Bugs and drugs: Predicting response to therapy

The human body is inhabited by 10 times more bacteria than the number of human cells, and there are 100 trillion bacteria in the gut alone that encode 100-fold more unique genes than the human genome.1, 2 The gut microbiome in humans consists predominantly of two phyla—Firmicutes and Bacteroidetes—which together account for about 90% of the microbiome.3 Proteobacteria and Actinobacteria contribute the remaining 10% of the luminal gut microbiome. Research over the past decade has highlighted the central role of the microbiome in intestinal inflammation in Crohn's disease (CD) and ulcerative colitis (UC). Initial research has been primarily from cross-sectional studies, comparing the microbiome of those with active CD or UC with healthy controls. These studies have revealed three broad findings. First, patients with CD or UC have a reduced microbiome diversity compared with healthy controls.4 Second, particularly in CD, there may be a depletion of bacterial species with anti-inflammatory activity such as butyrate production.5 Elegant studies have demonstrated that Faecalibacterium prausnitzii (among others), a butyrate producing Firmicute, is depleted in patients with CD. Abundance of this bacterium in the ileal resection specimen in patients with CD undergoing resection inversely correlate with risk of endoscopic recurrence. In mouse models, intragastric administration of F. prausnitzii relieves intestinal inflammation. Third, there may be an increased abundance of specific pro-inflammatory bacteria associated with specific phenotypes of Crohn's disease. Studies have demonstrated an increased occurrence of adherent invasive Escherichia coli in patients with ileal CD corresponding with the anti-microbial antibodies observed in this population.6 There is growing interest in the potential role of microbial markers to predict the complicated disease course or response to therapy. In the multicenter pediatric RISK cohort in North America, 913 children with CD were enrolled at diagnosis and prospectively followed for up to 3 years.7 A total of 9% of patients developed either stricturing (B2) or penetrating (B3) disease during the follow up. Abundance of Ruminococcus in the rectum or stool at baseline was associated with increased risk of B2 disease at follow up, while Veillonella abundance correlated with risk of B3 disease. Few studies have examined if microbial markers can predict response to therapy. A small multicenter pediatric cohort examined the outcomes of children with acute severe ulcerative colitis.8 At baseline, this population had a reduced diversity compared with healthy controls. Non-responders to intravenous steroids had significantly less microbial diversity than children who were steroid responsive. A study of 11 children initiating anti-tumor necrosis factor therapy demonstrated that microbial diversity at baseline (and similarity to healthy controls) predicted fecal calprotectin levels and treatment response 3 months after start of anti-tumor necrosis factor therapy.9 A prospective single center cohort of 85 patients with CD or UC initiating vedolizumab therapy used fecal metagenomic sequencing to predict clinical remission at 14 weeks with this treatment.10 Community alpha diversity was significantly higher in CD patients (at the species level) who achieved remission at week 14 compared with healthy controls. Pathway analysis revealed 13 pathways in CD and 5 in UC were differentially distributed at baseline in remitters compared with non-remitters, with many pathways involved in branched chain amino acid biosynthesis. A neural network model with a manually curated list of 40 microbiome variables provided the highest classifying power in separating remitters from non-remitters with an of 0.872 compared with a model comprising clinical covariates along (AUC 0.619). In summary, the microbiome is central to the pathogenesis of CD and UC, and emerging data suggest that it may be a promising tool to predict disease complications and response to therapy.