Introduction

Adolescence, a critical period of rapid social, emotional, and cognitive changes, increases vulnerability to psychiatric disorders, with major depressive disorder (MDD) being a significant global mental health issue affecting adolescent health and well-being1,2. The prevalence of MDD, initially low in children, sharply increases during adolescence and young adulthood, potentially due to social stressors, hormonal changes, and brain development1,3. Adolescent depression, marked by persistent sadness, hopelessness, and loss of interest, impairs academic performance, relationships, increases suicide risk, and predicts higher risk of depression in adulthood, with most individuals experiencing their first depressive episode during youth4,5. So early identification and intervention are crucial, especially in adolescence.

Depression is an etiologically complex, multifactorial disorder influenced by genetic and environmental factors. While large-scale genome-wide studies have revealed modest genetic heritability, the role of gut microbiota is emerging6,7. The gut-brain axis, involving immune, vagal, endocrine, and humoral pathways, affect emotional regulation and depression susceptibility8,9. Gut dysbiosis is proposed as a potential etiological factor in depression, with altered microbiota observed in both children and adults10,11,12,13,14,15,16,17,18. Radjabzadeh et al. identified thirteen microbial taxa, including Eggerthella, Coprococcus, and Ruminococcaceae, linked to depressive symptoms in two large adult cohorts10. In pediatric depression, Zhou et al. found a distinct metagenomic profile with fewer short-chain fatty acid-producing bacteria, and an overgrowth of bacteria like Escherichia-Shigella and Flavonifractor compared to healthy controls19. Our previous studies found higher Streptococcus and lower Faecalibacterium levels in depressed children13, and increased Enterobacteriaceae and Alistipes, alongside reduced Faecalibacterium in depressed adults12. Animal studies indicate that transferring gut microbiota from depressed patients to germ-free rats induces depressive-like behaviors and physiological changes, suggesting a role for gut microbiota in depression pathology20. Additionally, microbial metabolites, including short-chain fatty acids (SCFAs) and tryptophan derivatives, affect neuroinflammation and neurotransmitter signaling, which are crucial in depression pathophysiology21,22. Zhao et al. emphasizes the synergistic role of gut microbiota in regulating intestinal neurotransmitter production, with homovanillic acid administration notably improving depressive symptoms23. Moreover, chronic inflammation is linked to depression, with gut microbiota influencing immune regulation; gut dysbiosis may worsen inflammation and potentially aggravate or trigger depressive symptoms24. This intricate interplay sheds light on new pathways for comprehending the pathogenesis of depression.

Currently, most studies on the gut microbiota and depression have focused on children and adults, with relatively limited research on adolescents. Given the rapid brain development and pivotal emotional growth during adolescence, it is imperative to investigate how the gut microbiota influences depression in this age group. This research may reveal unique biological mechanisms and therapeutic targets specific to adolescents, improving our understanding of mental health during this critical developmental stage. In our current study, we enrolled 99 adolescents newly diagnosed with MDD and 106 age- and gender-matched healthy controls from Quzhou, China. We compared the fecal microbiota profiles using high-throughput sequencing targeting the hypervariable V3-V4 regions of the 16 S rRNA gene and investigated the circulating levels of 27 cytokines, growth factors, and chemokines using bead-based multiplex immunoassays. Additionally, we explored correlations between key functional bacteria related to adolescent MDD and inflammatory cytokines. The findings will reveal new insights into adolescent depression’s etiology and offer non-invasive diagnostics and microbiota-based treatments.

Materials and methods

Study participants

We recruited 99 newly diagnosed adolescent MDD patients (aged 13 to 18) and 106 age- and gender-matched healthy controls (HCs) from Quzhou Third People’s Hospital, China, between February and December 2022. The study was approved by the Ethics Committee of Quzhou Third People’s Hospital (reference no. SY-2022-27). Before enrollment, written informed consent was obtained from all participants or their parents. All procedures involving human participants were conducted in accordance with the relevant guidelines and regulations, including the Declaration of Helsinki. MDD diagnosis was based on criteria from the Hamilton Depression Scale (HAM-D), Diagnostic and Statistical Manual of Mental Disorders Fifth Edition (DSM-V), and Chinese Classification of Mental Disorder Third Edition (CCMD-3), assessed by two experienced psychiatrists. All adolescent MDD had an HAMDS score ⩾ 20. Detailed demographic data were collected using a series of questionnaires. The exclusion criteria for the study included ages < 10 or > 19 years, body mass index (BMI) exceeding 28 kg/m², active respiratory or intestinal infections, autism spectrum disorder, anorexia nervosa, bipolar disorder, attention-deficit/hyperactivity disorder, mania, recent administration of antibiotics, prebiotics, probiotics, or synbiotics within the past month, as well as use of antidepressants, mood stabilizers, or other psychiatric medications within the preceding month, and other diseases such as inflammatory bowel disease, irritable bowel syndrome or other autoimmune diseases.

Sample collection

Sample collection and processing followed standardized protocols. All sample collections were performed at the Department of Psychiatry, Quzhou Third People’s Hospital. Fresh fecal samples (approximately 2 g) were collected in sterile plastic cups and promptly stored at -80 °C within 15 min after collection for further microbiome analysis. Any samples that were left at room temperature for longer than 15 min were discarded. For cytokine analysis, whole blood samples were collected from participants’ fasting blood in the early morning. After centrifugation (1000×g, 10 min), the serum was divided into three equal 200 µl aliquots and immediately stored at − 80 °C for subsequent analysis.

Amplicon library construction and sequencing

Bacterial genomic DNA extraction, amplicon library construction and sequencing have been previously described in details8,25,26,27. Briefly, bacterial genomic DNA was extracted from 300 mg of homogenized feces using a QIAamp DNA Stool Mini Kit (Qiagen, Hilden, Germany) according to the manufacturer’s instructions. Amplicon library was constructed targeting the hypervariable V3-V4 regions of the 16 S rRNA gene. The sequencing library was prepared at Hangzhou KaiTai Bio-lab, and sequencing was performed using the NovaSeq™ 6000 system (Illumina).

Bioinformatic analysis

After sequencing, the raw data (> 100 bp) with error rate < 1% were processed and quality-controlled using QIIME2 (Quantitative Insights Into Microbial Ecology 2, v2020.11, http://qiime.org/) with default parameters8,25,26,27,28,29,30. The quality control involved trimming adapter sequences with Cutadapt v2.4, removing low-quality reads, sequences shorter than 100 bp, and chimeric sequences using the fastq_filter module of VSEARCH v2.13.6, which retains unique de novo sequence variants. Before proceeding with data analysis, reads from each sample underwent normalization to ensure even sampling depths. Operational taxonomic units (OTUs) were clustered at 97% similarity using VSEARCH 2.8.1, and taxonomic annotation utilized the SILVA database via RDP Classifier and UCLUST v1.2.22 within QIIME2. Bacterial α-diversity metrics—Shannon, Simpson, observed species, ACE, and Chao1 estimator—were computed at the 97% similarity level. β-diversity was evaluated using unweighted UniFrac, weighted UniFrac, Jaccard, and Bray-Curtis distances in QIIME2, with visualization through principal coordinate analysis (PCoA)31. Low-abundance taxa exhibiting an appearance frequency of less than 50% across samples were excluded for subsequent analysis. Microbiota composition variations were analyzed using Statistical Analysis of Metagenomic Profiles (STAMP) software package v2.1.3 32. Correlation analysis employed the sparse compositional correlation (SparCC) algorithm on the complete OTU table at the genus level, and networks were visualized using Cytoscape v3.6.1. SparCC was chosen for its ability to reveal potential ecological interactions within microbial communities, such as mutualism or competition. Functional predictions from the closed OTU-table generated by QIIME were compared with the Kyoto Encyclopedia of Genes and Genomes (KEGG) database using PiCRUSt v1.0.0. This approach utilized ancestral-state reconstruction to infer microbial community functional potentials based on phylogenetic composition, categorizing them into KEGG pathways at levels 1–3 33,34.

Multiplex cytokine analysis

The 27-plex Human Group I Cytokine Assay Kit (catalog #M50-0KCAF0Y, Bio-Rad, CA, USA) was used to evaluate participants’ systemic immune function in a single batch, following the methodology previously described8,25,26,27,35. These assays, utilizing Luminex® xMAP® technology, allow for the simultaneous detection of 27 targets, including 16 cytokines, 6 chemokines, and 5 growth factors. Specifically, the cytokines assessed comprised interleukin-1β (IL-1β), IL-1 receptor antagonist (IL-1ra), IL-2, IL-4, IL-5, IL-6, IL-7, IL-8, IL-9, IL-10, IL-12 (p70), IL-13, IL-15, IL-17, interferon gamma (IFN-γ), tumor necrosis factor-alpha (TNF-α). The chemokines quantified were Eotaxin, interferon gamma-inducible protein 10 (IP-10), monocyte chemotactic protein-1 (MCP-1), macrophages inflammatory protein-1α (MIP-1α), MIP-1β, regulated upon activation normal T-cell expressed and secreted (RANTES). The growth factors analyzed included Fibroblast growth factor-basic (FGF-basic), granulocyte colony-stimulating factor (G-CSF), granulocyte-macrophages colony-stimulating factor (GM-CSF), platelet-derived growth factor (PDGF-bb), and vascular endothelial growth factor (VEGF). The assays were conducted using the Luminex® 200™ system (Bio-Rad), where fluorescence values were recorded. The collected serum samples were diluted 4-fold with the sample diluent buffer and standard curves were generated using various concentrations of assay standards. and data acquisition was done using the Bio-Plex Array Reader system 2200. Results were expressed as picogram per milliliter (pg/mL) using standard curves integrated into the assay and Bio-Plex Manager v5.0 software with reproducible intra- and inter-assay CV values of 5–8%. The detection limits of the assay for each cytokine are provided in the manufacturer’s protocol, typically within the pg/mL range. To ensure data integrity, several quality control (QC) measures were implemented, including validation of standard curves, dynamic range checks, and the use of both positive and negative controls to confirm assay performance and specificity. Samples with outlier values were flagged for review, and cytokine concentrations below the limit of detection (LOD) were assigned half the LOD value to minimize bias. These QC steps helped maintain the reliability and accuracy of the results throughout the analysis.

Statistical analysis

Statistical analysis utilized various tests depending on the data type: White’s nonparametric t-test, independent t-test, or Mann-Whitney U-test for continuous variables such as α-diversity indices, taxonomic abundance, and cytokines; Pearson’s chi-square or Fisher’s exact test for categorical variables; Spearman’s rank correlation for analyses correlating microbial abundances with cytokines and inferred functions. Statistical analyses were performed with SPSS v24.0 (SPSS Inc., Chicago, IL) and STAMP v2.1.3, while graphical representations were created using R packages and GraphPad Prism v6.0. Random Forest was used for microbial classification, with Mean Decrease Gini to assess variable importance. The predictive power was evaluated using receiver operating characteristics (ROC) and area under the curve (AUC) analysis to assess the ability of differential bacteria to predict adolescent depression. ROC curve analysis is a reliable method for performance characterization, with AUC values commonly used to evaluate discriminatory power. Standard deviation was calculated to assess the variability in classifier performance. ROC analysis was performed using the OECloud tools at https://cloud.oebiotech.com. All tests were two-sided, and p-values were adjusted for multiple comparisons using the Benjamini-Hochberg method to control the False Discovery Rate (FDR). A threshold of FDR < 0.05 indicated statistical significance.

Accession number

The sequence data from this study are deposited in the GenBank Sequence Read Archive with the accession number PRJNA1137557.

Ethics approval statement

All our human experiments were conducted according to the ethical policies and procedures approved by the Ethics Committee of Quzhou Third People’s Hospital (reference no. SY-2022-27). Before enrollment, written informed consent was obtained from all participants or their parents. All procedures involving human participants were conducted in accordance with the relevant guidelines and regulations, including the Declaration of Helsinki.

Results

Changed overall structure of the fecal microbiota in adolescent depression

Based on our questionnaire data, there were no significant differences between depressed adolescents and controls in terms of age, gender, BMI, birth mode, past medical history, or medication history (p > 0.05). However, there was a significant difference in the family history of depression between the two groups (p < 0.05). All adolescent MDD patients met criteria for MDD according to the HAM-D, DSM-V, and CCMD-3 criteria, with HAM-D scores exceeding 20 (Table S1).

After sequencing, we obtained 15,750,206 high-quality sequence reads (8,725,847 for healthy controls and 7,024,359 for adolescents with MDD) from 18,204,232 raw sequence reads, with an average of 76,830 reads per sample. To ensure consistent sequencing depth across all samples, we normalized each sample to 38,100 reads for the subsequent analysis. In total, we identified 10,841 OTUs (unique bacterial phylotypes) among the fecal microbiota, attaining a Good’s coverage of 98.30%, indicating that most of the fecal bacteria had been detected. In comparing α-diversity indices between adolescents with MDD and controls, we noted lower bacterial diversity (Shannon and Simpson indices; Fig. 1A, B) and reduced richness (ACE, Chao1, and observed OTUs; Fig. 1C–E) among those with MDD (p < 0.05). A Venn diagram analysis, illustrating shared OTUs between adolescents with MDD and healthy controls, indicated 7,403 OTUs common to both groups. Additionally, there were 763 OTUs unique to MDD adolescents and 2,675 OTUs unique to controls (Fig. 1F). Despite notable individual variations, PCoA utilizing several algorithms (Bray–Curtis, Jaccard, unweighted UniFrac, and weighted UniFrac) consistently clustered adolescents with MDD and healthy controls into distinct groups (ADONIS test: p < 0.01; Fig. 1G-J). Based on the OTUs analysis, the rank-abundance curves depicting bacterial communities in both the healthy control and MDD groups revealed comparable abundance levels. However, the MDD group exhibited slightly lower evenness compared to the healthy group (Fig. 1K). The findings collectively show significant changes in both α- and β-diversity, indicating a clear structural alteration in fecal microbiota between adolescents with MDD and healthy controls.

Fig. 1
figure 1

Comparison of the overall structure of the fecal microbiota between adolescents with major depressive disorder (MDD) and healthy controls. (A–E) α-diversity indices (Shannon and Simpson) and richness indices (ACE, Chao1, and observed species) were utilized to assess the overall structure of fecal microbiota. Data are presented as mean ± standard deviation. Unpaired two-tailed t-tests were employed for inter-group comparisons. (F) A Venn diagram illustrates the overlap of OTUs between microbiota associated with adolescent MDD and healthy controls. (G–J) Principal coordinate analysis (PCoA) plots illustrate β-diversity of individual fecal microbiota based on Bray–Curtis, Jaccard, unweighted UniFrac, and weighted UniFrac distances. Each symbol represents a sample. (K) Rank abundance curves depict the distribution of bacterial OTUs in the two groups.

Full size image

Altered composition of fecal microbiota in adolescents with MDD

We conducted a comparative analysis of the fecal microbiota between adolescents with MDD and controls, categorizing sequencing data into 15 phyla, 97 families, and 323 genera using the RDP classifier. Figure 2A-C illustrates the overall microbiota composition across various taxonomic levels, including phylum, family and genus. Specifically, Firmicutes, Bacteroidetes, Actinobacteria, and Proteobacteria emerged as the predominant phyla, collectively representing more than 99.4% of the total sequences analyzed. Notably, the fecal microbiota can be classified into two enterotypes, E1 and E2 (Fig. 2D). The E1 enterotype was dominated by Faecalibacterium and Bacteroides, while the E2 enterotype was dominated by Blautia and Bifidobacterium (Fig. 2E). Approximately 69.8% of the samples (74/106) in the control group belonged to the E1 enterotype, whereas only 16.2% of the samples (16/99) in the MDD group were classified as E1 enterotype, suggesting a significant difference in microbiota composition between the two groups. Figure 3 further highlights these differences, presenting significant variations in microbiota composition between adolescents with MDD and healthy controls, as analyzed across various taxonomic levels using MetaStats 2.0. At the phylum level, Actinobacteria were more abundant in adolescents with MDD, while Bacteroidetes were less prevalent compared to controls (p < 0.05; Fig. 3A). At the family level, adolescents with MDD showed higher levels of Lachnospiraceae, Bifidobacteriaceae, and Streptococcaceae, and lower levels of Ruminococcaceae and Veillonellaceae compared to controls (p < 0.05; Fig. 3B). At the genus level, adolescents with MDD exhibited higher levels of Blautia, Bifidobacterium, and Lachnospiraceae_ge, alongside lower levels of Faecalibacterium and Bacteroides compared to controls (p < 0.05; Fig. 3C). Furthermore, we employed the SparCC algorithm to construct correlation-based networks that elucidate microbial interactions based on OTU relative abundance across different groups (Fig. 4). SparCC is particularly useful for identifying keystone species, which play a role in maintaining overall gut eubiosis. The analysis revealed a network of 36 nodes at the genus level, representing the major bacterial genera, with 23 genera enriched in controls and 13 enriched in the MDD group. Notably, the adolescent MDD group exhibited significantly weaker positive and negative correlations among these bacterial taxa compared to controls, suggesting a disruption in microbial interactions within the gut ecosystem of those with MDD. Together, these findings highlight significant dysbiosis in the fecal microbiota associated with adolescent MDD.

Fig. 2
figure 2

Taxonomic compositions of fecal microbiota in depressed adolescents and healthy controls. (A) Phylum; (B) Family; (C) Genus; (D) PCoA plot showing two enterotypes; (E) Relative abundances of representative genera of enterotypes.

Full size image
Fig. 3
figure 3

Comparisons of the relative abundance of abundant bacterial taxa in the fecal microbiota between depressed adolescents and healthy controls. (A) Key differential functional phyla. (B) Key differential functional families. (C) Key differential functional genera. The data are presented as the mean ± standard deviation. Data were analyzed using Mann–Whitney U-tests to assess differences between depressed adolescents and healthy controls. * p < 0.05 compared with the control group.

Full size image
Fig. 4
figure 4

Co-occurrence network of abundant fecal genera in depressed adolescents and controls. Co-occurrence network was constructed using the SparCC algorithm applied to relative abundance data at the genus level, illustrating potential ecological interactions within the microbial community. Cytoscape version 3.6.1 was used for network construction. The red and blue lines represent positive and negative correlations, respectively.

Full size image

Using key functional bacterial taxa identified, we assessed their potential to distinguish adolescents with MDD from controls using Random Forest and ROC analysis. Random Forest is a machine learning classification model that can efficiently and accurately classify microbial community samples. Mean Decrease Gini is used to compare the importance of variables by calculating their effect on the heterogeneity of the classification model. A higher value indicates that the key genus is more important (Fig. 5A). Bacteroides, Faecalibacterium, Blautia, and Roseburia emerged as the most important microbiota features distinguishing adolescents with MDD. ROC curves were employed to evaluate the diagnostic performance of these taxa at the genus level. The area under the curve (AUC) values served as a metric for the accuracy of each genus in distinguishing between MDD adolescents and controls. Notably, Bacteroides (AUC = 0.832), Roseburia (AUC = 0.768), Faecalibacterium (AUC = 0.760), Blautia (AUC = 0.734), Alistipes (AUC = 0.726), Streptococcus (AUC = 0.649), and Bifidobacterium (AUC = 0.622) showed significant discriminatory ability (Fig. 5B). Furthermore, combinations of these genera demonstrated even stronger discriminatory power. For instance, combinations such as Blautia plus Faecalibacterium (AUC = 0.802), Blautia plus Bacteroides (AUC = 0.824), and Blautia, Streptococcus, Bifidobacterium, Faecalibacterium plus Bacteroides (AUC = 0.868) exhibited enhanced diagnostic accuracy (Fig. 5C). Our results suggest that these key functional genera, both individually and in combination, could serve as potential biomarkers for distinguishing adolescents with MDD from healthy controls.

Fig. 5
figure 5

Diagnostic potential of differential genera in adolescent MDD. (A) Random Forest analysis with Mean Decrease Gini indicating genus importance. (B) Receiver-operating characteristic (ROC) curves for individual genera (Bacteroides, Roseburia, Faecalibacterium, Blautia, Alistipes, Streptococcus, Bifidobacterium) to distinguish depressed adolescents from healthy controls. (C) ROC curves for combined genera to discriminate between groups. AUC represents the area under the ROC curve.

Full size image

Inferred functional profile of microbiota in adolescent MDD

We employed the PiCRUSt algorithm to examine the functional characteristics of microbiota linked to adolescent MDD. PiCRUSt utilizes closed-reference OTU picking to predict functional categories from KEGG orthology database, facilitating the identification of metabolic and functional changes in fecal microbiota. The comprehensive functional profile of adolescent MDD-associated fecal microbiota is detailed in Fig. 6. Analysis of 64 level-2 KEGG pathways revealed significant differences in seven categories between adolescents with MDD and healthy controls (p < 0.05). Specifically, four pathways—carbohydrate metabolism, amino acid metabolism, xenobiotics biodegradation and metabolism, and biosynthesis of other secondary metabolites—were enriched in the MDD group, while three pathways—metabolism of cofactors and vitamins, glycan biosynthesis and metabolism, and folding, sorting and degradation—showed depletion. At level 3, 25 KEGG pathways exhibited statistically significant differences in activity between the two groups (p < 0.05). Notably, pathways such as starch and sucrose metabolism, glycolysis/gluconeogenesis, and biosynthesis of unsaturated fatty acids showed increased activity in adolescents with MDD, whereas pathways including citrate cycle (TCA cycle), glyoxylate and dicarboxylate metabolism, and glycosaminoglycan degradation displayed reduced activity. Overall, these findings suggest that the altered functional potential of bacterial communities in the fecal microbiota associated with MDD during adolescence may contribute to the pathogenesis and progression of this disorder.

Fig. 6
figure 6

Functional analysis of fecal microbiota in adolescents with MDD compared to healthy controls using PiCRUSt. Differences in bacterial functions were assessed using two-sided Welch’s t-test. Percentage comparisons for each KEGG functional category (levels 2 and 3) between groups are presented. Multiple testing correction based on false discovery rate (FDR) was applied using the Benjamini-Hochberg method through STAMP.

Full size image

Correlations between differentially abundant genera and host cytokines levels

We utilized the Bio-Plex Pro™ human cytokine group I panel 27-plex to identify significant changes in cytokine concentrations between healthy controls and adolescents with MDD (Figure S1). Specifically, adolescents with MDD showed elevated levels of inflammatory cytokines such as IL-5 and TNF-α, as well as increased concentrations of chemokines including Eotaxin, RANTES, IP-10, and MCP-1 compared to controls (Fig. 7). These findings highlight significant immune system dysregulation in MDD, particularly involving inflammatory and chemotactic responses.

Fig. 7
figure 7

Adolescent MDD-associated immune dysfunction. Concentrations (mean ± SEM, pg/ml) of 27 pro- and anti-inflammatory cytokines and chemokines measured in adolescent MDD patients and healthy controls using Bio-Plex immunoassays. The results indicate significant increases in the levels of IL-5 (A), Eotaxin (B), IP-10 (C), MCP-1 (D), RANTES (E), and TNF-α (F) in adolescent MDD patients relative to healthy controls (*p < 0.05; **p < 0.01; ***p < 0.001).

Full size image

To explore potential relationships between altered host immunity and predominant bacterial genera in adolescents with MDD, Spearman’s correlation analysis was conducted. The results were visualized using heatmaps based on Spearman’s correlation coefficient (Fig. 8). Notably, the genus Blautia, which exhibited increased abundance in adolescents with MDD, showed positive correlations with the upregulated cytokines mentioned earlier. Conversely, genera with reduced abundance in MDD patients—such as Bacteroides, Faecalibacterium, and Dialister—demonstrated negative correlations with chemokines like Eotaxin, RANTES, IP-10, and MCP-1. Specifically, higher levels of Eotaxin and RANTES were positively correlated with genera showing increased abundance and negatively correlated with those showing decreased abundance in adolescents with MDD. Our data suggest a complex interplay between cytokine profile changes and alterations in fecal microbiota composition among adolescents with MDD. This correlation suggests that microbiota-host interactions may contribute to the pathophysiology of MDD in adolescence, highlighting the complex interplay between gut microbiota and immune system dysregulation in adolescent MDD.

Fig. 8
figure 8

Correlation between key functional differential genera and pro- and anti-inflammatory cytokines and chemokines. Heatmap illustrating Spearman’s correlation coefficients between specific genera of fecal microbiota and pro- and anti-inflammatory cytokines and chemokines in adolescents with MDD. Statistical significance was assessed using Spearman’s correlation (r) and significance denoted by *p < 0.05; **p < 0.01; ***p < 0.001.

Full size image

Discussion

Adolescence, a phase of significant neural, behavioral, and biological changes, is marked by increased brain plasticity, pubertal maturation, and shifts in behavior, including greater independence and focus on peer relationships36. These factors collectively increase the risk of psychiatric disorders, particularly depression, which has surged during adolescence to become the second leading cause of death in this age group, and adolescent depression also predicts poorer long-term outcomes, including a three-fold higher risk of adult depression37,38,39. Given the limited effectiveness of current antidepressant treatments for adolescents, exploring new therapeutic approaches is crucial. The gut microbiota, often termed the “second brain”, has emerged as a key area of interest due to its potential impact on brain function and behavior40,41,42. This diverse microbiome residing the digestive tract interacts bidirectionally with the central nervous system through neural, endocrine, and immune pathways, known as the gut-brain axis. During adolescence—a crucial period for brain development and emotional maturation24,43—disruptions in this axis could potentially contribute to the onset or exacerbation of psychiatric disorders like depression12,13,44. Exploring how changes in the gut microbiota may impact adolescent depression offers promising avenues for understanding its underlying mechanisms and developing novel therapeutic strategies.

Unlike our previous research on depression in children and adults12,13, this is the first cross-sectional case-control study to comprehensively analyze the fecal microbiota and host cytokine profiles in Chinese adolescents with MDD, addressing a crucial gap by elucidating microbiota-host interactions at this developmental stage. We found significant changes in the fecal microbiota of adolescents with MDD compared to controls, including reduced α-diversity, increased richness (ACE, Chao1, and observed OTUs), and alterations in β-diversity metrics. LEfSe analysis identified that adolescents with MDD have higher levels of pro-inflammatory genera like Streptococcus and lower levels of anti-inflammatory genera such as Faecalibacterium, which could potentially serve as novel candidate non-invasive biomarkers for differentiating them from healthy individuals. Additionally, altered cytokine profiles were observed, with increased levels of pro-inflammatory cytokines such as IL-5 and chemokines like Eotaxin, showing significant correlations with changes in gut microbiota composition. Overall, these findings suggest that gut dysbiosis is associated with a complex pro-inflammatory response in adolescent MDD, offering insights into the potential mechanisms underlying depression during adolescence.

Our findings on bacterial diversity in adolescents with MDD contrast with previous studies involving various age groups—children, adolescents, and adults—with depression. Unlike our findings, earlier studies did not consistently identify significant differences in bacterial α- and β-diversity between depressed patients and healthy controls45,46. For instance, Thapa et al. found no significant differences in α- or β-diversity among older adolescents with MDD in USA47, and a study of Chinese female adolescents with depression found no notable differences in α-diversity indices like Shannon, Simpson, and Chao1 16. Meta-analyses of adult MDD patients and our previous research on school-aged children with MDD also revealed no changes in α-diversity13,15,46,48. However, findings on β-diversity in relation to MDD were relatively consistent. Systematic reviews indicate that 87% of β-diversity analyses show differences in gut microbiota composition in MDD patients49. Another comprehensive systematic review and meta-analysis also consistently reported changes in β-diversity, including decreased anti-inflammatory bacteria and increased pro-inflammatory bacteria among MDD patients15. Consistent with our prior research on school-aged children with MDD, PCoA revealed significant separation between adolescents with MDD and controls13. However, some studies have reported no significant β-diversity differences between MDD patients and healthy subjects, with PCoA analyses failing to differentiate the groups in Chinese female adolescents with MDD16,45,47. Variability in gut microbiota diversity linked to MDD may be influenced by age, as Chen et al. observed differences between young adults (18–29 years) and middle-aged adults (30–59 years) with MDD50, and baseline microbiota profiles change with age due to factors like diet, lifestyle, and gut development51,52,53,54. Psychosocial stress and gender differences also impact gut microbiota composition, with higher MDD prevalence among adolescent females and sex-specific microbiota changes improving diagnostic accuracy55,56,57. Overall, this suggests that gut dysbiosis during adolescence may influence the onset and progression of MDD.

Regarding gut microbiota composition in MDD, studies consistently link specific bacterial taxa with MDD, revealing that adolescents with MDD typically have elevated pro-inflammatory bacteria, like Escherichia, and reduced anti-inflammatory bacteria, including Bacteroides and Faecalibacterium. Consistent with findings in both adult and pediatric depression, depressed adolescents also exhibit reduced abundance of Bacteroides12,13. Bacteroides, known for its health-promoting effects such as enhancing pathogen resistance and influencing physiology, metabolism, and immune system development58,59,60, is increasingly recognized for its positive role in managing depression. Research indicates that reduced levels of Bacteroides in adults with depression are associated with lower quality of life and higher prevalence of depression11. Rhee et al. reported a negative correlation between Bacteroides abundance and total HAMD scores in adult MDD patients, suggesting that higher Bacteroides levels are associated with lower depression severity61. Strandwitz et al. found an inverse relationship between Bacteroides and brain activity patterns related to depression62. However, Hu et al. observed increased Bacteroides in moderate and severe MDD cases but not in mild depression63. The effects of Bacteroides on depression may vary by species or strains. For example, colonization by B. fragilis, B. uniformis, and B. caccae—but not B. ovatus—mimicked MDD-associated microbiome on behavior and hippocampal neurogenesis64, while B. ovatus, with its anti-inflammatory properties, affects intestinal SCFAs levels and neurotransmitter production, including γ-aminobutyric acid (GABA)65,66. Bacteroides are major producers of GABA, and reduced GABA levels have been linked to depressive-like behavior67. The exact mechanisms through which Bacteroides might influence behavioral changes remain unclear. Interestingly, our study associates Bacteroides abundance with systemic inflammation, suggesting it may influence host immunity in adolescent depression.

Faecalibacterium, especially F. prausnitzii, is considered a key indicator of gut and overall health, with reduced levels are commonly observed in individuals with inflammatory bowel disease, colorectal cancer, dermatitis, and depression68. The decreased Faecalibacterium in adolescents with MDD align with findings in children and adults with MDD, as well as in neuropsychiatric disorders like multiple sclerosis (MS), Alzheimer’s disease, and Parkinson’s disease12,13,15,26,69. Valles-Colomer et al. reported that higher Faecalibacterium levels are consistently associated with higher quality of life indicators11. Accumulating evidence indicates that Faecalibacterium is crucial for immune regulation, gut barrier integrity, and microbiota modulation, potentially through metabolites like butyrate, salicylic acid, shikimic acid, and raffinose, as well as anti-inflammatory molecules and extracellular matrix components68,70. Butyrate, a key metabolites of Faecalibacterium, directly influences serotonin and gut hormones release in the enteric nervous system, stimulating the vagus nerve and influencing endocrine signaling, which impacts brain function71. Our previous study identified an inverse relationship between Faecalibacterium abundance and the pro-inflammatory cytokine IL-17 in children with MDD13. These suggest that Faecalibacterium may contribute to the development of adolescent depression through the gut-brain axis.

Another key functional genus, Roseburia, was notably reduced in adolescents with MDD, consistent with findings in both adolescent and adult MDD patients16,23,72. We have previously observed a significant increase in Roseburia levels in adult patients with depression after effective treatment12. Zhou et al. identified a marked decrease in Roseburia in depressed adolescents, suggesting its potential as a depression predictor, and showed that R. intestinalis alleviates depression in mouse models by modulating tryptophan-derived neurotransmitter metabolism and supporting synaptogenesis and glial maintenance16. Furthermore, Zhao et al. demonstrated that R. intestinalis facilitated the synthesis of homovanillic acid, restored synaptic function by mitigating autophagic cell death, ultimately improving depressive symptoms in depressed mice23. Roseburia is crucial for colonic cell energy, barrier integrity, inflammation reduction, and the gut-brain axis through butyrate production8,73,74. R. hominis inhibits microglial activation, reduces brain inflammatory markers like IL-1α, IFN-γ, and MCP-1, and alleviates depressive behaviors through propionate and butyrate production75.

Interestingly, we observed elevated levels of Blautia, Bifidobacterium, and Escherichia_Shigella in adolescents with MDD. Although Blautia species are generally non-pathogenic, their increased presence in MDD patients has been documented12,44,72,76, and our prior study also found higher Blautia levels in patients with MS77. Hou et al. observed diminished B. massiliensis, virtually absent B. argi, and elevated B. faecis in children with Down syndrome, with higher B. faecis levels correlating with poorer cognitive function, highlighting the varying effects of different Blautia species on cognitive function78. Zhang et al. discovered that elevated Blautia levels in chronic hepatitis B patients treated with nucleos(t)ide analogues were positively correlated with advanced fibrosis, but not in treatment-naïve patients79. Consistent with findings in children and adults with depression, Bifidobacterium was elevated in adolescents with MDD13,72,80,81, challenging the assumption that Bifidobacterium universally benefits hosts and suggesting its impact may depend on specific species or strains82. Silva et al. found that B. longum exhibited strong antagonistic properties, while B. bifidum and B. pseudolongum promoted cytokine production and extracellular matrix proteins83. Metagenomic studies reported increased relative abundances of B. longum and B. dentium in adult MDD patients72,80. In addition, elevated Escherichia_Shigella levels were observed in both pediatric and adolescent patients with depression13. Studies by Bashir et al. and Chen et al. linked higher Escherichia_Shigella abundance to elevated depressive symptoms and anxiety severity, respectively72,84. Administering Lepidium meyenii Walp (Maca)-derived extracellular vesicles reduces Escherichia_Shigella levels and improves depressive behaviors in animal models85. Escherichia_Shigella, as a potential pro-inflammatory factor, is associated with increased inflammatory markers like IL-1β and NLRP3 86,87. It also contributes to intestinal inflammation by infiltrating epithelial cells, inducing macrophage apoptosis, and triggering IL-1β release88. Moreover, Escherichia_Shigella elevates lipopolysaccharide release into the bloodstream, potentially enhancing blood-brain barrier permeability and promoting chronic neuroinflammation, which may exacerbate depression89.

Consistent with previous studies, genera like Blautia, Faecalibacterium, Bacteroides, Roseburia, Alistipes, Streptococcus, and Bifidobacterium emerged as potential diagnostic markers distinguishing adolescents with MDD from healthy controls13,26,35,77. To enhance diagnostic accuracy, we employed multiple logistic regression analysis to identify optimal combinations of these genera (Blautia, Streptococcus, Bifidobacterium, Faecalibacterium, and Bacteroides), achieving an AUC of 0.868 in adolescent depression diagnosis. These findings suggest that targeting the gut microbiota composition and specific bacterial taxa may help to better understand the etiology of adolescent MDD and provide new diagnostic tools and therapeutic strategies for managing depression.

An ever-increasing body of evidence indicates that inflammatory processes have been implicated in the pathophysiology of depression90,91. Multiple meta-analyses have consistently shown that pro-inflammatory cytokines and acute-phase proteins, such as IL-6, TNF, and C-reactive protein (CRP), are elevated in MDD patients compared to healthy controls92,93,94. Our research also found elevated levels of pro-inflammatory cytokines (e.g., IL-5, TNF-α) and chemokines (e.g., Eotaxin, RANTES, IP-10, MCP-1) in depressed adolescents. A meta-analysis of twenty-two studies confirms the depression-inflammation link in children and adolescents95, and Cao et al. suggest early-life inflammation can lead to adolescent depressive symptoms96. Notably, TNF-α, a key inflammatory mediator, is consistently elevated in depressed individuals and associated with various neuropsychiatric disorders12,13,97. We also observed increased IL-5 in adolescents with MDD, which may affect brain function through neural plasticity pathways98. Chemokines are increasingly recognized in depression. Elevated serum levels of Eotaxin (CCL11) linked to cognitive decline and found in schizophrenia, bipolar disorder, and depression99,100,101,102. IP-10 is high in depression, and its levels decrease following antidepressant use103. MCP-1 (CCL2) is elevated in depression and associated with sleep disturbances and anxiety104. Research underscores the importance of MCP-1/CCR2 signaling in neuroinflammation in depression105. RANTES is significantly elevated in depressed patients and associated with inflammatory and neuroinflammatory disorders, highlighting its involvement in depression106. Collectively, these emphasize the importance of inflammation in depression and suggest potential targets for new treatments. Interestingly, our study revealed that elevated inflammatory biomarkers were significantly correlated with the gut microbiota: negatively with butyrate-producing genera and positively with acetate-producing genera. It has been hypothesized that alterations in the gut microbiota may contribute to the inflammatory processes observed in MDD107: gut dysbiosis may disrupt epithelial tight junctions, facilitating bacterial translocation and lipopolysaccharide (LPS) release. This activates toll-like receptor 4 (TLR-4), triggering nuclear factor-kappa B (NF-κB) signaling and promoting the expression of inflammatory cytokines (e.g., TNF-α, IL-1β, and IL-6). Peripheral inflammation can alter the blood-brain barrier (BBB) or be mediated by the vagus nerve, impacting the brain. In the brain, microglia and astrocytes amplify the immune response, leading to oxidative stress, mitochondrial dysfunction, and damage to neurons and oligodendrocytes, all of which are associated with depressive symptoms. In contrast, a healthy microbiota produces neuroactive mediators, including serotonin (5-HT) and SCFAs such as butyrate, which help maintain a healthy gut barrier. This indicates that modulating gut microbiota or targeting specific bacterial groups could be an effective strategy for managing inflammation in adolescent depression.

However, our study has several limitations. Firstly, the study’s exclusive focus on adolescent limits comparisons of gut microbiota profiles with those of children and adults, restricting insights across age groups. Secondly, our present case-control study primarily identified an association between gut microbiota composition and adolescent depression but did not investigate causal relationships. Future research should incorporate longitudinal follow-up studies, microbiota-targeted interventional trials, and mechanistic investigations using animal models to elucidate these causal mechanisms. Thirdly, although initial indications suggested a link between gut microbiota-derived metabolites and adolescent depression, a comprehensive investigation was not conducted; further metabolomics analyses are needed to explore this association. Fourthly, antidepressants can alter gut microbiota composition and function, so including drug-naïve subjects or comparing microbiota changes pre- and post-treatment is important, though recruiting drug-naïve patients poses challenges. Despite this, our study provides valuable foundational knowledge for future research, with existing evidence showing persistent associations between gut microbiota and depressive symptoms even after accounting for psychotropic drug use21. Lastly, the demographic homogeneity of participants from a specific region may limit the generalizability of our findings; future studies should include more diverse cohorts to validate and extend our results.

In summary, consistent with our previous studies on children and adults with depression, this study firstly observed gut dysbiosis and immune dysfunction in Chinese adolescents with MDD. Structural and functional disruptions in gut microbiota may influence host inflammation, contributing to the onset and progression of adolescent depression. Notably, we observed decreased butyrate-producing bacteria and increased acetate-producing bacteria, which were significantly correlated with elevated inflammatory cytokines and chemokines. These findings reveal a complex interaction between gut microbiota and the immune system, shedding light on the biological mechanisms underlying adolescent depression. Our present study has bridged the gap in depression research across different ages, providing novel, non-invasive diagnostic tools and potential therapeutic targets for adolescent depression. By uncovering the role of gut dysbiosis and immune dysfunction in adolescent depression, we offer new insights into the biological mechanisms of the disorder and pave the way for interventions aimed at restoring gut microbiota to improve mental health outcomes.