Intestinal Microbes in Patients With Schizophrenia Undergoing Short-Term Treatment: Core Species Identification Based on Co-Occurrence Networks and Regression Analysis

Abstract

Schizophrenia, a common mental disorder, has a tremendous impact on the health and economy of people worldwide. Evidence suggests that the microbial-gut-brain axis is an important pathway for the interaction between the gut microbiome and the development of schizophrenia. What is not clear is how changes in the gut microbiota composition and structure during antipsychotic treatment improve the symptoms of schizophrenia. In this study, 25 patients with schizophrenia were recruited. Their fecal samples were collected before and after hospital treatment for 14-19 days. The composition and structure of the intestinal microbiota were evaluated by 16S rRNA sequencing analysis, and the results showed significant differences in fecal microbiota before and after treatment. Firmicutes (relative abundances of 82.60 and 86.64%) and Gemminger (relative abundances of 14.17 and 13.57%) were the first dominant species at the phylum and genus levels, respectively. The random forest algorithm and co-occurrence network analysis demonstrated that intestinal flora (especially the core species ASV57) could be used as biomarkers to distinguish different clinical states and match treatment regimens accordingly. In addition, after fecal microbiota transplantation, antibiotic-treated recipient mice showed multiple behavioral improvements. These included decreased psychomotor hyperactivity, increased social interaction, and memory. In conclusion, this study suggests that differences in the composition and structure of gut microbiota after treatment are associated with the development and severity of schizophrenia. Results may provide a potential target for the treatment of this disorder.

Keywords: 16S rRNA sequencing; fecal microbiota transplantation; gut microbiota; gut-brain axis; schizophrenia.

Figure 1

Alpha, Beta diversity, and gut microbial composition. (A) Rarefaction curves for intestinal microbiota observed ASV richness. The color lines indicate BT and AT samples, the dashed black line indicates the selected rarefaction depth used to generate the box plots below each curve, 16,589 seqs/sample for microbial communities. (B) Community richness of intestinal microbiota in BT and AT samples (Richness); (C) Community diversity of intestinal microbiota in BT and AT samples (Shannon diversity), ns, no significant. (D) Principal coordinate analysis (PCoA) of intestinal microbiota of mice under different group. (E) Relative abundance of dominant phylum taxa (>1%) in intestine from each group. (F) Relative abundance of dominant genus taxa (>1%) in intestine from each group.

Figure 2

Comparison of intestinal difference groups between BT and AT groups. LEfSe analysis of key taxa of gut microbiota in mice between the BT group and AT group. (A) LDA score plot of bacterial taxa with LDA scores higher than 4. (B) Taxonomic cladogram derived from LEfSe analysis of 16S rRNA sequences. (C-H) Boxplots showing differences in the relative abundance of specific taxa between BT and AT patient. Specific taxa in figure (C–H) were Faecalibacterium (C), Butyricicoccus (D), Oscillospira (E), Roseburia (F), Prevotella (G), and Rothia (H). *p < 0.05, **p < 0.01.

Figure 3

Ecological processes shaping the gut community assembly in BT and AT patients. (A,B) Predicted frequency of occurrence of gut microbiota in the BT group (A) and AT group (B). The solid blue line is the best fit to the neutral community model (NCM), and the dashed blue line expresses 95% confidence intervals around the NCM prediction. ASVs that occur more or less frequently than predicted by the NCM are shown in lightgreen and lightred, respectively. R² indicate the fit to this model. (C) Boxplot shows the niche width index distribution of microbial community of gut in BT and AT samples. (D) Relative contributions of generalized and specialized species to BT group and AT group of intestinal microbiotas. (E) C-score metric using null models. The values of observed C-score (C-score_obs) > simulated C-score (C-score_sim) represent non-random co-occurrence patterns. ns, no significant.

Figure 4

Identification of core ASVs and microbial network analysis. (A,B) Random forest showing the top 20 features of the intestinal microenvironment distinguished in the pre- and post-treatment datasets. Each feature (ASV) was assigned two importance score of mean decrease accuracy (A) and mean decrease Gini (B). (C) Bipartite network showed ASVs specific to before or after treatment identified using indicator species analysis in gut microbial communities. The circle indicates that single species are positively and significantly correlated (p < 0.05) with one or more gut microenvironments (the link given by the line). Different ASVs belonging to same phylum level are given the common color. Co-occurrence sub-networks of (D) before treatment or (E) after treatment. Circles for microbial ASVs are represented by nodes. And partial nodes are colored by association with different modules, which include indicator species corresponding to different gut microenvironment. (F) Common sub-networks analysis of BT group and AT group. All nodes belonging to the same community were randomly assigned similar colors. Gray nodes represented ASVs appearing in both the variable group. Big and red node can be considered as “drive microbiota” Red line indicate only in the AT group of sub-networks of interactions. Green line indicates only in the BT group of sub-networks of interactions. And blue line expresses the interactions in both sub-networks.

Figure 5

(A) ROC according to 40 samples of the testing set calcualted by cross-validated random forest model. Area under ROC (AUC) and the 95% confidence intervals are also shown. (B) All-subsets Regression. The horizontal axis represents the predictive variables, and the vertical axis represents the accuracy of the regression (adjusted R²), which shows how the accuracy of the regression when different combinations of predictive variables are considered. (C) Non-metric multidimensional scaling (NMDS) based on the Bray–Curtis distances. Core ASVs (DelBet > 0) are labeled blue and projected as biomarkers to different locations pre- or post-treatment.

Figure 6

Gut microbiota from patients with SCZ (after treatment) induced a lower level of hyperkinetic behavior and improved sociability, social novelty, and memory in mice than before treatment but had no effects on anxious state and depressive behavior. (A) Schematic diagram of mouse treatment and behavioral testing. Mice were given a cocktail of antibiotics to eliminate gut microbiota and received oral gavage of the microbiota suspension from before or after treatment patients with SCZ. A series of behavioral tests were carried out 48 h after the last FMT. OFT open field test, EPMT elevated plus maze test, TCST three-chamber social test, NORT novel object recognition test, FST forced swimming test. (B) In the 10-min open field test, both groups of mice spent similar amounts of time in the central arena. (C) The cumulative distance (meters) traveled in the different zones, the representative traces of mouse activity. (D) Time in open arms of the elevated plus maze during 5-min exploration did not vary between two groups. (E) AT mice had obvious changes in sociability compared with BT mice. (F) AT mice had obvious changes in social novelty compared with BT mice. (G) Novel object recognition test. AT mice spent significantly more time exploring the novel object compared with BT mice, indicating improvement of impaired recognition memory. (H) Immobility in the forced swimming test (FST, p = 0.138) was similar between BT and AT mice. Means ± SD were presented in (B–H) and each circle represents individual mouse (n = 12 per group). P-values were determined by the Independent Student T-test. ns, no significant. *p < 0.05, **p < 0.01.