Major depressive disorder (MDD) is a prevalent and debilitating mental illness, significantly impacting daily life and incurring substantial societal costs. Endocrine dysfunction, particularly hyperactivity of the hypothalamic-pituitary-adrenal (HPA) axis leading to elevated cortisol levels, is implicated in MDD pathophysiology. Elevated cortisol levels have been linked to alterations in brain morphology and function in MDD patients, affecting regions like the hippocampus and orbitofrontal cortex. While voxel-based morphometry (VBM) and diffusion tensor imaging (DTI) have been employed to study brain structural abnormalities in MDD, they have limitations in capturing interregional connectivity. Source-based morphometry (SBM), a multivariate technique, offers a data-driven approach to identify unpredicted patterns of gray matter (GM) covariance across brain regions. This study aimed to investigate the relationship between brain structural networks and serum cortisol levels in first-episode, drug-naïve MDD patients using SBM, addressing the gap in research on this specific population.

The literature extensively documents the association between MDD and endocrine dysfunction. Studies have shown abnormal responses to thyroid-stimulating hormone and thyrotropin-releasing hormone (TRH), elevated TRH in cerebrospinal fluid, and increased antithyroid antibodies in depressed individuals. Subclinical hypothyroidism is also associated with increased frequency or severity of depressive symptoms. Furthermore, hormonal changes during puberty, menopause, and postpartum periods are linked to a higher prevalence of mood disorders. Oxytocin levels have been inversely associated with depressive symptoms, while cortisol, a key stress hormone regulated by the HPA axis, is consistently shown to be dysregulated in MDD. Studies have reported cortisol hypersecretion, reduced glucocorticoid receptor mRNA expression, and impaired glucocorticoid-induced feedback in MDD patients. Animal studies have demonstrated that glucocorticoids can increase excitotoxic brain injury and impair neuroplasticity. Prior neuroimaging research using VBM and DTI has shown associations between elevated cortisol and reduced hippocampal volume, orbitofrontal cortex thickness reduction, and white matter abnormalities. However, these methods provide limited insight into network-level changes.

This study involved 42 first-episode, drug-naïve MDD patients (21 males, 21 females; mean age 48.1 ± 14.3 years) and 39 healthy controls (26 males, 13 females; mean age 43.3 ± 11.6 years). Participants were recruited via interviews using the Structured Clinical Interview for DSM-IV-TR. Exclusion criteria for MDD patients included past Axis I disorders, medical or neurological illnesses, or use of drugs that could cause depression. Controls were recruited from the community and screened for psychiatric and neurological history. Morning serum cortisol levels were measured 1 hour after awakening using a direct radioimmunoassay. High-resolution T1-weighted brain MRI scans were acquired on a 3.0T MRI system using 3D fast spoiled gradient-recalled acquisition (3D-FSPGR). Image preprocessing involved correction for distortion and intensity inhomogeneity using Grad Warp and N3 functions, respectively. A radiologist confirmed the absence of gross abnormalities. Source-based morphometry (SBM) analysis using the GIFT toolbox was performed on the preprocessed GM images. Independent component analysis (ICA) identified independent components (ICs), with artifacts excluded based on established criteria. The remaining ICs were selected based on a prior review article on depression and anxiety. Two-sample t-tests compared loading coefficients between groups after Bonferroni correction. Spearman's rank correlation assessed the relationship between cortisol levels and loading coefficients. Statistical analyses were performed using EZR software, correcting for age, sex, and total intracranial volume as covariates. All analyses were corrected for family-wise error (FWE).

MDD patients had significantly higher serum cortisol levels than controls (p=0.03). Voxel-based morphometry (VBM) analysis did not reveal significant differences in regional GM volume between groups. SBM analysis identified 10 independent components (ICs) after excluding artifacts. Two ICs showed significant differences between MDD patients and controls after Bonferroni correction (p<.01): a prefrontal network and an insula-temporal network. The mean loading coefficients were significantly lower in the MDD group for both networks. In the MDD group, serum cortisol levels showed a statistically significant negative correlation with the loading coefficients of the prefrontal network (r=-0.354, p=0.02) but not with the insula-temporal network (r=-0.294, p=0.0588). No significant correlation was found in the healthy control group.

The findings suggest that alterations in GM volume, specifically within the prefrontal network, are associated with increased serum cortisol levels in early-stage, drug-naïve MDD. The negative correlation implies that higher cortisol levels are linked to reductions in the prefrontal network's GM volume. This prefrontal cortex (PFC) involvement aligns with established literature on the PFC's role in circuits underlying depression and anxiety, including the default mode network (DMN), positive affect circuit, frontoparietal attention circuit, and cognitive control circuit. The PFC's crucial role in emotional and cognitive processing suggests that these structural alterations might represent a causative factor in MDD rather than solely a consequence. The study's use of SBM, a multivariate approach, offers advantages over univariate methods like VBM by considering interregional connectivity and identifying unpredicted patterns of GM covariance. This aligns with the understanding of MDD as a heterogeneous syndrome involving dysregulation across multiple brain regions.

This study utilized SBM to demonstrate alterations in GM networks in first-episode, drug-naïve MDD patients, specifically within the prefrontal and insula-temporal networks. Notably, only the prefrontal network showed a significant association with serum cortisol levels. These results support the hypothesis that the PFC is a key target for the negative feedback effects of cortisol on the HPA axis during the early development of MDD. SBM offers a valuable alternative to univariate approaches for analyzing structural MRI data in psychiatric research. Further research with larger sample sizes is needed to validate these findings and explore the underlying mechanisms.

The study's relatively modest sample size is a limitation that could introduce sampling bias. While SBM is a powerful multivariate method, its robustness increases with larger datasets. The focus on first-episode, drug-naïve patients, while crucial for isolating the effects of MDD itself, limits the generalizability of the findings to other MDD populations.