Neural foundation of the diathesis-stress model: longitudinal gray matter volume changes in response to stressful life events in major depressive disorder and healthy controls

Stressful life events (SLEs) are a strong predictor of onset and recurrence of major depressive disorder (MDD) and are conceptualized within the diathesis-stress model, which posits that vulnerabilities (e.g., genetic predispositions or childhood maltreatment, CM) interact with adult stress to precipitate depression. Prior work shows not all individuals exposed to SLEs develop MDD, implying individual neurobiological differences in stress responses. Cross-sectional MRI studies link recent SLEs with gray matter volume (GMV) alterations in regions including insula, anterior cingulate, and medial prefrontal/orbitofrontal cortices, with differing patterns in MDD vs healthy controls (HCs). CM is a robust early environmental risk factor and has been associated with reduced GMV in multiple regions. However, longitudinal evidence is scarce, leaving unclear whether observed GMV patterns are consequences of stress or predictors of future SLEs. Two prior longitudinal studies in HCs report GMV reductions with recent SLEs in anterior cingulate, hippocampus, parahippocampal, and medial prefrontal regions. This study investigates, for the first time, longitudinal relationships between SLEs and GMV changes over two years in a large sample of MDD patients and HCs to ground the diathesis-stress model neurobiologically. Hypotheses: (1) HCs would show greater GMV reductions in response to SLEs than MDD patients across the interval; (2) CM would moderate SLE–GMV associations in MDD but not HCs. Exploratory aims included evaluating effects in MDD patients with greater severity or at least one depressive episode during follow-up and testing baseline high-sensitivity C-reactive protein (hsCRP) as a potential moderator given its links to stress, depression, and GMV changes.

Cross-sectional MRI studies in healthy adults have associated recent SLEs with GMV alterations in insula, anterior cingulate, and medial prefrontal/orbitofrontal cortices. In MDD, patterns differ, including fewer orbitofrontal GMV alterations than HCs. Independently, CM has been linked to smaller GMV in dorsolateral prefrontal cortex (DLPFC), anterior cingulate, supplementary motor area, postcentral gyrus, amygdala, and hippocampus in both HCs and MDD. Prior cross-sectional work suggested that CM might moderate SLE–GMV relationships in MDD but not in HCs, hinting at a neural substrate for the diathesis-stress model. Longitudinal evidence remains limited: two studies in HCs reported GMV reductions associated with recent SLEs (anterior cingulate, hippocampus, parahippocampal gyrus, medial prefrontal gyrus). The adaptive vs maladaptive nature of such reductions is unresolved, and longitudinal comparisons including MDD cohorts were lacking prior to this study.

Design and cohort: Longitudinal cohort study over approximately 2 years using data from the Marburg-Münster Affective Disorder Cohort Study (MACS; FOR2107 consortium). Participants: 754 adults aged 18–65 at baseline (T1): 392 healthy controls (HCs) and 362 MDD patients. Assessments occurred at University of Marburg and University of Münster, Germany. Follow-up (T2) mean interval 2.22 years (SD 0.31; range 1.9–4.3 years). Inclusion/exclusion: Excluded neurological/general medical disorders, current substance dependence, IQ ≤ 80. Additional HC exclusions: current/past DSM-IV-TR Axis I disorders (SCID-I) and lifetime psychotropic medication. Ethics: Approved by local committees; informed consent obtained. Clinical and psychosocial measures: Semi-structured interviews captured illness course (number/duration of depressive episodes, hospitalizations), remission status (SCID criteria), psychopathology (HAM-D 17-item; STAI-S), social functioning (GAF), and medication. Self-report/rater scales: familial risk (first-degree relatives with MDD, bipolar disorder, schizophrenia, or schizoaffective disorder), perceived stress (PSS), neuroticism (NEO-FFI), resilience (RS-25), social support (F-SozU), attachment style (RSQ). Childhood maltreatment (CM) assessed with Childhood Trauma Questionnaire (CTQ; sum score of emotional/physical/sexual abuse and emotional/physical neglect). Stressful life events (SLEs) assessed using the Life Events Questionnaire (LEQ), capturing cumulative impact of positive and negative events occurring between T1 and T2; total events score used as primary SLE index (measured at T2 referring to T2–T1 interval). Baseline hsCRP measured in a subset (n=509). MRI acquisition: 3T scanners with standardized sequences and QA. T1-weighted 3D MP-RAGE, 1 mm isotropic voxels, FOV 256 mm. Marburg: Siemens Tim Trio, 12-channel coil (TR 1.9 s, TE 2.26 ms, TI 900 ms, flip 9°). Münster: Siemens Prisma Fit, 20-channel coil (TR 2.13 s, TE 2.28 ms, TI 900 ms, flip 8°). Pre-processing: CAT12 (v1720) longitudinal pipeline in SPM12 on MATLAB R2017a: realignment, bias correction, tissue classification, spatial normalization (Geodesic Shooting), modulation to compute GMV, smoothing with 8 mm FWHM Gaussian kernel, normalization to MNI space, total intracranial volume (TIV) computed. QC included visual inspection and CAT12 homogeneity checks. Scanner hardware changes at Marburg (body coil June 2016; gradient coil August 2018) were modeled as dummy covariates. Statistical analyses: - Primary whole-brain longitudinal analysis: 2×2 repeated-measures ANCOVA (flexible factorial) with factors subject, time (T1, T2), and group (HC vs MDD). Covariate of interest: interaction LEQ total events score × group. Covariates of no interest at T2 (set to zero at T1): age, sex, interscan interval; plus dummy variables for body/gradient coil changes. TIV not included (within-subject design). Absolute GM threshold 0.1. Cluster-level FWE correction p<0.05 (one-tailed). - Extraction: Significant clusters’ weighted mean intensities (eigenvariates) extracted for visualization and secondary analyses. - Positive vs negative events: Differences in correlations of GMV change with LEQ positive vs negative scores tested using Steiger’s Z. - Moderation analyses (confirmatory): Multiple linear regression on extracted GMV change (T2–T1) values testing three-way interaction LEQ × CTQ × group (MDD vs HC), adjusting for age, sex, interscan interval, and scanner variables; included all main effects and two-way interactions. One-tailed p<0.05 Bonferroni-corrected. - Exploratory moderation in recurrence subgroups: Whole-brain 3×2 ANCOVA (groups: MDD with ≥1 episode during T2–T1; MDD without episode; HCs; time: T1, T2) with same covariates, followed by moderation analyses on extracted clusters. Additional severity indices considered: number/duration of episodes, hospitalization number/duration, remission status at T2 (time-adjusted by interscan interval where applicable). Two-tailed p<0.05 Bonferroni-corrected. - Control analyses: Tested associations and potential moderations by clinical (episodes, hospitalization, remission, depression severity), familial risk, psychosocial (state anxiety, perceived stress, neuroticism, social support, resilience, attachment), global functioning, and medication at T2. - Cross-sectional predictive/retrospective analyses: ANCOVAs (between-subjects) testing interactions of LEQ (T2–T1) × group on baseline (T1) GMV (predictive) and follow-up (T2) GMV (retrospective). Covariates: age, sex, TIV, scanner hardware dummies. Cluster-level FWE p<0.05 (two-tailed), k≥10 voxels.

- Primary longitudinal whole-brain results: Two significant clusters showed differential SLE–GMV change associations between HCs and MDD patients over ~2 years: 1) Left middle frontal gyrus (k=1588 voxels; peak MNI x/y/z = -46/27/28; FWE cluster-level p=0.005; Cohen’s d=0.29). In HCs, higher SLEs predicted larger GMV reductions (β = -0.18, t = -3.45, p<0.001); MDD showed no significant change (β = 0.10, t = 1.77, p=0.077). 2) Left precentral/postcentral gyri (k=1364 voxels; peak MNI x/y/z = -52/-22/44; FWE cluster-level p=0.009; Cohen’s d=0.33). In HCs, higher SLEs predicted larger GMV reductions (β = -0.21, t = -4.20, p<0.001); MDD showed no significant change (β = 0.07, t = 1.21, p=0.227). No clusters showed larger SLE-related GMV reductions in MDD vs HCs. - No significant associations of cluster GMV change with clinical/psychosocial variables at T2 (e.g., duration/number of episodes, remission, familial risk, STAI-S, PSS, NEO-FFI, HAM-D, CTQ, GAF, medication). No differences between LEQ positive vs negative event correlations with GMV change. - Confirmatory moderation: No three-way interaction LEQ × CTQ × group (MDD vs HC) on GMV change in either cluster (middle frontal: β = -0.02, t = -0.41, p=0.681; pre/postcentral: β = 0.03, t = 0.52, p=0.604). Other candidate moderators (familial risk, STAI-S, PSS, HAM-D, NEO-FFI, RSQ secure, RS-25, F-SozU) were also non-significant. - Exploratory recurrence subgroup analyses: Significant three-way interaction LEQ × CTQ × recurrence group on GMV change in middle frontal, precentral, and postcentral gyri (F ≈ 30.11, t = 2.67, p=0.008). In MDD patients with ≥1 depressive episode during the interval, higher SLEs were associated with GMV increases with higher CTQ (β = 0.23, t = 2.83, p = 0.005); not observed in MDD without episode (β = -0.12, t = -1.79, p = 0.076) or HCs (β = 0.03, t = 0.53, p = 0.596). Effect strengthened with increasing number of depressive episodes (β = 0.12, t = 3.19, p = 0.001). Elevated baseline hsCRP similarly moderated SLE–GMV increases in MDD with an episode (exploratory), not in overall groups. - Other clinical variables (remission status, hospitalizations) did not show significant LEQ × CTQ × factor interactions on GMV change; familial risk, remission status, attachment, resilience, and social support did not moderate LEQ × recurrence group effects. - Predictive cross-sectional (T1): Significant LEQ × group interaction on baseline GMV in right precentral/postcentral gyri (k=2167 voxels; peak MNI 38/-32/65; FWE cluster-level p=0.002). As future SLEs increased (T2–T1), HCs had larger baseline GMV whereas MDD had smaller baseline GMV in this area. ROI analysis showed overlap between predictive (T1) and longitudinal (T2–T1) pre/postcentral clusters (k=276 voxels; peak MNI 52/-10/40; t=3.97; FWE p=0.006). Retrospective (T2) analysis found no significant clusters.

Findings demonstrate distinct longitudinal neural responses to SLEs in HCs versus MDD. HCs showed significant SLE-related GMV reductions in middle frontal and pre/postcentral regions, consistent with an adaptive redistribution of neural resources in response to stress. In contrast, MDD patients did not exhibit such reductions overall, suggesting impaired adaptive mechanisms. Critically, within MDD, those who experienced an episode during follow-up and had higher CM showed SLE-related GMV increases in the same regions, consistent with a maladaptive response and supporting a neural basis of the diathesis-stress model for depressive recurrences. Predictive analyses indicate that baseline GMV in pre/postcentral gyri differentially relates to future SLE exposure in HCs (larger GMV predicts more SLEs) versus MDD (smaller GMV predicts more SLEs), pointing to differing vulnerability profiles and potentially explaining subsequent longitudinal trajectories. Mechanistically, observed GMV increases in MDD with episodes may reflect neuroinflammatory processes (supported by hsCRP moderation), glial proliferation, and vascular changes that could precede longer-term GMV loss, aligning with literature on synaptic and glial contributions to depression-related structural alterations. These effects persisted after accounting for medication, clinical severity, psychosocial variables, and familial risk, underscoring their robustness.

This study provides the first longitudinal evidence, across a large MDD and HC cohort, for differential GMV changes in response to SLEs and identifies a subgroup of MDD patients (those with episodes and higher CM) showing SLE-related GMV increases in middle frontal and sensorimotor cortices. Results suggest adaptive stress-related GMV reductions in HCs and potentially maladaptive increases in vulnerable MDD patients, offering a neural foundation for the diathesis-stress model in MDD recurrences. Future research should employ longer follow-up with multiple timepoints, replicate findings across other psychiatric disorders, and further probe inflammatory mechanisms and temporal dynamics underlying observed GMV changes.

SLEs and CM were assessed by self-report (LEQ, CTQ), introducing potential recall bias and measurement error, though these instruments are relatively state-independent and stable over time. Despite adjustment for numerous confounders, unmeasured factors (e.g., physical health, lifestyle) may influence associations, limiting causal inference.