Anatomical insights into disrupted small-world networks in pediatric posttraumatic stress disorder: a diffusion tensor imaging tractography study
Xueling Suo,Du Lei,Fuqin Chen,Min Wu,Lei Li,Ling Sun,Xiaoli Wei,Hongyan Zhu,Lingjiang Li,G. Kemp,Qiyong,Gong
2016-01-01
Abstract:Purpose: To use DTI and graph theory approaches to explore the brain structural connectome in pediatric posttraumatic stress disorder (PTSD). Materials and Methods: This study was approved by the relevant research ethics committee, and all subjects’ parents/guardians provided informed consent. Twenty-four pediatric PTSD patients and 23 trauma-exposed non-PTSD controls were recruited after the 2008 Sichuan earthquake. The structural connectome was constructed using diffusion tensor imaging tractography, by thresholding the mean fractional anisotropy of 90 brain regions to yield 90×90 partial correlation matrixes. Graph theory analysis was used to examine the group-specific topological properties, and nonparametric permutation tests were used for group comparisons of topological metrics. Results: Both groups exhibited small-world topology. However, PTSD showed increase in the characteristic path length (Lp) (P=0.0248), and decrease in local efficiency (Eloc) (P=0.0498) and global efficiency (Eglob) (P=0.0274). Furthermore, PTSD showed reduced nodal centralities mainly in the default mode, salience, central executive and visual regions (P < 0.05, false discovery rate corrected). The Clinician-Administered PTSD Scale score was negatively correlated with nodal efficiency of left superior parietal gyrus (P=0.043). Conclusion: The structural connectome showed a shift toward ‘regularization’, providing a structural basis for functional alterations of pediatric PTSD. These abnormalities suggest that PTSD can be understood by examining the dysfunction of large-scale spatially distributed neural networks. Introduction Brain structure can be interpreted as an integrated network using concepts of graph theory which quantify the whole brain as a single graph, comprising nodes linked by edges (1). Specifically, the small-world network pattern, which seems to have evolved to mediate high-efficiency parallel information transfer (2), has been used to define pathology in psychiatric disorders, including posttraumatic stress disorder (PTSD) (3-5). PTSD is a traumaand stressor-related disorder characterized by four symptom clusters: re-experience, avoidance, negative cognitions and mood, and arousal (6). Pediatric PTSD is not uncommon, the prevalence among children (12–17 years) being 3.7% for boys and 6.3% for girls (7). Childhood trauma is a severe stressor with multiple neurochemical and hormonal effects which can lead to lasting changes in brain structure and function (8). Children are particularly susceptible to PTSD (9), which may adversely influence brain development. Early interventions may help prevent brain changes. Most early structural neuroimaging studies investigating the impact of childhood trauma on white matter (WM) integrity in children (10, 11) used manual tracing or volumetric morphometry. Diffusion tensor imaging (DTI) has since emerged as a powerful technique to assess WM tracts by exploiting the diffusion of tissue water. Using DTI we identified whole-brain WM microstructural abnormalities in pediatric PTSD (12), but we did not explore the neurocircuitry directly using connectivity analyses. We also applied graph theory to functional magnetic resonance imaging (MRI) data, finding that the functional connectome in pediatric PTSD is shifted toward ‘regularization’ (from a small-world to a more regular network) (4). As functional interaction is constrained by brain cortical anatomy (13), to fully understand function one must study its structural substrate directly. Based on our findings in the functional connectome (4), we hypothesized that in pediatric PTSD the structural connectome would show a similarly disrupted topological organization. Our purpose was to use DTI and graph theory approaches to explore the brain structural connectome in pediatric posttraumatic stress disorder (PTSD). Materials and Methods Participants This study was approved by the local research ethics committee. Each child’s parent/guardian was given a detailed information sheet, and then gave written consent. A total of 4,200 earthquake survivors were screened by M.W. and X.W. 8-15 months after the 8.0 magnitude earthquake in Sichuan in May 2008. Each participant was interviewed and screened using the PTSD checklist (PCL) (14); those scoring >35 on PCL were given the Clinician-Administered PTSD Scale (CAPS) (15) by a psychiatrist (L.L., 31 years’ experience), of which those scoring >50 on CAPS were diagnosed with PTSD; those scoring <30 on PCL were considered non-PTSD trauma-exposed controls and were not assessed using CAPS (16). Inclusion criteria for all participants were: personal experience of the earthquake; personal witness of death, serious injury, or building collapse; age <18 years; IQ >80. This identified 161 PTSD patients and 99 trauma-exposed non-PTSD controls with similar demographic characteristics, lifestyle and earthquake experiences. Exclusion criteria were: psychiatric co-morbidities assessed using the Structured Clinical Interview for DSM-IV (17); history of psychiatric or neurological disorders (n=42); MRI contraindication (n=30); recent medication that might affect brain function (n=24); unavailability of key data (n=12); left-handedness (n=10); CAPS score >35 but <50 (n=8); history of or current brain injury (n=7). Twenty-eight drug-naïve first-episode PTSD patients and 26 trauma-exposed non-PTSD controls underwent MRI scanning. Head-motion artifacts excluded data from 4 PTSD and 3 controls. MRI data from 24 PTSD and 23 controls went forward for analysis. We have reported elsewhere some other MR data from some of these subjects. In (4) we reported resting state functional MRI data of 24 PTSD and 24 controls: that work investigated the brain functional connectome, while our current work explores the brain structural connectome. In (12) we reported DTI data of 27 PTSD and 24 controls: that work investigated the microstructural networks using voxel-based analysis, while our current work explores the structural network directly using connectivity analyses. Data Acquisition MRI data were acquired on a 3T MRI system (EXCITE; General Electric) using a single-shot spin-echo echo planar image (SE-EPI) sequence, and included one high-resolution T1 scan and one DTI data scan. Foam padding was used to minimize head motion. A whole-brain high-resolution T1-weighted image was acquired using a sagittal three-dimensional spoiled gradient recall (SPGR) sequence with repetition time (TR) = 8.5 ms, echo time (TE) = 3.4 ms, inversion time (TI) = 400 ms, slice thickness = 1 mm, no inter-slice gap, 156 axial slices, matrix size = 256×256, field of view (FOV) = 24×24 cm and flip angle = 12°. The diffusion sensitizing gradients were applied along 15 non-collinear directions (b-value = 1000 s/mm) together with an acquisition without diffusion weighting (b = 0). Imaging parameters were TR = 12000 ms, TE = 71.6 ms, number of excitations (NEX) = 2, slice thickness = 3 mm, 50 slices, 128×128 matrix and 24×24 cm FOV. The protocol included susceptibility-weighted imaging (SWI) which will be analyzed in a future study and fluid attenuated inversion recovery (FLAIR) sequences which were evaluated for clinical abnormalities by a neuroradiologist. A radiologist (L.S., with 3 years experience) evaluated and verified image quality. Data Pre-Processing and DTI-Based Structural Network Construction All the image preprocessing and analyses were implemented using a pipeline tool for diffusion MRI (PANDA) (18). We (X.S. and D.L.) extracted the fractional anisotropy (FA) map of each subject in 3 steps: BET (skull removal), eddy correct and DTIFIT (building diffusion tensor models). We then registered the FA maps with the FMRIB FA template in standard MNI space using nonlinear registration. The automated anatomic labeling atlas (90 regions) was used to define the nodes of the WM network. PANDA uses the procedure proposed by Gong et al (19). Briefly, each of the individual FA images in native space was co-registered to its corresponding T1-weighted image using an affine transformation. Then the transformed T1-weighted images were non-linearly registered to the MNI space. The inverse transformations were obtained to the above two steps to transform the automated anatomic labeling atlas from MNI space to DTI native space. Thus, the individual cerebrum in native space was divided into 90 nodes corresponding to the automated anatomic labeling atlas. Each node represents a region of the DTI-based structural brain network. Deterministic tractography was performed to reconstruct whole brain WM tracts using the Fiber Assignment by Continuous Tracking algorithm (20). A tract was terminated if the turn angle was >45° or the fiber entered a voxel with FA <0.2 (21). We defined the averaged FA of the linking fibers for each connection. For each individual, we generated a symmetric 90×90 network matrix in which each row/column represents a brain node/region and each element represents the averaged FA of the linking fibers between nodes. Using the GRETNA toolbox we investigated the topological properties of brain networks at both the global and nodal level. The global level properties were of two kinds: small-world parameters [for definitions see (2)], including the clustering coefficient Cp, characteristic path length Lp [calculated as the harmonic mean distance between all possible pairs of regions to address the disconnected graphs dilemma(22)], normalized clustering coefficient γ, normalized characteristic path length λ, and small-worldness σ; and network efficiency parameters [for definitions see (23)], including the local efficiency Eloc and global efficiency Eglob. The nodal level properties were the nodal degree, nodal efficiency, and nodal betweenness. For each network metric we calculated the area under the curve (AUC) over the sparsity range from S1 to Sn with an interval of ΔS, where S1 = 0.10, Sn = 0.34 and ΔS = 0.01. The AUC provides a summarized scalar for t