Network construction of gastric microbiome and organization of microbial modules associated with gastric carcinogenesis

Abstract

In addition to Helicobacter pylori infection, nitrosating/nitrate-reducing bacteria and type IV secretion system (T4SS) protein gene-contributing bacteria have been proposed as potential causes of gastric cancer development. However, bacterial modules related with gastric carcinogenesis have not been clarified. In this study, we analyzed gastric microbiome using the gastric mucosal samples obtained from the Hanyang University Gastric Microbiome Cohort by 16S rRNA gene sequencing. Weighted correlation network analysis was performed to construct a microbiome network and to identify microbial modules associated with gastric carcinogenesis. At the family level, 420 bacterial taxa were identified in the gastric microbiome of 83 participants. Through network analysis, 18 microbial modules were organized. Among them, two modules-pink and brown-were positively correlated with a higher-risk of gastric cancer development such as intestinal metaplasia with no current H. pylori infection (correlation coefficient [γ]: pink module, 0.31 [P = 0.004], brown module, 0.26 [P = 0.02]). At the family level, twenty-two and thirty-two bacterial taxa belonged to the pink and brown modules, respectively. They included nitrosating/nitrate-reducing bacteria, T4SS protein gene-contributing bacteria, and various other bacteria, including Gordoniaceae, Tsukamurellaceae, Prevotellaceae, Cellulomonadaceae, Methylococcaceae, and Procabacteriaceae. The blue module, which included H. pylori, was correlated negatively with intestinal metaplasia (γ = -0.49 [P < 0.001]). In conclusion, intragastric bacterial taxa associated with gastric carcinogenesis can be classified by network analysis. Microbial modules may provide an integrative view of the microbial ecology relevant to precancerous lesions in the stomach.

Conflict of interest statement

The authors declare no competing interests.

Figures

Figure 1

Sample dendrogram and trait heatmap. The dendrogram plotted by hierarchical clustering for gastric microbiome composition in 83 included participants. The heatmap presented below the dendrogram indicates clinical traits for the corresponding participants. ABCD group indicates categorization by H. pylori infection and atrophic gastritis as follows: (1) Group A: no H. pylori infection and no atrophic gastritis, (2) Group B: H. pylori infection and no atrophic gastritis, (3) Group C: H. pylori infection and atrophic gastritis with intestinal metaplasia, and (4) Group D: atrophic gastritis with intestinal metaplasia and no H. pylori infection. BMI, body mass index; PG, pepsinogen.

Figure 2

Module membership identification using dynamic tree. (A) Bacterial taxa dendrogram and module colors, (B) Clustering of module eigenvalues. In the dendrogram of panel A, each leaf, shown as a short vertical line, corresponds to a bacterial taxon. Branches of the dendrogram grouped together densely and interconnected represent highly co-occurring bacterial taxa. The highly co-occurred bacterial taxa can be classified into 18 different modules (from pink to green modules). In the clustering dendrogram of module membership, with dissimilarity based on the topological overlap, the pink, yellow, and brown modules show higher similarity to each other compared to the other modules. The gray module indicates unassigned bacterial taxa.

Figure 3

Correlation between module eigenvalue and clinical trait. Heatmap shows the correlation coefficient between module eigenvalues and clinical traits. The pink and brown modules are significantly correlated with an advanced stage of gastric carcinogenesis, or higher ABCD group (pink module, correlation coefficient [γ] = 0.31 [P = 0.004]; brown module, γ = 0.26 [P = 0.02]). These modules are also correlated with other clinical traits including age (both modules), Charlson comorbidity index (pink module only), pepsinogen I (both modules), and H. pylori infection (brown module only). ABCD group indicates categorization by H. pylori infection and atrophic gastritis as follows: (1) Group A: no H. pylori infection and no atrophic gastritis, (2) Group B: H. pylori infection and no atrophic gastritis, (3) Group C: H. pylori infection and atrophic gastritis with intestinal metaplasia, and (4) Group D: atrophic gastritis with intestinal metaplasia and no H. pylori infection, BMI, body mass index; PG, pepsinogen.

Figure 4

Visualization of the eigenvalue network representing relationships between the modules and ABCD groups. The dendrogram shows hierarchical clustering of the eigenvalues. In this dendrogram, the pink, yellow, and brown modules are closely related. The relationship between the higher ABCD group and the pink module is stronger than that among the pink, yellow, and brown modules. Heatmap shows the eigenvalue adjacency between modules and the higher ABCD group. The heatmap indicates that the higher ABCD group is positively correlated with the brown, yellow, and pink modules, while it is negatively correlated with the other modules. ABCD group indicates categorization by H. pylori infection and atrophic gastritis as follows: (1) Group A: no H. pylori infection and no atrophic gastritis, (2) Group B: H. pylori infection and no atrophic gastritis, (3) Group C: H. pylori infection and atrophic gastritis with intestinal metaplasia, and (4) Group D: atrophic gastritis with intestinal metaplasia and no H. pylori infection.

Figure 5

Visualization of full weighted networks in two modules associated with an advanced stage of gastric carcinogenesis. (A) pink module and (B) brown module. Strong co-occurrences among bacterial taxa other than type IV secretion system protein gene-contributing bacteria are identified in both the pink and brown modules (i.e., Gordoniaceae, Tsukamurellaceae, and Prevotellaceae in the pink module, and Cellulomonadaceae, Methylococcaceae, and Procabacteriaceae in the brown module). Gray zones represent the type IV secretion system protein gene-contributing bacteria. Thicker lines between bacterial taxa indicate high co-occurrence. Prefix “O_” and “C_” represent the order and class names, respectively. All taxa without a prefix are the family names.