Sequence Processing
We processed raw sequence data with the Quantitative Insights Into Microbial Ecology (QIIME2 version 2019.1) pipeline (Caporaso et al. 2010; Bolyen et al. 2019). In QIIME2, following standard demultiplexing and quality filtering, we generated amplicon sequence variants (ASVs) using the Divisive Amplicon Denoising Algorithm (DADA2) (Callahan et al. 2016). DADA2 statistically infers sample sequences and implements quality control elements including exclusion of singletons, chimera removal, and sequence error elimination. We trimmed all sequences outside base pair positions 13 and 145 base pairs to remove the primers. We classified ASV taxonomies using the Silva reference database (Quast et al. 2012, version 132). We identified bacterial contaminants using a frequency-based algorithm in the R package Decontam (Davis et al. 2018). We removed contaminants and negative controls from subsequent analyses.