Differential co-expression networks of the gut microbiota are associated with depression and anxiety treatment resistance among psychiatric inpatients

Abstract

Background: Comorbid anxiety and depression are common and are associated with greater disease burden than either alone. Our recent efforts have identified an association between gut microbiota dysfunction and severity of anxiety and depression. In this follow-up, we applied Differential Co-Expression Analysis (DiffCoEx) to identify potential gut microbiota biomarker(s) candidates of treatment resistance among psychiatric inpatients.

Methods: In a sample of convenience, 100 psychiatric inpatients provided clinical data at admission and discharge; fecal samples were collected early during the hospitalization. Whole genome shotgun sequencing methods were used to process samples. DiffCoEx was used to identify clusters of microbial features significantly different based on treatment resistance status. Once overlapping features were identified, a knowledge-mining tool was used to review the literature using a list of microbial species/pathways and a select number of medical subject headlines (MeSH) terms relevant for depression, anxiety, and brain-gut-axis dysregulation. Network analysis used overlapping features to identify microbial interactions that could impact treatment resistance.

Results: DiffCoEx analyzed 10,403 bacterial features: 43/44 microbial features associated with depression treatment resistance overlapped with 43/114 microbial features associated with anxiety treatment resistance. Network analysis resulted in 8 biological interactions between 16 bacterial species. Clostridium perfringens evidenced the highest connection strength (0.95). Erysipelotrichaceae bacterium 6_1_45 has been most widely examined, is associated with inflammation and dysbiosis, but has not been associated with depression or anxiety.

Conclusion: DiffCoEx potentially identified gut bacteria biomarker candidates of depression and anxiety treatment-resistance. Future efforts in psychiatric microbiology should examine the mechanistic relationship of identified pro-inflammatory species, potentially contributing to a biomarker-based algorithm for treatment resistance.

Keywords: Anxiety; Biomarkers; Brain-gut Axis; Depression; Gastrointestinal microbiome; Metagenomics.

Fig. 1.

Differential Co-Expression (DiffCoEx) plots for clinical depression treatment resistance and anxiety treatment resistance. Patients were stratified based on their treatment resistance status for each of the two conditions, depression, and anxiety. For each condition, bacteria species and pathways were clustered into modules with strong correlation within module and patient group, but with differential correlation between the distinct treatment resistant patient groups. Modules are depicted by arbitrarily chosen colors. Inside each module, pathways and species are significantly differentially co-expressed between the treatment resistant (indicated with 1) and lack of treatment resistance (indicated with zero) for each clinical condition. Significance was achieved for p < 0.05 based on dispersion statistics.

Fig. 2.

Microbial features networks for the pathways and bacteria species identified by DiffCoEx. A. Correlation was assessed for each pair of the 43 overlapping features between the turquoise modules identified; only feature pairs with correlation cutoff >0.3 were indicated by edges. B. Detailed view of the microbial subnetwork of strongly correlated features. Node names correspond to either species or pathways. Edge thickness properties represent the weights of the edges; connected components of nodes are labeled with the same color.