Metabarcoding
The bioinformatic pipeline applied is described in Brandt et al. (2019). Primers and leftover adapters were removed with the program Cudadapt (Martin, 2011). The basic pipeline was then based on DADA2 using stringed error correction (Callahan et al., 2016), including fragment size selection with an expected length of 300-500 bp for both 16SV4 and 18SV1V2, followed by a chimera removal step. Sequences were then clustered into operational taxonomic units (OTUs) using the program swarm2 (Mahé, Rognes, Quince, de Vargas, & Dunthorn, 2015) with an iterative local threshold (d) of 4 for 18SV1V2 and 1 for 16SV4. The ribosomal sequences were taxonomically assigned against the Silva database (release 132; Pruesse et al., 2007) with the MegaBlast algorithm (Camacho et al., 2009).
A decontamination step was applied after clustering, using extraction and PCR negative controls using decontam (Davis, Proctor, Holmes, Relman, & Callahan, 2018), by the prevalence method with a threshold of 0.8. Finally, all OTU counts were adjusted using an R-based script (Wangensteen, Palacín, Guardiola, & Turon, 2018) to renormalize potential tag switches during library preparations (Schnell, Bohmann, & Gilbert, 2015).