Commentary on Morey et al. (2020): Can applying neuroanatomical precision advance our models and treatment of PTSD?

Neurobiological models of posttraumatic stress disorder (PTSD) have long centered on the structure and function of the amygdala given its key role in stress-related processes such as fear conditioning. Efforts to establish the structural characteristics of the amygdala in individuals with PTSD, including consortium-based collaborative efforts that capitalize on larger imaging datasets, find conflicting evidence with regard to size or shape abnormalities (1). With the evolution of neuroimaging techniques comes increased accessibility of tools to capture detailed, spatially-precise views of brain structure. Humans display specialization of amydgala subnuclei that imaging studies have historically pooled when reporting amygdala size or shape - in part due to the logistical challenge of examining subnuclei in vivo. Rodent studies, however, reveal differential relationships between amygdala subnuclei size and function with facets of stress responding. Looking beyond gross amygdala volume and instead analyzing its anatomically and functionally distinct subcomponents (and associated behavioral correlates) holds potential for reconciling discrepancies in extant research, and for translating knowledge obtained from animal models to human stress and trauma reactions.

Taken together, the empirical literature in this area points to several questions: First, can advances in high resolution imaging resolve questions about neural features that characterize pathological reactions to stress in humans? A goal of applying neuroimaging techniques in psychiatry is to discover neuroanatomical features relevant to disease etiology and maintenance - that is, to elucidate biomarkers that could help in providing effective, personally-tailored psychiatric treatment. Thus, a second clinical translational question is, can a more nuanced understanding of amygdala characteristics among individuals with PTSD shed light on this complex and heterogeneous disorder in a treatment-relevant way?

In this issue of Biological Psychiatry, Cognitive Neuroscience and Neuroimaging, Morey and colleagues (2) begin to address these questions by testing whether or not structural features in amygdala subnuclei and amygdala shape differ in individuals with and without PTSD. The authors analyzed structural Magnetic Resonance Imaging (MRI) scans and symptom data of 355 trauma-exposed military veterans with (n=149) and without PTSD (n=206). The goals of the study were to identify potential between-group structural abnormalities (volumetric and morphological), drawing upon recent developments in FreeSurfer software methods that permit the application of an amygdala subnuclei atlas to anatomical datasets, and to examine how these structural variables relate to PTSD symptom severity across re-experiencing, avoidance, and hyperarousal clusters. Automatic segmentation and labelling of nine amygdala substructures, yielding volumes for each subregion and whole amygdala volume, as well as estimation of total intracranial volume was performed using FreeSurfer v6.0. Protocol and analysis pipelines established by the Consortium for Enhancing Imaging Genetics through Meta-Analysis-Psychiatric Genomics Consortium (ENIGMA-PGC) were used for quality control and to measure vertex-based amygdala shape (incorporating FreeSurfer labels and segmentation). Data revealed larger volumes in the left and right central, medial, accessory basal, and cortical nuclei in the PTSD group compared to the trauma-exposed controls, as well as smaller volumes in the left and right lateral and paralaminar nuclei in the PTSD cohort (see Figure 1). Shape analysis of whole amygdala volume showed concave aspects of anterior amygdala surface, and posterior aspects of amygdala surface were more contracted bilaterally in individuals with PTSD relative to trauma-exposed controls. Significant associations were observed between re-experiencing symptoms and left central, medial, and cortical nuclei volumes. Avoidance symptoms were associated with left lateral, central, medial, and bilateral accessory basal and cortical volumes. Hyperarousal symptoms were associated with left lateral, central, medial, cortical, and bilateral accessory basal volumes. The shared relationships between cluster-level symptoms and subnuclei could mean that these symptoms– though distinct phenotypically – arise from common neurobiological sources. Steps were taken to account for trauma exposure variables, including a comparison between the PTSD and trauma controls with high versus low combat exposure. Group differences were most robust in comparisons between high combat trauma-exposed controls and the PTSD group and raise the intriguing possibility of neuroanatomical indicators of resilience.

Figure 1.

Depiction of significant group volume differences between individuals with PTSD versus trauma-exposed controls (3)

The article by Morey et al. (2) reflects progress in neuroimaging psychiatry research both in study design (e.g., increasingly large samples available via collaborative multisite consortium projects) and analytic techniques (utilizing a novel probabilistic atlas of subnuclei acquired with high-resolution imaging). As noted by the authors, these data are also subject to limitations that impact collaborative multi-study projects, where data sources and participant phenotyping must be distilled to key elements. The comparator group was limited to individuals with trauma exposure (which may also be associated with structural brain changes relative to non-exposed comparators (4)), and the impact of comorbidities cannot be disentangled within the current study design. For example, preliminary evidence suggests alterations in amygdala subnuclei related to depression, which commonly co-occurs with PTSD (5). The dataset was restricted to predominantly male Veterans, limiting generalizability of these findings to other demographic groups. Nonetheless, for the open question of amygdala structural differences in PTSD, Morey et al.’s (2) findings of directional group effects across nuclei in individuals with and without PTSD offer a potential explanation for discrepant results in studies that previously measured the amygdala as a whole. These data deliver an exciting glimpse into the potential for new imaging applications to elucidate links between clinical symptoms and detailed structural brain characteristics.

Morey et al. (2) set the groundwork for multiple directions of future work to better understand how the identified structural differences relate to clinical presentation of individuals with PTSD. A better understanding of the temporal onset of these structural changes relative to traumatic stress is needed to know if observed structural differences reflect a pre-morbid vulnerability to PTSD, a consequence of traumatic stress, or are secondary to the chronic experience of PTSD. Inclusion of samples across development would be useful because trauma in the developing brain may look different from that in adults (6). Rodent studies demonstrate that smaller basolateral amygdala volume is associated with enhanced fear conditioning and glucocorticoid stress response (7). Findings from Morey et al. (2) offer new insights into the translation of these multidirectional morphological changes within amygdala subnuclei to humans with PTSD and posit the idea that humans with smaller basolateral amygdala volume may also demonstrate abnormalities in fear and stress responding, given that abnormal fear learning is observed in PTSD. Collection of detailed onset and chronicity data over time would also facilitate comparison of human data to other rodent data suggesting that chronic stress can increase basolateral amygdala size via dendritic growth and spinogenesis (8). It remains to be determined if dendritic growth occurs in some or all humans in response to chronic stress, and how this may interact with PTSD development (for example, it is possible that individuals with trauma exposure have larger nuclei volume than non-exposed comparators).

PTSD is an incredibly diverse disorder in terms of its polythetic diagnostic algorithm (as evidenced by the innumerous combinations that can yield the disorder (9)), heterogeneity in traumatic events and sequelae that precede the diagnosis (e.g., trauma type, age and chronicity of trauma), and subjective experiences that diversify its clinical picture (e.g., guilt, fear, disgust, dissociation). A key next step for eventual clinical translation of these findings will involve determining relationships between neurobiological variables and intermediary processes that give way to these diverse symptoms. Morey et al. (2) logically propose that because amygdala subnuclei possess differential relationships with fear extinction, this could be one mechanism through which biology translates to symptoms. Yet amygdala subnuclei also possess interrelationships with brain regions responsible for non-fear based functions (e.g., reward and motivation (10)); the exact nature of how these relationships play out in PTSD, including how structural differences relate to function, is fodder for much work to come. These intermediary processes are not specific to PTSD per se, so one might also speculate that the observed structural differences may covary with constructs or dimensions that are observable across disorders. Replication with datasets that can address questions of whether structural variables align with transdiagnostic versus disorder-specific characteristics (e.g., by associating structural variables with putative mechanistic processes, such as those outlined in the Research Domain Criteria initiative) and how biological features relate to behavioral phenotypes, will be needed. Continued collection of large-scale collaborative datasets using unified measures will facilitate studies that can truly address the heterogeneity of this disorder at a detailed neuroanatomical level. Advances in standardization of analytic pathways, made public via collaborations like the ENIGMA-PGC group, in which Morey and colleagues participated, are making strides towards addressing reproducibility and validation of this type of automated segmentation at standard voxel resolutions. This work will move us even further toward clinical translation to help us understand whether neuroanatomical markers are global or identify specific kinds of patients with PTSD (perhaps those with abnormal fear conditioning and exaggerated physiological reactivity, or those with anhedonia and disrupted social functioning), and eventually whether there are differential treatment responses that relate to anatomical characteristics. These steps will be crucial to better identify types of patients and/or improve care by delivering information about specific biobehavioral targets that could be amenable to intervention.