Statistical analyses.

Alpha and beta diversity statistics, Kruskal-Wallis test, and ordination via principal-coordinate analysis were performed using mothur. All other statistics were calculated using XLSTAT (version 2015.01.0; Addinsoft, Belmont, MA). Shannon indices and abundance-based coverage estimates were calculated to evaluate parametric and nonparametric alpha diversity. Bray-Curtis dissimilarity matrices were used for evaluation of beta diversity and ordination (48). The corr.axes command in mothur using the Spearman method was used to determine OTUs significantly affecting axis position. Analysis of similarity (ANOSIM) was used to evaluate differences in beta diversity (49), and analysis of molecular variance (AMOVA) was used to determine the significance of sample clustering on ordination plots (50). Kruskal-Wallis test was used to determine differences in relative abundances of OTUs among sample groups (51). Differences in alpha diversity and relative abundances of taxa were determined using analysis of variance (ANOVA). All statistics were evaluated with α = 0.05.