Association of the Biopsychosocial Factors Adverse Childhood Experiences, Adult Attachment Style, Emotion Regulation, and Mitochondrial Density in Immune Cells with Major Depressive Disorder Open Access

Major depressive disorder (MDD) is a severe mental condition that is characterized by depressed mood, negative thinking, and altered physiological functions hindering the ability of affected individuals to cope with daily life (DSM-V; [1]). A World Health Organization (WHO) report [2] states that approximately 3.8% (280 million people) of the global population suffers from depression, with a 9.2% estimated 12-month prevalence in Germany [3]. Epidemiological research demonstrated that mental, social, and biological factors are associated with MDD [4‒6]. However, they are often investigated separately. As a result, MDD research frequently focuses on single behavioral [7], cognitive [8], or biological variables [9, 10]. By contrast, updated models attempt to implement multiple factors into predictive models for the onset and manifestation of MDD. For instance, the integrative explanatory model (IEM) incorporates social and environmental factors, as well as psychological and biological variables together, promoting a more holistic understanding of the development and manifestation of MDD [11]. In line with this, some reviews [10, 12] present a multifaceted conception of MDD, considering biological, neurobiological, psychological, social, and environmental vulnerability factors. The diathesis-stress model [11] illustrates how intraindividual and social factors interact to contribute to the development of depression. Similarly, psychological MDD vulnerability factors include low self-esteem, dysfunctional thought patterns, attachment disorders, and specific personality traits [10]. Environmental factors like adverse experiences in childhood and adolescence as well as negative social relationships further increase the probability for depression. Given the multitude of variables associated with MDD [10] (for review, see [12]) in the IEM, it seems necessary to investigate these variables conjointly to identify their unique contributions.

This study thus focuses on three candidate categories of variables related to MDD: social, psychological, and biological variables were analyzed collectively in a single regression model, unlike previous studies [13, 14] that investigated them separately. To do so, we examined the association of social, psychological, and biological variables [15‒18] with the likelihood of an MDD diagnosis in a logistic regression model. Figure 1 shows a schematic representation of the biopsychosocial model of depression based on the IEM.

Adverse childhood experiences (ACEs) and possible effects on attachment patterns in adolescence and adulthood have been reported as important social risk variables linked to MDD. Brinker and Cheruvu [19] identified an association of depression with ACE, particularly with emotional neglect (EN) as a highly predictive subtype [20‒22]. Salokangas and colleagues [23] also found an association of EN with depression and anxiety. Hanson and colleagues [24] found that decreases in reward-related ventral striatum activity in adolescents were related to EN, which in turn predicted depressive symptoms. A meta-analysis by Nelson and colleagues [25] revealed that 45.59% of individuals with MDD experienced at least one subtype of ACE, with 43.2% reporting EN as the most strongly associated subtype. Our study therefore focused on the association of EN with the presence of a diagnosis of MDD. A second important social risk variable associated with MDD is attachment style, denoting a pattern of relational expectations, emotions, and behavior originating from parent-child interactions in childhood [14]. Wilhelm and colleagues [26] proposed that early parent-child relationships shape emotions and expectations in subsequent adult relationships [26]. Bartholomew and Horowitz [27] proposed four prototypical adult attachment styles (AASs) that combine self-image and view of others: secure, preoccupied, dismissing, and fearful. Securely attached individuals have positive self-perceptions and confident relationships. In contrast, insecure attachment styles (preoccupied, dismissing, and fearful) increase the risk of developing psychopathology in adulthood [28, 29], including depression [30, 31] compared to a secure attachment style.

As psychological variables, the two emotional regulation strategies, thought suppression [32] and cognitive reappraisal [33], have been previously found to be associated with MDD. Thought suppression, defined as avoiding unpleasant thoughts [34], intensifies the persistence of intrusive thoughts [35] and correlates with the severity of depressive symptoms [36]. It is considered to be a depression risk factor [37], exacerbating depressive symptoms [38]. Conversely, cognitive reappraisal is seen as a protective factor against developing MDD [33], involving construing emotional situations nonemotionally [39]. It leads to downregulated negative emotions, reduced negative affect expression, and decreased physiological responses [40]. Higher cognitive reappraisal abilities are associated with more positive and less negative emotions [41], demonstrating a negative association between cognitive reappraisal and depression in some studies [15, 42].

Numerous biological variables at different levels have been related to MDD (for a review, see [12, 43]). Among these, the functioning and regulation of the hypothalamic-pituitary-adrenocortical axis have been a focus, playing a central role in the body’s neuroendocrine response to stress [44, 45]. On a biomolecular level, mitochondrial biogenesis and bioenergetics have been identified as promising biological factors associated with depression [46, 47]. Mitochondria supply cells with essential biochemical energy in the form of adenosine triphosphate (ATP). According to the process of fusion and fission, mitochondria are stress-adaptive and reduce their distribution of the mitochondrial network within cells. Stress mediators like cortisol have been described to affect the regulation of the intracellular mitochondrial network [48]. As the amount of mitochondria is linked to the total amount of cellular oxygen consumption, the regulation of the mitochondrial network has to be co-considered when investigating mitochondrial bioenergetics. Our recent findings [49] demonstrated that a reduced density of mitochondria in immune cells seems to be the key driver of impaired mitochondrial function. Therefore, we reduced the number of variables related to mitochondrial function to mitochondrial density (reflecting the regulatory state of biogenesis) as the biological variable of choice. We want, however, emphasize at this place that it remains currently unclear whether the energy metabolism of peripheral blood mononuclear cells (PBMCs) accurately reflects that of neurons and glial cells. To date, studies involving nonhuman primates are the only ones to compare mitochondrial respiration in monocytes with that of mitochondria isolated from the prefrontal cortex, and the findings have been inconclusive [50, 51]. This discrepancy may be due to metabolic plasticity of PBMCs during activation [52‒54]. Consequently, it is plausible to speculate that depression induces alterations in PBMC energy metabolism due to variations in immune cell activation. However, we are now only able to report changes in peripheral cells in MDD, which does not necessarily correlate with mitochondrial bioenergetics in the brain. Nevertheless, alterations in PBMC energy metabolism would show that MDD is a multisystemic disease also affecting mitochondrial function in PBMCs. It is important to note that also other factors affect mitochondrial function and density. For example, oxidative stress can cause allostatic load on a biomolecular level by damaging mitochondrial DNA (mtDNA), which encodes for most of the proteins of the electron transport chain system. Damage of mtDNA caused by reactive oxygen species and oxidative stress can affect the mitochondrial density in cells confronted with biological stressors [55, 56].

Previous research indicated that social, psychological, and biological variables are related to depression. However, these variables have often been investigated separately leaving open the question of their joint contribution when included in a common analysis. Our study was therefore aimed to fit logistic regression models to examine the association between MDD diagnosis (defined as binary outcomes, i.e., the presence or absence of MDD diagnosis) and social, psychological, and biological variables. Participants included inpatients with current MDD and an MDD-free healthy control group. Selected social variables were EN and attachment style. Psychological variables were intrusive thoughts as a sign of failed thought suppression and cognitive reappraisal as a cognitive emotion regulation strategy. Citrate synthaseactivity (CSA), an enzymatic marker for mitochondrial density in PBMCs, was selected as a biological variable. Due to the cross-sectional character of our study, it cannot be determined whether the measured psychological constructs or biological variables precede, co-occur with, or are consequences of MDD. We hypothesized the following associations with MDD diagnosis: Individuals who had experienced more severe EN or who had a more insecure attachment style would be more likely to be in the MDD group than those without such experiences (hypotheses [h] 1 [h1] and 2 [h2]). Individuals reporting more frequent intrusive thoughts would have higher odds of belonging to the MDD group compared to those with less intrusive thoughts (h3). Individuals who report less frequent habitual use of cognitive reappraisal would have higher odds of belonging to the MDD group (h4). Finally, lower mitochondrial density in PBMCs is assumed to be associated with higher odds of belonging to the MDD group (h5).

All participants provided written informed consent to a protocol approved by the Local Ethics Committee at Ulm University (approval number 169/12) before participating in the study. All aspects were conducted in full accordance with the newest Declaration of Helsinki and its amendments. Inclusion and exclusion criteria (detailed in the section “Participants”) were assessed using a custom-designed demographic questionnaire. Additional data on age, sex, body mass index (BMI), and school education level were collected. MDD diagnoses were determined by an experienced psychiatrist using a DSM-IV-based structured interview, and medication history was recorded. After completing the self-report questionnaires (see the section Questionnaires), a 35-mL non-fasting blood sample was collected by venipuncture at the Department of Psychiatry and Psychotherapy III at Ulm University. Samples were sent to the Laboratory of the Department of Clinical & Biological Psychology at Ulm University for PBMC isolation, cryopreservation, and analysis. Mitochondrial density in PBMCs was assessed at the Institute of Anaesthesiological Pathophysiology and Process Development using the CSA assay as reported previously [18, 57]. All analyses adhered to Good Scientific Practice (GSP) and Good Laboratory Practice (GLP). The average time spent on study procedures was 2.5 h. Control participants received either four credits or EUR 40 for their participation, while the participation in the patient group was compensated with EUR 40.

The participants comprised two groups: a group of inpatients with current MDD (patient group) and a group of MDD-free healthy control participants (control group). Patients were recruited from 2015 to 2017 at the Department of Psychiatry and Psychotherapy III at Ulm University, meeting the criteria of a main diagnosis of MDD without comorbid mental disorders (e.g., posttraumatic stress disorder or anxiety disorders) or somatic illnesses (e.g., cardiovascular diseases or diabetes mellitus). Patients were diagnosed for MDD using a structured interview based on the criteria outlined in the DSM-IV by an experienced psychiatrist. Participants of the control group were recruited through public advertisements. They were also screened for mental disorders in the same way as the inpatients, using the Mini-DIPS interview [58] by a trained psychiatrist or psychotherapist. Inclusion criteria for all participants included age between 19 and 59 years, no current substance dependence, and no severe physical illnesses or medication affecting the immune system. We did not recruit participants aged 60 years and older to avoid the inclusion of participants with aging-related decline or aging-related disorders. Notably, only inpatients with a primary diagnosis of MDD were included in the present study, as we applied this strict criterion for our study to exclude a possible confounding influence of comorbid mental disorders. This is one of the reasons why our sample size was relatively small. Participants in the control group did not have a history of mental disorders or somatic illnesses according to self-report. The study initially involved 71 participants (n = 33 patient group, n = 38 control group). Twelve participants (9 from the patient group and 3 from the control group, who did not complete questionnaires and/or left the hospital before the completion of the present study) were excluded, leaving 59 participants for statistical analyses (n = 24 patient group, n = 35 control group). The sex distribution within the groups was not statistically different between the two groups (χ² = 0.487, df = 1, p = 0.485). The average age was 39.78 years, with no significant differences in age, sex, education, BMI, or smoking between groups (p > 0.05). As expected, we found significant differences in the current severity of depressive symptoms as measured by the self-rating scale Beck depression inventory (BDI-II) between the two groups (Mann-Whitney U tests = 0.814, p < 0.0001, see Table 1) (BDI-II, 59; Cronbach’s α = 0.98). Most patients (92%) received antidepressants based on clinical judgment and according to international clinical guidelines for depression psychopharmacotherapy. Additional sociodemographic and medication details are given in Table 1.

The German version [59] of the BDI-II [60] was utilized to assess the severity of depressive symptoms [59] by self-report. ACEs were assessed with the German version of the Childhood Trauma Questionnaire (CTQ) [61] developed by Bernstein and Fink [62]. The CTQ covers experiences of abuse, maltreatment, and neglect, comprising 28 items across 5 subscales. Emotional abuse (Cronbach’s α = 0.86), EN (Cronbach’s α = 0.90), physical abuse (Cronbach’s α = 0.88), and sexual abuse (Cronbach's α = 0.85) exhibited good to high reliability, while physical neglect showed lower reliability (Cronbach’s α = 0.56). All items are evaluated on a five-point Likert scale (1 = “not at all” to 5 = “very often”). ACE severity was measured using CTQ subscale sum scores. AASs were classified with the German translation [63] of the Relationship Scales Questionnaire (RSQ) [64], consisting of 4 subscales. They included fear of separation (Cronbach’s α = 0.77), fear of closeness (Cronbach’s α = 0.80), lack of trust (Cronbach’s α = 0.84), and desire for independence (Cronbach’s α = 0.68) and demonstrated acceptable to high reliability. Attachment styles (secure and insecure [preoccupied, dismissing, and fearful]) were determined based on criteria established by Steffanowski et al. [65]. The RSQ comprises 30 items rated on a 5-point Likert scale (1 = “not applicable” to 5 = “very applicable”). A secure attachment style is defined as a mean sum score of 2.88 or lower for the fear of separation subscale and 2.75 or lower for the fear of closeness subscale. The fearful attachment style is marked by higher scores on both subscales (fear of separation >2.88 and fear of closeness >2.75). Individuals with higher fear of separation scores (>2.88) and lower fear of closeness (≤2.75) were classified as preoccupied. The dismissive attachment style is characterized by a low fear of separation (≤2.88) and a high fear of closeness (>2.75). For further analyses, the three attachment styles (preoccupied, dismissive, and fearful) were considered instances of an insecure attachment style and contrasted with the secure attachment style. Intrusive thoughts were assessed with the German version [66] of the White Bear Suppression Inventory (WBSI) [67], featuring intrusion (Cronbach’s α = 0.89) and suppression (Cronbach’s α = 0.86) subscales. The questionnaire comprises 15 items across two subscales, with response options ranging from 1 (“not at all true”) to 5 (“completely true”) on a five-point scale. Cognitive reappraisal was measured using the German version [68] of the Emotion Regulation Questionnaire (ERQ) [41], with subscales for cognitive reappraisal (Cronbach’s α = 0.89) and expressive suppression (Cronbach’s α = 0.74). This questionnaire consists of ten items that are answered using a 7-point Likert scale (1 = “not at all correct” to 7 = “completely correct”) for each item.

A total volume of about 35 mL of non-fasting blood was taken by venipuncture into EDTA-buffered S-monovettes (Sarstedt, Nümbrecht, Germany) by the medical staff of the Department of Psychiatry and Psychotherapy III at Ulm University. The time of collection was fixed twice a day (morning and afternoon). After collection, the blood samples were sent by courier to the laboratory of the Department of Clinical and Biological Psychology at the University of Ulm, where the PBMCs were isolated using a Ficoll-Hypaque gradient according to the manufacturer’s protocol (GE Healthcare, Chalfont St Giles, UK). Isolated PBMCs were stored in a standardized cryoprotective freezing medium (ethyl sulfoxide: Sigma-Aldrich, St. Louis, MO, USA; fetal calf serum: Sigma-Aldrich; dilution: 1:10) at −80°C until respirometric analysis with the O2K high-resolution oxygraph (Oroboros Instruments, Innsbruck, Austria; see [49]). Subsequently, and as reported previously [18, 57, 69], a cell suspension volume including 1 mio living cells was shock-frozen with liquid nitrogen to measure CSA after respirometry (data not shown, in preparation for a different manuscript) in cell suspensions after thawing. Samples were measured in duplicates blinded to one executive experimenter following the protocol and recommendations by Eigentler and colleagues [70]. Mean values were used to determine CSA for each individual. The assessment of cell viability is a mandatory preanalytical step in the process of measuring mitochondrial physiology in PBMCs as reported previously [18, 57, 69, 71]. To do so, PBMCs are thawed in prewarmed (37°C), sterile phosphate-buffered saline (PBS). Next, 10 µL of the cell suspension is mixed with 10 µL of Trypan blue solution (0.4%, Sigma-Aldrich, USA) to count the total amount of cells and to co-assess the amount of dyed cells, representing dead cells in the sample, using a bright-field microscope (Nikon, Japan) in combination with a standard Neubauer chamber.

In general, circadian rhythm-based biological fluctuations can affect various physiological regulatory systems of the body. As described above, there were two blood sampling times (morning and afternoon). Due to limitations of holding blood collection times constant in our study, of relevance mainly in the control group, we tested for potential effects of the daytime of blood collection in this group by median splitting: Based on midnight as a reference time point, we estimated the time interval in hours (h) to blood collection and calculated the median to compare the following two subgroups: Individuals from the control group who provided a blood sample before 1:25 p.m. (n = 16) and control participants with a blood donation after 1:25 p.m. (n = 19). Next, we tested for potential group differences regarding age, BMI, the severity of depressive symptoms (BDI-II sum scores), and CSA values. The between-group comparison revealed no significant differences for age (t(33) = −0.714, p = 0.48), BMI (t(33) = −1.069, p = 0.293), and BDI-II sum scores (t(33) = 0.961, p = 0.343). Finally, we compared the CSA values in the two groups (blood collection before vs. after 1:25 p.m. according to the median split) and again found no significant differences (t(33) = −1.148, p = 0.259).

Statistical analyses were performed using the statistical software R (version 4.0.5) [72] for Windows (Microsoft, USA). Figures were generated using the package “ggplot2” [73]. First, to assess group differences, we conducted a t test for independent samples for all continuous variables, whereas Mann-Whitney U tests were used for the BDI-II sum score and the ordinally scaled variable of school education. Pearson’s χ² test was used for the comparison of the sex of the participants between two groups as a nominally scaled variable. Second, to determine the social, psychological, and biological variables for their association with a higher likelihood of being in the MDD group, multiple binary logistic regression analyses were performed using the package “lme4” [74].

Before testing our 5 hypotheses (h1−h5) (see the section “Introduction”), we conducted preliminary checks on the variables to determine their statistical suitability for logistic regression analysis. First, we assessed the absence of multicollinearity between the variables as a prerequisite for binary logistic regression using Spearman’s correlation. If the bivariate correlation exceeded r = 0.8, these predictors were excluded from each model. Second, categorical variables (e.g., sex, attachment styles) may result in lower power and weaker model fit, when there are less than 20% or less than 5 observations in the sample for each category in logistic regression. Therefore, the dichotomous control variable AASs were created and dummy-coded. The variable sex was omitted due to the small number of males. However, the distribution of females and males was similar in the control and patient groups (χ²(1) = 0.487, p = 0.485). Third, considering the relatively small sample size (n = 59) in the present study, the number of predictors in the binary logistic regression was limited to a maximum of six [75, 76]. For each predictor category, only the variable with the greatest effect size (Cohen’s d, see Table 2) in the statistical comparison between the patient group and control group was included.

The model tested in the regression analysis included the following five biopsychosocial variables and age as a control variable as predictors: Two social variables (emotional neglect[EN] and attachment style), two psychological variables (cognitive reappraisal and intrusive thoughts), and mitochondrial density in PBMCs (CSA) as a biological variable. EN demonstrated a higher effect size compared to the other CTQ subscales (see Table 2A) and has been most robustly associated with depression in the previous research [25]. Attachment style was included in the model as a dummy-encoded dichotomous variable with two levels: “secure” (1) and “insecure” (0). The goodness of fit of the logistic regression model was evaluated using Nagelkerke’s R² [77] and the Hosmer-Lemeshow test [78]. Adjusted odds ratio and 95% confidence intervals were reported for each variable in the model. We estimated the predicted probability of being in the patient group for the variables in the model. The statistical significance level was set at p < 0.05.

To assess whether the regression model including five variables as predicted was superior compared to models including less predictors, simpler hierarchical models were calculated and compared to the more complex model reported here. The model comparisons are described in the online supplementary material (supplementary data analyses; for all online suppl. material, see https://doi.org/10.1159/000544833).

Mean scores of the five selected biopsychosocial variables of the patient group and the health control group, respectively, are displayed in Table 2. There was a significant difference between the groups on the CTQ subscales of EN (t(57) = −3.026, p = 0.004) and emotional abuse (t(57) = −3, p = 0.004). The control group displayed low to moderate scores on the scale EN of the CTQ (M = 10.11, SD = 5.26), while the patient group exhibited moderate to severe scores (M = 14.22, SD = 4.91) based on the classification of Häuser and colleagues [79]. For emotional abuse, the patient group showed moderate scores (M = 12.54, SD = 5.08), while the control group reported low to moderate scores (M = 8.89, SD = 4.26). Attachment style in adulthood (RSQ) also significantly differed between groups (χ²(1) = 4.390, p = 0.036, φ = −0.308): In the control group, n = 17 participants reported insecure attachment, while n = 18 reported secure attachment. In the patient group, n = 19 participants displayed an insecure attachment style, while n = 5 participants showed a secure attachment style (see Table 2). The patient group and the control group also differed on the WBSI scale intrusive thoughts (t(57) = −7.1, p < 0.001, d = 1.88). The patient group (M = 19.74, SD = 4.42) reported more frequent intrusive thoughts compared to the control group (M = 11.90, SD = 4.72). Regarding the ERQ scale cognitive reappraisal, there was also a significant difference between the groups (t(57) = 3.60, p < 0.001, d = 0.957). The control group (M = 28.26, SD = 7.42) employed the cognitive reappraisal strategy more frequently than the patient group (M = 21.32, SD = 7). The results of the remaining psychological variables are reported in Table 2B. Mitochondrial density as indexed by CSA is reported in Table 2C (see also online suppl. Fig. S2 for spread and degree of the values). Mean CSA (nmol citrate/min per mio cells), measured as an index of mitochondrial density in PBMCs, was significantly lower (t(57) = 2.306, p = 0.024, d = 0.611) in the patient group (M = 2.7, SD = 1.08) compared to the control group (M = 3.49, SD = 1.42).

Spearman’s correlation analyses were performed among the predictor variables included in the model to examine the relationships among different biopsychosocial variables and the control variable age. There was a strong negative correlation (r = −0.560, p < 0.001) between intrusive thoughts and cognitive reappraisal as psychological variables. Other variables did not significantly correlate with age, social, and other biological parameters (all p > 0.05).

In the binary logistic regression analysis, the diagnosis of MDD was determined as the dependent variable, while EN, attachment style, intrusive thoughts, cognitive reappraisal, and CSA were used as predictor variables. The control group was used as the reference category. This analysis revealed that EN, intrusive thoughts, and CSA were significantly associated with the likelihood of being in the MDD group. Attachment style and cognitive reappraisal were not significantly associated with MDD. The regression results are summarized in Table 3. The tested regression model exhibited a very good fit according to several parameters (see Table 3). The predicted probabilities of belonging to the MDD group for the significant predictor variables in the regression model are shown in Figure 2.

Model comparisons of the tested regression model including the five social, psychological, and biological variables with simpler models including less predictor variables revealed a significantly better fit and lower information criteria of this more complex model. The results of these additional analyses are reported in the Supplementary Material (see online suppl. Table S1; Fig. S1 for receiver-operating characteristic curves).

According to the biopsychosocial model, variables at different levels of analysis are related to MDD. Previous research mainly considered the relation of single variables in isolation. Due to their partial interdependence, it is important to identify the unique contribution of each variable within the context of others. The present study thus aimed to assess the association of the social (EN, AAS), psychological (intrusive thoughts, cognitive reappraisal), and biological (mitochondrial density in PBMCs by CSA assessment) variables with the presence of a diagnosis of MDD in binary logistic regression analyses. This analysis showed that EN, intrusive thoughts, and CSA were significantly associated with the likelihood of an MDD diagnosis.

EN in childhood was associated with a higher likelihood of a diagnosis of MDD, indicating a higher risk of MDD. This finding is in line with several previous studies, showing an association of EN in childhood with the development of depression in adolescence and adulthood [80‒82]. Children with experiences of EN often develop low self-esteem and a sense of powerlessness, which can significantly contribute to the development and manifestation of depression in adulthood [17]. Understanding that the risk of developing a treatment-resistant disorder is twice as high in people with ACEs [25] will enable individual risk assessments that can inform holistic and therefore more personalized therapies. Häuser et al. [79] also suggested that EN is one of the most common forms of ACEs. Intrusive thoughts were also associated with a diagnosis of MDD as previously shown [36, 83]. They often arise due to cognitive load or stress when attempting to suppress unpleasant thoughts [84]. Watkins and Moulds [36] found lower WBSI scores in acutely depressed patients compared to remitted patients with MDD. However, the remitted patients still had significantly higher WBSI values than control participants without a history of depression. Another study [83] also demonstrated the association between depression severity and thought suppression. The persistent presence of intrusive thoughts in depression provides a crucial target for therapeutic intervention. Recognizing dysfunctional emotion regulation based on intrusive thoughts resulting from heightened stress [36] serves as a foundation for interventions such as competence training in flexible and diverse emotion regulation, stress management, and cognitive resource enhancement within psychotherapy. CSA measured in PBMCs was also associated with the probability of an MDD diagnosis in our regression model. Lower CSA in PBMCs was linked to a higher probability of belonging to the patient group, indicating reduced mitochondrial energy metabolism in MDD [85]. This finding is important for interpreting associations between CSA and diagnosis of MDD, specifically, as the patient group and the control group were matched for age and BMI in the present study. Inconsistent results regarding the character of mitochondrial biogenesis in different body cells across health and disease in previous studies might be related to an insufficient matching of these possibly confounding variables [86]. As the storage time of samples in the freezer might affect the CSA results, future studies should also analyze interindividual differences by measuring samples not in batches as conducted and reported here, but also measuring these values at distinct time intervals after freezing to be able to better characterize the effects on sample quality and CSA results.

Unlike earlier research [30, 31, 87, 88] and despite differences between patient and control groups, insecure attachment was not significantly related to MDD in the present study. In the presence of other statistically related predictors in the regression model, insecure attachment may not contribute strongly to the likelihood of being diagnosed with MDD, compared to EN. Similarly, contrary to earlier findings [15, 42], cognitive reappraisal was not a significant predictor in our regression analyses. This is likely due to the relatively high negative correlation between intrusive thoughts and cognitive reappraisal (r = −0.560, p < 0.001), indicating that cognitive reappraisal does not have sufficient unique explanatory power in the regression model. Based on the present regression results, it seems that intrusive thoughts are more closely related to depression diagnosis compared to cognitive reappraisal.

Overall, this study shows a complex, holistic picture with variables at different levels of analysis to be related to MDD diagnosis. Understanding the multifaceted nature of this mental disorder with different levels of potential action also enables appropriate and personalized therapy with numerous advantages. These range from relatively rapid recovery from an acute illness to reducing the risk of relapse as well as developing chronic or recurrent depression. In addition, the concept of predictive, preventive, and personalized medicine (3 PM) is being introduced into clinical psychology and psychiatry. In this way, it would lead to benefits for both the individual and the society as a whole.

The interpretation of our findings is limited by factors that need to be considered. First, our sample size was relatively small with unequal numbers of participants in the control group and the patient group. As a result, we decided to restrict the number of predictors for the logistic regression analyses to six variables [75, 76]. Given this relatively small sample size, our findings therefore have a rather preliminary character, mainly aimed at stimulating future studies with larger sample sizes to test the robustness and generalizability of our findings. Also due to the pronounced overrepresentation of female patients with MDD in our sample, which reflects the preponderance of female patients with MDD [89], sex differences could not be analyzed in the present study. Future highly powered studies could strive for a more balanced distribution of sexes. EN, adult attachment, intrusive thoughts, and cognitive reappraisal were assessed by self-report questionnaires, which might be prone to biases including social desirability. More theoretical conceptualization is needed to further explain the model. The present results could be validated using other instruments to measure these variables such as observation or external assessment. Finally, CSA in PBMCs was measured as a marker for mitochondrial density. Therefore, it remains open, whether the present observation of reduced CSA in PBMCs from patients with MDD generalizes to other cells and tissues (e.g., platelets, muscle fibers, brain tissue). As the enzymatic assessment of CSA is sensitive, but also requires relatively high resources, the question remains whether alternative laboratory measures might be able to replace CSA in future biomarker approaches. For example, the assessment of mitochondrial DNA (mtDNA copy numbers) by polymerase chain reaction seems to be a promising alternative [90, 91]. Moreover, although the control analysis did not show effects of the timing of blood sampling on CSA, further studies could either standardize the time of blood sampling or systematically test for circadian fluctuations of CSA in both individuals with and without MDD. Another limitation is related to the presence of medication. All patients with MDD included in this study received antidepressive medication, which might have influenced mitochondrial density [92‒94]. There are some signs of evidence that antidepressive medication can affect mitochondrial biogenesis and bioenergetics [93]. For example, there are indications that selective serotonin reuptake inhibitors (SSRIs) may change the functionality of the mitochondria [93]. Connections between the SSRIs and the mitochondria-related apoptosis of the cells are also discussed, although a dose-time effect can be assumed [95]. Moreover, 66.67% of the depressed patients in the present sample received an SSRI, and some participants in the patient group received two or more substances. We therefore cannot rule out that the association between mitochondrial biogenesis and MDD diagnosis observed in the present study is potentially biased by medication effects in the patient group. Due to the small sample size in our study, an equivalence dose of antidepressive medication could not be included in the analyses as a control variable. Thus, future studies should include measures of medication as a control variable in the model. Alternatively, medication-free or medication-naive patients could be investigated to replicate the present findings.

As our main goal was to predict MDD diagnosis, we treated depression as a categorical variable in a logistic regression analysis. Similarly, as outlined in the introduction section, the five factors selected in our study (EN, insecure attachment style, intrusive thoughts, cognitive reappraisal, and CSA) are known from previous studies to be frequently associated with depressive symptoms in general. However, the association of these factors with specific symptoms of depression has not been systematically assessed to the best of our knowledge. As a consequence, the assessment of such symptom-specific associations would be an interesting question for future research.

In sum, the present logistic multiple regression analysis determined EN, intrusive thoughts, and CSA in PBMCs as important variables associated with an MDD diagnosis. This result thus supports integrative biopsychosocial models of MDD and highlights the multifaceted nature of this mental disorder. The present multivariate approach including variables at different levels of analysis revealed unique contributions of social, psychological, and biological variables to be associated with MDD diagnosis. Our study also has implications for the diagnosis and therapy of depression by emphasizing the role of EN, intrusive thoughts, and mitochondrial readouts (CSA or mtDNA copy number a future easily applicable biomarker test) as potential factors related to the development or maintenance of depression. Nonetheless, future highly powered studies need to replicate these first findings in independent cohorts, but the arguments to do so are underlined here.

The authors thank Eva-Maria-Kögel for her help during data collection.

The study protocol was reviewed and approved by the Local Ethics Committee (Approval No. 169/12) of the University of Ulm, Germany. All aspects of the study were conducted in accordance with the tenets of the Declaration of Helsinki. All participants gave written informed consent before participation.

Prof. Asst. Alexander Karabatsiakis was a member of the journal’s Editorial Board at the time of submission. The other authors have no conflicts of interest to declare.

This study was not supported by any sponsor or funder.

K.S.: data curation, formal analysis, and writing – original draft. E.-J.S.: conceptualization, formal analysis, methodology, project administration, writing – original draft, and writing – review and editing. K.W.: investigation, writing – original draft, and writing – review and editing. C.S.-L.: conceptualization, project administration, supervision, writing – original Draft, writing – review and editing, and validation. P.R.: formal analysis, resources, investigation, and writing – original draft. A.K.: formal analysis, resources, investigation, methodology, writing – original draft, writing – review and editing, and supervision. M.K.: conceptualization, writing – original draft, writing – review and editing, supervision, project administration, and validation.

Katharina Strecker, Eun-Jin Sim, Alexander Karabatsiakis, and Markus Kiefer contributed equally to this work.

The data that support the findings of this study are not publicly available due to local data policy regulations but are available from the corresponding author, M.K., upon reasonable request.