Bacterial DNA extraction and sequencing
We extracted total DNA from feces using a ZymoBIONICS DNA kit and sent DNA extractions to the University of Connecticut Microbial Analysis, Resources and Services for sequencing with an Illumina MiSeq platform and v2 2x250 base pair kit (Illumina, Inc.). We also sequenced a laboratory extraction blank to control for kit contamination. We conducted bacterial inventories via amplification of the V4 region of the 16S rRNA gene using primers 515F and 806R and with Illumina adapters and dual indices (Kozich et al., 2013). We used the DADA2 (v. 1.22.0) pipeline (Callahan et al., 2016) in R (v. 4.2.0) to process sequence data. After quality assessment, we trimmed sequences to remove low quality read areas and chimeric reads. We classified amplicon sequence variant (ASV) taxonomies using RDP’s Naive Bayesian Classifier (Q. Wang et al., 2007) with the Silva reference database (v. 138.1) (Quast et al., 2012). After classification, we removed sequences identified as chloroplast and mitochondria from the dataset. We identified and removed likely bacterial contaminants with the package decontam (Davis et al., 2018) in R using the kit extraction blank that was processed in parallel with the other samples as a control. Sequences were aligned using the DECIPHER package (v. 2.22.0) in R (Wright, 2015), and a generalized time-reversible maximum likelihood tree of the remaining ASVs was constructed with the phangorn package version 2.9.0 (Schliep, 2011). The ASV table, taxonomic information, phylogeny, and sample metadata were joined for bacterial community analyses using the package phyloseq (McMurdie & Holmes, 2013). We filtered the feature table to retain samples with at least 1,500 total reads which reduced the dataset to 58 samples containing 2,526 unique ASVs. The resulting data set had an average of 51,056 ± 6,178 reads per sample (min: 1,732; max: 294,720). For alpha and beta diversity analyses, we rarefied samples to the sample with the lowest read count (1,732). The filtered dataset contained 1,448 ASVs after random subsampling.