This cross-sectional study involved a total of 60 subjects (40 females). The age range of the subjects was between 18 and 54 years (mean ± SD = 33.5 ± 8.512). Participants diagnosed with MDD were recruited from the Affective Disorder Unit of the Department of Psychiatry and Psychotherapy, Clinical Centre, University of Pécs. Participants with MDD were categorized into two subgroups based on their history of childhood maltreatment (CM). The MDD+CM group included those with moderate to severe CM (N = 21; 14 females), while the non-maltreated MDD group consisted of individuals with a low incidence of childhood maltreatment (N = 18, 12 females). Additionally, a healthy control (HC) group was formed, consisting of subjects matched in age and IQ, with no history of mental disorders (N = 21, 14 females). In the HC group, the Symptom-Checklist-90-R (58, 59) was applied to rule out subthreshold psychiatric symptoms. The detailed demographic data of the three experimental groups are presented in Table 1 .

The exclusion criteria for participation were as follows: current substance use (abstinence for < 2 years); IQ < 85; a history of head injury; a history of any neurological or psychiatric disorders (non-excluding psychiatric disorders are summarized below); experience of traumatic life events meeting DSM-5 post-traumatic stress disorder (PTSD) criterion A; or any contraindications for MRI (i.e. claustrophobia or the presence of metal objects in the body).

In our study, non-excluding, co-morbid psychiatric disorders were: anxiety disorders (panic disorder N = 3; generalized anxiety disorder N = 3; social phobia N = 2; specific phobias N = 4); cluster C personality disorders (dependent N = 2, avoidant N = 2); obsessive-compulsive disorder in the past 6 years, and never treated when symptomatic before (N = 1); lifetime sedatives, hypnotics, and anxiolytics use disorder (N = 2) in full remission for more than 2 years; mild and non-chronic alcohol use disorder (N = 2).

In patients with MDD, the mean age of disease onset was 25.49 ± 9.47 years. The mean duration of illness was 7.15 ± 7.74 years (range 0.2–26 years). Thirty-six (97%) patients with MDD were treated with antidepressant medication: SSRIs (N = 25); SNRIs (N = 3); NaSSAs (N= 7); agomelatine (N = 4); trazodone (N= 2); combined with mood stabilizer (N = 2); combined with low-dose atypical antipsychotics (N = 5).

The local Research Ethics Committee of the University of Pécs approved the study design and protocol (Ethical Approval Nr.: 2015/5626). All participants were Hungarian speaking Caucasians, living in the urban and suburban area of Pécs, and gave written informed consent.

All participants underwent comprehensive screening for any current or past psychiatric disorders, along with an assessment of general intelligence and emotional coping strategies and recognition using standardized neuropsychological measures.

Participants diagnosed with MDD fulfilled the DSM-5 diagnostic criteria for MDD (60), as evaluated by a trained psychiatrist (MS) using the Structured Clinical Interviews for DSM-5 disorders (SCID-5-CV and SCID-5-PD; 61, 62). The 21-item version of the Hamilton Depression Rating Scale (63) and the Beck Depression Inventory (64) were employed to evaluate the severity of depression, while the Beck Anxiety Inventory (65) was used to examine the severity of anxiety. Participants were also evaluated using the 11-item General Traumatic Experiences subscale of the 21-item Self Report Early Trauma Inventory (66) to measure causal traumatic childhood life events. Individuals with random trauma were excluded from this study. Four-subtest version of Hungarian adaptation of the Wechsler Adult Intelligence Scale-Revised was applied to test the General Intelligence Quotient (IQ; 67–69). See data for the results of the psychiatric assessments in Table 1 .

IBM SPSS Statistics Software version 29.0.2 was applied for data analyses. The assumptions were tested in all cases. For parametric data, one-way ANOVA with Bonferroni or Dunnett’s T3 (in case of unequal variance) post hoc test was used to compare groups for demographic, IQ and clinical variables, as well as for CERQ, DERS, TAS and RMET scores. Nonparametric data and datasets with skewed distributions were compared with the Mann-Whitney U test, as well as with the Kruskal-Wallis H test followed by Dunn’s pairwise post hoc comparison.

Given that no sexual dimorphism of hippocampal size is observed after adjusting for head size (85) and considering that the Point-Biserial Correlation revealed a relatively strong multicollinearity between estimated total intracranial volume (eTIV) and gender (r_pb= -0.519, N = 60, p = 0.00002), age and eTIV were selected as regressors for subsequent statistical analyses.

Volumetric data were analyzed with one-way ANCOVA followed by Bonferroni or Dunnett’s T3 (in case of unequal variance) post hoc tests, and with Kruskal-Wallis H test followed by Dunn’s pairwise post hoc test to identify differences between groups, while controlling for age and eTIV as covariates.

The interaction effect was examined using a general linear model to determine whether there is a general between-group difference in the association of psychological parameters and brain volumes. Group, age, eTIV, psychological parameters, and the group × psychological parameter interaction term were included as independent variables. Brain volumes were used as dependent variables. If any significant findings arose, further within-group multiple linear regression analyses were conducted, while controlling for age and eTIV, to explore the relationship between volumetric data and psychological parameters. The assumptions of multiple linear regressions were satisfied, as judged by testing for linearity, normality assumptions of the residues, outliers, independence of errors, homoscedasticity, and multicollinearity (86).

The level of significance was set at two-tailed p ≤ 0.05 for all statistical tests. Uncorrected p-values are reported to facilitate comparisons with other studies. However, to address the issue of multiple comparisons, Benjamini-Hochberg correction was applied with q = 0.15 when analyzing volumetric and psychological data. P-values that survived this correction for multiple comparisons are indicated in the tables.

The demographic data and the results of the psychiatric assessments are listed in Table 1 . Since the individuals were carefully selected for this study, age, gender ratio, and IQ values were comparable in the three experimental groups. Maltreated MDD patients spent significantly shorter time in education compared to controls (post hoc test: p = 0.036), but they were similar to the MDD group ( Table 1 ).

As a result of the group assignment, the CTQ scores of maltreated MDD patients (MDD+CM group) were significantly higher than those of both the non-maltreated MDD patients (MDD group) and the healthy control (HC) group across all subscales, while CTQ scores of the MDD and HC groups were very similar ( Table 1 ). Maltreated and non-maltreated depressed patients showed much higher scores in the Beck Depression and Anxiety Inventories compared to control group. The age at illness onset, the length of illness and the number of depressed episodes were comparable in the two sub-groups of depressed patients.

Representative examples of T1-weighted images and the results of the automatic segmentation of the hippocampal subfields and the amygdala are depicted in Figure 1 . The results of the volumetric analysis are summarized in Table 2 . We observed a consistent trend for volume reduction, i.e. a shrinkage of 1-10% in nearly all hippocampal subregions of the depressed patients. However, none of these reductions were statistically significant when compared to the control group. Notably, patients who had been exposed to childhood maltreatment generally displayed smaller volumes compared to both the control group and the non-maltreated MDD group, but again, these differences did not reach statistical significance. The largest observed volume reduction (-10%) occurred in the right presubiculum, right hippocampal tail, and left amygdala of the MDD+CM group ( Table 2 ). The least affected hippocampal region was the CA3 area of both hemispheres.

Results of the CERQ, DERS, TAS and RMET tests are presented in Table 3 . In general, depressed patients, especially the ones with the history of childhood maltreatment, displayed numerous difficulties with emotion processing.

Assessment with the Cognitive Emotion Regulation Questionnaire revealed that compared to the other two groups maltreated individuals reached the highest scores in the CERQ maladaptive sum and the lowest scores in the CERQ adaptive sum ( Table 3 ). Maltreated patients had significantly different scores compared to controls in almost all subscales. Similarly, non-maltreated depressed individuals had significantly different scores compared to controls in most of the subscales and they also reached significantly higher scores in the CERQ maladaptive sum and significantly lower scores in the CERQ adaptive sum compared to controls ( Table 3 ).

Results of the Difficulties in Emotion Regulation Scale indicated that maltreated individuals had the most severe difficulties with emotion regulation. They had the highest scores in DERS sum, and in all subscales, and they were always significantly different compared to controls ( Table 3 ). Non-maltreated depressed patients also had difficulties in emotion regulation as they reached significantly higher scores in DERS sum and in most of the subscales compared to controls ( Table 3 ).

Results of the Toronto Alexithymia Scale indicated that maltreated individuals reached the highest scores, and they were always significantly different compared to controls ( Table 3 ). Eleven maltreated depressed individuals had equal to or greater than 61 TAS sum scores indicating alexithymia. Non-maltreated depressed patients had significantly higher TAS sum scores compared to controls and in one TAS subscale (Difficulty Identifying Feelings), they were also different compared to controls. Among the non-maltreated depressed patients, there were two subjects who had 61≤ TAS sum scores indicating alexithymia.

The Reading the Mind in the Eyes Test revealed no difference between the three experimental groups ( Table 3 ).

The positive correlations between brain area volumes and results of the psychological tests are presented in Table 4 . We report here only the statistically significant findings.

In maltreated depressed patients, the scores reached in the rumination subscale of the CERQ test showed positive correlation with the volume of the right CA3 area. Notably, the right CA3 area had no volume shrinkage in these patients ( Table 2 ). Furthermore, in maltreated patients, the low scores of the positive reappraisal subscale of the CERQ test were associated with the volume reduction of the left presubiculum ( Table 4 ).

In non-maltreated depressed patients (MDD group), results of recognizing positive emotions in the RMET test showed positive correlation with the volume of the left presubiculum. Furthermore, results of the emotional neglect subscale of the CTQ test showed positive correlation with the volume of the right CA3 area. However, none of these two results remain significant after controlling for multiple comparisons using the Benjamini-Hochberg method with q=0.15 ( Table 4 ).

In the control group, the results of recognizing neutral faces in the RMET test showed positive correlations with several hippocampal subfield volumes, namely with the volumes of the right CA1, right subiculum, left molecular layer, and left subiculum ( Table 4 ).

The negative correlations between brain area volumes and results of the psychological tests are presented in Table 5 . We report here only the statistically significant findings.

In maltreated depressed patients, we did not find any negative correlations between the results of the psychological tests and brain area volumes.

In the non-maltreated MDD patients, the scores reached in the rumination subscale of the CERQ test showed negative correlations with the volumes of several subfields of the right hippocampus i.e. the GCL-ML-DG, molecular layer, CA4, CA1, and the entire right hippocampus ( Table 5 ).

In the control group, results of the awareness subscale of the DERS test showed negative correlation with the volumes of the right hippocampal tail and left subiculum. Furthermore, in control subjects results of the acceptance subscale of the CERQ test showed negative correlation with the volume of the right CA3, but this result did not remain significant after controlling for multiple comparisons using the Benjamini-Hochberg method with q=0.15 ( Table 5 ).

In case of the amygdala, we found very few significant correlations. In maltreated individuals, the scores of the emotional abuse subscale in the CTQ test were positively associated with volume of the bilateral amygdala. However, this result did not remain significant after controlling for multiple comparisons using the Benjamini-Hochberg method with q=0.15 ( Table 4 ). Furthermore, in non-maltreated depressed patients, the scores of the rumination subscale in the CERQ were associated with greater volume shrinkage of the right amygdala ( Table 5 ).

Numerous studies have used in vivo magnetic resonance imaging to examine hippocampal volume changes in subjects suffering from major depressive disorder, or in relation to adverse childhood experiences (17, 25, 30, 87–91). While the meta-analytic studies reveal a significant hippocampal volume loss of 8-10%, the results of the individual studies have been inconsistent and negative findings are not without precedent (87, 88, 92). The exact cellular changes responsible for the hippocampal volume loss are not fully understood, neuronal loss, dendritic reorganization, reduced adult neurogenesis and glial changes have all been implicated (93). A widely held view is that the stress-induced activation of the HPA-axis results in elevated glucocorticoid levels, which then initiates a cascade of neurotoxic – or at least neuroplastic – events in the brain, resulting in gross volume decrease (94–96). Based on this line of thinking, one may conclude that the hippocampal volume shrinkage is linked to the maladaptive emotional and/or cognitive behavior typical for depressed patients. Indeed, there is evidence for such a consequence, for example a study reported that in depressed patients the hippocampal volume loss correlated with executive dysfunctions (97). A more recent longitudinal study of depressed youth could also link hippocampal volume with emotion regulation and episodic memory impairment (34). Results of our present study could not however, substantiate these earlier findings.

The notion that volumetric changes in limbic structures can be linked to functional impairments has a long tradition (e.g. 15, 97, 98). Nevertheless, studies that have directly examined the structure-behavior relationship have so far yielded ambiguous results (18, 99, 100). The factors contributing to changes in brain area volume may be more complex or variable, rendering the correlation between brain structure and behavior challenging. Therefore, correlating functional neuroimaging data with complex psychological functioning may yield more consistent results.

Several factors may explain our inability to demonstrate an association between maladaptive emotional behavior and hippocampal subfield or amygdala volumes. One significant consideration is neuroplasticity, a key characteristic of the human brain. Volumetric changes may occur more rapidly and with greater variability than previously assumed (101). Furthermore, individual variability in brain development (102), hippocampal volume (21, 103), and emotional brain network topology (104) may provide another explanation. Additionally, functional reorganization (105) and structural resilience (106) following traumatic experiences may further elucidate the absence of significant results in our study.

Subjects of the present study were assessed with five psychological tests and the rumination subscale of the CERQ test was the one which had the largest number of correlations with hippocampal subfield volumes. This was reassuring since the hippocampus has been implicated in the regulation of stress-coping strategies (107, 108). The CERQ test has been constructed to identify the cognitive coping strategies in response to stressful life events (35). A recent study found a few positive correlations with some subscales of the CERQ test (e.g. catastrophizing, rumination, refocus on planning and positive refocusing) in healthy individuals, but none these correlations remained significant after correction for multiple analyses (19).

We also found a few negative correlations between results of the awareness subscale of the DERS test and hippocampal subfield volumes of control subjects. Only a few studies investigated the relationship between hippocampal, or amygdala volumes and emotion regulation difficulties. One study reported a strong relationship between emotion regulation and hippocampal volume (34), while another found that prolonged orphanage rearing was associated with atypically large amygdala volume and difficulties in emotion regulation (109).

In case of the RMET test, we found no difference between the groups, but within-group correlation analysis revealed that in the control group there were numerous positive correlations between hippocampal subfield volumes and scores of the RMET-neutral-faces subscale. Numerous MRI studies have been performed to relate performance in the RMET test with results of structural brain imaging and there is evidence that larger amygdala/hippocampal volumes are associated with better performance in the RMET test (52, 54).

In case of the TAS test, we could not find any correlations with brain area volumes. In the literature, there are numerous imaging studies which investigated the association between alexithymia and gray matter volumes, but most studies yielded inconsistent findings (40). Results of a recent meta-analysis indicates that the volumes of the left insula, left amygdala, orbital frontal cortex and striatum is consistently smaller in people with high levels of alexithymia (40). However, our present data could not replicate these findings.

Our present study has yielded a few counterintuitive correlations. Notably, we observed that individuals who reported higher instances of emotional abuse exhibited larger volumes of their bilateral amygdalae. Although this result did not remain significant after controlling for multiple comparisons, it stands in contrast to previous research, which typically associates childhood emotional abuse with reduced amygdala volumes (110–112). As discussed in chapter 4.1, potential explanations for these contradictory results may include individual differences in brain development, sample variability, or the presence of statistical noise.

As with the majority of studies, the current research is subject to several limitations. A major limitation is the relatively low sample size. To circumvent this issue, participants were meticulously selected to ensure matching groups in terms of age, gender, depression severity, and IQ. The relatively small sample size is likely a contributing factor to our inability to detect statistically significant volumetric differences among the three examined groups. A formal power analysis calculated by G*Power (version 3.1.9.4) indicated that a minimum total sample size of 153 participants (51 per group) would be required to detect statistically significant volumetric differences across the three groups in hippocampal subfields and the amygdala, based on assumed effect size of 0.322 and a statistical power of 0.9508. Further limitations are the cross-sectional design of the study and the retrospective assessment of childhood maltreatment, which was conducted with a self-report questionnaire that lacks complete objectivity. The cross-sectional nature of our study coupled with the retrospective self-reporting of childhood abuse, limits our ability to establish causal relationships between childhood maltreatment, brain morphology, and emotional regulation. Longitudinal studies are necessary to clarify such causal relationships, and indeed, there have been efforts to verify such causalities. A recent longitudinal study demonstrated that childhood maltreatment is associated with a persistent reduction of hippocampal volume in children and adolescents (113). Similarly, longitudinal studies could prove that experiencing childhood maltreatment is related to emotion dysregulation (114, 115).

The segmentation of the hippocampal subfields and the amygdala was based on only T1-weighted images, as employed in previous studies (116–118). For this reason, finding related to the volumes of GCL-ML-DG, CA4 and molecular layer must be interpreted with caution. Additionally, a limitation is that the amygdala was not segmented into nuclei. This limitation arose because we began the data analysis with the latest available version of the Freesurfer software (version 6.0), which does not include the segmentation of amygdala subregions. As a result, segmentation was confined to the entire amygdala without any division into further subregions. Future studies that employ high-resolution T2-weighted images and more recent versions of FreeSurfer software (beyond version 6.0) are warranted to achieve more detailed segmentation of the amygdala. Finally, this study relied only on structural neuroimaging data, whereas correlating functional neuroimaging findings with strategies of emotion regulation in maltreated and non-maltreated patients with major depression may yield more robust and informative results.

We report here that depressed patients with or without childhood maltreatment exhibit a modest reduction in hippocampal volume. Moreover, these individuals also display pronounced difficulties in emotion regulation. We demonstrate here a few associations between hippocampal subfield and amygdala volumes and disturbances in emotional processing. However, we could not detect the expected negative correlations between maladaptive behavior and hippocampal/amygdala volume shrinkage. Our present data suggest that correlating volumes of specific brain regions with complex psychological functions may not yield convincing results. Consequently, future research should prioritize functional neuroimaging methods, such as assessments of neural activity or functional connectivity, over structural data, like volumetric measurements, when investigating complex emotional phenomena.