Integration of transcriptome profiling to identify key genes involved in the interplay between oxidative stress and mitophagy in major depressive disorder, followed by multidimensional phenotypic validation

Abstract

Background:

Major depressive disorder (MDD) is recognized as a pressing global public health burden. However, its molecular mechanisms remain incompletely understood.

Methods:

In this study, an integrative analysis of transcriptome datasets from the GEO database was conducted. GEO2R and the R programming language were used to identify differentially expressed genes (DEGs) related to oxidative stress and mitophagy. Key hub genes, such as EEF2, CCT3, EIF3I, and RPS5, were further identified through enrichment analysis and protein–protein interaction (PPI) network construction. Following validation using an independent human dataset, we established a corticosterone-induced C8-D1A cell model. Reactive oxygen species and mitochondrial membrane potential were measured via flow cytometry. The results demonstrated that this model reliably recapitulates key pathological features of elevated oxidative stress and mitochondrial dysfunction in MDD. Finally, using an in vivo mouse model, we assessed synapse-associated proteins and mitophagy markers using Western blotting and measured the mRNA expression levels of candidate genes by qPCR to comprehensively validate the associations between the expression of the aforementioned genes and oxidative stress, mitophagy, and synaptic damage.

Results:

This study combined bioinformatics screening and multidimensional phenotypic validation to construct an MDD-specific molecular regulatory network focused on carbon metabolism, thereby elucidating the interplay between four genes and oxidative stress and mitophagy. Although CCT3 and RPS5 demonstrated modest diagnostic utility in the independent validation dataset (AUC ≈ 0.6, Padj < 0.05), subsequent in vivo experiments revealed that the mRNA expression levels of these genes were significantly downregulated in MDD models (EEF2: P < 0.05; CCT3: P < 0.005; EIF3I: P < 0.05). Furthermore, the expression levels of these genes were positively correlated with those of synaptic proteins and negatively correlated with those of mitophagy markers. The downregulation of these genes may impair protein synthesis and folding, which acts in synergy with oxidative stress and mitochondrial dysfunction to perpetuate the vicious cycle of bioenergetic crisis and proteostasis collapse in MDD.

Conclusion:

Although this study did not experimentally validate the regulatory functions of the target genes or identify highly specific diagnostic biomarkers, it offers a novel molecular perspective for deciphering the complex pathology of MDD. Notably, this highlights the synergistic interaction between translational regulation and metabolic homeostasis. Further validation in larger independent cohorts is warranted to assess the viability of these genes as mechanistic therapeutic targets.

1 Introduction

Major depressive disorder (MDD) is a prevalent global health concern with high recurrence and disability rates. It is clinically characterized by persistently depressed mood, anhedonia, and cognitive impairment, along with diverse somatic symptoms, leading to significant deficits in patients’ social functioning and quality of life (1, 2). Epidemiological data reveal that more than 350 million individuals worldwide suffer from MDD, resulting in a substantial disease burden. In terms of years lived with disability (YLDs), MDD ranks as the second leading cause of global disability, while in terms of disability-adjusted life years (DALYs), it ranks 15th among causes of the global disease burden (3). The dual burden of high prevalence and substantial economic loss stands in stark contrast to the severe imbalance in the global allocation of mental health resources, resulting in significant geographic disparities in the treatment gap. A 2021 study examining service coverage for major depressive disorder (MDD) across 204 countries and regions revealed that only 9.1% of patients globally received minimally adequate treatment, with coverage rates decreasing to as low as 2.0% in low-income regions (4, 5). Current diagnosis relies mainly on the taxonomic criteria of the DSM-5-TR and ICD-11. Major depressive disorder (MDD) has a highly heterogeneous pathological mechanism, and a total of 227 distinct symptom combinations can meet the diagnostic thresholds of the DSM-5-TR, thereby leading to low clinical recognition rates. Recently, both diagnostic systems have increased their emphasis on biological dimensions. The DSM-5-TR incorporates specifiers such as “with mixed features” to refine diagnostic subtypes, whereas the ICD-11 employs a semantic knowledge framework to facilitate the future integration of biomarker data (4, 6). The translation−regulatory genes identified in this study provide molecular evidence for refining such diagnostic systems, thereby aligning with the evolving trends in diagnostic criteria.

Abnormal activity of the hypothalamic–pituitary–adrenal (HPA) axis is among the core biological characteristics of major depressive disorder (MDD) (1). HPA axis hyperactivity has been established as a crucial pathophysiological mechanism underlying MDD (7). Juruena (8) reported that stressful life events, such as childhood emotional abuse, physical abuse, and sexual abuse, can trigger hyperactivity of the HPA axis and an increase in circulating cortisol levels. Hypercortisolemia plays a central role in stress−related neuropathologies, including depression (9). Astrocytes play pivotal roles in the mechanisms underlying stress-related neuropathology, depressive susceptibility, and disease persistence. Specifically, astrocytic dysfunction acts as a crucial mediator between chronic stress and impaired neural plasticity (10, 11). Growing evidence from studies on major depressive disorder (MDD) indicates that disruption of the astrocyte–neuron glutamate–glutamine cycle, particularly a homeostatic imbalance within the medial prefrontal cortex, may precipitate the accumulation of reactive oxygen and nitrogen species and ultimately induce depressive-like behaviors (12). Furthermore, large-scale clinical studies have confirmed that patients with MDD generally exhibit elevated oxidative stress levels, manifested by increased lipid peroxidation and oxidative protein damage, as well as significantly reduced levels of antioxidant enzymes and nonenzymatic antioxidants (13–15). An imbalance in the oxidative−antioxidant system is among the key pathological hallmarks of major depressive disorder (MDD) (16, 17). Furthermore, the de novo synthesis of glutamine in astrocytes during the glutamate–glutamine cycle consumes a substantial amount of energy through the generation of α−ketoglutarate from the TCA cycle (18). Glutamate uptake is also an energy-intensive process (19). Consequently, mitochondrial dysfunction severely compromises astrocytes within the glutamate-glutamine (Glu-Gln) cycle. Mitochondrial quality control forms the fundamental regulatory network for maintaining mitochondrial function. Among these processes, mitophagy serves as a pivotal mechanism for eliminating damaged mitochondria under stress and sustaining the cellular energy supply. Mitophagic dysregulation plays a pivotal role in the pathogenesis of depression. Recent studies have demonstrated that drugs targeting mitophagy hold considerable potential for treating neurodegenerative diseases and aging, thereby positioning them as promising therapeutic targets with substantial value for research and clinical translation (20). The specific types of crosstalk among major depressive disorder (MDD), oxidative stress and mitophagy require further elucidation.

By mining public databases and performing bioinformatic analyses, this study identified differentially expressed genes (DEGs) between patients with MDD and healthy controls. Venn diagram analysis was employed to identify overlapping DEGs associated with oxidative stress and mitophagy. Subsequent analyses included functional enrichment analysis and the identification of hub genes within the protein–protein interaction (PPI) network, which was subsequently visualized using Cytoscape software (version 3.10.3) (21). The candidate genes were identified via the CytoHubba plugin (22). These genes may contribute to the pathophysiological mechanisms of MDD and act as potential biomarkers linked to oxidative stress and mitophagy. Candidate genes were validated using an independent human validation cohort and a corticosterone-induced depressive mouse model. This study aims to identify key predictive factors for the development and progression of MDD and to explore novel multitarget synergistic intervention strategies informed by these findings.

2 Materials and methods

2.1 GEO dataset

The dataset was obtained from the NCBI GEO database (https://www.ncbi.nlm.nih.gov/geo/) with the search terms “major depressive disorder” and “RNA expression”. The GSE52790 dataset, deposited by Li et al., was downloaded from the GEO database and served as the training set. This dataset was generated using the Affymetrix Human hGlue_3_0_v1 Array platform and included 10 patients with major depressive disorder (MDD) and 12 age- and sex-matched healthy controls. The original metadata of this dataset did not provide data on the medication history of the participants. A total of 22 samples were included in this dataset and acquired from public databases. The GSE98793 dataset, generated using the Affymetrix Human Genome U133 Plus 2.0 Array platform, was used as the validation set and comprises 168 patients with major depressive disorder (MDD) and 64-well-matched healthy controls. Given the retrospective nature of the study, informed consent and ethics committee approval were not needed.

2.2 DEG identification

Differentially expressed genes (DEGs) in the training dataset GSE52790 were identified using the online tool GEO2R (23). Genes whose ∣Log2FC∣>1 and P value<0.05 were classified as differentially expressed for subsequent analysis. Gene annotation files were downloaded from the Ensembl database (24) (https://www.ensembl.org/), and four lncRNAs were excluded. Analyses were performed using R version 4.2.1, and the R package ggplot2 (version 3.4.4) was employed for subsequent analyses and the visualization of PCA plots. In addition, phenotype-related gene sets related to mitophagy and oxidative stress were retrieved on September 8, 2025, from GeneCards: The Human Gene Database (https://www.genecards.org/) (25) using the keywords “mitophagy” and “oxidative stress”. The relevance score was thresholded at ≥ 1, and the category was restricted to “protein coding”. After these criteria were applied, a total of 1716 mitophagy−related genes and 3056 oxidative stress−related genes were retained. These gene lists were cross-referenced with the differentially expressed genes (DEGs) from the GSE52790 dataset to identify DEGs linked to mitophagy and oxidative stress. The R packages ggplot2 (v3.4.4), VennDiagram (v1.7.3), and ComplexHeatmap (v2.13.1) were employed to generate Venn diagrams and heatmaps, respectively.

2.3 Functional enrichment analysis

Functional enrichment analyses of the DEGs were conducted via the Xiantao Academic Online Platform (https://www.xiantaozi.com). Venn diagrams were generated from the DEGs associated with mitophagy, oxidative stress, and MDD to identify common coexpressed genes. Subsequent GO enrichment analysis (covering biological processes [BP], cellular components [CC], and molecular functions [MF]) and Kyoto Encyclopedia of Genes and Genomes (KEGG) pathway analysis, along with their visualization, were performed using the following parameters: Min Overlap = 3, P value cutoff = 0.05, and Min Enrichment = 1.5.

2.4 Protein–protein interaction network analysis

The list of differentially expressed genes linked to mitophagy and oxidative stress was submitted to the STRING database for protein–protein interaction analysis (https://string-db.org) (26). Construction of the protein–protein interaction (PPI) network: The search was restricted to “Homo sapiens” within the STRING database interface, and the minimum interaction confidence score was set to the medium threshold (≥ 0.400). This threshold filters out low-confidence predicted interactions, retaining only those pairs with moderate or higher confidence derived from multiple sources, such as experimental validation, database predictions, and literature mining. This approach ensures sufficient network coverage while minimizing background noise. Following data retrieval, the interaction tables and network files (in TSV format) were exported and subsequently imported into Cytoscape (version 3.10.3) for PPI network visualization and topological analysis. The MCC algorithm, available within the CytoHubba plugin, was employed to identify the top 10 hub genes on the basis of their degree scores; these genes were then selected as candidate targets for subsequent functional enrichment and mechanistic validation (27).

2.5 Candidate gene diagnosis model construction

In this study, preliminary screening and diagnostic efficacy evaluation of candidate genes were conducted via the diagnostic ROC module of the online XianTaoZi platform (https://www.xiantaozi.com). After confirming consistent expression trends, an adjusted P value (P.adj) < 0.05 was set as the threshold for statistical significance. Additional validation was performed using the independent external dataset GSE98793 (containing peripheral whole-blood transcriptome data from 128 patients with major depressive disorder (MDD) and 64 healthy controls). Batch effect correction was performed using the ComBat algorithm from the R “sva” package. Multiple testing correction was performed using the Benjamini–Hochberg (BH) method with m=4, which was applied exclusively to the testing of candidate genes. Then, the pROC package was employed to generate receiver operating characteristic (ROC) curves for each candidate gene exhibiting significant differential expression in the validation set to calculate the area under the curve (AUC) and its 95% confidence interval (CI) to evaluate the diagnostic performance and generalizability of the findings. (The detailed code is provided in the Supplementary Material).

2.6 Experimental validation

2.6.1 Cell experiments

2.6.1.1 Grouping and viability of the C8-D1A cell line

Astrocytes function as pivotal metabolic regulators in the brain. Their crucial roles in maintaining the physiological homeostasis of the nervous system are mediated primarily by the regulation of glutamate excitotoxicity, antioxidant defense, and cerebral energy metabolism (28–30). The C8-D1A cell line (Pricella, China) is a spontaneously immortalized cell line derived from mouse cerebellar astrocytes. After cell resuspension, the cells were cultured under standard conditions. Cells were maintained in high-glucose DMEM supplemented with 10% fetal bovine serum (FBS) and 1% penicillin-streptomycin (P/S) and incubated at 37 °C in a humidified incubator with 5% CO₂ and 95% ambient air. Cells in the logarithmic growth phase were used for all experiments. Cells were seeded into each well of a 96−well plate (10,000 cells/well) and incubated at 37 °C. After 24 h, 200 μL of the corticosterone solution at concentrations ranging from 0 to 800 μmol/L (0, 50, 100, 200, 400, 800 μmol/L) was added to each well, and the cells were incubated for an additional 48 h. Cell viability was evaluated via the CCK−8 assay. Briefly, 10 μL of CCK−8 solution was added to each well. After incubation at 37 °C for 40 min, the absorbance was measured at 450 nm with a microplate reader (BioTek Instruments, Inc., USA). Cell viability was calculated using the following formula: cell viability (%) = (experimental − blank)/(control − blank) × 100%. Each group consisted of five replicate wells, and all experiments were performed in triplicate.

2.6.1.2 Determining intracellular reactive oxygen species (ROS) levels and mitochondrial membrane potential

Mitochondria are known as the “powerhouses” of cells, and a vicious cycle of bidirectional regulation occurs between mitochondrial damage and reactive oxygen species (ROS) production. Dysfunction of the mitochondrial electron transport chain, particularly Complexes I and III, induces electron leakage, leading to the excessive generation of ROS such as superoxide anions (O₂^-) and hydrogen peroxide (H₂O₂). Excess ROS further damage mitochondrial inner membrane lipids, respiratory chain proteins, and mitochondrial DNA (mtDNA), resulting in the disruption of membrane integrity, a decrease in mitochondrial membrane potential, impaired ATP synthesis, and exacerbated mitochondrial dysfunction. Therefore, ROS levels serve as a core quantitative indicator of mitochondrial damage (31). Intracellular ROS levels were measured using a DCFH−DA detection kit (Beyotime, China). C8-D1A cells in the logarithmic growth phase were collected and resuspended in DMEM at a concentration of 2.5×10⁵ cells/mL. The cell suspension was seeded into a 6-well plate at 2 mL per well and incubated overnight at 37 °C under 5% CO₂ in a humidified atmosphere. The cells were assigned to the normal group or the corticosterone−treated model group. The cells were rinsed twice with PBS and then centrifuged at 1500 rpm for 5 min, after which the supernatant was discarded. Subsequently, 1 mL of DCFH solution diluted with serum-free medium at a ratio of 1:1000 was added, and the cells were incubated at 37 °C for 20 min. The suspension was gently agitated every 3 min during incubation. Following three washes with serum-free medium, the cells were resuspended in 500 μL of PBS. The intracellular ROS level was ultimately determined by flow cytometry (Beckman Coulter, USA).

Changes in mitochondrial membrane potential were assessed using a JC-1 detection kit (Beyotime, China). Healthy C8-D1A cells in the logarithmic growth phase were collected and prepared as single-cell suspensions in DMEM, with the cell concentration adjusted to 2.5 × 10⁵ cells/mL. Cells were seeded uniformly into a 6-well plate at 2 mL per well and incubated overnight at 37 °C in a humidified atmosphere containing 5% CO₂. The cells were divided into a normal group and a corticosterone-treated model group. Subsequently, 0.5 mL of JC-1 staining working solution was added, and the mixture was gently inverted several times to ensure homogeneous mixing. The cells were incubated at 37 °C for 20 min. Concurrently, 1× JC-1 staining buffer was prepared by diluting the 5× stock with distilled water at a 1:4 (v/v) ratio and keeping it on ice. After incubation at 37 °C, the cells were centrifuged at 1200 rpm for 3 min at 4 °C, and the supernatant was discarded. The cells were subsequently washed twice with 1× JC-1 staining buffer. For each wash, the cells were resuspended in 1 mL of buffer and centrifuged at 1200 rpm for 3 min at 4 °C, after which the supernatant was removed. This resuspension and centrifugation step was repeated once 1 mL of fresh 1× JC-1 staining buffer was used under identical conditions, after which the supernatant was removed. Following resuspension in 500 μL of 1× JC-1 staining buffer, changes in JC-1 fluorescence were measured using a flow cytometer (Beckman Coulter, Brea, CA, USA).

2.6.2 Animal experiments

2.6.2.1 Experimental group allocation

This study was conducted in accordance with the internationally accepted ARRIVE 2.0 guidelines regarding experimental design and reporting, thereby fulfilling all 10 essential items. A total of 20 4-week-old C57BL/6 mice (weighing 12 ± 1 g) consisting of an equal number of males and females were purchased from Henan Sikebeisi Biotechnology Co., Ltd. (License No.: SCXK(Yu) 2025-0005). The animal facility environment was maintained at 24 ± 2 °C with a relative humidity of 55 ± 5%. The animals were kept on a 12-hour light/dark cycle. All the mice were provided ad libitum access to purified water and standard laboratory chow. The sample size was calculated a priori using GPower 3.1 (Heinrich Heine University Düsseldorf, Germany). For the independent two−sample t−test, the two−tailed significance level (α) was set at 0.05, and the target statistical power was set to 0.8. On the basis of data from preliminary sucrose preference tests, the mean sucrose preference index was 94.5% in the normal control group and 69.7% in the depression model group, yielding a pooled standard deviation (σ) of 0.17. A Cohen’s d effect size of 1.46 was calculated, indicating a required sample size of 6 animals per group. To account for an anticipated 15% rate of modeling failure and animal attrition, an initial sample size of 10 mice per group was determined. The mice were randomly assigned to either a normal control group or a depression model group induced via subcutaneous corticosterone injection. Following a 7-day acclimatization period, control mice were group-housed, whereas model mice were individually housed under controlled experimental conditions. The Experimental Animal Ethics Committee of Guizhou Medical University approved this study (No. NO2502768).

2.6.2.2 Establishment of a model of depression-like behavior

Corticosterone solution was prepared with 0.1% DMSO, 0.3% Tween−80 and normal saline. Briefly, corticosterone was dissolved in a minimal volume of DMSO to prepare a stock solution to a final concentration of 25 mg/mL. The working solution was then diluted to 2.5 mg/mL with a mixed vehicle consisting of 40% PEG300, 5% Tween−80 and 45% normal saline. Mice were subcutaneously injected with exogenous corticosterone at a daily dosage of 20 mg/kg body weight for 3 consecutive weeks to establish a chronic stress state and induce depression−like behaviors, thereby mimicking hypothalamic–pituitary–adrenal (HPA) axis dysfunction caused by chronic stress in humans (1, 32, 33). Body weight was measured every 3 days, and dorsal injection sites were examined daily; animals exhibiting redness, swelling, or induration with a diameter exceeding 3 mm were excluded. The mice in the healthy control group were subcutaneously injected with an equal volume of vehicle containing 0.1% DMSO, 0.3% Tween−80 and normal saline. Behavioral tests were performed before and after the experimental intervention. All the mice were euthanized under deep anesthesia. Anesthesia was induced via inhalation of 3%–5% isoflurane. Once a stable plane of anesthesia was achieved, it was maintained with 0.7–1.0% isoflurane via inhalation (34). After cardiac perfusion with ice-cold normal saline, the whole brain was rapidly excised. The hippocampi were dissected on an ice-cold dissection platform, gently rinsed with ice-cold normal saline, and stored at −80 °C for subsequent analysis.

2.6.2.3 Behavioral testing

The sucrose preference test (SPT) was used to assess stress-induced an hedonic behavior in the mice. (35). This behavioral paradigm for assessing depression-like phenotypes is predicated on the strong innate preference of mice for sucrose. Each mouse was given access to two bottles: one filled with plain water and the other filled with a 2% sucrose solution. After a 24-hour adaptation period, the positions of the two bottles were swapped every 12 hours to mitigate potential biases arising from side preference. To align with the nocturnal activity patterns of rodents, the experimental phase was conducted from 21:00 to 09:00. Bottle weights for water and sucrose solution were measured pre- and post-experiment. Fluid intake for each mouse was determined by calculating the weight loss of each bottle. Sucrose preference was calculated as the percentage of sucrose intake relative to total fluid intake.

The tail suspension test (TST) is a behavioral paradigm used to assess depression-like behaviors. It is based on the observation that mice, when suspended by their tails, initially attempt to escape but eventually discontinue active escape attempts and exhibit a characteristic immobile state interpreted as a measure of despair. One-third of the mouse tail was suspended in the tail suspension apparatus using adhesive tape to maintain the head approximately 20 cm above the ground. Mouse behavior was videotaped for 5 minutes. After each trial, the feces were removed, and the bottom of the suspension apparatus was disinfected with an ethanol-based solution. All the data were analyzed using EthoVision XT software from Noldus Information Technology.

The forced swimming test (FST) is a well-established paradigm for assessing depression-like behavior. After repeatedly unsuccessful attempts to escape within an inescapable water tank, mice gradually cease actively struggling and enter a state of floating immobility, which is indicative of depression-like despair behavior. The mice were individually placed in a transparent cylindrical container (approximately 20 cm in diameter and 27.5 cm in height). The water depth was maintained at approximately 20 cm to prevent the tail from providing support, and the water temperature was maintained between 23 and 25 °C. Each mouse underwent a 5-minute swimming trial. The water in the cylinder was replaced before each trial to eliminate olfactory interference. The behavior of the mice was videotaped from a lateral view. All the data were analyzed using EthoVision XT software from Noldus Information Technology.

2.6.2.4 Western blotting analysis

Western blotting was employed to assess the expression levels of specific proteins in the mouse prefrontal cortex, namely, brain-derived neurotrophic factor (BDNF), postsynaptic density protein-95 (PSD-95), and mitophagy-related proteins, namely, PTEN-induced kinase 1 (PINK1), Parkin RBR E3 ubiquitin protein ligase (Parkin), microtubule-associated protein 1 light chain 3 beta (LC3B), and sequestosome 1 (SQSTM1/P62). After mouse brain tissue samples were collected, total protein was extracted using RIPA lysis buffer. The protein concentration was measured with a BCA protein quantification kit (Beyotime, China). Protein samples were heat-denatured at 100 °C. Equal amounts of protein were subsequently separated by 10% sodium dodecyl sulfate–polyacrylamide gel electrophoresis (SDS–PAGE) (Yamay, China). After electrophoretic separation, the proteins were transferred onto PVDF membranes and subsequently blocked with 5% skim milk for 1 h. The membranes were incubated with primary antibodies overnight at 4 °C. The primary antibodies and their working dilutions were as follows: anti-BDNF (recombinant rabbit monoclonal, 1:10,000; Huabio), anti-PSD95 (rabbit monoclonal, 1:2,000; Zenbio), anti-PINK1 (rabbit monoclonal, 1:2,000; Huabio), anti-Parkin (rabbit monoclonal, 1:1,000; ABMART), anti-LC3B (rabbit polyclonal, 1:3,000; Huabio), anti-P62 (rabbit monoclonal, 1:20,000; Huabio), and anti-GAPDH (rabbit monoclonal, 1:5,000; Zenbio). The membranes were subsequently washed with TBST and incubated with horseradish peroxidase (HRP)-conjugated secondary antibodies at room temperature for 1 h. After treatment with ECL luminescent substrate (Proteintech, China), chemiluminescence was detected using a ChemiScope imaging system (Shanghai Junsci Scientific Instrument Co., Ltd., China). The grayscale intensity of the protein bands was quantified using ImageJ software. GAPDH served as the internal reference protein for normalization. Relative expression levels of target proteins were then compared across groups.

2.6.2.5 Quantitative polymerase chain reaction

CCT3, EEF2, EIF3I, and RPS5 mRNA expression levels were primarily quantified. Following the manufacturer’s instructions, total RNA was extracted from the hippocampal tissue homogenates using TRIzol reagent (MagZol Reagent, China) and a chloroform substitute (Buffer BCP, China). Reverse transcription was carried out using Hifair III 1st Strand cDNA Synthesis SuperMix for qPCR (China) on a ProFlex thermal cycler. Quantitative real−time PCR (qPCR) was performed on a SLAN−96P Real−Time PCR System using SYBR Green Master Mix and gene−specific primers (Sangon Biotech, China), with 20 ng of cDNA used per reaction. The thermal cycling conditions were as follows: initial denaturation at 95 °C for 2 min, followed by 40 amplification cycles of 95 °C for 10 s and 60 °C for 30 s. All genes were analyzed in three technical replicates. Relative quantification was performed, and the data were normalized to those of β−actin, which served as the internal reference gene. The sequences of the primers used for qPCR in this study are listed in Table 1.

2.7 Statistical analysis

The mice were randomly assigned to experimental groups, while the outcome assessors remained blinded to the group assignments. Statistical analyses were performed using SPSS v. 29.0 (IBM Corp., Armonk, NY, USA) and GraphPad Prism v. 10.1.2 (GraphPad Software, San Diego, CA, USA). An independent two-sample t test was used to compare the two groups. All the data are presented as the mean ± standard deviation (SD). P < 0.05 was considered to indicate statistical significance.

3 Results

3.1 Identification of DEGs in MDD patients versus healthy controls

The gene expression dataset GSE52790, comprising 10 patients with MDD and 12 healthy controls, was downloaded from the GEO database for comprehensive gene expression analysis. Principal component analysis (PCA) of the dataset revealed that PC1 and PC2 accounted for 22.6% and 12.2% of the variance, respectively, indicating a distinct separation between the two groups and validating the suitability for subsequent differential expression analysis (Figure 1A). On the basis of the criteria of adjusted P value < 0.05 and |Log2FC| > 1, a total of 239 differentially expressed genes (DEGs) were identified in the comparison between the MDD and control groups. After excluding 4 long noncoding RNAs (lncRNAs), we identified 235 DEGs (data summarized in Table 2), with 214 downregulated and 21 upregulated (Figures 1B, C). We retrieved two gene sets comprising 1716 mitophagy-related genes and 3056 oxidative stress-related genes from the GeneCards database. Intersecting these gene sets with the DEGs revealed 44 overlapping DEGs associated with oxidative stress and 57 linked to mitophagy (Figures 1D, E).

Figure 1

3.2 Functional enrichment analysis of DEGs associated with oxidative stress and mitophagy

Intersecting oxidative stress-related genes with differentially expressed genes (DEGs) identified 44 genes (Figure 1D). GO enrichment analysis was conducted to elucidate functional attributes across three domains: biological process (BP), cellular component (CC), and molecular function (MF). The BP terms were predominantly associated with host–pathogen interaction, symbiosis, and microenvironment regulation, suggesting that these overlapping genes may maintain homeostasis of the neural microenvironment by modulating symbiotic interactions. The CC module is predominantly linked to protein secretion pathways, integral outer mitochondrial membrane proteins, and functional core complexes, underscoring their roles in subcellular localization and mitochondrial function. The MF module is significantly enriched for terms related to the assembly and regulation of transcription initiation complexes as well as transcriptional coregulation, indicating that these genes likely function as hubs in the transcriptional regulatory network to orchestrate the fine-tuning of gene expression across multiple signaling pathways. KEGG pathway enrichment analysis revealed significant enrichment of the 44 overlapping genes involved in carbon metabolism, oxalic acid and dicarboxylic acid metabolism, pyruvate metabolism, and the tricarboxylic acid (TCA) cycle (Figures 2A, B). These findings indicate that the identified overlapping genes are involved in the reprogramming of cellular energy metabolism, offering molecular support for the hypothesis linking energy metabolism disorders to depression.

Figure 2

The subsequent intersection of mitophagy-related genes and DEGs revealed 57 common genes (Figure 1E). GO enrichment analysis revealed that the biological process (BP) category was significantly enriched in processes associated with protein synthesis and quality control, as well as genomic stability maintenance, which are potentially related to the cellular stress response and global regulation of gene expression. The cellular component (CC) category was significantly enriched in cytoplasmic translation machinery and nuclear transcription machinery, highlighting the core subcellular structures mediating the flow of genetic information (DNA→RNA→protein). The molecular function (MF) category was enriched for functions involving transcriptional regulation and modulation of translational efficiency, suggesting that these genes play a central role in maintaining gene expression homeostasis. KEGG pathway enrichment analysis revealed that the 57 overlapping genes were significantly enriched in the ribosome, ubiquitin-mediated proteolysis, and central carbon metabolism pathways (Figures 2C, D), suggesting their role in the maintenance of protein homeostasis and energy reprogramming.

3.3 PPI network analysis and hub gene screening

Protein–protein interaction (PPI) networks for DEGs associated with oxidative stress and mitophagy were independently constructed using the STRING 12.0 database, and distinct networks were generated and subsequently visualized using Cytoscape 3.10.3 (Figures 3A, B). The MCC algorithm within the CytoHubba plugin was employed to identify hub genes in each network. The top 10 hub genes among the identified oxidative stress-related DEGs were FBL, TRIM28, HNRNPM, RPS5, EEF2, ILF2, EIF3I, CCT3, CTCF, and DDB1 (Figure 3C), all of which were downregulated. The top 10 hub genes among the mitophagy-related DEGs were EEF2, RPS5, EIF3G, EIF3D, EIF3I, EIF3L, CCT7, CCT3, CCT4, and RPS17 (Figure 3D), all of which were downregulated. We identified four candidate genes (EEF2, CCT3, EIF3I, and RPS5) common to both the mitophagy and oxidative stress gene sets. (Figure 3E).

Figure 3

3.4 Validation of candidate gene expression

Four candidate genes were identified within the GSE52790 dataset through data mining analysis. The differential expression of these genes between the two groups was subsequently examined using the Xiantao platform. The expression levels of EEF2, CCT3, EIF3I, and RPS5 were significantly lower in MDD patients than in controls (Figure 4A). To verify the generalizability of the aforementioned differential expression patterns, we observed that the four candidate genes exhibited a consistent downregulation trend in the validation dataset GSE98793. Following Benjamini–Hochberg (BH) correction for multiple testing (m=4, applied exclusively to the candidate gene set), the adjusted P values were as follows: EEF2 (P.adj = 0.78), EIF3I (P.adj = 0.12), CCT3 (P.adj = 0.027), and RPS5 (P.adj = 0.012). Notably, compared with those in the healthy control group, the expression levels of CCT3 and RPS5 were significantly lower. The candidate genes exhibited consistent directions of expression changes between the training set and the validation set, despite the absence of perfect concordance. Nevertheless, their downregulated expression pattern suggests a stable molecular signature of MDD. (Figure 4B).

Figure 4

3.5 Molecular characterization of candidate genes

On the basis of the GSE52790 dataset, the area under the curve (AUC) values for (Figure 5A), CCT3 (Figure 5B), EIF3I (Figure 5C) and RPS5 (Figure 5D) were 0.967 (95% CI: 0.896–1.000), 0.925 (95% CI: 0.790–1.000), 0.933 (95% CI: 0.836–1.000), and 0.967 (95% CI: 0.905–1.000), respectively. In the validation dataset GSE98793, ROC curve analysis revealed AUC values of 0.513 (95% CI: 0.427–0.598) for EEF2(Figure 5E), 0.615 (95% CI: 0.531–0.698) for CCT3(Figure 5F), 0.577 (95% CI: 0.494–0.660) for EIF3I(Figure 5G), and 0.635 (95% CI: 0.552–0.718) for RPS5 (Figure 5H). The AUC values of the four genes in the validation set did not meet the threshold required for clinically applicable diagnostic performance (AUC ≥ 0.7). Nevertheless, this screening approach retains exploratory value. The stable downregulation of the four genes in the training set is consistent with the energy metabolism hypothesis of major depressive disorder (MDD)—encompassing mitochondrial autophagy, oxidative stress, and protein quality control—suggesting their potential as promising candidate targets for subsequent mechanistic research. The sustained downregulation of CCT3 and RPS5 expression also suggests that these two genes may serve as potential molecular markers for MDD. Further validation is warranted in larger, more homogeneous cohorts using peripheral blood or cerebrospinal fluid samples. The AUC of the combined model on the training set was markedly high, indicating substantial overfitting and significantly impaired generalization performance.

Figure 5

3.6 Results of cell experiments

3.6.1 Corticosterone treatment affected the viability of C8−D1A cells

Following exposure to corticosterone, cell viability decreased in a dose-dependent manner at concentrations ≥100 μmol/L, reaching approximately 70% at a concentration of 400 μmol/L (Figure 6A). Therefore, this concentration and the corresponding treatment duration were selected for subsequent experiments.

Figure 6

3.6.2 Flow cytometric analysis of the corticosterone-induced increase in ROS generation and reduction in mitochondrial membrane potential in C8-D1A cells

Intracellular ROS levels were measured using the nonfluorescent probe DCFH-DA, which is oxidized to fluorescent DCF following oxidation by ROS (Figure 6B). Compared with untreated control cells, Corticosterone-treated C8-D1A cells exhibited significantly higher DCF fluorescence, with a mean fluorescence intensity (MFI) of 98740.83 ± 1367.05, indicating excessive corticosterone-induced ROS production.

Mitochondrial membrane potential (MMP) was measured by flow cytometry using a JC-1 fluorescent probe. JC-1 forms red fluorescent aggregates at high mitochondrial membrane potential, whereas it exists as green fluorescent monomers at lower potentials. The red-to-green fluorescence intensity ratio reflects the mitochondrial membrane potential. Compared with that in the control group, the JC−1 red/green fluorescence ratio was significantly lower in the corticosterone-treated C8-D1A cells. The percentage of depolarized cells in the LR quadrant increased to 25.89% ± 0.44% (Figure 6C), suggesting that corticosterone reduced the mitochondrial membrane potential, which is a hallmark of early mitochondrial damage.

3.7 Animal experimental results

In general, we found that the depression-model mice exhibited gradual and significant anorexia and weight loss, and their body weights were recorded at baseline (15.95 ± 0.70 g), 1 week (16.25 ± 0.56 g), 2 weeks (16.18 ± 0.14 g), and 3 weeks (16.03 ± 0.45 g). The body weight began to decrease at the second week (P < 0.05), and the reduction became more significant at the third week (P < 0.005) (Figure 7A).

Figure 7

All behavioral assessments were conducted in a double-blind manner both before and after the induction of the corticosterone model, with the results presented in Figures 7B–F. After 3 weeks of corticosterone exposure, the treated mice exhibited significant depression-like behaviors. Compared with that of the control group, the sucrose preference rate of the depression group in the SPT decreased significantly to 70 ± 18% (P < 0.0001) (Figure 7B). In the TST, the struggling duration of the depressed group was significantly shorter (39.87 ± 5.64 s, P < 0.005), whereas the immobility time was significantly longer (260.13 ± 5.64 s, P < 0.005) (Figures 7C, D). In the FST, the depression group also exhibited reduced struggling time (62.77 ± 4.19 s, P < 0.005) and increased immobility time (232.50 ± 4.81 s, P < 0.005) (Figures 7E, F). Western blot results revealed that compared with control treatment, corticosterone (Cort) treatment significantly downregulated the expression of the synapse-related proteins BDNF and PSD-95 in murine brain tissue (P < 0.005; Figures 8A–C).

Figure 8

These findings indicate impaired synaptic function. The expression levels of the mitophagy-related proteins PINK1, Parkin, LC3B-II, and p62 were significantly upregulated in the Cort-treated group (P < 0.05 or P < 0.005 versus the control group; Figures 8D-H). The upregulation of PINK1 and Parkin suggests that corticosterone-induced mitochondrial damage activates the PINK1/Parkin-dependent mitophagy initiation pathway, thereby facilitating the elimination of damaged mitochondria via ubiquitin tagging. Increased levels of LC3B-II, a core marker protein of the autophagosome membrane, indicate increased autophagosome formation. Furthermore, the accumulation of the autophagic substrate p62 suggests an impairment in autophagosome–lysosome fusion and subsequent degradation, thereby impairing autophagic flux. The concurrent shifts in these parameters suggest that the model is characterized by dysregulated mitophagy, marked by excessive initiation coupled with impaired clearance. Consequently, damaged mitochondria are not degraded in a timely manner, potentially exacerbating synaptic injury.

Quantitative real−time PCR revealed that corticosterone treatment significantly reduced the mRNA expression of EEF2 (CT: 19.77 ± 0.04; P < 0.05), CCT3 (CT: 23.86 ± 0.06; P < 0.005), and EIF3I (CT: 25.29 ± 0.14; P < 0.05) in mouse brain tissue. In contrast, the RPS5 mRNA level (CT: 29.39 ± 0.22) did not significantly differ from that in the control group (P > 0.05; Figure 9).

Figure 9

4 Discussion

This study investigated the core molecular regulatory network underlying the crosstalk between oxidative stress and mitophagy in major depressive disorder (MDD). By integrating an approach involving transcriptome screening based on the training dataset from the GEO database, validation in an independent verification cohort, phenotypic detection in cell models, and multidimensional validation in animal models, we first identified four key genes, namely, EEF2, CCT3, EIF3I and RPS5, which are significantly correlated with both oxidative stress and mitophagy. Although CCT3 and RPS5 exhibited limited diagnostic performance in the independent clinical validation cohort (AUC ≈ 0.6), which fell short of the threshold required for high-precision diagnostic biomarkers, EEF2 and EIF3I lacked independent clinical diagnostic utility. Nevertheless, using a Cort-induced C8−D1A cell model, we found that the expression levels of these genes correlated negatively with reactive oxygen species (ROS) levels and positively with mitochondrial membrane potential (as assessed by a JC−1 assay), confirming their significant association with elevated oxidative stress and mitochondrial dysfunction. Further validation in animal models revealed that the mRNA expression levels of EEF2, CCT3, and EIF3I were significantly decreased in the Cort-treated group (P<0.05, P<0.005, P<0.05, respectively). Concurrent analysis demonstrated that the expression of the synapse-related proteins PSD-95 and BDNF decreased in the model group, whereas the levels of markers of the mitochondrial mitophagy pathway, including PINK1, Parkin, LC3-II, and the autophagic substrate p62, markedly increased. These findings suggest that abnormal expression of core translational regulatory genes, mediated by alterations in protein synthesis efficiency, may synergistically contribute to the pathology characterized by dysregulated mitochondrial autophagy (excessive initiation and impaired clearance) and reduced synaptic plasticity. On the basis of these results, in this study, a specific molecular regulatory network focusing on translational regulation was developed.

All four core genes identified in this study are involved in protein translation and folding (36–38). These results provide novel molecular evidence supporting the pathological model suggesting a vicious cycle linking energy crisis and proteostasis collapse in MDD. Dysregulated translational control impairs the synthesis of synaptic structural proteins, thereby reducing synaptic plasticity, a key mechanism driving cognitive impairment and emotional dysregulation in MDD (38–40). Specifically, as eukaryotic elongation factor 2, EEF2 downregulation directly inhibits protein translational elongation (41–43). This finding is consistent with the decreased expression of synaptic markers and neurotrophic factors (PSD-95 and BDNF) observed in this study, suggesting that impaired translational elongation may act as a critical mechanistic link between oxidative stress, dysregulated mitophagy, and synaptic damage.

Consistent with the findings of previous studies, our findings indicate that the diagnostic performance of CCT3 and RPS5 was suboptimal in the independent validation cohort (AUC ≈ 0.6). These results align with a well-established principle in biomarker research: the predictive performance of a single gene is often compromised by sample heterogeneity, variability in detection methodologies, demographic variability, and other confounding factors, thereby precluding attainment of the diagnostic threshold for clinical application (44). Although EEF2 and EIF3I did not demonstrate independent diagnostic utility, their significantly altered expression levels in animal models suggest that they may serve as mechanistic therapeutic targets rather than purely diagnostic ones. These findings are consistent with the evolving paradigm in MDD biomarker research, which increasingly prioritizes mechanistic targets over purely diagnostic ones (45, 46).

Oxidative stress is a pathological state caused by an imbalance between intracellular ROS generation and the antioxidant defense system (13, 47). The human brain is particularly vulnerable to oxidative stress owing to its high oxygen consumption, high lipid content, and relative weakness of the antioxidant defense system (48, 49). Molecular evidence has indicated that the expression of the antioxidant enzyme-encoding genes SOD1, SOD2, and GPX1 is significantly downregulated in the brains of patients with MDD (50). Further critical dysregulation is observed in the nuclear factor erythroid 2−related factor 2 (Nrf2) antioxidant pathway. The expression levels of Nrf2 and its downstream antioxidant enzymes (such as heme oxygenase-1 and HO-1) are decreased in the peripheral blood of MDD patients, suggesting that the impaired function of the Nrf2 antioxidant pathway compromises the ability of cells to counteract oxidative stress (16, 51). Nrf2 and mitophagy (PINK1/Parkin axis) pathways exhibit significant reciprocal regulation (52). Decreased Nrf2 activity further exacerbates mitochondrial dysfunction and ROS leakage, thereby establishing a vicious cycle between oxidative stress and mitochondrial damage. Studies have demonstrated that the pathogenesis of major depressive disorder (MDD) involves the interplay among the pathological triangle of neuroimmunity–metabolism–oxidative stress (NIMETOX). This triad establishes a vicious cycle through the “inflammation–oxidative stress loop” and “metabolism–mitochondrial dysfunction axis”, ultimately leading to impaired neural plasticity and impaired emotional regulation (53). Oxidative stress, which serves as a common core mechanism in these conditions, induces reactive oxygen species (ROS) accumulation driven by thyroid hormone deficiency in patients with hypothyroidism and Hashimoto’s disease. Moreover, autoantibodies such as thyroid peroxidase antibodies and anti-α-enolase antibodies in individuals with Hashimoto’s disease can traverse the blood–brain barrier, thereby triggering neuroinflammation and compromising hippocampal synaptic plasticity (54). Patients with coronary heart disease and unstable angina exhibit an integrated phenotypic profile characterized by the metabolism–immunity–oxidation–opioid pathway. Cardiac metabolic abnormalities activate the NLRP3 inflammasome and lead to excessive ROS accumulation, which not only aggravates atherosclerosis but also inhibits hippocampal BDNF expression and subsequently triggers depression (55). Such cross-disease evidence strongly supports the growing consensus that depression is a systemic disease and that oxidative stress serves as a common convergent pathway across multiple diseases. Through bioinformatics analysis, we identified 44 genes associated with oxidative stress in MDD patients who were predominantly enriched in the tricarboxylic acid (TCA) cycle. Kisty reported that elevated ROS levels are not only a consequence of TCA cycle dysfunction but also disrupt cellular energy metabolism by oxidatively modifying cysteine residues of key TCA cycle enzymes and thereby impairing their catalytic activity. It is speculated that DEGs related to oxidative stress may contribute to the pathogenesis and progression of MDD through modulation of the tricarboxylic acid (TCA) cycle pathway (56). We identified 57 DEGs associated with mitophagy, which were significantly enriched in the ribosome pathway, the ubiquitin-mediated proteolysis pathway, and core carbon metabolism modules, which are closely linked to energy reprogramming and translational burden. Enrichment analysis of the mitophagy and oxidative stress pathways revealed that the energy metabolism (carbon metabolism) pathway is involved in both biological processes.

Reported that excessive ROS not only are a consequence of TCA cycle dysfunction but can also disrupt cellular energy metabolism by oxidatively modifying the cysteine residues of key TCA cycle enzymes and impairing their oxidative activity. Further functional and pathway enrichment analyses of these DEGs revealed that oxidative stress-related DEGs are involved in the regulation of carbon metabolism, oxalic acid and dicarboxylic acid metabolism, pyruvate metabolism and the TCA cycle pathway. These findings indicate that oxidative stress-related DEGs may be involved in the occurrence and development of MDD through the regulation of the TCA cycle. Research has indicated that NADPH and α-keto acids detoxify ROS during carbon metabolism (57). Key metabolites involved in carbon metabolism, including acetyl-CoA, pyruvate, and lactate, regulate mitophagy (58). Cellular energy and redox status, as the core output signals of carbon metabolism, serve as the common molecular basis for the regulation of oxidative stress and mitophagy. These findings suggest that carbon metabolism may mediate functional crosstalk between these two pathways and that dysregulation of these pathways may collectively contribute to the pathophysiological alterations associated with MDD. This study confirmed via flow cytometry that the corticosterone (CORT)-induced C8-D1A cell model consistently recapitulates the hallmarks of elevated oxidative stress (increased ROS) and mitochondrial dysfunction (decreased membrane potential) in MDD. This study addresses the lack of phenotypic validation in existing MDD cell models and provides a reliable experimental system for subsequent mechanistic research on MDD at the cellular level.

The core goal of this study is to reveal the correlations between translation regulation-related genes and the interactive network of oxidative stress and mitophagy in MDD, providing a novel molecular perspective for understanding the vicious cycle linking energy crisis and proteostasis collapse in MDD. Downregulated gene expression leads to translational disorders. On the one hand, it leads to reduced synthesis of synaptic proteins, directly compromising synaptic plasticity; on the other hand, it exacerbates mitochondrial dysfunction and ROS accumulation by disrupting the synthesis of mitophagy-related proteins, thus perpetuating a vicious pathological cycle.

Although this study did not identify high-precision diagnostic biomarkers, the screened genes can serve as core components for subsequent combined multigene diagnostic models. The construction of a multiomics diagnostic model through the integration of epigenetic biomarkers and metabolomic data holds promise for improving the accuracy and specificity of MDD diagnosis. Furthermore, as key molecules in the translational regulatory pathway, these genes represent promising targets for the development of novel antidepressants that modulate the translational process. Notably, activating EEF2 kinase or modulating the EIF3 complex can alleviate disruptions in proteostasis and impairments in synaptic plasticity in MDD.

The scope of this study is restricted to correlation analysis. The regulatory roles of the aforementioned genes in oxidative stress, mitophagy, and synaptic function have not been validated via intervention experiments, such as gene knockdown or overexpression; consequently, their causal relationships remain unestablished. Future studies should employ gain- and loss-of-function experiments in cell models to further investigate these functional roles. With the integration of cutting-edge research directions, such as the PINK1/Parkin-dependent mitophagy pathway and the Drp1-mediated endoplasmic reticulum–mitochondria contact (MERCs) regulatory axis, future work must further elucidate the specific regulatory mechanisms of these genes in the pathogenesis of MDD. The sample sizes of the training and validation sets in this study are limited, and MDD patients with different subtypes and disease courses were not represented, which constrains the generalizability of the research findings. Further validation of the expression profiles and diagnostic efficacy of these genes is warranted in larger, independent, multicenter cohorts. Furthermore, a multiomics diagnostic model could be developed by integrating epigenetic biomarkers and metabolomic data to improve the diagnostic accuracy and specificity of MDD. The animal experiments in this study failed to elucidate the direct interaction between the above genes and both the endoplasmic reticulum–mitochondria contact (MERCs) regulatory axis and the PINK1/Parkin pathway. Furthermore, the full regulatory cascade of “molecular–phenotype–behavior” remains incompletely understood. Future research can further investigate the effects of traditional Chinese medicine compounds and small-molecule compounds on alleviating the pathological phenotypes of MDD through the regulation of these genes, thereby promoting the clinical translation of basic research findings.

References

• 1

  MarxWPenninxBWJHSolmiMFurukawaTAFirthJCarvalhoAFet al. Major depressive disorder. Nat Rev Dis Primers. (2023) 9:44. doi:  10.1038/s41572-023-00454-1

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 2

  OtteCGoldSMPenninxBWParianteCMEtkinAFavaMet al. Major depressive disorder. Nat Rev Dis Primers. (2016) 2:16065. doi:  10.1038/nrdp.2016.65

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 3

  GBD 2021 Diseases and Injuries Collaborators. Global incidence, prevalence, years lived with disability (ylds), disability-adjusted life-years (dalys), and healthy life expectancy (hale) for 371 diseases and injuries in 204 countries and territories and 811 subnational locations 1990-2021: a systematic analysis for the global burden of disease study 2021. Lancet. (2024) 403:2133–61. doi:  10.1016/S0140-6736(24)00757-8

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 4

  HarrisonJEWeberSJakobRChuteCG. Icd-11: an international classification of diseases for the twenty-first century. BMC Med Inform Decis Mak. (2021) 21:206. doi:  10.1186/s12911-021-01534-6

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 5

  SantomauroDFVosTWhitefordHAChisholmDSaxenaSFerrariAJ. Service coverage for major depressive disorder: estimated rates of minimally adequate treatment for 204 countries and territories in 2021. Lancet Psychiatry. (2024) 11:1012–21. doi:  10.1016/S2215-0366(24)00317-1

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 6

  American Psychiatric Association. Diagnostic and Statistical Manual of Mental Disorders, Fifth Edition, Text Revision(DSM-5-TR). Arlington: American Psychiatric Association Publishing (2022).

  • Google Scholar
  • View reference in article

• 7

  CuiLLiSWangSWuXLiuYYuWet al. Major depressive disorder: hypothesis, mechanism, prevention and treatment. Signal Transduct Target Ther. (2024) 9:30. doi:  10.1038/s41392-024-01738-y

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 8

  JuruenaMF. Early-life stress and hpa axis trigger recurrent adulthood depression. Epilepsy Behav. (2014) 38:148–59. doi:  10.1016/j.yebeh.2013.10.020

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 9

  GeorgeMYMageedSSAMansourDEFawziSF. The cortisol axis and psychiatric disorders: an updated review. Pharmacol Rep. (2025) 77:1573–99. doi:  10.1007/s43440-025-00782-x

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 10

  DingRGaoLWangXYangJHanXFeiSet al. High-fat diet and chronic restraint stress exacerbate anxiety-depressive behaviors via astrocytic a1 phenotype transformation. Sci Rep. (2025) 15:15031. doi:  10.1038/s41598-025-99355-4

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 11

  Miguel-HidalgoJJ. Astroglia in the vulnerability to and maintenance of stress-mediated neuropathology and depression. Front Cell Neurosci. (2022) 16. doi:  10.3389/fncel.2022.869779

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 12

  KangJSBaekJHParkHRehmanNUChungHJLeeDKet al. Disrupted astrocyte-neuron glutamine-glutamate cycling in the medial prefrontal cortex contributes to depression-like behaviors. Int J Biol Sci. (2026) 22:553–65. doi:  10.7150/ijbs.123740

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 13

  CarpitaBBettiLPalegoLBartolommeiNChicoLPasqualiLet al. Plasma redox and inflammatory patterns during major depressive episodes: a cross-sectional investigation in elderly patients with mood disorders. CNS Spectr. (2021) 26:416–26. doi:  10.1017/S1092852920001443

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 14

  Jimenez-FernandezSGurpeguiMDiaz-AtienzaFPerez-CostillasLGerstenbergMCorrellCU. Oxidative stress and antioxidant parameters in patients with major depressive disorder compared to healthy controls before and after antidepressant treatment: results from a meta-analysis. J Clin Psychiatry. (2015) 76:1658–67. doi:  10.4088/JCP.14r09179

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 15

  KatrencikovaBVavakovaMPaduchovaZNagyovaZGaraiovaIMuchovaJet al. Oxidative stress markers and antioxidant enzymes in children and adolescents with depressive disorder and impact of omega-3 fatty acids in randomized clinical trial. Antioxidants (Basel). (2021) 10:1256. doi:  10.3390/antiox10081256

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 16

  BaiYLiuWZhangFZhengYGuoQHuHet al. Dysregulation of peripheral oxidative stress and the nrf2 antioxidant system in major depressive disorder. J Affect Disord. (2025) 382:336–45. doi:  10.1016/j.jad.2025.04.066

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 17

  Jimenez-FernandezSGurpeguiMGarrote-RojasDGutierrez-RojasLCarreteroMDCorrellCU. Oxidative stress parameters and antioxidants in adults with unipolar or bipolar depression versus healthy controls: systematic review and meta-analysis. J Affect Disord. (2022) 314:211–21. doi:  10.1016/j.jad.2022.07.015

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 18

  AndersenJVSchousboeA. Milestone review: metabolic dynamics of glutamate and gaba mediated neurotransmission — the essential roles of astrocytes. J Neurochem. (2023) 166:109–37. doi:  10.1111/jnc.15811

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 19

  McKennaMC. Glutamate pays its own way in astrocytes. Front Endocrinol (Lausanne). (2013) 4:191. doi:  10.3389/fendo.2013.00191

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 20

  GhoshDKumarA. Harnessing mitophagy for therapeutic advances in aging and chronic neurodegenerative diseases. Neuroglia. (2024) 5:391–409. doi:  10.3390/neuroglia5040026

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 21

  ShannonPMarkielAOzierOBaligaNSWangJTRamageDet al. Cytoscape: a software environment for integrated models of biomolecular interaction networks. Genome Res. (2003) 13:2498–504. doi:  10.1101/gr.1239303

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 22

  ChinCChenSWuHHoCKoMLinC. Cytohubba: identifying hub objects and subnetworks from complex interactome. BMC Syst Biol. (2014) 8 Suppl 4:S11. doi:  10.1186/1752-0509-8-S4-S11

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 23

  BarrettTWilhiteSELedouxPEvangelistaCKimIFTomashevskyMet al. Ncbi geo: archive for functional genomics datasets--update. Nucleic Acids Res. (2013) 41:D991–5. doi:  10.1093/nar/gks1193

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 24

  YatesADAllenJAmodeRMAzovAGBarbaMBecerraAet al. Ensembl genomes 2022: an expanding genome resource for nonvertebrates. Nucleic Acids Res. (2022) 50:D996–D1003. doi:  10.1093/nar/gkab1007

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 25

  StelzerGRosenNPlaschkesIZimmermanSTwikMFishilevichSet al. The GeneCards Suite: from gene data mining to disease genome sequence analyses. Curr Protoc Bioinformatics. (2016) 54:e51. doi: 10.1002/cpbi.5

  • CrossRef
  • Google Scholar
  • View reference in article

• 26

  SzklarczykDNastouKKoutrouliMKirschRMehryaryFHachilifRet al. String: functional protein association networks[db/ol. (2022). The STRING Database in 2023: Protein–protein Association Networks and Functional Enrichment Analyses for Any Sequenced Genome of Interest. NUCLEIC ACIDS RES. 51:D638–D646. doi: 10.1093/nar/gkac1000

  • CrossRef
  • Google Scholar
  • View reference in article

• 27

  MatthewsBW. Comparison of the predicted and observed secondary structure of t4 phage lysozyme. Biochim Biophys Acta. (1975) 405:442–51. doi:  10.1016/0005-2795(75)90109-9

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 28

  BarrosLFSchirmeierSWeberB. The astrocyte: metabolic hub of the brain. Cold Spring Harb Perspect Biol. (2024) 16:a041355. doi:  10.1101/cshperspect.a041355

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 29

  DringenRArendC. Glutathione metabolism of the brain-the role of astrocytes. J Neurochem. (2025) 169:e70073. doi:  10.1111/jnc.70073

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 30

  Yun-FangJWiningerKHoAMPeytonLBakerMChoiD. Astrocytic glutamate transporter 1 (glt1) deficiency reduces anxiety- and depression-like behaviors in mice. Front Behav Neurosci. (2020) 14. doi:  10.3389/fnbeh.2020.00057

  • Pubmed Abstract
  • CrossRef
  • Google Scholar
  • View reference in article

• 31

  MurphyMP. How mitochondria produce reactive oxygen species. Biochem J. (2009) 417:1–13. doi:  10.1042/bj20081386

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 32

  BurkeHMDavisMCOtteCMohrDC. Depression and cortisol responses to psychological stress: a meta-analysis. Psychoneuroendocrinology. (2005) 30:846–56. doi:  10.1016/j.psyneuen.2005.02.010

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 33

  ZajkowskaZGullettNWalshAZoncaVPedersenGASouzaLet al. Cortisol and development of depression in adolescence and young adulthood - a systematic review and meta-analysis. Psychoneuroendocrinology. (2022) 136:105625. doi:  10.1016/j.psyneuen.2021.105625

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 34

  DuanYJiaQLuoJWangYLiQLvSet al. High-precision neural information detection of multiple brain regions in mice under different concentrations of isoflurane anesthesia based on microelectrode arrays. Microsyst Nanoeng. (2025) 11:119. doi:  10.1038/s41378-025-00944-0

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 35

  LiuMYYinCYZhuLJZhuXHXuCLuoCXet al. Sucrose preference test for measurement of stress-induced anhedonia in mice. Nat Protoc. (2018) 13:1686–98. doi:  10.1038/s41596-018-0011-z

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 36

  AutryAEAdachiMNosyrevaENaESLosMFChengPet al. Nmda receptor blockade at rest triggers rapid behavioral antidepressant responses. Nature. (2011) 475:91–5. doi:  10.1038/nature10130

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 37

  IgnacioZMReusGZArentCOAbelairaHMPitcherMRQuevedoJ. New perspectives on the involvement of mtor in depression as well as in the action of antidepressant drugs. Br J Clin Pharmacol. (2016) 82:1280–90. doi:  10.1111/bcp.12845

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 38

  ZhangXEladawiMARyanWGFanXPrevoznikSDevaleTet al. Ribosomal dysregulation: a conserved pathophysiological mechanism in human depression and mouse chronic stress. Pnas Nexus. (2023) 2:pgad299. doi:  10.1093/pnasnexus/pgad299

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 39

  GaoMWuYYangXWangJHuXZhouJ. Advances of proteostasis network and its stability maintenance mechanism. Sheng Wu Gong Cheng Xue Bao = Chin J Biotechnol. (2024) 40:434–45. doi:  10.13345/j.cjb.230537

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 40

  LongoFManciniMIbraheemPLAryalSMesiniCPatelJCet al. Cell-type-specific disruption of perk-eif2α signaling in dopaminergic neurons alters motor and cognitive function. Mol Psychiatry. (2021) 26:6427–50. doi:  10.1038/s41380-021-01099-w

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 41

  ChengJWuCLiJYangQZhaoMZhangX. Capturing eukaryotic ribosome dynamics in situ at high resolution. Nat Struct Mol Biol. (2025) 32:698–708. doi:  10.1038/s41594-024-01454-9

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 42

  JanaSBrahmaSAroraSWladykaCLHoangPBlinkaSet al. Transcriptional-translational conflict is a barrier to cellular transformation and cancer progression. Cancer Cell. (2023) 41:853–70. doi:  10.1016/j.ccell.2023.03.021

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 43

  MarinPNastiukKLDanielNGiraultJACzernikAJGlowinskiJet al. Glutamate-dependent phosphorylation of elongation factor-2 and inhibition of protein synthesis in neurons. J Neurosci. (1997) 17:3445–54. doi:  10.1523/jneurosci.17-10-03445.1997

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 44

  LiuMWangDSunYXuDCaoYQiXet al. Cd63 as a critical biomarker associated with immune infiltration in depression through bioinformatics analysis and experimental validation. J Affect Disord. (2025) 389:119724. doi:  10.1016/j.jad.2025.119724

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 45

  CaiXSunTFengMChenGZhouJZhuangHet al. Taurocholic acid is associated with disturbed functional connectivity in the hippocampus of patients with depression. Advanced Sci (Weinheim Baden-Wurttemberg Germany). (2025) 13:e8693. doi:  10.1002/advs.202508693

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 46

  RoweJNogovitsynNMazurkaRSquiresSDHasselSPoppenkJet al. Latent profile analysis of childhood maltreatment and neural markers in depression. JAMA Netw Open. (2025) 8:e2525147. doi:  10.1001/jamanetworkopen.2025.25147

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 47

  NoormohammadiMEtesamFAminiAKhosravianDPMohammadzadehMShidfarF. Impact of nano-selenium supplementation add-on sertraline on depressive symptoms and oxidative stress in patients with major depressive disorder: a triple-blind randomized controlled trial. Biometals. (2025). doi: 10.1007/s10534-025-00750-1

  • CrossRef
  • Google Scholar

• 48

  BhattSNagappaANPatilCR. Role of oxidative stress in depression. Drug Discov Today. (2020) 25:1270–1276. doi: 10.1016/j.drudis.2020.05.001

  • CrossRef
  • Google Scholar

• 49

  RossettiACPaladiniMSRivaMAMolteniR. Oxidation-reduction mechanisms in psychiatric disorders: a novel target for pharmacological intervention. Pharmacol Ther. (2020) 210:107520. doi: 10.1016/j.pharmthera.2020.107520

  • CrossRef
  • Google Scholar

• 50

  ChandleyMJSzebeniASzebeniKWang-HeatonHGarstJStockmeierCAet al. Markers of elevated oxidative stress in oligodendrocytes captured from the brainstem and occipital cortex in major depressive disorder and suicide. Prog Neuropsychopharmacol Biol Psychiatry. (2022) 117:110559. doi: 10.1016/j.pnpbp.2022.110559

  • CrossRef
  • Google Scholar

• 51

  ZuoCCaoHSongYGuZHuangYYangYet al. Nrf2: an all-rounder in depression. Redox Biol. (2022) 58:102522. doi:  10.1016/j.redox.2022.102522

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 52

  GumeniSPapanagnouEManolaMSTrougakosIP. Nrf2 activation induces mitophagy and reverses parkin/pink1 knock downmediated neuronal and muscle degeneration phenotypes. Cell Death Dis. (2021) 12:671. doi:  10.1038/s41419-021-03952-w

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 53

  MaesMAlmullaAFYouZZhangY. Neuroimmune, metabolic and oxidative stress pathways in major depressive disorder. Nat Rev Neurol. (2025) 21:473–89. doi:  10.1038/s41582-025-01116-4

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 54

  Al-BaldawiSQAl-HakeimHKKhémiriIMaesM. Neuro-immune, metabolic, and oxidative pathways in depression due to hypothyroidism and hashimotos thyroiditis. J Affect Disord. (2026) 402:121224. doi:  10.1016/j.jad.2026.121224

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 55

  MousaRFSmesamHNQazmoozHAAl-HakeimHKMaesM. A pathway phenotype linking metabolic, immune, oxidative, and opioid pathways with comorbid depression, atherosclerosis, and unstable angina. CNS Spectr. (2021) 27:676–90. doi:  10.1017/s1092852921000432

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 56

  KistyEASaartECWeerapanaE. Identifying redox-sensitive cysteine residues in mitochondria. Antioxidants (Basel). (2023) 12:992. doi:  10.3390/antiox12050992

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 57

  ShimizuKMatsuokaY. Redox rebalance against genetic perturbations and modulation of central carbon metabolism by the oxidative stress regulation. Biotechnol Adv. (2019) 37:107441. doi:  10.1016/j.bioteChadv.2019.107441

  • Pubmed Abstract
  • CrossRef
  • Google Scholar

• 58

  ZhangTLiuQGaoWSehgalSAWuH. The multifaceted regulation of mitophagy by endogenous metabolites. Autophagy. (2022) 18:1216–39. doi:  10.1080/15548627.2021.1975914

  • Pubmed Abstract
  • CrossRef
  • Google Scholar