Fig. 2

The saliva and esophageal microbiota in subjects with GERD and metaplasia. Full-length 16S rRNA amplicon sequencing was performed on a PacBio platform. A Comparison of alpha diversity measures between saliva (S) and esophageal (E) samples. Significant differences between sample types (Richness: Pseudo-F: 69.7, P = 0.001, df = 92; Evenness: Pseudo-F: 48.7, P = 0.001, df = 92; Shannon’s diversity: Pseudo-F: 83.2, P = 0.001, df = 92) were observed across all measures using linear models that corrected for age, sex, PPI, BMI, reflux symptoms, and disease. B Principal coordinate analysis of Bray-Curtis resemblance matrix generated from square-root-transformed pOTU relative abundances. Distance-based linear models corrected for all variables identified significant differences in composition between sample types (Pseudo-F: 5.1, P = 0.001, df = 92). C The same principal coordinate analysis as above incorporating the relative abundance of Streptococcus pOTU1. D Esophageal species richness stratified according to disease. E Non-metric multidimensional scaling plot of Bray-Curtis resemblance matrix generated from square-root-transformed OTU relative abundances. Only esophageal samples were plotted. F Heatmap of genera and OTUs identified by LEfSe to be significantly different (LDA > 3) between normal and GERD (Additional file 1: Table S9) or normal and MET (Additional file 1: Table S10). G Mean relative abundance of Campylobacter stratified according to disease. Errors are SEM. H Significant changes in the esophageal microbiota occur by the time of diagnosis of GERD. I A significant co-exclusion relationship between the relative abundance of Campylobacter and NAPSB was present. Non-parametric correlations were identified through MINe and confirmed with Spearman’s correlation