Bacterial community composition and diversity
Across all samples, bacteria from 37 phyla were detected. Firmicutes, Proteobacteria, Bacteroidetes, and Actinobacteria composed 91.13% of the total reads; 5.8% of the reads belonged to the 33 remaining phyla and 3.07% of reads did not align to any known bacterial phyla (Figure 1A). Clostridia (Phylum Firmicutes), Gammaproteobacteria(Phylum Proteobacteria), and Bacteroidia (Phylum Bacteroidetes) were the most abundant classes, representing 70.16% of all reads. The mean abundance of most phyla and classes differed between initial sampling in The Bahamas and subsequent samplings in Michigan (Figure 1B, Table S2). The birds shifted from a Firmicutes dominated microbiome in The Bahamas (mean abundance per individual 39.82% [SD, ±13.97%]) and Michigan following arrival (38.12% [SD, ±16.41%]) to Proteobacteria as the most abundant phylum in the second Michigan recapture period (47.07% [SD, ±27.90%]). Bacteroidetes and Actinobacteria were also proportionally more abundant in The Bahamas than in the second Michigan recapture period. Notably, Cyanobacteria represented 1.91% (SD, ±5.93%) of the total microbiota in The Bahamas, but decreased to 0.05% (SD, ±0.23%) by the second recapture period in Michigan.
Alpha diversity was not significantly affected by year, host age or host sex (Table S3). However, the three sampling periods significantly differed (Type III ANOVA with Satterthwaite’s method; Observed richness: F2,116.34 = 14.76, P <0.0001; Shannon Diversity: F2,126.91 = 29.22, P<0.0001). All Bonferroni corrected pairwise comparisons on the fitted values from the linear model were significantly different from each other (Observed: Bahamas vs. each recapture period both P<0.0001, first vs. second recapture period, P = 0.002; Shannon Diversity: all comparisons: P <0.0001). Birds in The Bahamas showed increased bacterial diversity compared to either recapture period in Michigan, demonstrated through a comparison of all samples (Figure 2A) as well as with paired sampling of the same individuals (Figure 2B). In the birds sampled in triplicate, alpha diversity varied between first and second Michigan recaptures (Figure 2C). A comparison of tagged and randomly captured birds in the first Michigan sampling period of 2018 revealed no significant differences in alpha diversity (pairwise t-test; Observed: p = 0.13, Shannon Diversity: p = 0.22). We observed decreased alpha diversity through the first three days after arrival in Michigan followed by a slight increase over the following six days (Figure 3).
Our results indicate that beta diversity was not significantly affected by age or sex of the birds within the full dataset or individual sampling periods (Table 2), with the exception of age in the second Michigan resampling period (unweighted UniFrac: PERMANOVA p = 0.0128, PERMDISP p = 0.2213). Community composition of the microbiota significantly differed by year in the full dataset and at each sampling period (Table 2, Figure 4A). Additionally, our PERMANOVA results suggest that sampling period exerted a significant effect on the microbiota composition (Bray-Curtis: p = 0.0002, unweighted UniFrac: p = 0.0001), though the significant unweighted UniFrac result can be explained through variation in spread of the sample composition, rather than with significantly different centroids such as with the Bray-Curtis dissimilarity matrix (PERMDISP; Bray-Curtis p = 0.7104, unweighted UniFrac p = 3.71e-6). This indicates that although the abundances of microbiota are significantly different during sampling periods, the taxonomic variation of bacterial lineages present are not. The effect of sampling period on the gut microbiota explained 2.5% and 2.4% of the variation in microbiota composition for Bray-Curtis and unweighted UniFrac respectively. Taken together, all variables tested (Sampling period, Year, Sex, Age) explained less than 5% of the total variation in the microbiome (Bray-Curtis: 4.91%, unweighted UniFrac: 4.6%). No consistent changes were observed in the beta diversity of the birds sampled in triplicate (Figure 4B, Figure S3).
Across the three sampling periods, six ASVs and six genera were identified by ANCOM as differentially abundant. GeneraSolirubrobacter , Nocardioides , and Rubrobacter were significantly associated with The Bahamas, Endobacter andCandidatus Hamiltonella associated with first recapture period in Michigan, and Serratia with second recapture period. ASVs within families Beijernickiaceae and Pseudomonadaceae were significantly more abundant during first recapture in Michigan and families Enterobacteriaceae , Beijernickiaceae , andSynergistaceae contained ASVs associated with the second recapture in Michigan. No ASVs were statistically associated with The Bahamas.