Associations between geography, host genetics, and the skin microbiome
There were 845 microeukaryotic OTUs and 303 bacterial OTUs recovered across all samples after filtering and rarefaction. Microeukaryotic diversity was positively correlated with bacterial diversity across all samples (number of OTUs; Spearman’s ρ = 0.25, p < 0.001). Mantel tests revealed significant positive associations between geographic distance and beta diversity in both eukaryotic and bacterial microbes: populations that were geographically closer showed significantly more similar microbiome community structure (eukaryotic taxa: r = 0.18, p < 0.05; bacterial taxa: r = 0.40, p < 0.01, Fig. 2). Neither MHC IIB genetic distance nor neutral genetic distance (from microsatellite data published in Duryea et al. 2015) were associated with microbiome community structure (p > 0.1 for all Mantel tests between 18S or 16S beta diversity and FST matrices for MHC IIB or microsatellites).
In the 16S bacterial dataset, Proteobacteria were dominant across all samples, both by number of OTUs and sequence reads (Fig. 3A&B). Proteobacteria also formed the core bacterial microbiome across samples (Fig. 3C). Among the eukaryotic microbiota, fungi were dominant by both number of OTUs and sequence reads (Fig. 3D&E). No core group of eukaryotic taxa was recovered, though some fungal OTUs were found in approximately 50% of samples (Fig. 3F). These common fungal OTUs included Ascomycota, Basidiomycota, and unidentified fungi.
Bacteria diversity was similar across site types (Mann-Whitney tests, p > 0.05), but island frogs had fewer eukaryotic OTUs than mainland frogs (85.5 OTUs on average on islands vs. 110.5 on average on the mainland; Mann-Whitney test, U = 2,604, p = 0.001). However, site type was associated with variation in composition of both bacteria and microeukaryotes (Fig. 4A&B). Eight bacterial groups were significantly associated with site type: Cyanobacteria and Proteobacteria were statistically associated with coastal mainland sites, while six bacterial groups were statistically associated with island sites. Among the microeukaryotes, Rhizaria, Nucleariids, Ichthyosporeans, and Apusozoans were statistically associated with coastal sites while Fungi and Apicomplexans were statistically associated with island sites.
While the number of bacterial OTUs was not significantly different between MHC IIB homozygotes and heterozygotes (t-test, p > 0.05), MHC IIB homozygotes possessed fewer microeukaryote OTUs than MHC IIB heterozygotes (16.6 on average on homozygotes vs. 29.7 on heterozygotes; Mann-Whitney U, U = 3,729.5, p < 0.01). However, this result could be confounded by differences across island and mainland sites: both microbial community composition (Fig. 4A&B) and the number of MHC IIB heterozygotes and homozygotes (Table 1) vary across site types. Therefore, the analysis of microeukaryote diversity against MHC genotypes was repeated on a subset of the data only including individuals from mainland sites. MHC IIB homozygotes possessed significantly fewer microeukaryote OTUs than heterozygotes when only mainland frogs were included in the analysis (19.2 on average on homozygotes vs. 39.1 on heterozygotes; Mann-Whitney U, U = 681.5, p < 0.01).
Microbiome community composition varied between MHC IIB heterozygotes and homozygotes for both bacteria and microeukaryotes when compared with null expectations based on genotype randomizations. MHC IIB heterozygotes hosted significantly more unidentified Bacteria, Bacteroidetes, and Firmicutes, but fewer Actinobacteria and Proteobacteria OTUs than homozygotes (Fig. 4C). In terms of microeukaryotes, MHC IIB heterozygotes hosted significantly more OTUs belonging to the Ciliates, Rhizaria, and Stramenopiles, but significantly fewer Fungi and Algae OTUs than homozygotes (Fig. 4D).