Major depressive disorder (MDD) is a prevalent psychiatric disorder affecting millions globally, posing significant social and economic burdens. Treatment outcomes are significantly influenced by initial symptom severity, with mild-to-moderate depression often responding to conservative treatment while severe depression requires more aggressive interventions. Misdiagnosis can lead to ineffective treatment and disease exacerbation. Identifying biomarkers to stratify MDD severity is crucial for early intervention and improved patient care. Growing evidence suggests a significant role for gut microbiota in the development and progression of mental disorders, including MDD. The gut-brain axis highlights the potential for gut microbiota alterations to impact brain function and behavior. Microbial biomarkers offer a promising avenue for developing novel diagnostic tools and therapeutic targets. While previous studies using 16S rRNA sequencing have explored the association between altered gut microbiota and MDD, these studies had limitations in resolution, only identifying bacteria to the genus level. This study aimed to utilize metagenome sequencing to provide a more comprehensive analysis of gut microbial composition and function in MDD patients with varying severities, seeking to identify potential microbial markers associated with disease severity and test their diagnostic performance.

Previous research has explored the link between gut microbiota and MDD, with studies utilizing 16S rRNA sequencing reporting variations in bacterial composition among MDD patients compared to healthy controls. Some studies have focused on specific bacterial genera like *Bacteroides*, *Prevotella*, *Alistipes*, and *Anaerostipes*, finding alterations in their abundance in MDD patients. However, these studies often lacked the resolution to identify bacteria at the species level, and the findings have been inconsistent across studies. The inconsistencies may be due to variations in study populations, methodologies, and the limited resolution of 16S rRNA sequencing. This study sought to address these gaps by using a higher-resolution metagenomic sequencing approach to investigate the gut microbiome in MDD patients stratified by disease severity.

This study utilized a cross-sectional design. A total of 155 healthy controls (HCs) and 138 treatment-naive MDD patients were recruited. MDD severity was assessed using the Hamilton Depression Rating Scale (HAMD-17), stratifying patients into mild, moderate, and severe groups. Patients with bipolar disorders, schizophrenia, other Axis I disorders, serious chronic somatic diseases, alcohol or substance abuse, or those who had changed diet or used antibiotics within a month of sampling were excluded. Fecal samples were collected, and DNA was extracted using the E.Z.N.A. Soil DNA Kit. Shotgun metagenomic sequencing was performed using Illumina NovaSeq. Bioinformatic analysis included quality control using Sickle, alignment to the human genome using BWA, contig assembly using MEGAHIT, open reading frame prediction using MetaGene, and gene clustering using CD-HIT. Gene annotation was performed using Diamond against the NCBI database, and KEGG database annotation was used to determine the abundance of KOs. Alpha-diversity indices (Dominance, Simpson, Shannon, Evenness) were calculated using PAST 4.0. Beta-diversity was analyzed using Principal Coordinate Analysis (PCoA) based on Bray-Curtis distance, and PERMANOVA was used to test group differences. Dirichlet multinomial mixtures (DMM) were used to identify enterotypes. Linear discriminant analysis effect size (LEfSe) was used to identify differentially enriched bacteria and KOs (LDA score > 2.5, p < 0.05). A random forest classifier with 5-fold cross-validation was used to assess the diagnostic performance of potential microbial markers, and receiver operating characteristic (ROC) curves were used to estimate diagnostic efficacy. SparCC correlation analysis was performed to determine co-occurrence relationships between differentially enriched bacteria and KOs (p < 0.05, r2 > 0.25). Statistical analysis was performed using SPSS 22.0.

Alpha-diversity analysis revealed significantly decreased Simpson indices in moderate and severe MDD groups compared to HCs. PCoA showed distinct microbial compositions in moderate and severe MDD groups compared to HCs, but not significantly different among the MDD subgroups themselves. *Bacteroides* was significantly enriched in moderate and severe MDD compared to HCs, while *Ruminococcus* and *Eubacterium* were depleted primarily in the severe group. LEfSe analysis identified 14 differentially enriched bacteria in the mild group, 60 in the moderate group, and 74 in the severe group. A Venn diagram showed 99 differentially enriched bacteria across all groups. SparCC analysis revealed a strong negative correlation between enriched *Bacteroides* and depleted bacteria (*Blautia*, *Ruminococcus*, *Eubacterium*) in moderate and severe MDD, suggesting a competitive relationship. The analysis also identified potential correlations between altered bacterial abundance and specific KOs, particularly K12373 and K21572, which showed a positive correlation with *Bacteroides* and a negative correlation with *Blautia* and *Ruminococcus*. A random forest model using 37 bacterial species showed excellent diagnostic performance in distinguishing between MDD subgroups (AUC 0.992–0.998). Gender-specific analysis revealed similar trends in microbiota alterations in both male and female MDD patients, although males showed a higher number of differentially enriched bacteria, most belonging to *Bacteroides*.

This study demonstrated a clear association between gut microbiota composition and the severity of MDD. The findings support the notion that gut dysbiosis is not just a feature of MDD but that its nature and extent change with disease progression. The increase in *Bacteroides* and decrease in potentially beneficial bacteria like *Ruminococcus* and *Eubacterium* in severe MDD warrant further investigation into their potential roles in disease pathogenesis. The identified microbial marker panel, with high diagnostic accuracy, offers a promising non-invasive tool for MDD stratification, potentially aiding in personalized treatment strategies. Future research should explore the functional implications of these bacterial alterations and their potential interaction with the host immune system and metabolic pathways. The observed gender differences in the number of differentially enriched bacteria in MDD merits further investigation with a larger cohort to confirm these findings.

This study provides compelling evidence for a link between gut microbiota composition and MDD severity. The identified 37-species microbial marker panel offers a potential non-invasive diagnostic tool for stratifying MDD patients based on their disease severity. Future research focusing on mechanistic studies and larger, longitudinal cohorts is needed to validate these findings and explore the therapeutic potential of targeting the gut microbiome in MDD.

The study had several limitations, including a relatively small sample size for each MDD subgroup and the exclusion of medicated patients. The cross-sectional design limits the ability to establish causality, and the sample was collected from a single clinical center, potentially limiting generalizability. Future studies should address these limitations through larger, longitudinal studies encompassing diverse populations and including the effects of medication on gut microbiota.