Ascitic Bacterial Composition Is Associated With Clinical Outcomes in Cirrhotic Patients With Culture-Negative and Non-neutrocytic Ascites

Abstract

Ascites bacterial burden is associated with poor clinical outcomes in patients with end-stage liver disease. However, the impact of ascitic microbial composition on clinical course was still not clear. In this study, the ascitic microbiota composition of 100 cirrhotic patients with culture-negative and non-neutrocytic ascites were researched with 16S rRNA pyrosequencing and enterotype-like cluster analysis. Results: By characterizing the ascitic microbial composition, two distinct microbial clusters were observed, Cluster 1 (86 patients) and Cluster 2 (14 patients). Cluster 1 showed lower microbial richness than Cluster 2. At the phylum level, Cluster 1 had greater abundance of Bacteroidetes and Firmicutes, but less abundance of Proteobacteria and Actinobacteria than Cluster 2. At the family level, family Bacteroidales S24-7 group, Prevotellaceae, Lachnospiraceae, Lactobacillaceae, Rikenellaceae, and Vibrionaceae were found over-represented in Cluster 1. And family Acetobacteraceae, Erysipelotrichaceae, Rickettsiaceae, and Streptococcaceae were found enriched in Cluster 2. The levels of plasma cytokine IL-17A, IL-7, and PDGF-BB were found significantly higher in Cluster 1 than in Cluster 2. There were four OTUs closely correlated with plasma cytokines, which were OTU 140 and OTU 271 (both from Bacteroidales S24-7 group), OTU 68 (Veillonellaceae), and OTU 53 (Helicobacteraceae). Patients from Cluster 1 showed significant higher short-term mortality than patients from Cluster 2. Conclusion: Our study demonstrated that the microbial composition of culture-negative and non-neutrocytic ascites in cirrhotic patients is associated with short-term clinical outcomes. The results here offer a rational for the identification of patients with high risk, and provide references for selective use of prophylactic methods.

Keywords: ascitic fluids; bacterial translocation; cytokines; end-stage liver disease; microbiome.

Figure 1

Two clusters were observed in ascitic fluids microbiota. (A) The principal coordinate analysis of the Jensen-Shannon distance generated from the OTU-level relative abundance profiles. Samples are colored by clusters identified by the partitioning around medoids clustering algorithm. Dark blue, Cluster 1; red, Cluster 2. (B) Two clusters were supported with the highest Calinski-Harabasz (CH) pseudo F-statistic value, as the optimal number of clusters. (C) Boxplot comparison of the number of observed OTUs between Cluster 1 and Cluster 2. (D) Boxplot comparison of the Chao1 index between Cluster 1 and Cluster 2. **p < 0.01 based upon Mann-Whitney U-tests with Benjamini & Hochberg correction.

Figure 2

Compositional analysis of ascitic microbial clusters. (A) Pie chart comparison of bacterial phyla represented in two clusters (upper: Cluster 1, lower: Cluster 2). (B) Histograms showing differentially enriched bacterial classes between Cluster 1 and Cluster 2. Blue histograms, Cluster 1; red histograms, Cluster 2. (C). LEfSe analysis revealed differentially enriched bacterial families associated either with Cluster 1 (blue) or Cluster 2 (red). (D) LEfSe analysis revealed differentially enriched bacterial genus associated either with Cluster 1 (blue) or Cluster 2 (red).

Figure 3

Plasma cytokine levels with statistical difference between Cluster 1 and Cluster 2. Blue histograms, Cluster 1 (n = 68); red histograms, Cluster 2 (n = 10). Significance values are indicated: * p < 0.05 **p < 0.01 based upon Mann-Whitney U-tests with Benjamini & Hochberg correction.

Figure 4

Correlation analysis of ascitic bacterial OTUs and plasma cytokines. Spearman rank correlation was performed. Only correlations with a coefficient r > 0.40 are displayed. The colors of OTU nodes and lines indicate bacterial families as labeled on the lower right.

Figure 5

Ninety-days Kaplan–Meier survival curves for different ascitic microbiota clusters. P-values were calculated by the Log Rank-test.