Effects of stressful life events on human brain structure: A longitudinal voxel-based morphometry study

Abstract

Although stressful life events (SLEs) have been associated with an increased risk of illness and mental disorder, their impact on brain anatomy remains poorly understood. Using a longitudinal design, we tested the hypothesis that SLEs are significantly associated with changes in gray matter volume (GMV) in brain regions previously implicated in post-traumatic stress disorder (PTSD) in a group of clinically healthy adults. Magnetic resonance imaging was used to acquire an anatomical scan from 26 subjects (13 males and 13 females; mean age ± SD: 25.2 ± 4.3 years), with no psychiatric diagnosis, at two time points with a 3-month interval. Voxel-based morphometry was used to examine an association between SLEs and gray matter changes during this period. The number of SLEs was associated with a decrease in GMV in the anterior cingulate, hippocampus, and parahippocampal gyrus (p < 0.001). In contrast, there were no areas where the number of SLEs was associated with an increase in GMV. These results provide evidence that, in adults with no formal psychiatric diagnosis, SLEs are associated with GMV decreases in a subset of regions implicated in PTSD, and that these alterations can be observed within a period as short as 3 months.

Keywords::

• Gray matter volume
• hippocampus
• magnetic resonance imaging
• post-traumatic stress disorder
• stressful life events
• voxel-based morphometry

Introduction

Stress is associated with an increased risk of morbidity and mortality (Dube et al. 2009; McLaughlin and Hatzenbuehler 2009). In particular, stressful life events (SLEs) have been reported to increase risk for psychopathology, substance abuse, criminal activity, and autoimmune diseases (McEwen 2007; Anisman 2009; Dube et al. 2009; McLaughlin and Hatzenbuehler 2009). SLEs are thought to have both short- and long-term pathophysiological effects that are mediated by several factors including genetics, cognitive style, and social support (Park 2010). A characterization of how SLEs affect brain's structure and function may lead to a better understanding of stress-related disorders and inform the development of more effective treatments.

A number of studies have examined the impact of early life stress on brain structure in humans (Bremner 2002; Cohen et al. 2006; McEwen 2007); these studies have reported significant alterations in brain regions implicated in post-traumatic stress disorder (PTSD), particularly the hippocampus, amygdala, anterior cingulate cortex, subcallosal cortex, and medial frontal gyrus. However, the vast majority of them have been carried out in adults with symptoms of PTSD or other psychopathology (Bremner et al. 2008; Kasai et al. 2008; Bluhm et al. 2009). This raises the possibility that at least some of the observed alterations associated with early life stress reflect the manifestation of PTSD or psychopathology rather than the impact of SLEs per se.

Only few studies have examined the relationship between life stress and brain structure in adult participants without a history of psychopathology or brain disorder (Cohen et al. 2006; Gianaros et al. 2007; Ganzel et al. 2008). However, these studies used a cross-sectional rather than a longitudinal design; this may have limited statistical sensitivity to stress-related changes in regions with pronounced inter-subject variability. Furthermore, brain scans were typically acquired several years after exposure to stress; thus, it was not possible to establish whether the observed structural changes had become evident within weeks, months, or years.

The aim of the present investigation was to examine for the first time the relationship between brain structure and SLEs in healthy adults using magnetic resonance imaging (MRI) in a longitudinal design. Based on the recent evidence that experience-related neuroplastic changes can be observed using MRI within weeks or even days (Draganski and May 2008), we hypothesized that SLEs would produce reductions in gray matter volume (GMV) that could be detected with a 3-month interval between scans. We expected these alterations to be manifested in regions implicated in mediating the body's response to stress and in PTSD, particularly the amygdala, medial frontal gyrus, anterior cingulate cortex, hippocampus, and parahippocampal gyrus.

Materials and methods

Participants

Twenty-six healthy volunteers (13 males and 13 females) with a mean age of 25.2 (SD: 4.3) years were recruited through local advertisement. None of the participants reported a history of psychiatric or neurological disorder or was taking psychiatric medication at the time of scanning. Written informed consent was obtained from all participants and the study was approved by the Research Ethics Committee at King's College London. Mean IQ score as measured with the Wechsler Abbreviated Scale of Intelligence was 115.12 (SD: 8.19; range: 102–135).

List of threatening experiences questionnaire

The list of threatening experiences (LTE) questionnaire, also known as the brief life event questionnaire, assesses the incidence of life stressors including illness or injury, death of a close friend or relative, unemployment, financial loss, and end of important relationships (Brugha and Cragg 1990). This 12-item questionnaire has been shown to have robust test–retest reliability and validity and has been used in several previous studies to assess the incidence of SLEs (Brugha and Cragg 1990). On the day of the first MRI scan, we used the LTE questionnaire to examine the number of SLEs experienced by each participant in the previous 1 month, 6 months, and 5 years, respectively; this allowed us to establish the base rate of SLEs in our sample. On the day of the second MRI scan, we used the LTE questionnaire to assess the number of SLEs experienced by each participant between the first and second MRI scans.

Somatic and Psychological Health Report questionnaire

In order to assess the presence of somatic and/or psychological symptomatology that might indicate underlying psychiatric disorders, we used the Somatic and Psychological Health Report (SPHERE) questionnaire (McFarlane et al. 2008) 3 months after the second MRI scan. The SPHERE questionnaire is a reliable and valid screening measure of mental disorders; it has been used extensively in large community samples (McFarlane et al. 2008).

MRI acquisition

Two whole brain T1-weighted structural brain images were acquired from each participant with a 3-month interval (mean: 92.19 days; SD: 4; range: 87–102) using a 1.5T GE (General Electronics, Milwaukee, WI, USA) MRI scanner at the Institute of Psychiatry, London, UK. The number of sagittal slices acquired was 180 and image matrix was 256 × 180 (Read × Phase), resulting in a final resolution of 0.93 × 0.93 × 1.2 mm³, a repetition time of 8.592 ms, and an echo time of 3.8 ms. No upgrade or modification of the scanner was performed between the two acquisitions.

MRI data pre-processing

The pre-processing and statistical analyses of the MRI images were performed using voxel-based morphometry (VBM) implemented with the Statistical Parametric Mapping software (SPM8, Wellcome Department of Imaging Neuroscience; http://www.fil.ion.ucl.ac.uk/sm) running under MATLAB 7.0 r14 (MatWorks, Natick, MA, USA). VBM is an unbiased, semi-automated technique that allows the characterization of regional differences across the whole-brain (Mechelli et al. 2005). First, early and late images were manually reoriented and the anterior commissure was placed at the origin of the 3D Montreal Neurologic Institute coordinate system. A rigid registration of the follow-up to baseline images was performed within each subject and the co-registered images were then segmented in order to extract gray and white matter. The “diffeomorphic anatomical registration through exponentiated lie algebra” algorithm (Ashburner 2007) was used to spatially normalize the segmented images. This procedure yields definite improvements for VBM studies both in terms of localization and sensitivity by registering individual structural images to an asymmetric “custom” T1-weighted template derived from the subjects' structural images themselves (Ashburner 2007). Once the pre-processing was completed, we computed the difference between the early and late images for each subject; the resulting “difference image” represented a measure of the changes in GMV which occurred between the early and the late scans.

Statistical analysis

To examine the impact of SLEs on brain structure, we performed a voxel-wise correlation analysis between the number of SLEs and the difference in GMV between early and late scans as encoded in the “difference images”. Gender and age were modeled as covariates of no interest to ensure that the results were not affected by these potential confounds. Given our strong a priori hypotheses based on the present literature on the impact of SLEs in patients with PTSD (Sapolsky 2000; Bremner 2002; Cohen et al. 2006; McEwen 2007) and healthy individuals (Cohen et al. 2006; Gianaros et al. 2007; Ganzel et al. 2008), we constrained our search to a pre-defined mask which consisted of a total of 30,490 voxels. The mask included the following regions bilaterally: (1) amygdala, (2) medial frontal gyrus, (3) anterior cingulate, (4) hippocampus, and (5) parahippocampal gyrus. These regions of interest were defined using the automated anatomical labeling as implemented in PickAtlas toolbox. The use of a mask allowed us to focus on a priori regions of interest, thereby minimizing the risk of false positive results. Statistical inferences were made at p < 0.001 (uncorrected) with an extent threshold of five voxels.

Results

Before the first MRI scan, the average number of SLEs experienced by participants in the previous 1 month, 6 months, and 5 years was 0.96 (SD: 0.95; range: 0–2), 1.23 (SD: 1.1; range: 0–4), and 2.88 (SD: 2.00; range: 0–6), respectively.

Before the second MRI scan, more than half of the subjects (57.7%) reported at least one SLE since the first scan. Specifically, 4 subjects reported 3 or more SLEs, 7 subjects reported 2 SLEs, 4 subjects reported 1 SLE, and finally 11 subjects reported 0 SLEs. On average, participants reported 1.37 SLEs (SD: 1.65; range: 0–7). None of the participants reported somatic and/or psychological symptomatology, as revealed by the SPHERE questionnaire (mean: 0.06 SD: 0.15; range: 0–0.48).

The number of SLEs was associated with a decrease in GMV in four regions including the left parahippocampal gyrus (x = − 16, y = − 49, z = 3; Z score = 3.54), left anterior cingulate (x = − 16, y = 39, z = 14; Z score = 3.40), right anterior cingulate (x = 16, y = 31, z = 22; Z score = 3.22), and right hippocampus (x = 34, y = − 6, z = − 25; Z score = 3.19) (Figure 1). In contrast, there were no areas where the number of SLEs was associated with an increase in GMV.

Figure 1.  Brain regions where the number of SLEs was associated with decreased GMV at follow-up relative to baseline. Images are from sagittal MRI scans; regions showing decreased volume are in yellow. These regions included (a) the parahippocampal gyrus (Z score: 3.54); (b) the left anterior cingulate (Z score: 3.40); (c) the right anterior cingulate (Z score: 3.22); and (d) the right hippocampus (Z score: 3.19). GMV was measured in terms of mm³ of gray matter per voxel from MRI scans. For illustrative purpose, statistical threshold for the figure was set to p < 0.01 (uncorrected).

In order to examine the possibility that the identified decreases might be driven by differences in IQ, we repeated the statistical analysis after including this variable as covariate of no interest; all significant results were replicated (p < 0.001), indicating that the observed decreases in GMV were not dependent on IQ. In addition, in order to examine the possibility that the identified decreases might be explained by SLEs that occurred before the 3-month period under investigation, we performed three additional statistical analyses with a covariate of no interest which encoded the number of SLEs experienced by each participant 1 month, 6 months, or 5 years before the first MRI scan. In each case, all significant effects were replicated (p < 0.001), indicating that the observed decreases in GMV could not be explained by the base rate of SLEs.

Discussion

The aim of the present investigation was to examine for the first time the relationship between brain structure and SLEs in healthy adults with MRI using a longitudinal design. We focused on a distributed bilateral network including the hippocampus, parahippocampal gyrus, medial frontal gyrus, anterior cingulate, and amygdala, a network of regions believed to be sensitive to stress exposure (Cohen et al. 2006; McEwen 2007; Bremner et al. 2008). Our results indicate that exposure to SLEs is associated with GMV reductions in the bilateral anterior cingulate, right hippocampus, and left parahippocampal gyrus. These alterations are unlikely to be due to SLEs that occurred before the 3-month period under investigation, since the results were replicated when the base rate of SLEs (estimated over a period of 1 month, 6 months, or 5 years) was modeled as covariate of no interest; furthermore, they cannot be explained by possible differences in gender, age, and IQ, since these variables were also modeled in the statistical analysis. Importantly, none of the subjects showed somatic and/or psychological symptomatology which might indicate the presence of an underlying psychiatric disorder. This suggests that the impact of SLEs on GMV lies along a continuum that extends to individuals without a clinical syndrome.

Previous studies on the impact of SLEs on brain structure could not establish whether the observed structural changes had manifested within weeks, months, or years (Cohen et al. 2006; Gianaros et al. 2007; Bremner et al. 2008; Ganzel et al. 2008; Kasai et al. 2008; Bluhm et al. 2009;); our results suggest that the impact of SLEs on GMV within a subset of regions implicated in PTSD can be detected even with an interval between scans of only 3 months. This is consistent with recent reports that training-related neuroplastic changes can manifest in the adult brain within weeks or even days (Draganski and May 2008). We speculate that the structural changes detected in the present investigation may be associated with a functional reorganization, consistent with the results of a recent study on brain function in individuals who had witnessed a traumatic event (Lui et al. 2009).

Several animal and human studies point to glucocorticoid hypersecretion as one possible mechanism mediating stress-related brain atrophy (Sapolsky 2000). For example, exposure to excessive glucocorticoids results in regression of pre-existing dendritic processes as well as the inhibition of the genesis of new neurons in the hippocampus of rodents; these effects can be detected following only a few weeks of glucocorticoid overexposure (Sapolsky 2000). In addition, hippocampal atrophy has been reported in a range of conditions characterized by glucocorticoid dysfunction in humans, such as PTSD, depression, and Cushing's syndrome (Sapolsky 2000). Thus, some of the neuroplastic changes identified in the present investigation may be mediated by glucocorticoid hypersecretion, particularly in the hippocampus, which is a primary glucocorticoid target.

The observation of reduced GMV in anterior cingulate and hippocampal regions replicates previous studies on the impact of SLEs in humans (Cohen et al. 2006; Gianaros et al. 2007; Ganzel et al. 2008). These regions are also part of a distributed network which mediates the integration of cognitive, affective, and autonomic responses (Critchley 2009). In contrast, we found no evidence for SLEs-related alterations in the amygdala and the medial frontal gyrus. This aspect of our findings may be surprising, since both the amygdala and the medial frontal gyrus have been consistently implicated in PTSD by structural and functional neuroimaging studies (McEwen 2007). However, the results in clinically healthy subjects have been inconsistent, with only one study reporting stress-related structural alterations in the amygdala (Ganzel et al. 2008). One possibility is that alterations in the amygdala and the medial frontal gyrus might represent specific markers associated with psychopathology; this hypothesis could be tested by extending the present investigation to include both individuals who did and did not develop PTSD. The identification of alterations specific to the manifestation of the disorder could inform the development of new psychopharmacologic and psychotherapeutic approaches.

Limitations

Although the present findings are consistent with the wider stress literature, a number of methodological limitations should be noted. First, the sample was rather small, although sensitivity was enhanced by the use of a longitudinal design and restrictive inclusion criteria. Future studies would benefit from a larger sample which would also allow the examination of a possible moderating effect of IQ, gender, and cognitive style on the impact of SLEs, consistent with the behavioral literature. Second, our statistical threshold (p < 0.001) was not corrected for the total number of statistical comparisons; however, by focussing on regions implicated in stress and PTSD by previous studies, we were able to minimize the risk of false positive results. Third, it was not possible to establish whether the GMV reductions found in anterior cingulate and hippocampal regions were directly related to stress or reflected changes in lifestyle, emotional regulation, or cognitive style which may have resulted from the experience of SLEs (Park 2010). Finally, the present investigation used only a quantitative measure of SLE; future studies would benefit from using a qualitative approach.

Despite these limitations, the present longitudinal investigation provides evidence that, in adults with no formal psychiatric diagnosis, SLEs are associated with GMV decreases in a subset of regions implicated in PTSD, and that these alterations can be observed within a period as short as 3 months.

Acknowledgement

This work was supported by a research grant (2007/R2) from the Royal Society to Dr Andrea Mechelli.

Declaration of interest: The authors report no conflict of interest. The authors alone are responsible for the content and writing of the paper.