Patients with schizophrenia (SZ) are at increased risk of engaging in violence toward others (1, 2) and account for around 10% of all homicide offenders (3, 4). A 38-year total population study in Sweden showed that 10.7% of men with SZ and related non-affective psychoses were convicted of violent offenses within 5 years from their initial diagnosis (5). Given that violence is a multifaceted phenomenon (6, 7), the neurobiological basis of violent offenses in patients with SZ is still unknown. Thus, the exploration of neurobiological underpinnings of violence in patients with SZ can help inform the biological evaluation of the risk for violence as well as develop individualized disease prediction and intervention for violent behavior.

Numerous studies have reported abnormal morphological changes in violent offenders with schizophrenia (vSZ), which involved the frontal and temporal lobes, as well as subcortical nuclei (8–10) compared to non-violent patients with schizophrenia (nvSZ). For instance, studies suggested that reduced gray matter volume (GMV) in the frontal lobe, temporal lobe (10–12), and hippocampus (13–15) were associated with violent behavior in patients with SZ. Despite the above discoveries, inconsistent results were apparent. A crucial region belonging to the limbic system, the amygdala, has been involved in controversial results (9, 16), as some studies reported increased volume of the amygdala in vSZ relative to nvSZ (13, 17, 18), while others reported opposite results (19, 20). Another region exhibiting controversial findings was the orbitofrontal cortex. For instance, Hoptman et al. found that larger gray and white matter volumes were associated with higher levels of aggression in patients with SZ (21). Furthermore, a few neuroimaging studies reported no significant differences in the brain structure between vSZ and nvSZ (17, 22–24). The mixed findings might be derived from the heterogeneous sample characteristics and various definitions of violence (20, 25, 26). Regarding these issues, more study exploring the intrinsic brain structure of violence in patients with SZ is needed.

Reasons for violent behavior in SZ patients can be classified into intrinsic and external factors, such as young age, deviant personality, childhood trauma, and substance abuse (27), which may contribute to neurobiological variability of violence in SZ. An explicit viewpoint is that those domains that exist in different populations might be associated with the pathology of violence (28), suggesting that intrinsic factors, such as the deviant personality traits, were more reflective of the neurodevelopmental basis of violence. A series of previous epidemiological studies have recognized that psychopathy, which referred to a personality type characterized by emotional callousness, lack of empathy, grandiose interpersonal style, impulsivity, and persistent antisocial behaviors (29), was significantly associated with violence in patients with SZ (28, 30). For instance, Fullam et al. found that 20–30% of patients with SZ were comorbid with psychopathy in forensic psychiatric settings (31, 32). In such a context, patients with SZ who have psychopathic traits might be at higher risks for violence; thus, the exploration of the anatomical basis of psychopathic traits may help reveal the neurological basis of violence in the population with SZ.

Literatures have revealed that the anatomical basis of psychopathic traits might be more specifically associated with regions involved in emotional processing and integration of emotion into cognition, including the limbic and paralimbic regions (33–35), it supports the finding that high levels of psychopathic traits might be resulted from the dysfunction of the frontotemporal-limbic network (36). However, relevant studies examining the neural basis of psychopathic traits in violent patients with SZ are rare, especially on a brain morphological level. A previous study on white matter (WM) integrity showed extensive reduction of WM in vSZ, which was positively associated with the level of psychopathic traits (37); another functional MRI study revealed a blunted amygdala response to fearful faces in male vSZ who had high levels of psychopathic traits (32). Some other comparative studies examining the differences in brain structure between vSZ and those patients with ASPD/psychopathy found common structural changes in these populations involving the hippocampus (13, 15), anterior cingulate cortex (22), and sensorimotor cortices (38). The previous findings demonstrated that the structural profiles underlying violent offenses in psychopathic individuals and vSZ were partly overlapped and dominantly involved in emotion processing and high-order cognition.

Based on the above evidence, we hypothesized that the vSZ might have structural deficits in the frontotemporal-limbic network as compared to nvSZ; furthermore, the network abnormalities might be associated with a higher level of psychopathic traits, which lead to a propensity to violence in SZ. The aim of the present study was to investigate the neural correlates of violence in patients with SZ and their clinical characteristics, and to explore the association between the neural correlates and psychopathic traits in the vSZ group. A whole brain GMV analysis derived from the voxel-based morphology (VBM) was conducted between violent patients, non-violent patients and healthy control groups (vSZ vs. nvSZ, vSZ vs. HCs, and nvSZ vs. HCs), which was followed by a correlation analysis to further reveal whether these structural deficits were associated with high levels of psychopathic traits. Hopefully, this study would shed light on the mechanisms of violence in SZ and further facilitate early intervention against violence in patients with SZ.

The present study found a more pronounced decrease of GMV in vSZ, which involved the right STG extending to the TP and INS, as well as the left FFG and bilateral IFG, as compared to HCs; while decreased GMV in nvSZ was specifically located in the left IPL, PHG, and bilateral mOFG compared to HCs. Furthermore, the vSZ group showed a larger GMV in the left IPL than the nvSZ group. These results suggested that structural deficits in the frontotemporal lobe, particularly the temporal lobe, might be predispositions to violent behaviors in patients with SZ. The significant correlations between psychopathic traits and temporal structural deficits in the vSZ group indicate that the neurobiological underpinnings of violence in patients with SZ might be attributed to high levels of psychopathic traits and represent more of the characteristics of violence per se. The study provided the advantage of using the whole-brain approach to detect any regions associated with violence, as specific isolated regions (ROI-based analysis) were insufficient to reveal the mechanisms of complex phenotypes, especially violence (43). Furthermore, to improve the homogeneity of data and reduce the impact of confounders on neurobiological variability in relation to violence, all the patients were free of substance abuse and were comparable for clinical symptoms and medications.

Extensive research has elucidated the involvement of the temporal gyrus in violence, not only in patients with SZ (10, 44), but also in non-psychotic criminals (45, 46). For instance, Kuroki et al. (11) found great losses in the volume of the STG and TP, Liu et al. (47) also found significantly reduced GMVs in widespread frontal, temporal, and limbic regions in vSZ compared to HCs. A recent work demonstrated that violence in male violent patients with SZ is related to reduced cortical thickness in the temporal region and FFG (48). Our findings were consistent with these studies and support that violence in SZ is associated with morphological deficits in the right STG and its surrounding areas, which are involved in the processing of emotion-related information. Similar to these findings in vSZ, abnormal GMV was also prominent in violent offenders with psychopathy from forensic facilities (46, 49, 50). These results suggested a unique role of the STG in predisposing to violence in SZ by interfering with visual information processing and higher-order cognition related to emotional processing (51). Intriguingly, correlation analysis in this study further revealed a negative correlation between psychopathic traits (reflected by the antisocial score) and structural damage in the STG in the vSZ group, suggesting that abnormal GMV of the STG might predispose to violent behavior in patients with SZ, which might be fostered by the antisocial trait of psychopathy. By investigating the effect of psychopathy on the brain structure and function in vSZ, the present study has provided primary evidence that abnormal GMV in the STG might contribute to interpersonal violence in SZ.

In addition to the STG and TP, we also found significantly decreased GMV in the FFG in vSZ rather than nvSZ. This result is accord with other studies, which supported the malfunction of the FFG in the etiology of violence in patients with SZ (8, 17, 47). For instance, Liu (47) and Del Bene (17) found reduced GMV in the right FFG in vSZ compared to HCs, as well as reduced cortical thickness of this region compared to those with nvSZ. However, no correlation was reported between the aberrance in the FFG and any clinical characteristics. It is believed that the FFG plays an important role in face perception and visual processing, especially when viewing negative faces (52). Evidence on the neurodevelopmental abnormality of the FFG in patients with SZ (53, 54) has indicated that impairments in face recognition might be a prevalent deficit in SZ. However, the present study found a volumetric reduction in the FFG in the vSZ group only, and bridged the relationship between psychopathic traits (reflected by the antisocial score) and volumetric reduction in the FFG in the vSZ group, suggesting that a pronounced GMV reduction in the FFG might predispose to a persistent risk for violence in SZ due to deficits in emotion perception, especially for negative faces. Some functional imaging studies in individuals with psychopathy showed reduced FFG activation, which related to face processing (55, 56), supporting the association between impairments of the FFG-related face processing and violence in SZ.

In addition to the right STG and FFG, the right INS, as a hub of the emotional salience network (57), might be involved in violent behavior in patients with SZ (11) through socio-emotional information processing and experiencing other's pains and inner states (58). Furthermore, we have consistently found reduced GMV in the OFG in both patient groups. However, significantly reduced GMV in the IFG, which consists of the bilateral orbital and right opercular frontal gyri, has been found specific to violent patients. As part of the ventral prefrontal gyrus, the IFG is explicitly engaged in salient information integration and emotional regulation (59), in addition to inhibitory control (60). Previous studies have demonstrated that structural alterations of the IFG were related to aggressive behaviors in patients with SZ (61, 62). Therefore, our results further highlighted the association between the IFG and persistent violence in patients with SZ, especially a potential correlation of the IFGoper with psychopathic traits.

It is noteworthy that, on the one hand, a significant GMV reduction in the left IPL was observed in the nvSZ group only, indicating that the morphological deficits in the left IPL might be specific to the disease itself (63); on the other hand, the significant correlation between GMV in the left IPL and psychopathic traits in the present study might be explained by a more mild impairment in cognition in the vSZ group, as compared to the nvSZ group. For instance, Abu-akel et al. found that violent patients had more difficulties than non-violent patients in tasks involving empathic inference and better abilities in inferring cognitive mental states in others (64). Given the sparse findings of structural abnormality in the IPL associated with violence in SZ in prior studies, further studies are needed to determine whether the differences in the IPL between the patient groups are resulted from the impairment in cognition.

Given that GMV is the product of both cortical thickness and surface area (65, 66), it was shown that reduced cortical volume was associated with less numbers of neuronal and glial cells as well as reduced SA, sulcal depth, and region-specific cortical folding complexity. Our findings suggest there would be neurodevelopmental deficits in the frontotemporal network in violent patients with SZ. Inversely, increased cortical volume in the IPL might be associated with more neuronal and glial cells as well as increased microstructure of the cortical volume.

Our results extended the evidence that the structural abnormality of the frontotemporal network, especially the temporal lobe, which was involved in the regulation of emotion recognition, response and decision-making, was related to greater risk for violence in patients with SZ [for reviews (9, 44, 51)]. Our findings also provided knowledge about the association between psychopathic personality traits and the STG and FFG in patients with SZ who had violent behavior, suggesting that psychopathic traits, as a risk factor of violence, might be a predisposition to violent behaviors in patients with SZ through the frontotemporal network. In future studies, we need to take a closer look at psychopathic traits for better understanding of the mechanism of interpersonal violence in patients with SZ and to explore whether the imaging findings from this study can serve as a biomarker to predict future violent behaviors and community living.

The raw data supporting the conclusions of this article will be made available by the authors, without undue reservation.

The studies involving human participants were reviewed and approved by Second Xiangya Hospital, Central South University. The patients/participants provided their written informed consent to participate in this study.

NG and JZ designed the study. NG analyzed the data and drafted the manuscript. XW and JZ revised the manuscript. NG, SZ, JL, XL, HG, and QS collected the clinical and MRI data. All authors contributed to and have approved the final manuscript.

This study was supported by the National Natural Science Foundation of China (ID: 82171509 to XW, 82071543 to JZ), the Hunan Province Innovation Province Construction Project (Grant Number: 2019SK2334 to XW and JZ), the Natural Science Foundation of Hunan Province (2019JJ40424 to JZ), and the Health Committee of Hunan (202103091470 to JZ).

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.

All claims expressed in this article are solely those of the authors and do not necessarily represent those of their affiliated organizations, or those of the publisher, the editors and the reviewers. Any product that may be evaluated in this article, or claim that may be made by its manufacturer, is not guaranteed or endorsed by the publisher.