Fifty-nine right-handed subjects who all experienced the same Wenchuan 8.0-magnitude earthquake were recruited (Table 1). The acquisition of neuroimaging and clinical data from survivors took place in December 2009, 17 months after this disaster. All participants were children or adolescents, who were between 8 and 18 years of age (35 PTSD with a mean age of 14.74 ± 2.08 years and 24 non-PTSD controls with a mean age of 14.58 ± 1.79 years) and didn’t suffer any physical head injury or any loss of consciousness > 5 min in the catastrophe. They were drawn from a large-scale PTSD survey of survivors 8–15 months after the earthquake. They were first carefully screened through the PTSD checklist (PCL) (32, 33). Subjects with a PCL score > 35 further participated in interviews led by two experienced psychiatrists. The Clinician-Administered PTSD Scale (CAPS) (34) was used to confirm the PTSD diagnosis in these individuals with suspected PTSD, and the structured clinical interview for DSM-IV (SCID) was used to exclude any psychiatric comorbidities. Those who scored > 50 on the CAPS were eligible for further evaluation for inclusion in the PTSD group. An age-and sex-matched trauma-exposed non-PTSD group was formed from those with a PCL score < 30 points and a CAPS score < 35 points. The exclusion criteria for the PTSD group were: (1) history of other nervous system diseases and psychiatric disorders; (2) use of psychotropic medications and drug treatment in the past 2 months; (3) any significant medical or history of head injury; (4) left-handedness; (5) IQ < 80; (6) magnetic resonance imaging contraindications. The traumatized control participants were similarly screened using the SCID and CAPS scale. These subjects who experienced the same traumatic events but with CAPS scores below 35 were included in the non-PTSD group. The non-PTSD control group used the same exclusion criteria as the PTSD group. An experienced neuroradiologist inspected brain scans and ruled out clinical abnormalities. This research was approved by the Medical Ethics Committee of West China Hospital, Sichuan University, and all subjects and their guardians offered written informed consent after being provided a complete description of this study.

Images were acquired on a Siemens 3.0T Trio TIM MRI scanner at the West China Hospital. All subjects underwent high resolution three-dimensional T1-weighted anatomical brain scans (MPRAGE, 176 slices, TR/TE = 1900/2.26°msec, flip angle = 9°, acquisition matrix = 256 × 256, resolution = 1 × 1 × 1°mm³).

The T1-weighted MRI data were preprocessed in FreeSurfer (Version 7.1.0¹). The image processing includes the following main steps: motion correction, skull strip, Talairach transform computation, subcortical grey matter (GM)/white matter (EM) segmentation, intensity normalization, WM-GM boundary tessellation, automatic topology fixer, and spherical registration. Then, the cortical surface of individuals is registered to a standard spherical map after inflation by the spherical registration method. And sulcal and gyral features were recognized automatically by the software. The CTh data were smoothed by a 10-mm FWHM Gaussian kernel for reducing measurement noise and improving statistical power.

We predefined twenty-five region-of-interests (ROIs) according to our previous research (35) in Table 2. The specific strategy was followed: each candidate seed was originally identified in at least two individual published neuroimaging meta-analyses on PTSD. These ROIs were modeled by spheres of 20°mm diameter around each peak coordinate and chosen as seeds to create structural covariance connectivity maps for a hypothesis-driven approach.

All cortical statistical analyses were performed using the SurfStat toolbox² within MATLAB www.mathworks.com. And for the problem of multiple comparisons, each cluster was corrected with a random-field-theory (RFT) (36) at ap < 0.05 level of significance, limiting the probability of reporting a family-wise error to below 0.05. Two-sample t-test and chi-square analyses in SPSS software³ were used to compare the demographic characteristics between trauma-exposed non-PTSD and PTSD participants.

The comparisons of demographic characteristics between trauma-exposed non-PTSD and PTSD participants are provided in Table 1. The average gender and age of PTSD individuals were alike to the control’s, as were education and height, and weight. In addition, the groups differed in CAPS scores.

The Surface-based morphometry analysis of CTh highlighted that PTSD patients displayed a significantly thinner cluster of the left medial orbitofrontal cortex (mOFC) than non-PTSD control subjects (P = 0.0265, corrected with RFT) (Figure 1). No difference was seen in PTSD patients> non-PTSD subjects.

The seed-based structural connectivity analyses between PTSD and non-PTSD control groups are presented in Figure 2 and Table 3. Findings showed that there were three types of alteration of cortical structural covariation for seed regions of interests: positive contrast (PTSD groups > non-PTSD controls), negative contrast (PTSD groups < non-PTSD controls), and both positive and negative contrasts. When considering positive contrast, there were two seed regions located in the auditory association cortex [left superior temporal gyrus (STG)] and motor cortex (right precentral gyrus) that showed increased covariance with other brain vertices in PTSD patients compared with non-PTSD participants between CTh. When considering the negative contrast, the results showed that lesser connectivities of ten seed regions located in the default mode network (DMN) regions [including left posterior cingulated cortex (PCC), right precuneus, right medial prefrontal cortex/anterior cingulate cortex (mPFC/ACC) and right inferior parietal lobule (IPL)], midline cortex structures [including bilateral medial frontal gyrus (MFG), right mid-cingulate cortex (MCC)], visual cortex [including left middle occipital gyrus (MOG) and left inferior temporal gyrus (ITG)], and limbic system (right amygdala) were found in PTSD patients compared with non-PTSD participants between CTh. When both positive and negative sides were considered, there were two seed regions including the left insula and precentral gyrus.

In summary, covariance networks found were centered on six key regions, and representative images are shown in Figure 3. These were connectivities: (1) of seed regions including the right mPFC/ACC, right precuneus, right MCC, left insula, left precentral gyrus, and bilateral MFG with the bilateral entorhinal (Figure 3A); (2) of seed regions including the left MFG, left STG, right MCC, and bilateral precentral gyrus with the left parahippocampal (Figure 3B); (3) of seed regions including the right amygdala, left MOG, left precentral gyrus, and left ITG with the bilateral superior parietal gyrus (SPG) (Figure 3C); (4) of seed regions including the left MFG, right IPL, right precentral gyrus, and left ITG with the bilateral supramarginal gyrus (SMG) (Figure 3D); (5) of seed regions including the left STG, right IPL, and left ITG with the bilateral PCC (Figure 3E); (6) of seed regions including the left insula and precentral gyrus with the bilateral precuneus (Figure 3F). All these findings survived the p-RFT < 0.05 correction. For RFT-corrected maps depicting SCNs of all reported findings, please see Supplementary Figures 1–18.

Post-traumatic stress disorder (PTSD) patients highlighted significantly thinner cortex in the left mOFC than non-PTSD groups. One of the leading factors in PTSD is the failure of fear extinction, and the reported mOFC which plays an important role in extinction learning is closely associated with the fear loop (38–40). The mOFC in classical experiments (41–43) has been widely considered to be involved in the retention or recall of extinction learning. In addition, the brain activity in the mOFC region decreased during experiencing a trauma (44), which seemed to reflect the lack of the “switch-off” mechanism of fear at this point (45). Furthermore, evidence from neuroimaging research suggested a positive correlation between the cortex thickness of mOFC and extinction recall (46, 47). And Morey et al. found that there were smaller orbitofrontal volumes in minors with PTSD (48). Taken together, individuals will maintain a better extinction ability to respond to fear when mOFC regions are activated more. In contrast to that, the thinner cortex in the mOFC of PTSD patients found in this research may reflect the reduced flexibility to control fear and may make children and adolescents more vulnerable to traumatic events. Therefore, the abnormalities in mOFC areas might be the outcome of disrupted extinction recall mechanisms for fear in PTSD.

Covariance networks in the current study were significantly correlated with brain activations in areas centered on six key regions: the bilateral entorhinal, the left parahippocampal, the bilateral SPG, the bilateral SMG, the bilateral PCC, and the bilateral precuneus. Previous studies on the SCNs of PTSD have shown the structural integrity of the cingulate region, the cingulum bundle, and/or the amygdala or amygdala and other frontal regions in the adult population (49–52). At the same time, there are only a few reports about the structural connectivity changes derived from cortical thickness for minors with PTSD, which reported abnormalities in the left PCC, the right inferior frontal cortex, and the left ACC (53, 54). The current study discovered abnormal integrations of the structural connectivity of the entorhinal cortex, parahippocampal, and parietal cortex for children and adolescents with PTSD. These extensive abnormalities of integrated changes of structural connectivity may reveal why children and adolescents might be more vulnerable to the harmful effects of disastrous events, which will continuously threaten the health and welfare of current and future generations calling for more post-disaster policy attentions to the trauma problems faced by children in the process of growth and development after the earthquake and giving them long-term interventions and support (55, 56).

The entorhinal, parahippocampal, PCC, and precuneus are important components of the episodic memory network which have been proved to be often activated during episodic recall in PTSD patients by imaging studies (57–62). The entorhinal cortex and parahippocampal work together for spatial positioning and the formation of declarative memory (63). As the primary interface for information flow between the neocortex and the parahippocampal formation, the entorhinal cortex receives multimodal information from other cortical regions and then transmits them to the parahippocampal (64). We observed reduced structural connectivities in DMN regions (including right mPFC and right precuneus), midline cortex structures (including bilateral MFG and right MCC), the motor cortex (right precentral gyrus) with the left entorhinal and in left insula with the right entorhinal, which may be a result of inhibition of information flow between entorhinal cortex and neocortex associated with unique symptoms of emotional processing, re-experiencing memory and awareness of bodily states in PTSD youths. Additionally, the disturbance of structural connectivity in the parahippocampal might be the basis for sensory and contextual memory defects in posttraumatic patients (65, 66). The results that there were decreased structural connectivities in the midline cortex structures (including left MFG and right MCC)) and increased structural connectivities in the motor cortex (right precentral gyrus) and auditory association cortex (left STG) with the left parahippocampal gyrus supported the dual representation theory of PTSD (67), which suggest a dissociation between sensory and contextual memory representations in PTSD. This abnormal alteration of connectivity in the left parahippocampal gyrus is consistent with the previous structural covariance network studies showing that there was an increased correlation between the limbic system and the visual-related cortex (51) and impaired integration of the prefrontal limbic network with other parts of the brain in PTSD patients (68). Similar dissociation between hyperactive sensorimotor regions and hypoactive memory-associated regions was reported during PTSD-related flashbacks (69, 70). Viard et al. indicated decreased within-DMN connectivity of PCC in adolescents with PTSD (71), so the decreased structural connectivity in the visual cortex (left ITG) and DMN regions (right IPL) is known for their roles in visual mental imagery and contextual cue processing (67, 72) with the left PCC and the increased structural connectivity between auditory association cortex (left STG) and the right PCC might relate with distorted images and sounds, dysfunctional autobiographical memory retrieval. The precuneus which responds to multiple cognitive processes could be divided into regions involved in cognition, sensorimotor, and visual processing (73). Our results that seed regions in the left insula and precentral gyrus were structurally connected with the bilateral precuneus confirm this separation and might help explain the complex clinical manifestations of PTSD patients. From the above, our findings of altered structural connectivity of the four areas in the episodic memory network may be the foundation of trauma re-experiencing and recalling in children and adolescents with PTSD.

In addition, there were different connections in the bilateral SPG and SMG for PTSD patients compared with controls. The SPG is known for a crucial role in the early integration of visuospatial information carried by somatosensory, proprioceptive, and auditory stimuli (74, 75). And SMG is the secondary somatosensory cortex which can integrate exogenous and internal-sensory information (76). The precentral gyrus has an increased structural connection with SPG and SMG may indicate that PTSD patients are more difficult to process tactile and proprioceptive visuospatial sensory information related to external cues to carry out somatic drive. The decreased connection in the amygdala, MOG, and ITG with SPG suggests that PTSD patients have a reduced ability to integrate visuospatial information. Besides, the decreased connection between the inferior temporal gyrus and supramarginal gyrus may reveal the decreased spatial perception of PTSD patients. Consequently, these connection changes in SPG and SMG may reflect the reduction of visuospatial information integration ability in the youth population with PTSD.