Association between abnormal default mode network activity and suicidality in depressed adolescents

All study participants were Chinese Han people. Fifty-three patients with depression who were between 15 and 25 years old were recruited from the inpatient and outpatient units of the First Affiliated Hospital of Chongqing Medical University and University-Town Hospital of Chongqing Medical University from April 2012 to December 2013. The patient cohort was divided into two groups: a suicidal depressed group (SD group) composed of 35 patients with depression and a history of at least one suicide attempt and a non-suicidal depressed group (NSD group) composed of 18 patients with depression and no history of attempted suicide. A suicide attempt was defined as a self-injurious act associated with at least some intent to die as a result of the act, regardless of whether the attempt resulted in actual injury. Evidence that an individual intended to kill him/herself could be explicit or inferred from the individual’s behavior or circumstances [35]. The Structured Clinical Interview for DSM-IV-TR Axis I disorders (SCID-I/P) was used to identify adolescents with depressive symptoms and to develop a healthy control (HC) group [36]. The enrolled suicidal and non-suicidal depressed patients refrained from the use of antidepressants for at least 2 weeks prior to the start of the study and from the use of electroconvulsive therapy for at least 4 weeks prior to the start of the study. A total of 47 sex- and age-matched individuals were recruited from the local community to serve as the HC group. The following exclusion criteria were employed: history of neurological illness; severe traumatic brain injury; current or previous psychiatric disorder other than depression; heart, liver, or kidney disease or other serious physical illness; any type of contraindication for MRI; and substance dependence or abuse. The Intelligence Quotient (IQ) of each subject was measured using the Chinese Combined Raven’s Test, Copyright 2 (CTR-C2) [37]. The Hamilton Depression Rating Scale (HDRS, [38]) and Beck Depression Inventory (BDI, [39]) were used to evaluate depression severity. The Suicide Attitude Questionnaire (SAQ, [40]), a 29-item-long self-reported checklist, was used to describe attitudes related to suicide along 4 dimensions: (1) view of suicidal behavior, (2) attitude towards suicide survivors, (3) family history of suicide attempts, and (4) perspective on euthanasia. The Scale for Suicide Ideation (SSI, [41]) was also employed; higher scores on this scale indicate a stronger intention to commit suicide. The Barratt Impulsiveness Scale-11 (BIS-11, [42]) was used to measure impulsive personality traits, and the Beck Hopelessness Scale (BHS, [43]) was used to measure negative attitudes about the future. All subjects personally agreed to participate in this study. Subjects who were older than 18 years signed written informed consent before the experiment. For subjects younger than 18 years, our group contacted family members to obtain written informed consent from the subjects’ legal guardians. The study was approved by the Ethics Committee of the First Affiliated Hospital of Chongqing Medical University in China.

Images were acquired using a GE Signa Hdxt 3.0-Tesla scanner (General Electric Medical Systems, Milwaukee, WI, USA) with a standard eight-channel head coil; the scanner was located at the First Affiliated Hospital of Chongqing Medical University. Foam padding was used to minimize head motion and machine noise. Each participant was instructed to lie down and relax, to keep their eyes closed, to stay awake and to avoid performing specific cognitive tasks. Conventional T1-weighted images (using fast spin echo (FSE), repetition time/echo time (TR/TE) = 500 ms/14 ms, thickness/gap = 5.0/0 mm, flip angle = 45°, NEX = 1, field of view (FOV) = 24×24 cm, and matrix = 256×126) and high-resolution 3D T1 images (using fast gradient echo (FGRE), TR/TE = 24 ms/9 ms, flip angle = 90°, thickness/gap = 1.0/0 mm, FOV = 24×24 cm, and matrix = 256×256) were acquired. rs-fMRI images were obtained using an echo-planar image (EPI) pulse sequence with the following parameters: 33 axial slices, TR/TE = 2000/40 ms, matrix = 64×64, flip angle = 90°, FOV = 240×240 mm², thickness/gap = 4.0/0 mm. A total of 240 time points were axially recorded over 8 min. No obvious structural damage was found, and none of the subjects felt discomfort during or after the procedure.

All resting-state functional images were analyzed using Data Processing Assistant for Resting-State fMRI Advanced Edition (DPARSF; http://​www.​restfmri.​net) [44] and the REST 1.9 (http://​www.​restfmri.​net) [45] toolbox in Matlab version 7.8.0 (MathWorks, Inc., CA). The first 10 time points were discarded to allow for scanner calibration and participant adaptation to the scanning environment. The preprocessing steps included slice timing, head motion correction, spatial normalization in Montreal Neurological Institute (MNI) space (resampling with 3 × 3 mm3 resolution), smoothing (a smoothing kernel with FWHM of 4x4x4 mm³) [46], linear trend removal and filtering (0.01–0.08 Hz). Six parameters associated with head motion signals were regressed out (3 for shift and 3 for rotation), and the overall head motion was calculated as the average magnitude of the head motion. No significant differences were found in the root mean square (RMS) head movement among the three groups (p = 0.35). There was less than 1 mm maximum displacement in subject head movement in any direction (x, y, and z) and less than 1° maximum displacement in any angular dimension. The preprocessed images of each subject were subjected to group ICA to identify functional networks and their temporal activities using a MICA toolbox (http://​www.​nitrc.​Org/​projects/​cogicat) [47]. Three-stage principal component analysis (PCA) was used to reduce the dimensionality of the data to 40. The Infomax algorithm [48] was used for ICA decomposition; this analysis was repeated 100 times to achieve robust and accurate results. Then, spatially independent components (ICs) were back-reconstructed for each subject, and subject-specific spatial maps and time courses were acquired and z-score transformed. The ICs were grouped into different brain networks of interest, including the DMN, based on visual inspection, previous investigations and reported spatial patterns [49, 50].

Statistical analyses of clinical and demographic subject characteristics were conducted using SPSS 17.0 software. One-way analysis of variance (one-way ANOVA) was used to evaluate group differences in age, years of education, IQ scores, and scores on the six included clinical scales. Post hoc pairwise comparisons were performed using least significant difference (LSD) t tests. Gender differences among the three groups were assessed using the chi-squared test. The brain rs-fMRI maps of the three groups were analyzed using REST 1.9 software. First, to characterize a selected DMN, individual spatial maps from the SD, NSD and HC groups were separately subjected to a group-specific one-sample t-test with AlphaSim multiple comparisons correction within a whole-brain mask using a voxel-wise threshold of p < 0.05 and a cluster threshold of ≥85 voxels (1000 Monte Carlo simulations) [51]. Then, the brain regions significantly connected within the DMN at the group level in the three groups were combined and investigated using one-way ANOVA for between-group comparisons, which revealed whether there were several regions in which there was a main effect or group difference. Clusters >324 mm³ (12 voxels) with p < 0.01 (as determined by multiple comparisons correction using the AlphaSim tool in the REST software package with rmm = 5) were deemed potential ROIs. The resultant ROIs were used to create a mask for further analysis. To further reveal the directions of the group differences, post hoc t-tests were applied to compare connectivity differences in the mask between any two groups. Post hoc two-sample t-tests were restricted to the voxels showing significant differences by one-way ANOVA. Subregions with corrected p < 0.01 and a cluster size >324 mm³ (12 voxels) were considered significantly different between groups. In the case of type I error, we extracted the average time course of each ROI in the mask for the three groups and compared them using one-way ANOVA and post hoc pairwise comparisons (LSD t-tests) (p < 0.01).

One-way ANOVA and chi-squared tests showed that the scores on all six clinical scales significantly differed among the three groups (all p < 0.001) (Table 1). There were no differences in gender, age, educational level or IQ score. Post hoc analysis showed that the scores on all six scales significantly differed between the HCs and the depressed patients with and without suicide history (p < 0.001), whereas no differences were found between the depressed patients with and without suicide history (see Table 1).

Compared with the HCs, the depressed patients with and without suicide history showed increased functional connectivity in the right MTG, right middle occipital gyrus and left middle frontal gyrus. Moreover, the suicidal patients showed increased connectivity in the left cerebellum [−6 −75 −18] and decreased connectivity in the right PCC [0 –48 3], whereas the non-suicidal depressed patients showed increased connectivity in the left superior frontal gyrus, left lingual gyrus, and right precuneus and decreased connectivity in the left cerebellum. Compared with the non-suicidal depressed patients, the suicidal group showed increased connectivity in the left cerebellum [−10 −75 −20] and left lingual gyrus [−13 −57 −1] and decreased connectivity in the right precuneus [18 −60 30] (see Fig. 1 and Table 2).

This study requires replication and further verification in a larger patient population. We found no differences in total scores on the BIS, SSI, SAQ and BHS scales between the suicidal and non-suicidal depressed groups; this result may be due to the severity of the depression in the enrolled patients and the medical interventions the patients received before they were enrolled. Additional factors, including degree of depression, number of suicide attempts, family history of suicide, and medication load before enrollment, should be controlled in future studies because these factors may result in structural and functional DMN changes.