Bioinformatic analyses
The adapter and primers from the raw sequence reads were trimmed using Cutadapt v2.10 (Martin et al., 2011). Bioinformatics pipeline was run using DADA2 v1.16 (Callahan et al., 2016) to infer amplicon sequence variants (ASVs) with single-nucleotide resolution. For quality trimming, a more relaxed filtering option was applied on the reverse reads as maxEE = c (2, 5), and the low-quality sequence tails were removed from the forward and reverse reads with truncLen = c (270, 210). Bacterial taxonomy was assigned to representative ASV sequences using DADA2 implementation of RDP naïve Bayesian classifier based on the EzBiocloud database (Yoon et al., 2017). Sequences matched to the Eukaryota, Archaea or Cyanobacteria were removed from the dataset. Sequences are available in the NCBI Sequence Read Archive (SRA) database under accession number PRJNA753257.