Small intestinal microbial dysbiosis underlies symptoms associated with functional gastrointestinal disorders

Abstract

Small intestinal bacterial overgrowth (SIBO) has been implicated in symptoms associated with functional gastrointestinal disorders (FGIDs), though mechanisms remain poorly defined and treatment involves non-specific antibiotics. Here we show that SIBO based on duodenal aspirate culture reflects an overgrowth of anaerobes, does not correspond with patient symptoms, and may be a result of dietary preferences. Small intestinal microbial composition, on the other hand, is significantly altered in symptomatic patients and does not correspond with aspirate culture results. In a pilot interventional study we found that switching from a high fiber diet to a low fiber, high simple sugar diet triggered FGID-related symptoms and decreased small intestinal microbial diversity while increasing small intestinal permeability. Our findings demonstrate that characterizing small intestinal microbiomes in patients with gastrointestinal symptoms may allow a more targeted antibacterial or a diet-based approach to treatment.

Fig. 1

The duodenal microbiome is altered in patients with GI symptoms. Principal coordinate axis (PCoA) plot showing beta diversity of patients with GI symptoms (n = 126, blue) and healthy controls (n = 38, red) based on a unweighted UniFrac and b Bray–Curtis distances (p = 0.001, PERMANOVA). Alpha diversity (within subject) of patients with GI symptoms (n = 126, blue) and healthy controls (n = 38, red) based on c phylogenetic distance d observed OTUs and e Shannon diversity index metrics (rarefied to 5000 sequences; p < 0.0001, t test). Tukey boxplots show the median with IQR and 1.5 IQR whiskers

Fig. 2

Symptom index differentiates healthy controls from patients with GI symptoms. a Symptom index indicating the probability (0–1) of being classified in the symptomatic patient group determined using Random Forests classification based on the individuals’ OTU profiles. b Tukey boxplots show differences in relative abundance of the 26 OTUs among symptomatic patients (n = 126, blue) and healthy controls (n = 38, red) that significantly contribute to the Random Forest classification performance based on Boruta feature selection (boxplots show the median with IQR and 1.5 IQR whiskers). The four factors contributing to the variance in the symptom index among healthy controls (red) and symptomatic patients (blue) are c age d history of antibiotic use within 3 weeks e history of GI surgery and f PPI

Fig. 3

Dysbiosis index identifies a subset of symptomatic patients with altered microbial communities. a Classification of healthy and symptomatic patients as dysbiotic or healthy-like based on the CLOUD test, which evaluates each sample by its distance to healthy controls’ microbiome distribution. b The log-transformed CLOUD statistic for samples (or dysbiosis index (DI)) from healthy controls (n = 38, red) and symptomatic patients (n = 126, blue). Tukey boxplots show the median with IQR and 1.5 IQR whiskers. c Heatmap of significantly different genus-level taxa from healthy control and symptomatic patients that contribute significantly to the DI identified using Boruta feature selection from the random forest model. Samples classified as dysbiotic or healthy-like are indicated at the top by orange and green, respectively. Pearson correlation of DI with alpha diversity based on d phylogenetic diversity (p < 0.0001, r = −0.47, Pearson correlation), e observed OTUs (p < 0.0001, r = −0.4, Pearson correlation), f Shannon index (p < 0.0001, r = −0.36, Pearson correlation)

Fig. 4

Quantitative small bowel culture does not reflect small bowel microbial composition. Distribution of microbial communities from symptomatic patients (blue) with and without SIBO based on Aitchison distance from healthy microbial communities (red). SIBO does not correlate with microbial community composition as summarized by dysbiosis classification (p = 0.33, Fisher’s exact test). Open circles represent small bowel microbiomes from individuals who tested negative for SIBO; closed circles represent those tested positive for SIBO by aspirate culture

Fig. 5

A subset of healthy individuals consuming high-fiber diet have SIBO. a DI and b distribution based on Aitchison distance of healthy controls without SIBO (green), symptomatic patients with (red) and without (orange) SIBO, and healthy individuals consuming a high fiber diet with (blue) and without (green) SIBO. ****, q < 0.0001; n.s., q > 0.05; pairwise t- test with FDR correction

Fig. 6

Diet change is associated with changes in host physiology, microbial diversity, and metabolites. a Spearman correlation of change in alpha diversity (PD whole tree) and change in duodenal permeability measured by FITC flux across duodenal biopsy in an Ussing chamber (n = 14, p = 0.02, rho = −0.61, Spearman correlation). Association of change in alpha diversity (PD whole tree) with b postprandial bloating (n = 14, p = 0.01, Mann–Whitney test) and c abdominal pain relieved by defecation (n = 14, p = 0.07, Mann–Whitney test). Tukey boxplots show the median with IQR and 1.5 IQR whiskers. Change in d acetate (n = 11, FDR q = 0.1, Wilcoxon signed rank test) in duodenal aspirates and e acetate (n = 14, FDR q = 0.03, Wilcoxon signed rank test), butyrate (n = 14, FDR q = 0.01, Wilcoxon signed rank test), and lysine (n = 14, FDR q = 0.04, Wilcoxon signed rank test) in stool measured before and after dietary intervention using ¹H-NMR