0.01,***means p < 0.001). (C) Statistical analysis of dynamic functional connections between the Right Anterior Insula with the Precuneus_R and the Frontal_Sup_Medial_L (**means 0.001 < p ≤ 0.01, *** means p < 0.001). (D) Statistical analysis of dynamic functional connections between the VTA and the Calcarine_R (* means p < 0.05, **means 0.001 ≤ p ≤ 0.01). (E) Statistical analysis of dynamic FC between the pre-SMA and the Inferior Frontal Gyrus (**means 0.001 ≤ p ≤ 0.01)

Relationship of functional connectivity dynamics of the Reward Network with clinical scores

Pearson’s correlation analysis was applied to investigate the relationship between dFC of ROIs in the Reward Network and clinical scores in the sMDD group and nMDD group. However, no significant correlations were observed between HAMD scores, SSI scores and the altered dynamic functional connectivity of ROIs in the Reward Network (P > 0.05, FWE corrected).

Discussion

It has been over two decades since Biswal proposed that resting local functional connections measure the correlation of BOLD signals between regions of the brain at rest [27]. This analysis method calculates the statistical relationships between brain regions over the entire scanning period, yet it fails to account for the potential temporal variations in interactions among functional regions. However, studies conducted in both animal models and human subjects have demonstrated that even at rest, the brain is a highly dynamic system, characterized by significant fluctuations in functional coupling between regions that occur over timescales ranging from seconds to minutes. Smith et al. (2011) [28]found that the time series of most voxels showed changes over time. In other words, the functional connectivity of the resting state is not static, and during the data collection period, the activity of the brain regions will also fluctuate over time [29]. Therefore, it is obvious that part of the information in the time dimension will be lost, and the relatively high time resolution of fMRI cannot be fully utilized if only relying on resting functional networks to assess overall brain activity. The researchers set out to study the non-resting part of rsfMRI across shorter timescales than previously examined to capture the unstable nature of non-task-related brain activity and its dynamic alterations in connectivity networks, collectively referred to as dynamic rsfMRI.

Understanding rsfMRI on a shorter time scale is critical because such periods may allow this resting approach to be used during intermission periods. The scan can be completed in less time while also capturing diverse signal characteristics [30]. Furthermore, if rsfMRI is able to effectively measure known non-stationary states of the brain, this insight could prove vital for constructing a comprehensive understanding of brain activity that might otherwise yield uniform results throughout its duration. Chang and Glover [29] first demonstrated that the strength and direction of functional connections changed over time. Since then, researchers have proposed a variety of analytical methods to measure time-resolved connections and identify biomarkers from dynamic connections. In recent years, sliding time window technology [31, 32] has gained widespread recognition. Relying on the sliding window method, researchers have observed the dynamic activity mechanism of spontaneous brain activity in primates [33] and humans [34]. Previous studies have shown that dFC analysis based on sliding time windows can more accurately reflect the time-varying patterns of individual brain neural activity [18, 19, 32], which has important clinical significance for understanding the potential mechanism of brain functional network changes in normal people or patients with diseases [14]. Based on the method of sliding time window, Du et al. [35] examined the correlation of low-frequency fluctuation amplitude and functional connectivity with time in schizophrenia patients. Allen et al. [36] also evaluated the dynamic functional connectivity in resting-state by calculating the temporal variability of functional connectivity using a data-driven sliding time window technique. These studies found that changes in dynamic functional connectivity can reflect dynamic, continuous interactions between brain regions. This method has been widely used to elucidate the abnormality of dynamic brain function in different neuropsychiatric diseases [37]. Zilverstand et al. [38] showed that higher ventral-striatal dFC variability predicts future substance use relapse, suggesting that excessive fluctuation marks a loss of network stability. Hu et al. [39]reported that reduced dFC variability between NAcc and mPFC relates to anhedonia in MDD, indicating that too little flexibility may also be maladaptive. Although many studies have examined the temporal dynamic characteristics of local indicators of spontaneous brain activity in MDD from various perspectives, no studies have systematically explored the dynamic functional connectivity of adolescent and young MDD patients with suicide attempts. In this study, we systematically investigated the abnormal temporal dynamic characteristics of reward network functional connectivity among young MDD patients who have made suicide attempts.

Previous study indicated that lower NAcc↔mPFC dFC variability was associated with higher suicidal ideation (SI) [40], which interpret this as “rigid coupling” that limits the adolescent’s ability to update reward value in changing contexts, potentiating hopelessness and suicidal ideation. Conversely, one study reported higher dFC variability in AMY↔insula was linked to greater SI, which interpret this as affective noise that disrupts reward evaluation and heightens threat sensitivity or pain perception [41]. Alterations in dynamic functional connectivity variability elevates suicide risk. In present study, our dynamic FC analysis of the reward network revealed significant differences in dynamic functional connectivity within the left ventral striatum (ROI1), left anterior insula (ROI4), right anterior insula (ROI5), ventral tegmental area (ROI7), and anterior supplementary motor area (ROI9) among youth depressed patients who attempted suicide, those who did not attempt suicide, and a normal control group. The results indicate that dFC between the left ventral striatum and putamen, the left anterior insula and the left medial superior frontal gyrus, the right anterior insula and the right anterior central and left medial superior frontal gyrus, and the anterior auxiliary motor area and the inferior frontal gyrus were significantly altered among the sMDD, nMDD and HC groups.

We found dFC between the left ventral striatum (l-VS) and putamen was significantly reduced in sMDD patients compared with nMDD patients. As a key region in regulating reward response, the VS plays an important role in conveying reward information, and is connected to both the cerebral cortex and limbic regions to facilitate the integration of emotion and cognition [42, 43]. Several studies suggested that the VS is crucial in the impaired goal-directed behavior and motivational processes (such as goal-oriented behavior, reward learning and pleasure response) in MDD [44]. Pan et al. [45] investigated the iFC of the VS within the reward network in a community sample of adolescents. The results revealed that increased ventral striatum node strength was associated a 50% increase in the odds of developing a depressive disorder after three years. Chen et al. (2017). found that altered variability in the dynamic FC between the inferior VS and mPFC is an MDD-specific alteration. These studies indicated that aberrant ventral striatum functional connectivity may contribute to youth depression and can predict future risk for depressive disorder. Oswald et al. [46] discovered that VS function is related to behavioral inhibition, and impaired VS may lead to suicidal behavior through reduced inhibition of impulsivity. The research showed that functional changes in the VS were associated with suicidal behavior characterized by an impulsive and aggressive disposition, particularly in adolescent MDD. Dombrovski et al. [47] found that VS lesions could contribute to suicidal behavior by increasing impulsivity. Individuals who attempt suicide with VS lesions tend to make impulsive decisions that prioritize superficial, immediate rewards over more strategic choices that consider the long-term benefits of rewards and the necessity for delayed gratification. McClure et al. [48] reported that during a suicide crisis, short-sighted decision-makers may fail to realize the far-reaching impact of their choices, and this change in decision-making may be related to VS abnormalities.

Previous evidence from brain structural and functional imaging has shown that abnormalities in insular structure and function are associated with suicidal behavior [49]. Hu et al. [50] have discovered that the baseline brain functional activity of the bilateral insula in depressed patients with suicide attempts was significantly enhanced compared to those without suicide attempts. Single photon emission computed tomography (SPECT) imaging revealed that the local cerebral blood flow in the right insula of the brain was increased in suicide attempters during the resting state [51], indicating that the insula of the brain in suicide attempters has abnormal function.The insula has been repeatedly demonstrated to be involved in processing various somato- and viscerosensory stimuli, e.g., pain perception. It is the integration center of sensory information processing, and plays a key role in the process of monitoring and filtering behavior-related stimuli. It is a key node of the brain prominent network, and plays an important role in the dynamic regulation of the whole brain large-scale network. Forrest et al. [52] suggest that people with suicidality have poor interoception and may be disconnected from their bodily selves. Interoception may be more disturbed among suicide attempters relative to suicide ideators and planners; and furthermore, interoception may be most disturbed among recent attempters relative to distant attempters. Given this, disturbed interoception could be a near-term risk factor for suicide attempts. The insula is the integration center of intersensory information. Evidence from fMRI shows that adolescents with suicide and self-injury have abnormal activation of insula when processing unpleasant physical tactile information [53]. In the current study, the sMDD group showed increased dFC of the right anterior insula with right precuneus and left medial superior frontal gyrus compared to the healthy controls and nMDD group. Besides, in comparison to the healthy controls, the dFC between the left anterior insula and the left medial superior frontal gyrus was decreased in sMDD and nMDD group. Thus, abnormal resting-state dFC of the insular may be involved in the neural mechanism of suicide attempts in adolescents and young adults with depression. In view of the central role of insula in intersensory processing and previous studies on intersensory processing and suicide, we believe that the abnormal dynamic functional connectivity between the insula and the left medial superior frontal gyrus in this study may be related to the dysfunction of sensory information processing and intersensory integration in patients with depression after a suicide attempt.

In this study, we also found that there was different dynamic FC between the ventral tegmental area (VTA) and the right occipital calcarine cortex in the sMDD group compared with the nMDD and HC group. VTA is a key node of the brain’s reward circuit and plays a pivotal role in the regulation of many important behaviors, such as motivation, reward, addiction, aversion, memory, learning and sleep-wakefulness cycles [54–56]. It has been widely recognized that the activity of dopamine (DA) neurons in the VTA and DA release to various brain regions regulate mood-related behaviors. Research has shown that the optogenetic inhibition of VTA leads to a depressive-like phenotype [57]. Previous task fMRI studies found decreased FC between the VTA and striatum in patients with MDD during reward outcome in monetary instrumental learning task compared to healthy individuals [58], which indicates a reward-related deficit in MDD and the underlying disrupted VTA-NAc circuit. Shi et al. reported increased resting-state FCs between the bilateral VTA and ventral striatum in MDD compared to healthy controls using an ROI-to-ROI analysis. Ding [59] discovered significant differences in functional connectivity of the NAc and VTA between patients with recurrent MDD and healthy controls. Other studies have also shown that under the action of addictive drugs such as morphine, dopamine (DA) neurons in the VTA are activated, and the level of DA in the NAc is significantly increased. The psychological pain theory of suicide considers that suicide is the only solution for an individual to escape unbearable psychological pain [60]. Suicide attempts can improve the patient’s mood and relieve psychological pain in the short term, the patients suffering from depression may resort to suicide as a means to alleviate their emotions and psychological pain [61]. Thus, we hypothesize that the aberrant connectivity of the VTA increases levels of pain avoidance during a major depressive episode which may be a dominant component of the motivation for suicide. Willeumier [51] found depressed patients with suicidal behavior exhibited reduced regional cerebral blood flow (rCBF) in the VTA region. In addition, cognitive psychological research of depressed adolescents with suicide attempt have demonstrated that suicidal behavior was associated with impairments in reward response, impaired top-down cognitive control of emotions, and selective overattention to negative emotional stimuli. The VTA plays critical roles in reward motivation and reappraisal, emotional regulation, cognitive control and reasoning, this suggests that the abnormal VTA connectivity within the reward network may cause the occurrence of suicidal behavior by influencing cognitive dysfunction.

Additionally, altered dynamic FC between the pre-SMA and the inferior frontal gyrus in the sMDD group was identified in this study. The function of supplementary motor area (SMA) is related to the planning of movement, primarily involved self-generated and controlled movements rather than those initiated by external stimuli. The SMA is involved in the sequence of movements generated from memory. Previous studies have shown that the SMA is involved in the occurrence of suicidal behavior [62]. Tymofiyeva [62] examined the whole-brain resting state functional connectivity (RSFC) of six putamen regions and found that depressed adolescents with suicide attempts exhibited lower RSFC with the superior frontal gyrus and supplementary motor area compared to those without a suicide attempt history. This study suggests that connectivity between putamen and the supplementary motor area may be used as a valuable biomarker of treatment efficacy and possibly prediction treatment outcomes for suicide in depressed adolescents. Another previous study conducted by Wagner and colleagues revealed relatives of suicide victims exhibited decreased RSFC of the fronto-cingulo-striatal network including the bilateral supplementary motor area compared with healthy subjects [63]. This suggests the existence of a heritable association with suicidal vulnerability. The sensorimotor area is an inhibitory structure which plays a critical role in exerting control over voluntary actions in situations of response conflict and impulsivity [64]. Thus, according to the research described above, we speculate that the altered dynamic functional connectivity of the inferior frontal gyrus with SMA, which is related to impaired response inhibition and high impulsivity, increases the vulnerability to suicide among adolescents and young adults.

Dynamic functional connectivity of the reward network was not significantly correlated with either HAMD or SSI scores, indicating that temporal fluctuations in reward network coupling do not bear a simple linear relationship to depressive severity or suicidal ideation in MDD. Thus, altered dynamic reward network coupling is unlikely to be the proximate driver of affective symptoms, instead, it may constitute a risk mechanism for suicide-related behavior or a compensatory process that emerges after suicide attempt. In addition, we did not observe any correlation between these aberrant connections in reward network and HAMD and SSI scores. One possible reason is that our patients had relatively high HAMD ratings, indicating severe illness. The relative small range of high HAMD scores of the enrolled patients in the present study might be one of the reasons. Second, there is a lack of clinical information such as illness duration or age at onset, which could influence both findings and their interpretation. Third, a cross-sectional design was employed in this study so that it is not possible to know whether the altered brain dFC in reward network reflects a state or trait effect in adolescent and young people with MDD. More studies are required in the future to further elucidate the characteristics of dFC of reward network and the associations between them in different stages and cohorts of MDD and suicide in adolescent and young adults.

There are several limitations that must be noted in this study. First, this study was a cross-sectional design, with no longitudinal imaging observed and it was impossible to observe changes in the dynamic FC of the reward circuit in children prior to the occurrence of suicidal behavior. Thus, the causality of the relationships between longitudinal changes in the reward circuit and the occurrence and development of suicidal behavior in depressive adolescents and young adults remains unclear. In future studies, we can conduct a follow-up study on a cohort of MDD patients to compare their neural activity differences before and after the onset of suicidal behavior in order to further verify our conclusions. Second, we could not fully control the differences in antidepressant treatment and illness duration between sMDD and nMDD patients. The heterogeneity in concomitant medication types, dosages, and durations among MDD patients may introduce potential confounding factors. This study found abnormal dynamic functional connectivity (dFC) in brain regions associated with the reward circuit in both sMDD and nMDD groups. However, it remains unclear whether this abnormality represents a drug-induced compensatory mechanism or inherent pathological characteristics of the patients, which needs further study. We can control the effect of drugs on the results by comparing magnetic resonance imaging (MRI) of MDD patients treated with the same drugs in future studies. Moreover, we did not conduct a direct comparison between static and dynamic connectivity. Consequently, although only dynamic connectivity achieved significance, we cannot determine which metric is more strongly associated with suicidality. Future studies should explicitly compare the explanatory power of both approaches. Additionally, this study did not provide a detailed classification of depression patients, such as distinguishing between first-episode and recurrent depression. Consequently, we could not exclude the possibility that the refractory nature of MDD, which has been reported to be associated with functional brain networks and suicidality [65, 66], may have influenced our results. Moreover, the severity of depressive symptoms was not evaluated, nor was the lethality of suicidal behaviors assessed and analyzed. In addition to the above limitations, the relatively modest group size may have influenced the statistical power, and future studies are needed to validate the findings in a larger sample size.

In summary, we investigated the dynamics of functional connectivity patterns within the reward network indepressed adolescents and young adults with and without suicide attempts using dFC analysis. Our dFC analysis revealed significant differences in dynamic functional connectivity between the left ventral striatum and putamen, left anterior insula and left medial superior frontal gyrus, right anterior insula and right anterior central gyrus, and anterior auxiliary motor area and inferior frontal gyrus, and the anterior auxiliary motor area and the inferior frontal gyrus among the sMDD, nMDD and HC groups. This study reveals that alterations in reward network function are associated with suicidality, which demonstrate that suicide is more closely linked to abnormal brain network function, rather than to local function. Together, these results provide objective imaging markers for suicide risk in depressed adolescents and young adults and may enhance our understanding of the neurobiological mechanisms underlying suicide and depression in this age group.

Author contributions

Jun Cao and Xiaorong Chen designed the study, collected data and analysed the data. Jianmei Chen, Xuemei Li, Ming Ai and Wo Wang tested the participants.Yao Gan and Jinglan He recruited and scheduled participants. Haitang Qiu and Li Kuang designed the study and revised the manuscript. All authors have read and approved the submission of this manuscript.

Funding

This study was supported by a grant from the Natural Science Fund of the Chongqing Science and Technology Commission (CSTB2023NSCQ-MSX0227, CSTC2018jcyjAX0164), the Special Key Project of Chongqing Technology Innovation and Application Development (CSTC2021jscx-gksb-N0002) and the Medicine Scientific Research Fund for Youth from the Chongqing Health and Family Planning Committee (2023QNXM006, 2018QNXM014). Additionally, funding was received from the Scientific Research and Cultivation Project of the First Affiliated Hospital of Chongqing Medical University (PYJJ2018-20) and the National Natural Science Foundation of China (81671360, 81901373).

Data availability

Not applicable.

Declarations

Ethics approval and consent to participate

The present study was approved through the Medical Ethics Committee of the First Affiliated Hospital of Chongqing Medical University, P.R. China (NO.2020-97-2), and the protocols were performed according to the standards of the Declaration of Helsinki. All the participants recruited into the study were fully informed of the details of the study and agreed to participate in this experiment. For individuals under the age of 16, their parents signed the informed consent after the study procedures were explained in detail.Written informed consent was obtained from each participant or parents (for under 16 years old).

Consent for publication

Not applicable.

Competing interests

The authors declare no competing interests.