Forty-seven right-handed patients with MMD (17 men; mean age 26.12 ± 5.06 years) and 43 age- and sex-matched HCs (13 men; mean age 26.27 ± 5.69 years) participated in the study. All patients satisfied the following inclusion criteria: (1) Diagnostic and Statistical Manual of Mental Disorders-Fourth Edition (DSM-IV) criteria for MDD diagnosis with an actual mild-to-moderate episode; (2) score 7 < HAMD-24 scores <24; (3) 18 years ≤ age ≤ 50 years; (4) right-handed (Edinburgh Handedness Inventory, EHI) (Oldfield, 1971); (5) no antidepressant treatment in the past 2 months; and (6) no comorbid Axis-I diagnosis, which comprises all psychological diagnostic categories except intellectual disability and personality disorder. It is worth noting that older patients have an increased risk of vascular disease, which may affect brain structure. To mitigate the influence of vascular factors on brain structure (Kendler et al., 2009), the participants in this study were adults under the age of 50. In fact, considering the vascular factors, we actively encouraged younger patients to participate in this study. Exclusion criteria included a history of neurological diseases, past or current history of alcoholism, elevated risk of suicide, and contraindications of MRI scans (e.g., implanted metal objects and claustrophobia). Forty-three age- and sex-matched, right-handed HCs were recruited from the community nearby and were accessed via the Structured Clinical Interview for DSM-IV, non-patient edition (SCID-I/NP) to exclude the diagnosis of MMD, as well as other axis-I psychiatric disorders. In addition to the above, the other criteria also included: (1) HAMD-24 scores <7 and (2) no family history of mental illness. (3) There were no psychological or psychopathological symptoms. The exclusion criteria for the HCs were identical to those for MMD. The participants signed an informed consent form to participate in this study. This study was approved by the Ethics Committee of Shanxi Medical University.

For patients with MMD, clinical assessments included age at illness onset and the severity of illness. Depressive symptoms and anxiety symptoms were independently assessed based on HAMD-24 and HAMA-14 by two research psychiatrists. The clinical assessment of HCs included the HAMD and HAMA. The HAMD was established in 1960 and adopts a 5-level rating scale from 0 to 4 points. The total score ranges from 0 to 78, and the depression level can be divided as follows: <8, no depression; 8–24, MMD; and > 24, severe depression. The HAMA was established in 1959 and adopts a 5-level rating scale from 0 to 4 points. The total score ranges from 0 to 56, and the anxiety levels can be divided as follows: <7, no anxiety; 7–14, possible anxiety; 15–21, anxiety; 21–29, obvious anxiety; and >29, severe anxiety. Both scales were confirmed to have good reliability.

After the pre-tests, patients included were instructed to take a Chinese herbal medicine named SG, which is mainly made up of Acanthopanax and H. perforatum, licensed and widely prescribed for depression in China since 2008. Patients with MMD received SG treatment for 8 weeks after the baseline MRI scans. According to the recommended medication dose, patients with MMD took 1.44 g of SG per day, including 0.72 g in the morning and 0.72 g in the evening for 8 weeks. During treatment, all participants were followed up weekly via telephone interviews to maintain good adherence. After treatment, all the participants completed the HAMD and HAMA assessments.

MR images were acquired using a 3.0-T Siemens Skyra MRI scanner (Siemens Medical, Erlangen, Germany) at the Shanxi Provincial People's Hospital. A magnetization-prepared rapid gradient echo sequence was used to acquire high-resolution T1-weighted anatomical images (repetition time = 2,530 ms, echo time = 2.01 ms, inversion time = 900 ms, flip angle = 9°, resolution matrix = 256 × 256, slices = 192, thickness = 1 mm, voxel size = 1.0 × 1.0 × 1.0 mm³, and duration = 7 min). The diffusion data for each subject were obtained using a diffusion-weighted, single shot, spin-echo, echo planar imaging (EPI) sequence (TR/TE = 10,200/91 ms, matrix = 96 × 96, FOV = 192 × 192 mm, voxel size = 2.0 × 2.0 × 2.0 mm³, 70 axial slices, 2.0 mm slice thickness, b-value 1 = 0 s/mm², and b-value 2 = 1,500 s/mm²) in 64 directions. Resting-state fMRI data were acquired using an EPI sequence (TR/TE, 2620/30 ms; flip angle: 90°, FOV 192 ×192 mm; matrix, 64 ×64; slice thickness, 3 mm; slices, 47; 220 volumes).

The GM volume was examined using VBM from the FSLVBM. All T1-weighted images were first extracted from the brain and then segmented into GM, WM, or cerebrospinal fluid. A GM template was generated by registering and averaging all GM images. The GM image for each participant was then registered to the template using a non-linear transformation. A voxel-wise permutation test was used to identify significant group differences between patients with MMD and HCs to a distribution generated from 5,000 permutations of the data for each voxel of the template, using a sigma filter of 3 mm for smoothing. P < 0.05 (FWE, using the TFCE method) was considered statistically significant.

We analyzed WM properties using voxel-wise TBSS from the FMRIB Software Library (FSL) (https://fsl.fmrib.ox.ac.uk/fsl/fslwiki/TBSS) (Smith et al., 2006). This method identified a core WM “skeleton” that was anatomically equivalent across participants. First, diffusion data were corrected using Eddy correction (eddy_correct) and brain-extracted using BET in FSL (Smith et al., 2002). The FA images were created by fitting a tensor model using DTI fit, and then FA data from all participants were aligned into a common standard space using the non-linear registration tool, FNIRT. Next, the mean FA image was thinned to create a mean FA skeleton representing the centers of all tracts common to the group. The FA images from individual subjects were then projected onto this skeleton and the resulting data were fed into voxel-wise cross-subject statistics. The significant regions were thickened for visualization using the TBSS fill script in the FSL. The Johns Hopkins University ICBM-DTI-81 WM labels atlas was used to locate anatomical structures in the MNI152 spaces. Independent-sample t-tests were conducted to examine FA differences between the two groups using the non-parametric permutation tool Randomize in FSL.

For the demographic data, group differences in age and education level were evaluated using independent sample t-tests, and sex differences were assessed using Pearson's chi-square test. The difference between HAMA and HAMD scores of patients before and after treatment was tested by paired sample t-test. Pearson's correlation analyses were performed to assess the relationship between structural deficits and depressive symptoms in the MDD group.

Furthermore, to assess the joint influence of structural deficits on depression, a multiple linear regression analysis based on a forward stepwise selection procedure was applied. In this analysis, depression symptom severity (overall HAMD scores) was used as the dependent variable, and structural features (GM volume and FA values) were used as explanatory variables. The forward stepwise selection procedure started by including the explanatory variable that could mostly and significantly explain the dependent variable (p < 0.05) and adjusted by repeatedly adding other explanatory variables (if any) that could significantly improve the fit of the model (p < 0.05), until none could improve the model. Before multiple linear regression analysis, all variables were checked for deviation from normality using the Kolmogorov-Smirnov test, and no significant deviation from normality was observed for any variable (p > 0.05).

Demographic (i.e., age, sex, and years of education) and clinical variables in patients with MMD and HCs are summarized in Table 1. Although sex, age, and years of education were not significantly different between the two groups, the clinical variables, as quantified by 24-item Hamilton Depression Rating Scale (HAMD) (t = 16.060, p < 0.001; Table 1 and Figure 1A) and 14-item Hamilton Anxiety Rating Scale (HAMA) (t = 12.220, p < 0.001; Table 1 and Figure 1A), were significantly larger in patients with MMD than in HCs. Patients with MMD received SG treatment for 8 weeks, and the symptoms were significantly reduced from baseline to the end of SG treatment. Specifically, 37 patients had HAMD scores <7, and 10 patients had HAMD scores between 7 and 14. At the same time, 32 patients had HAMA scores <7, and 15 patients had HAMA scores between 7 and 14. In addition, the mean HAMD and HAMA scores were <7 in patients with MMD after 8 weeks of SG treatment (Table 2 and Figure 1B). Therefore, it can be considered that after 8 weeks of SG-treatment, the condition of all patients has been effectively alleviated.

At the baseline, FA values in patients with MMD were significantly increased in the regions of the anterior limb of the internal capsule (ALIC) and posterior limb of the internal capsule (PLIC) compared to those in HCs (Figure 2).

After 8 weeks of SG treatment, FA values were significantly decreased in the regions of the ALIC (MNI coordinates: Peak_(x/y/z) = 102, 126, 77; t = 3.720, p < 0.001; Table 3) and PLIC (MNI coordinates: Peak_(x/y/z) = 109, 117, 87; t = 18.880, p < 0.001; Table 3) compared to those at baseline. TBSS revealed no significant differences in FA values between patients with MMD and HCs.

To further understand the association between FA values and clinical variables in patients with MMD, we conducted correlation analyses between FA values and depression and anxiety scores in the patient group. The results indicated that FA values were significantly positively correlated with HAMD scores (r = 0.297, p = 0.043; Figure 2), but not with HAMA scores (r = 0.062, p = 0.691; Figure 2) in patients with MMD.

At baseline, after controlling for the effects of age, sex, and total brain size, we observed that the GM volume of the dorsolateral prefrontal cortex (DLPFC), frontal cortex, occipital cortex, and precuneus was significantly smaller in patients with MMD than in HCs [p < 0.05, threshold-free cluster enhancement (TFCE) corrected; Figure 3].

After 8 weeks of SG treatment, GM volume was significantly increased in the regions of the Frontal gyrus (MNI coordinates: Peak_(x/y/z) = 50, 94, 43; t = 4.402, p < 0.001; Table 4), DLPFC (MNI coordinates: Peak_(x/y/z) = 28, 84, 50; t = 7.416, p < 0.001; Table 4), precuneus (MNI coordinates: Peak_(x/y/z) = 46, 33, 57; t = 5.766, p < 0.001; Table 4), and Frontal gyrus (MNI coordinates: Peak_(x/y/z) = 29,83,26; t = 12.650, p < 0.001; Table 4) compared to those at baseline. There were no significant differences in GM volume between patients with MMD and HCs.

To understand the effects of depression on structural GM volume, correlation analyses were conducted to determine the relationship between GM volume abnormalities and depression scores. We extracted GM volume in the DLPFC and precuneus, correlation analyses indicating that the GM volume of DLPFC was negatively correlated with HAMD scores (r = 0.300, p = 0.041, Figure 4), whereas the GM volume of precuneus was not correlated with HAMD scores (r = 0.074, p = 0.621; Figure 4) in patients with MMD.

It is worth noting that researchers have previously found that MDD patients with different severity have different neuroimaging changes in the default mode network (DMN) and the execution control network (ECN) (He et al., 2022). Thus, we divided patients with MMD into mild (N = 24) and moderate (N = 23) depression groups, and compared the GM volume and FA values of the two groups at baseline and 8 weeks of SG-treatment, respectively. At the whole brain level, there was no significant difference in structural characteristics between the two groups. In addition, we compared whether there were differences between mild and moderate depression groups in brain regions (FA: ALIC and PLIC; GMV: DLPFC, frontal cortex, occipital cortex, and precuneus) with structural differences between MMD and HCs. We found that there was no significant difference in those brain regions characteristics between the two groups at baseline (FA: t = 1.601, p > 0.05; DLPFC: t = 1.582, p > 0.05; frontal cortex: t = 1.032, p > 0.05; occipital cortex: t = 1.374, p > 0.05; precuneus: t = 0.403, p > 0.05) and 8 weeks of SG-treatment (FA: t = 0.442, p > 0.05; DLPFC: t = 0.997, p > 0.05; frontal cortex: t = 0.785, p > 0.05; occipital cortex: t = 0.893, p > 0.05; precuneus: t = 0.305, p > 0.05).

Alterations of GM volume in the DLPFC was found to be positively correlated with HAMD score changes in patients with MMD after 8 weeks of SG treatment (Figure 5; r = 0.335, p = 0.048). And GMV in other three brain regions was not related to the HAMD scores changes: frontal cortex (r = 0.127, p = 0.942); occipital cortex (r = 0.250, p = 0.147); precuneus (r = 0.275, p = 0.109).

There were no significant correlations between clinical symptoms and FA value alterations.

Structural properties were obtained using voxel-wise TBSS. The most commonly used parameter in DTI research is FA, which reflects the structural integrity and geometry of axonal fibers (Gulani and Sundgren, 2006). Greater FA may reflect greater myelination of WM fibers, an increased number of myelinated fibers, smaller axonal diameter, or reduced neural branches within the MRI voxel (Boero et al., 1999; Beaulieu, 2002). In this study, the pattern in patients with MMD was different from that in HCs; increased FA values were observed in the ALIC and PLIC in MMD.

The ALIC in human and non-human primates carries ascending and descending fibers from the OFC and ACC. The OFC and ACC are associated with several psychiatric disorders, including obsessive-compulsive disorder, MDD, posttraumatic stress disorder, and addiction (Mayberg et al., 1997; Volkow et al., 2011; Milad and Quirk, 2012; Haber and Heilbronner, 2013). Importantly, diseases linked to the OFC and ACC show abnormal volume, FA, and diffusivity in the internal capsule, and some studies have demonstrated these abnormalities, specifically in the ALIC (Lang et al., 2006; Wobrock et al., 2008; Zou et al., 2008; de Souza Duran et al., 2009; Jia et al., 2010; Togao et al., 2010; Upadhyay et al., 2010; Nakamae et al., 2011; Zhu et al., 2011; Koch et al., 2014; Kuan et al., 2015). In addition, the ALIC is a WM bundle between the head of the caudate nucleus and the lenticular nucleus, which plays an important role in motivation, decision-making, and evaluation of the saliency of emotional and reward stimuli (Henderson et al., 2013). Consistent with our findings, Hyett et al. found that FA was distinctly increased in the right anterior internal capsule in patients with melancholia compared to controls (Hyett et al., 2018), and Coloigner et al. reported patterns of increased FA in the splenium of the corpus callosum and PLIC in patients with depression (Coloigner et al., 2019). In addition, researchers also found consistent results in other mental diseases, such as schizophrenia, obsessive-compulsive disorder, and bipolar depression; that is, the FA values of ALIC and PLIC in patients were increased compared with those in HCs (Versace et al., 2010; Lochner et al., 2012; Alba-Ferrara and de Erausquin, 2013). In patients with depression, a higher FA in the ALIC and PLIC was associated with more severe overall clinical symptoms, which is consistent with previous studies that have shown a strong association between WM disruptions in these regions and the symptoms of depression. Since FA values were significantly positively correlated with HAMD scores in patients with MMD after 8 weeks of SG treatment, TBSS revealed that there were no significant differences in FA values between patients with MMD and HCs. Similarly, behavioral data showed no significant difference in HAMD scores between patients with MMD and HCs. The Chinese herbal medicine SG is mainly composed of Acanthopanax and H. perforatum. Acanthopanax, commonly known as Ciwujia or Siberian ginseng, is a traditional Chinese medicine that ameliorates cognitive dysfunction induced by cholinergic blockade (Li et al., 2016). H. perforatum, also known as St. John's wort, has been used as an antidepressant agent and cognitive enhancer in rodents (Avila et al., 2018). H. perforatum has been shown to induce positive effects on both mood and short-term verbal memory (Yechiam et al., 2019). The main drug components of SG have good therapeutic effects on the clinical symptoms of MMD. Therefore, after 8 weeks of SG treatment, the clinical symptoms of patients with MMD were reduced, and there was no significant difference compared with HCs, while the FA positively correlated with clinical symptoms was also reduced, and there was no significant difference compared with HCs.

After taking sex, age, and baseline volume as covariates, brain GM volume of the DLPFC, frontal cortex, occipital cortex, and precuneus were significantly smaller in patients with MMD than in HCs. Consistent with our findings, several previous studies have reported differences in GM volume between patients with depression and HCs. For example, Lyoo et al. found evidence of decreased GM volume in the anterior cingulate gyrus, left medial frontal gyrus, and right precentral gyrus (Lyoo et al., 2004). Most studies have identified decreased GM volume in a wide range of brain regions, including the pre-supplementary motor area, parietal-temporal regions, frontal cortex, temporal cortex, cingulate cortex, insular cortex, parahippocampal gyrus, hippocampus, cerebellum, and OFC (Tang et al., 2007; Cheng et al., 2010; Lai et al., 2010; Zou et al., 2010; Peng et al., 2011; Ma et al., 2012; Wang et al., 2012; Zhang X. et al., 2012; Kong et al., 2014). Neuroimaging studies have produced interesting findings on the structural effects of antidepressant treatment in the prefrontal areas. The medial frontal gyrus and DLPFC volumes predict remission after antidepressant treatment (Chen et al., 2007; Yucel et al., 2009; Lorenzetti et al., 2010). Furthermore, effective treatment with fluoxetine and sertraline results in enlargement of the middle frontal gyrus, DLPFC, and OFC (Smith et al., 2013; Kong et al., 2014). Accordingly, remission has been shown to correlate with more preserved volumes of the anterior cingulate gyrus, dorsomedial prefrontal gyrus, and DLPFC over time (Frodl et al., 2008; Gunning et al., 2009). Furthermore, correlation analyses indicated that the GM volume of the DLPFC was negatively correlated with HAMD scores. Depression is characterized by dysfunction in cognitive, emotional, and behavioral processes related to emotion processing and regulation (Joormann and Gotlib, 2010). Thus, the study of neurobiologically informed models of depression focuses on brain regions underlying normative functioning of these processes (Drevets et al., 2008), including examining structural differences in these regions between patients with depression and healthy. The DLPFC is a large and functionally heterogeneous brain region (Glasser et al., 2016) and is involved in emotional and cognitive processing, and is one of the key players in the pathophysiology of mood disorders (Matsuo et al., 2019). Meta-analyses of MR images from patients with MDD have shown that compared with healthy subjects, patients with MDD have decreased GM volumes in the right DLPFC (Bora et al., 2012). In one another previous meta-analysis study, patients with MDD were shown to have reduced activation in the DLPFC (Fitzgerald et al., 2008). Besides, excitatory repetitive transcranial magnetic stimulation (rTMS) over the left DLPFC (Gu and Chang, 2017) or anodal left/cathodal right DLPFC transcranial direct current stimulation (tDCS) (Valiengo et al., 2017) have been shown to reduce depressive symptoms in post-stroke depression patients. This provides evidence for the negative correlation between DLPFC and MDD symptoms from the perspective of non-pharmacological interventions. In fact, stimulation at the DLPFC is the FDA-approved antidepressant intervention (Lan et al., 2016), and is based on the observation that, in depression, this structure has impaired functional connectivity (Sheline et al., 2010; Liston et al., 2014) and that left prefrontal cortical strokes increase the risk of depression (Shimoda and Robinson, 1999). The above studies are in line with a common idea that the structural changes of DLPFC either predict the onset of depression or reflect the damage caused by repeated emotions in depression patients.

Multiple linear regression analysis showed that, as a typical mental disorder, MMD also has serious brain structural deficits, such as reduced GM volume, in addition to its etiology. Previous studies on depression have reported consistent findings regarding the association between the left DLPFC and clinical outcomes. Several lines of evidence suggest a role of the DLPFC in the phenotypes, onset, duration, and treatment efficacy of depression (Kinou et al., 2013; Tsujii et al., 2016). For example, previous studies have shown that patients with MDD have a significantly decreased GM volume in the left DLPFC, which plays a key role in positive emotional judgment. The significantly decreased GM volume we observed in the DLPFC of depression was in agreement with the clinical symptoms of diminished interest or pleasure in almost all activities in depression (Qi et al., 2022), supporting the view that DLPFC plays a crucial role in the symptomatology of depression (Koenigs and Grafman, 2009). This view is further supported by functional MRI study, which indicates that DLPFC activity was decreased and the functional connection (FC) between the amygdala and the DLPFC was reduced in untreated depression patients performing cognitive function tests (Siegle et al., 2007), and that in depression patients there was a widespread reduction in FC between the left DLPFC and other brain areas (Liston et al., 2014). Our results also provide evidence that DLPFC is a key brain region in the pathogenesis and rehabilitation of depression. At baseline, after controlling for the effects of age, sex, and total brain size, we observed that the GMV of the DLPFC, frontal cortex, occipital cortex, and precuneus was significantly smaller in patients with MMD than in HCs. And there were no significant differences in GMV between patients with MMD and HCs after 8 weeks of SG treatment. This means that after treatment, the structure of the above four brain regions has changed, that is, GMV has been improved. Then, we made a correlation analysis between the alterations of GMV in these four brain regions and the HAMD score changes. The results showed that only the alteration of GMV in DLPFC was positively correlated with the HAMD score change. We consider that the possible reason for this result is that as one of the key brain regions in the pathogenesis and rehabilitation of depression, DLPFC is a key hub involved in cognitive control, working memory, and emotion regulation. This means that DLPFC may be more sensitive to SG treatment than the other three brain regions. And in line with previous reports, alterations in GM volume after 8 weeks of SG treatment indicated a potential modulation mechanism in brain structural plasticity within the DLPFC in patients with MMD. In addition, PFC abnormalities observed in patients with MDD or rodent models are partly corrected by antidepressants. An 8-week escitalopram treatment indicated reduced irregular high functional connectivity in the bilateral dorsal medial PFC in patients with MDD, and a 2-week administration of a selective serotonin reuptake inhibitor, fluvoxamine, restored dendritic length, and spine densities, but not cortical thickness after early life stress exposure (Lyttle et al., 2015; Wang et al., 2015). Overall, our data suggest that the degree of depression in MMD largely depends on the GM volume of the DLPFC in a given patient.

First, the sample size of patients was limited and involved patients with varying severity and duration of disease, which limits the investigation of the detailed role of each mechanism in the development of MMD. Second, the causal relationship between pathophysiological factors and depression severity should be detected not only based on a large sample of patients but also using some effective experimental designs, for example, a longitudinal study to monitor the transition from MMD to major depression. Of course, such a study would be fraught with additional challenges based on time effects and the likely contaminating effects of medication or therapeutic interventions to address depression. Third, in addition to the severity of depression, other factors such as the duration of depression and the number of depressive episodes can regulate the brain structure. Unfortunately, we have not collected information on these factors. Finally, our study may have underestimated the extent of structural changes associated with depression. Insufficient sensitivity may have arisen because of the limited sample size. Thus, if there were mood-state-dependent changes in GM volume, we may not have detected them because of insufficient power.

Despite these limitations, this study provides further evidence that the PFC might play a key role in the regulation of depression in the pathophysiology of MMD. Future studies using a larger sample size or longitudinal studies are needed to replicate and generalize these results across patients with severe depression. More detailed insights into the structural and functional brain plasticity underlying MMD will shed light on the development of new and more targeted treatment options in clinical practice.