Assembly

Read data were processed using ATROPOS³² v.1.1.5 (parameters: -q 15 --minimum-length 90), after which the reads were mapped (BOWTIE2³³ v.2.3.4, parameters: --very-sensitive) against a draft bank vole genome (GCA_001305785.1) to filter out reads (from *.bam files) derived from the host (SAMTOOLS v.1.4 view, parameters: view -f 12 -F 256) (https://www.htslib.org/doc/samtools.html). Assembly of metagenome read data to obtain draft bacterial genomes followed the approach used to recover draft genomes from the TARA oceans metagenomics data³⁴. Individual samples were assembled using MEGAHIT³⁵ v.1.1.1-2 (parameters: default) to reduce memory requirements and in attempt to avoid bubbles (unresolvable branches) due to genetic diversity among strains. Using MEGAHIT, we assembled a total of 1,057 million paired end reads into 4,721,549 primary contigs (5,916,721,003 bp). These primary contigs were filtered to retain only those contigs ≥ 2 kbp in length, which were then passed through CD-HIT-EST³⁶ v.4.7.0 (parameters: -c 0.99 -n 11 -M 0) to merge the completely overlapping contigs (at 99% identity); this reduced set of primary contigs was then co-assembled using MINIMUS2 in AMOS³⁷ v.3.1.0 (parameters: -D REFCOUNT = 0 OVERLAP = 100 MINID = 95) to combine overlapping contigs. After this procedure, we were left with 171,806 secondary contigs (39,092 contigs and 132,714 singletons) for binning.