Multi-Omics Analysis Reveals Aberrant Gut-Metabolome-Immune Network in Schizophrenia

Abstract

Schizophrenia (SCZ) is associated with several immune dysfunctions, including elevated levels of pro-inflammatory cytokines. Microorganisms and their metabolites have been found to regulate the immune system, and that intestinal microbiota is significantly disturbed in schizophrenic patients. To systematically investigate aberrant gut-metabolome-immune network in schizophrenia, we performed an integrative analysis of intestinal microbiota, serum metabolome, and serum inflammatory cytokines in 63 SCZ patients and 57 healthy controls using a multi-omics strategy. Eighteen differentially abundant metabolite clusters were altered in patients displayed higher cytokine levels, with a significant increase in pro-inflammatory metabolites and a significant decrease in anti-inflammatory metabolites (such as oleic acid and linolenic acid). The bacterial co-abundance groups in the gut displayed more numerous and stronger correlations with circulating metabolites than with cytokines. By integrating these data, we identified that certain bacteria might affect inflammatory cytokines by modulating host metabolites, such as amino acids and fatty acids. A random forest model was constructed based on omics data, and seven serum metabolites significantly associated with cytokines and α-diversity of intestinal microbiota were able to accurately distinguish the cases from the controls with an area under the receiver operating characteristic curve of 0.99. Our results indicated aberrant gut-metabolome-immune network in SCZ and gut microbiota may influence immune responses by regulating host metabolic processes. These findings suggest a mechanism by which microbial-derived metabolites regulated inflammatory cytokines and insights into the diagnosis and treatment of mental disorders from the microbial-immune system in the future.

Keywords: cytokines; gut microbiota; metabolism; metagenomics; schizophrenia.

Figure 1

Identification of differential metabotypes associated with schizophrenia. (A) The orthogonal partial least squares discriminant analysis (OPLS-DA) model was used to discriminate between 63 SCZ subjects (blue boxes) and 57 HCs (yellow diamonds). (B) Permutation test showing the original R2 and Q2 values (top right) as significantly higher than corresponding permuted values (bottom left), demonstrating the OPLS-DA model’s robustness. (C) Average relative intensities of the 18 co-abundance metabotypes (clusters) between the two groups. Red and green represented relatively high and low densities, respectively. The asterisk represented a significant negative correlation between the metabotype and the duration of illness (P-value < 0.05). .

Figure 2

Identification of the important co-abundance groups (CAGs) that were strikingly different between two groups. (A) Metagenomic operational taxonomic unit (mOTU)-level network diagram showing the enrichments of mOTUs in the different groups based on significantly changed CAGs. Node size indicated the mean abundance of each mOTU. Lines between nodes represented correlations, with line width indicating correlation magnitude, purple representing positive correlation. Only lines corresponding to correlations with magnitudes greater than 0.4 were drawn. Different colors were used to distinguish CAGs. (B) Relative abundances of 12 CAGs with significantly differences identified by the Wilcoxon rank-sum test between the two groups (P-value < 0.05). The names of the CAGs comprising the schizophrenia-enriched and healthy control-enriched CAGs were highlighted in red and blue, respectively.

Figure 3

Interrelationship between gut microbiota composition, host metabolic profile, and inflammatory cytokines. (A) Visualization of the correlation network according to Spearman correlation analysis between the gut microbiota of significant co-abundance groups (CAGs) and the inflammatory cytokines was mediated by serum metabolites. Red connections indicated positive correlations (Spearman correlation test, FDR < 0.05), whereas blue connections showed correlations that were negative (Spearman correlation test, FDR < 0.05). All correlations were adjusted for age, sex, body mass index, and diet. In the gut microbiota column, the green stratum represented CAGs that were highly enriched in the healthy control (HC) group, and the stratum colored in blue was increased in the schizophrenia (SCZ) group. In the metabolomics column, the orange stratum represented HC-enriched metabotypes, and the claybank stratum represented SCZ-enriched metabotypes. In the cytokines column, the red stratum represented elevated inflammatory cytokines in the SCZ group, and the gray represented no difference between the two groups. (B) Scatter plot representing the association between C–C motif chemokine ligand 2 and CAG11 in all subjects. (C) The percentage of variance in Shannon index that was explained by each serum metabolite that was significantly associated with Shannon index (P-value < 0.05). For each metabolite, red bars correspond to a positive β-coefficient, whereas blue bars correspond to a negative β-coefficient. Significance was assessed using ordinary least squares regression.

Figure 4

Correlation analysis of multi-omics. (A) Pairwise comparisons of metabotypes and cytokines were shown, with a color gradient denoting Pearson’s correlation coefficients. Actinobacteria, Bacteroidetes, Firmicutes, and mOTUs were related to each metabotypes and cytokines by partial (age, sex, BMI and diet-corrected) Mantel tests. Edge width corresponded to the Mantel’s r statistic for the corresponding metabotypes and cytokines correlations, and edge color denoted the statistical significance based on 9,999 permutations. (B) Venn plot of the number of differentially abundant mOTUs that were associated with serum metabolites and cytokines, respectively. (C) Clostridium symbiosum contributed to C–C motif chemokine ligand 2 (CCL2) through cholic acid (P_mediation = 0.022, 20%). (D) Eggerthella lenta contributed to CCL2 through asparaginyl-valine (P_mediation = 0.012, 24%). (E) Eggerthella lenta contributed to CCL2 through cholic acid (P_mediation = 0.010, 24%). The gray lines indicated the Spearman correlations adjusting for age, sex, BMI and diet, with corresponding rho coefficients and false discovery rate values. The blue arrowed lines indicated the microbial effects on CCL2 mediated by metabolites, with the corresponding mediation P-values.

Figure 5

Metabolites classify schizophrenia from healthy control. (A) Distribution of five trials of 10-fold cross-validation error in random forest classification of schizophrenia as the number of features increases. The model was constructed using relative abundance of the 50 metagenomic operational taxonomic units and 133 metabolites from 70% of schizophrenia and healthy control samples (training set). The black curve indicates the average of the five trials (gray lines). The pink line marks the number of features in the optimal set. (B) Receiver operating characteristic analysis was performed to evaluate the diagnostic performance of these seven metabolite biomarkers, and the areas under the curve of the training set and test set were 99.17% and 99.45%, respectively. (C) The seven metabolites with most weight to discriminate schizophrenia and controls were selected by the cross-validated random forest models. The length of line indicated the importance of the metabolites for classification.

Figure 6

The summary of gut microbiota composition, metabolism and cytokines analysis between schizophrenia and healthy control.