Statistical Analyses
We used QIIME 2 version 2022.8 ; https://qiime2.org) to perform all downstream statistical analyses. We first implemented thecore-metrics-phylogenetic function with sampling depth set to 2500 and a midpoint rooted phylogeny created with the alignmentand phylogeny functions in order to calculate multiple measures of diversity. Alpha diversity metrics included observed features to measure richness , Pielou’s evenness to measure species equitability , and the Shannon Diversity Index as a combined measure of richness and evenness . Beta diversity metrics included Bray-Curtis Dissimilarity to measure differences in microbial abundance and the Jaccard Index to measure differences in species presence . We calculated these five diversity measures in order to tease apart which aspects of diversity (e.g. , bacterial presence vs. abundance) differed between our groups of interest. We performed multivariate significance testing using the longitudinal anova function for alpha diversity measures and the diversity adonis function (multivariate PERMANOVA; for beta diversity measures. These analyses contained the following variables: island, species, their interaction, and sequencing plate. We visualized beta diversity differences using principal coordinate analysis implemented through the EMPeror plug-in . All figures were created using the qiime2R and ggplot2 R packages.
We assigned taxonomy to representative ASVs by using a Naïve Bayes classifier trained on Greengenes 13_8 reference sequences , as described previously . We then characterized the taxonomic composition of island fox and island spotted skunk gut microbiomes using thetaxa barplot function and identified differentially abundant taxa within and between species using analysis of composition of microbiomes (ANCOM; as implemented through the ancom function. ANCOM overcomes statistical hurdles associated with compositional datasets by calculating pairwise log-ratios between all taxa present within sampling groups. It identifies which taxa consistently differ in abundance between groups by calculating the test statistic, W, as the number of times the null hypothesis of “no difference” is violated for a particular taxon. As numerous differences are expected to arise between host species, we performed this analysis at the taxonomic level of class.