After Surgically Induced Remission, Ileal and Colonic Mucosa-Associated Microbiota Predicts Crohn's Disease Recurrence

Abstract

Background & aims: Investigating the tissue-associated microbiota after surgically induced remission may help to understand the mechanisms initiating intestinal inflammation in Crohn's disease.

Methods: Patients with Crohn's disease undergoing ileocolic resection were prospectively recruited in 6 academic centers. Biopsy samples from the neoterminal ileum, colon, and rectosigmoid were obtained from colonoscopies performed after surgery. Microbial DNA was extracted for 16S rRNA gene sequencing. Microbial diversity and taxonomic differential relative abundance were analyzed. A random forest model was applied to analyze the performance of clinical and microbial features to predict recurrence. A Rutgeerts score ≥i2 was deemed as endoscopic recurrence.

Results: A total of 349 postoperative colonoscopies and 944 biopsy samples from 262 patients with Crohn's disease were analyzed. Ileal inflammation accounted for most of the explained variance of the ileal and colonic mucosa-associated microbiota. Samples obtained from 97 patients who were in surgically induced remission at first postoperative colonoscopy who went on to develop endoscopic recurrence at second colonoscopy showed lower diversity and microbial deviations when compared with patients who remained in endoscopic remission. Depletion of genus Anaerostipes and increase of several genera from class Gammaproteobacteria at the 3 biopsy sites increase the risk of further recurrence. Gut microbiome was able to predict future recurrence better than clinical features.

Conclusions: Ileal and colonic mucosa-associated microbiome deviations precede development of new-onset ileal inflammation after surgically induced remission and show good predictive performance for future recurrence. These findings suggest that targeted microbial modulation is a plausible modality to prevent postoperative Crohn's disease recurrence.

Keywords: Crohn’s Disease; Disease Recurrence; Microbiota; Prediction.

Figure 1.. Phenotypic features explaining microbiota variance in postoperative Crohn’s disease.

a. Inferred variance (R² as %) explained by each clinical variable as determined by the PERMANOVA method on amplicon sequence variance (ASV) after Benjamini-Hochberg correction (* FDR < 0.05). All the samples (n = 944 biopsy samples) were included; therefore, permutations were blocked within subject to account for repeated measures. b. Principal coordinate analysis (PCoA) with Bray-Curtis distance demonstrating differences between samples with and without ileal recurrence based on Rutgeerts score. Principal coordinates (PCo) 1 to 3 are shown.

Figure 2.. Microbial diversity of samples obtained from patients in surgically induced remission at first colonoscopy (Rutgeerts score i0-i1) stratified according to the future development of endoscopic recurrence.

Upper and middle panels: Comparison of Chao1 and Shannon index at each biopsy site (* p-vale < 0.05) adjusted for Rutgeerts score at first colonoscopy (i1 vs i0). Lower panels: Beta diversity using Bray-Curtis dissimilarity. PERMANOVA test comparing samples with and without remote endoscopic recurrence (Neoterminal ileum: R2 = 1.7%, p-value = 0.02; colon: R2 = 1.2%, p-value 0.2; and rectosigmoid: R2 = 1.2%, p-value = 0.1). Samples obtained from patients with postoperative exposure to antibiotics and those from colonic areas with endoscopic inflammation were removed for these analyses.

Figure 3.. Microbial taxa associated with future endoscopic recurrence in Crohn’s disease patients in postsurgical remission.

a) Microbial differential abundance analysis in samples in endoscopic remission (Rutgeerts score i0-i1) at first colonoscopy whether they developed endoscopic recurrence (Rutgeerts score ≥ i2) or remained in endoscopic remission (Rutgeerts score i0-i1) at second colonoscopy. Each biopsy site was analyzed separately and coefficients are based on MaAsLin2 algorithm centered log-ratio-transformed relative abundances adjusted by Rutgeerts score at first colonoscopy (i0 vs i1). Taxa with q-value < 0.25 are shown (* q-value < 0.05). b) Area under the curve for random Forest model to predict recurrence at second colonoscopy considering only clinical data (Rutgeerts score at first endoscopy, postoperative anti-TNF use and smoking status), only microbiome data and microbiome plus clinical data. Results from validation set (40% of the cohort) are shown. AUC, area under the curve.