Prediction of anastomotic insufficiency based on the mucosal microbiome prior to colorectal surgery: a proof-of-principle study

Abstract

Anastomotic leakage (AL) is a potentially life-threatening complication following colorectal cancer (CRC) resection. In this study, we aimed to unravel longitudinal changes in microbial structure before, during, and after surgery and to determine if microbial alterations may be predictive for risk assessment between sufficient anastomotic healing (AS) and AL prior surgery. We analysed the microbiota of 134 colon mucosal biopsies with 16S rRNA V1-V2 gene sequencing. Samples were collected from three location sites before, during, and after surgery, and patients received antibiotics after the initial collection and during surgery. The microbial structure showed dynamic surgery-related changes at different time points. Overall bacterial diversity and the abundance of some genera such as Faecalibacterium or Alistipes decreased over time, while the genera Enterococcus and Escherichia_Shigella increased. The distribution of taxa between AS and AL revealed significant differences in the abundance of genera such as Prevotella, Faecalibacterium and Phocaeicola. In addition to Phocaeicola, Ruminococcus2 and Blautia showed significant differences in abundance between preoperative sample types. ROC analysis of the predictive value of these genera for AL revealed an AUC of 0.802 (p = 0.0013). In summary, microbial composition was associated with postoperative outcomes, and the abundance of certain genera may be predictive of postoperative complications.

Figure 1

Graphical abstract of the study.

Figure 2

Microbial stability at different locations and time points. (A) Dendrogram of all sample underlying a Bray–Curtis resemblance measurement at Phylotype level. Sample of AS and AL displayed in blue and red, respectively. (B) Paired comparison of Bray–Curtis-Similarity of different groups. (C) Multivariate analysis between different groups underlying a Bray–Curtis resemblance measurement at genus level (ANOSIM displayed in blue, PERMANOVA displayed in black).

Figure 3

Reduced microbial diversity following surgery. (A) Dominance plot for all sample classified for their timepoint. Comparison of diversity measurements (B) Pilous evenness, (C) Simpson Index, (D) Shannon diversity and (E) Species richness.

Figure 4

Microbial differences between patients with and without AL. PCO underlying a Bray–Curtis resemblance measurement at genus level displaying groups (A) and with bubble-plot and vector overlay representing the abundance of genera across the sample (B). (C) Comparison of the average relative abundance of bacteria in AS and AL across all timepoints. (D) ROC-analysis for the occurrence of AL based on the abundance of certain bacterial genera. (E) Violin-plot of the score-index applied in the ROC-analysis, with the best suited cut-off displayed as dashed line (95% CI). (F) Comparison of the genus Prevotella between patients with and without stoma. (G) Timepoint specific relative abundance of bacteria in AS and AL. Arrow displays statistically significant trend over time in one group (trendyspliner-Test). Bracket displays statistically significant differences between two groups over the full-time course (permuspliner-Test).

Figure 5

Phylotype distribution of Enterococcus, Prevotella and Phocaeicola. Type strains from the platform NCBI shown with black background. The abundance of the phylotypes is displayed by the diameter of the black circles. Phylotype belonging to abundant species are labeled. Phylotype in heatmap follow the dendrogram order.