Characteristic alterations of gut microbiota and serum metabolites in patients with chronic tinnitus: a multi-omics analysis

Abstract

Chronic tinnitus is a central nervous system disorder. Currently, the effects of gut microbiota on tinnitus remain unexplored. To explore the connection between gut microbiota and tinnitus, we conducted 16S rRNA sequencing of fecal microbiota and serum metabolomic analysis in a cohort of 70 patients with tinnitus and 30 healthy volunteers. We used the weighted gene co-expression network method to analyze the relationship between the gut microbiota and the serum metabolites. The random forest technique was utilized to select metabolites and gut taxa to construct predictive models. A pronounced gut dysbiosis in the tinnitus group, characterized by reduced bacterial diversity, an increased Firmicutes/Bacteroidetes ratio, and some opportunistic bacteria including Aeromonas and Acinetobacter were enriched. In contrast, some beneficial gut probiotics decreased, including Lactobacillales and Lactobacillaceae. In serum metabolomic analysis, serum metabolic disturbances in tinnitus patients and these differential metabolites were enriched in pathways of neuroinflammation, neurotransmitter activity, and synaptic function. The predictive models exhibited great diagnostic performance, achieving 0.94 (95% CI: 0.85-0.98) and 0.96 (95% CI: 0.86-0.99) in the test set. Our study suggests that changes in gut microbiota could potentially influence the occurrence and chronicity of tinnitus, and exert regulatory effects through changes in serum metabolites. Overall, this research provides new perceptions into the potential role of gut microbiota and serum metabolite in the pathogenesis of tinnitus, and proposes the "gut-brain-ear" concept as a pathomechanism underlying tinnitus, with significant clinical diagnostic implications and therapeutic potential.IMPORTANCETinnitus affects millions of people worldwide. Severe cases may lead to sleep disorders, anxiety, and depression, subsequently impacting patients' lives and increasing societal healthcare expenditures. However, tinnitus mechanisms are poorly understood, and effective therapeutic interventions are currently lacking. We discovered the gut microbiota and serum metabolomics changes in patients with tinnitus, and provided the potential pathological mechanisms of dysregulated gut flora in chronic tinnitus. We proposed the innovative concept of the "gut-brain-ear axis," which underscores the exploration of gut microbiota impact on susceptibility to chronic tinnitus through serum metabolic profile modulation. We also reveal novel biomarkers associated with chronic tinnitus, offering a new conceptual framework for further investigations into the susceptibility of patients, potential treatment targets for tinnitus, and assessing patient prognosis. Subsequently, gut microbiota and serum metabolites can be used as molecular markers to assess the susceptibility and prognosis of tinnitus.Furthermore, fecal transplantation may be used to treat tinnitus.

Keywords: chronic tinnitus; gut microbiota; gut-brain-ear axis; multi-omics analysis; serum metabolites.

Fig 1

Gut microbiome dysbiosis in tinnitus group. (A) A Venn diagram demonstrating the existence of ASVs in the tinnitus and healthy groups. The area where two circles intersect indicates the number of ASVs shared between the two groups. (B) Significant differences existed in alpha-diversity between the two groups, including Chao1, observed_otus, and goods_coverage. Wilcoxon rank-sum test was used. (C) The PCoA was performed using an unweighted UniFrac matrix. (D) UPGMA cluster diagram. The UPGMA is based on unweighted UniFrac metrics.

Fig 2

Overview of the gut microbiome in tinnitus and HC groups. (A) Composition of these two groups at the phylum level. Dominant phyla and their relative abundance in each group were exhibited. (B) Relative abundance in tinnitus and control groups. Each row represents a bacterium (top 30), and each column represents groups. At the phylum level, the expression abundance of the same bacterium is normalized by Z-score. It is possible to compare the same microorganism between different groupings. (C) Composition of these two groups at the genus level. Dominant genus and their relative abundance in each group. (D) Genus-level microbial differential analysis. Mann-Whitney U-test. A bar chart displaying the top 30 bacterium with the highest relative abundance with P < 0.05. The horizontal axis represents the differential species (arranged from left to right in order of abundance), and the vertical axis represents relative abundance. (E) Relative abundance of aerobic bacteria predicted based on the BugBase database. P < 0.01. Mann-Whitney-Wilcoxon test with FDR corrected. (F) Distribution of aerobic bacteria at the phylum level in each group.

Fig 3

Biomarker of gut microbiota and their predicted function. (A) Cladogram of gut microbiota. The concentric circles radiating outward represent the seven taxonomic levels of genus, species, order, family, class genus, order genus, and species-genus, respectively, starting from the innermost circle. Each node represents a species classification at this level, with the species abundance represented by the node’s size. Each dot represents a taxonomic hierarchy and is marked for significant (LEfSe: P < 0.05) enrichment either in the tinnitus group (green) or in the HC group (red). (B) Histogram of distribution. Taxa that reached a LDA score >3.0 are listed. The colors in the bar chart indicate the relative abundance of distinct species across different groups. In contrast, the length of the bars reflects the LDA score, which quantifies the degree of significant differences among species between groups. (C) PICRUSt2 analysis based on KEGG pathway annotation. P < 0.05. The two groups calculated P-value using Stamp differential analysis (t-test).

Fig 4

Dysregulation of serum metabolic profile in tinnitus group. (A) Histogram reveals the counts of metabolites annotated by HMDB at the superclass level. The horizontal axis represents the superclass classification of metabolites. The vertical axis represents the number of metabolites contained in that classification. (B) Histogram reveals the counts of metabolites annotated by the KEGG pathway at level 3. The horizontal axis represents the level 3 classification. The vertical axis represents the number of metabolites in that classification. Different colors indicate different level 1 classifications. (C) The PCA performed to reduce dimension and discriminate serum metabolic profile between tinnitus (red) and control (purple) groups. The horizontal axis represents the first principal component. The vertical axis represents the second principal component. The numbers in parentheses indicate the degree to which each principal component explains the overall data. (D) The PLS-DA model discriminates individuals. Tinnitus patients were labeled in red. Healthy individuals were labeled in purple. The horizontal axis represents the first principal component. The vertical axis represents the second principal component. The numbers in parentheses indicate the degree to which each principal component explains the discriminative model. (E) Permutation test indicated that the model has not overfit. The R2 regression line was labeled in red and Q2 line was labeled in blue. The intercept of the Q2 regression line with the y-axis was shown in the bottle.

Fig 5

Differential serum metabolites and their enriched pathway. (A) Volcano plot showed the final metabolites with significant difference (P < 0.05, fold change >1.2 or <0.83 by t-test, and VIP > 1 calculated via PLS-DA) between tinnitus and control groups. The horizontal axis represents the log₂(fold change). The vertical axis represents the -log₁₀(P-value) obtained by t-test for the metabolite abundance in the two groups. Each point represents a metabolite, with red indicating a significant upregulation in the tinnitus group, blue indicating a significant downregulation, and gray indicating no significant difference. (B) GSEA and the Molecular Signatures Database are used for gene enrichment analysis. The top 30 significant pathways are shown in the plot, with the horizontal axis representing the NES value of the metabolite set and the vertical axis representing the pathway’s name. The color represents the NOM P-values. A significant difference between the two groups was determined by |NES| > 1 and NOM P-values of <0.05. (C) Scatter plot shows the KEGG pathway enriched by differential serum metabolites. The horizontal axis typically represents the enrichment score of the pathway. The vertical axis represents the name of the pathway. The size of the points represents the number of metabolites enriched, while the color indicates the P-value. (D) KEGG enrichment bar plot. The horizontal axis represents the number of metabolites enriched in the metabolic pathway. The vertical axis represents the name of the metabolic pathway. The length of each bar corresponds to the number of metabolites, and the color indicates the KEGG classification of the metabolic pathway.

Fig 6

Correlation analysis results between gut bacteria and serum metabolites. (A) The Spearman’s correlation coefficient for differential gut microbiota and serum metabolites between the two groups. The color represents the correlation coefficient (rho) in Spearman’s analysis. “*” is labeled for P < 0.05. “**” is labeled for P < 0.005. (B) Correlation analysis between tinnitus-related metabolic modules and bacterial modules generated by WGCNA. The left and bottom panels show associations between clusters and tinnitus. Blue indicates that modules enrich in tinnitus group, and pink indicates depletion. The right panel shows associations between the metabolomic clusters and bacterial clusters. Pearson’s test. The heatmap color represents the correlation coefficients, and the labels for the P-value are shown on the right. (C) Provided detailed information on partial gut bacteria and serum metabolites in tinnitus-related clusters generated by WGCNA. FDR-corrected P-value <0.05. Wilcoxon rank-sum test. (D and E) Inner pairwise comparisons of tinnitus-related metabolic clusters were shown, with a color gradient denoting Pearson’s correlation coefficient. Tinnitus-related bacteria clusters and specific gut microbiota are related to each metabotypes by Mantel tests. Edge width corresponded to the Mantel’s r statistic for the corresponding metabotypes, and edge color denoted the statistical significance.

Fig 7

Diagnostic performance of the predictive model. (A) Mean decrease in Gini of the selected variables. The panels display the selected alpha-diversity index, serum metabolites, and gut microbiota from left to right. The horizontal axis represents MeanDecreaseGini, which calculates each feature’s average decrease in the Gini Index when building a decision tree. Specific features are shown in the vertical axis. (B) The performance of the RF models were evaluated by calculating the AUC of the ROC curve.