Gut microbial CAZymes markers for depression

Abstract

Major depressive disorder (MDD) is a serious mental illness, characterized by disturbances of gut microbiome, it is required to further explore how the carbohydrate-active enzymes (CAZymes) were changed in MDD. Here, using the metagenomic data from patients with MDD (n = 118) and heath controls (HC, n = 118), we found that the whole CAZymes signatures of MDD were significantly discriminated from that in HC. α-diversity indexes of the two groups were also significantly different. The patients with MDD were characterized by enriched Glycoside Hydrolases (GHs) and Polysaccharide Lyases (PLs) relative to HC. A panel of makers composed of 9 CAZymes mainly belonging to GHs enabled to discriminate the patients with MDD and HC with AUC of 0.824. In addition, this marker panel could classify blinded test samples from the two groups with an AUC of 0.736. Moreover, we found that baseline 4 CAZymes levels also could predict the antidepressant efficacy after adjusted confounding factors and times of depressive episode. Our findings showed that MDD was associated with disturbances of gut CAZymes, which may help to develop diagnostic and predictive tools for depression.

Fig. 1. Compositional variation in human gut CAZymes.

a, b The overall characteristics of CAZymes were displayed by Pcoa (bray-curtis distance), internal differences were analyzed with permanova test (permutations = 999, Bonferroni p = 0.002). Also, the distance of first and second principal component distance was showed by boxplots (Mann–Whitney U-test, p = 0.004). c Difference of 6 gut CAZymes diversity index between MDD and HC. d 83 discriminated CAZymes were identified between two groups (Mann–Whitney U-test). e The fold change of CAZymes related to plant and animal carbohydrate utilization, plant carbohydrate utilization and mucin glycan and animal to plant carbohydrate utilization in the MDD and HC. The boxplots displayed the respective distributions of three types of CAZymes in metagenome data (Mann–Whitney U-test, two-sided). Box-Whisker plot, box = 25–75th percentiles, whiskers = 5–95th percentiles, horizontal line in box = median.

Fig. 2. Gut CAZymes features can distinguish MDD and HC.

Using random forest model, 9 CAZymes were identified with importance score >1%. a Volcano plot for differential CAZymes markers. Significantly regulated metabolites between groups determined by fold change and value of p (FC > | 1.5 | , p < 0.05). Gray dots represent increased CAZymes in HC; green dots represent increased CAZymes in MDD. b The box plots showed the differences of 9 CAZymes markers (Wilcoxon rank-sum test). c A random forest model was constructed and displayed by ROC. In the discovery set, individual signature could discriminate the two groups with area under the curve (AUC) at 0.824, the value in brackets is 95% CI. The diagnostic efficiency was confirmed by 5 fold cross validation test (accuracy: 65.24 ± 7.72%). d Correlation-based networks of co-occurring MDD-related CAZymes colored by node affiliation, a co-varying cluster was composed of 6 GHs in MDD subjects. A node stands for an CAZyme and a connection (i.e. edge) stands for a significant (pearson’s r > 0.2 or <-0.2, p < 0.05) pairwise correlation. Size of the nodes represents the rpkm of these variables. Edges between nodes indicate pearson’s positive (green) or negative (gray) correlation, edges thickness indicate range of p-value (p < 0.05).

Fig. 3. Prospective association of CAZymes markers with severity.

a, b Prospective association of baseline gut CAZymes with HAMD and QIDS score. A total of 45 participants were included in this analysis. Linear mixed-effects model was performed to assess the prospective association of gut CAZymes with the clinical score, adjusting for the demographic (age, gender and educational background), anthropometric (BMI), times of depressive episode and medication. A p-value < 0.05 was considered as statistically significant (estimated value β, 95%CI).