3.1 Comparison of general demographic information
No statistically significant differences were observed between the two groups in terms of age, gender, BMI, years of education, smoking, drinking, family history, marital status, or dietary habits (P > 0.05) as shown in Table 1. However, significant differences were found between the two groups concerning age at onset and duration of disease (P < 0.05), also detailed in Table 1.
3.1 Comparison of scale scores
The demographic information of both groups was detailed in Supplementary Table 1. For the HAMD-24, HAMA, somatic anxiety, TMT-A, and DSST scores, no statistically significant differences were observed between the two groups (P > 0.05) as indicated in Table 1. In contrast, significant differences were found in mental anxiety, DST, DST-forward, and DST-reverse scores between the two groups (P < 0.05), also presented in Table 1.
3.2 Microbiome analysis
3.2.1 Analysis of sequencing results
A total of 82 stool samples were sequenced using 16S rRNA. After initial sequencing, Flash software was employed for read splicing and quality control, yielding 6,247,810 high-quality clean reads. These were subsequently subjected to chimera filtering. OTUs were then clustered using QIIME software, resulting in 1,899 OTUs at 97% sequence similarity. At the OTU level, 836 OTUs were common to both BD and MDD groups, while 609 OTUs were unique to the BD group and 454 OTUs were unique to the MDD group, as depicted in Figure 1.
The Shannon-Wiener curve plateaued, indicating that the sequencing depth was sufficient to capture the majority of gut microbiota in the samples (Figure 2).
Fig.1 Venn diagram of species
Fig.2 Shannon-Wiener curve.
Note:The horizontal axis represents different sequencing depths in the sample, the vertical axis represents the Shannon Wiener index at the corresponding depth, and the curves of different colors represent the Shannon Wiener curves of different samples.
3.2.2 Analysis of intestinal flora diversity
Alpha diversity metrics, including Chao1, ACE, and Observed species indices, represented species richness, while Shannon and Simpson indices were used for species diversity, and the J index indicated community evenness. The Chao1, ACE, and Observed species indices were significantly lower in the BD group compared to the MDD group (P < 0.05) as shown in Figure 3. However, no significant differences were found in Shannon, Simpson, and J indices between the two groups (P > 0.05), also shown in Figure 3, suggesting similar species diversity and evenness but differing richness between the two groups.
In this study, PCA based on OTU abundance data was used to compare the microbial community structure between samples. Significant differences were observed in the β-diversity index between the BD and MDD groups (P = 0.001) as shown in Figure 4, indicating distinct intestinal microbial structures between the two groups.
Fig.3 Alpha diversity index box diagram.
Note: The abscissa represents the grouping, and the ordinate is Alpha Index.
Fig.4 PCA analysis results.
3.2.3 Analysis of differences in microbial composition
To identify distinct microbial species between the BD and MDD groups and isolate the dominant flora in each, LEfSe analysis was conducted with a set LDA threshold of greater than 3.8. This analysis revealed 17 microbial groups that were significantly different between the two groups (P < 0.05, LDA > 3.8), as shown in Figures 5a and 5b.
Fig. 5a Histogram of LDA effect value of species with significant difference (LDA > 3.8)
Fig. 5b Branch diagram of LDA effect value of species with significant difference (LDA > 3.8)
3.2.4 ROC curve analysis
ROC curve analysis was performed on the differential flora at the genus level to evaluate their discriminative power between the BD and MDD groups. The results are presented in Figure 6 and Table 2. The AUC for the combination of these four differential genera was calculated to be 0.925, suggesting that this combination holds strong diagnostic potential for differentiating BD from MDD.
Fig. 6 ROC analysis of genus level differential bacteria and combined model.
3.2.5 Analysis of correlation between differential intestinal genera and clinical symptoms
A correlation study was performed to explore the relationships between the four differential genera and various clinical symptom scores, the results of which are depicted in Figure 7. Specifically, the genus g__un_f_Muribaculaceae was found to be negatively correlated with HAMD-24, HAMA, and somatic anxiety scores in the BD group (r = -0.423, P < 0.01; r = -0.345, P < 0.05; r = -0.313, P < 0.05). In contrast, this genus was positively correlated with DST-reverse scores (r = 0.329, P < 0.05), DSST scores (r = 0.415, P < 0.01), and TMT-A time use (r = -0.423, P < 0.01) in the MDD group.
Fig. 7 Correlation analysis between different gut microbiota and clinical symptoms
Note: A positive correlation is red, and the deeper the red, the higher the positive correlation. A negative correlation is blue, and the deeper the blue, the higher the negative correlation. **<0.01,*<0.05.
3.3 Untargeted metabolome
3.3.1 Identification Results and Statistical Analysis
A total of 40 blood samples were analyzed—22 from the BD group and 18 from the MDD group. Utilizing an untargeted metabolomics approach based on UHPLC-MS, 1168 metabolites were identified across both groups. Of these, 514 were identified through negative ionization and 654 through positive ionization. These metabolites were subsequently categorized based on their chemical classifications, as illustrated in Figure 8.
Fig. 8 Proportion of identified metabolites in each chemical classification.
3.3.2 Validation of the model
OPLS-DA was employed to observe the overall sample distribution, depicted in Figures 9a and 9b. To confirm the model's stability, a 7-fold cross-validation was carried out. The analysis demonstrated that the metabolic data effectively distinguished between the BD and MDD groups (Positive ions: R2X=0.164, Q2=0.433; Negative ions: R2X=0.146, Q2=0.382). A permutation test was conducted to validate the model and prevent overfitting, as displayed in Figures 10a and 10b. The model was deemed robust when the R2 and Q2 values of the permutation model gradually decreased.
Fig. 9a OPLS-DA score of positive ion.
Fig. 9b OPLS-DA score of negative ion.
Note: t [1] represents principal component 1, to [1] represents principal component 2, and the ellipse represents a 95% confidence interval. Dots of the same color represent various biological duplications within a group, and the distribution of dots reflects the degree of difference between and within groups.
Fig.10a Positive ion OPLS-DA replacement test.
Fig.10b Negative ion OPLS-DA replacement test.
Note: The abscissa in the figure represents the degree of displacement retention, that is, the proportion consistent with the order of the Y variables in the original model, and the ordinate represents the values of R2 and Q2. The green dot represents R2, the blue dot represents Q2, and the two dashed lines represent the regression lines of R2 and Q2, respectively. The R2 and Q2 in the upper right corner indicate that the displacement retention is equal to 1, which is the R2 and Q2 values of the original model.
3.3.3 Differential metabolite screening
In this study, a total of 50 differential metabolites were identified using VIP scores greater than 1 and a P-value less than 0.05 as the criteria for significance (Figures 11a and 11b). In the positive ion mode, 13 differential metabolites were found, nine of which were elevated in the BD group. Meanwhile, in the negative ion mode, 37 differential metabolites were identified, 24 of which were elevated in the BD group.
Fig.11a Multiple analysis of significant difference in metabolite expression in positive ion
Fig.11b Multiple analysis of significant difference in metabolite expression in negative ion
Note: The abscissa in the figure represents the log2 FC value of the differential metabolite, that is, the logarithmic value of the differential multiple of the differential metabolite based on 2, and the ordinate represents the significant differential metabolite. Red indicates that the BD group upregulates differential metabolites, while green indicates that the BD group downregulates differential metabolites.
3.3.4 Metabolic pathway analysis of differential metabolites
KEGG pathway analysis was performed on the differential serum metabolites to identify the most relevant metabolic pathways between the BD and MDD groups. The analysis revealed six metabolic pathways that were most affected: Taurine and Hypotaurine Metabolism, Neuroactive Ligand-Receptor Interactions, Pyrimidine Metabolism, Vitamin B6 Metabolism, Purine Metabolism, and ABC Transporter (Figures 12a and 12b).
Fig. 12a KEGG enrichment pathway diagram (bubble diagram)
Fig. 12b KEGG enrichment pathway diagram (histogram)
3.3.5 ROC curve analysis
Given the significant differences in serum metabolic profiles between the BD and MDD groups, ROC curve analysis was conducted on the differential metabolites to identify specific biomarkers. The top 10 metabolites with VIP scores greater than 1 were selected for this analysis (Table 3 and Figure 13). Eight metabolites demonstrated an AUC greater than 0.7: adenosine, glycerophosphocholine, Phe-phe, 1-hydroxy-2-naphthoic acid, hypoxanthine, Dl-lactate, taurine, and isocitric acid (AUC values as follows: 0.818, 0.737, 0.889, 0.770, 0.720, 0.760, 0.730, 0.715). These metabolites appear to have strong diagnostic potential.
Fig.13 ROC analysis of the first 10 metabolites with high VIP value difference
3.3.6 Correlation analysis of differential metabolites and clinical symptoms
A correlation analysis was conducted between the serum levels of the eight metabolites with strong diagnostic value and the clinical symptom scores (Figure 14). In the MDD group, 1-hydroxy-2-naphthoic acid showed a positive correlation with DSST score (r=0.616, P<0.01) and a negative correlation with TMT-A (r=-0.633, P<0.01). Dl-lactate was positively correlated with both DSST score and DST-reverse (r=-0.614, P<0.01; r=0.486, P<0.05) and negatively correlated with TMT-A (r=-0.554, P<0.05). In the BD group, no differential metabolites were found to correlate with the severity of clinical symptoms.
Fig.14 Correlation analysis between differential metabolites and clinical symptoms
Note: A positive correlation is red, and the deeper the red, the higher the positive correlation. A negative correlation is blue, and the deeper the blue, the higher the negative correlation. **<0.01,*<0.05.
3.4 Integrated analysis of microbiome and metabolome
To investigate the potential relationship between differential intestinal genera and differential serum metabolites, a Spearman correlation analysis was conducted (Figure 15). The analysis indicated a substantial interaction between the differential metabolites and the differential gut microbiota.
Fig. 15 Correlation analysis of differential gut microbiota and serum differential metabolites
Note: A positive correlation is red, and the deeper the red, the higher the positive correlation. A negative correlation is blue, and the deeper the blue, the higher the negative correlation. **<0.01,*<0.05.