Page 97 - Read Online

Ambros et al. Microbiome Res Rep 2023;2:34 https://dx.doi.org/10.20517/mrr.2023.18 Page 5 of 19

change in the future.

Prophage annotation
Detection, first annotation and completeness evaluation were performed with the web tool “Phage search
tool enhanced release” (PHASTER) [36,37] . For the lysogenic L. curvatus strains with inducible prophages,
PHASTER-derived annotations were additionally checked manually by BLASTing translated amino acid
sequences against UniProt’s virus database (uniprot.org; BLASTp; E-Threshold: 10; Matrix: Auto; Filtering:
None; Gapped: yes) and replaced whenever reasonable. Prophage visualisation and gene colour allocation
were performed with the program SnapGene Viewer (Insightful Science; snapgene.com).

Isotype and anticodon prediction of phage tRNA genes (derived from NCBI annotation) were performed
via tRNAscan-SE version 2.0 [38,39] . Codon usage propensities were analysed via CLC Main Workbench
version 8.1.4 (Qiagen). The secondary structure predictions derived from tRNAscan-SE 2.0 and CLC Main
Workbench version 8.1.4 (Qiagen) were checked manually.

The nucleotide sequences of potential methyltransferase genes were BLASTed (BLASTN) against the
REBASE nucleotide sequences database using default parameters (date: 23rd February 2023) at
[40]
rebase.neb.com to verify a correct annotation and type of methylation.
Assigned (pro)phage names consist of their host strain name and a capitalised “p” for “(pro)phage”, as well
as a number to distinguish different intact (pro)phages within one host. According to the “International
Committee on Taxonomy of Viruses” (ICTV) , all (pro)phage names were written in non-italic, even
[41]
when including host identifiers (e.g., L. curvatus phage TMW 1.591 P1).

The position of each prophage in Figure 1 is based on a whole genome alignment pairwise comparison
(prophages with higher sequence similarity were depicted next to each other), using CLC Main Workbench
version 8.1.4 (Qiagen). The phylogenetic tree in Figure 1 is depicted as a cladogram. Sequence similarity
bars between each prophage are based on BLASTn, using Easyfig 2.2.5 with the following settings: Min.
[42]
length: 0; Max. e-value: 0.001; Min identity value: 0; Outline blast hits in black: yes; Filter small blast hits/
annotations: yes. Genome alignment: Left side.

Phylogenetic analysis
Using default settings, alignments were created with CLC Main Workbench version 8.1.4 (Qiagen).
[43]
Construction of phylogenetic trees was performed using MEGA X version 10.2.6 with the neighbor-
joining method (distance measure: Maximum Composite Likelihood method). For the bootstrap analysis,
using the Jukes-Cantor model, 1,000 replicates were used. All marker genes were extracted with the “extract
annotation” feature of CLC Main Workbench and checked manually. The integrase outgroups used
(lactobacilli phages infecting other species) are listed with their respective accession number in the
supplemental material.

RESULTS
Prophage prediction in L. curvatus
To analyse the prophage incidence and distribution within L. curvatus, 45 genomes were searched for
prophages with PHASTER (phaster.ca) [36,37] and all identified phage-related sequences were classified by it as
intact, questionable, or incomplete [Supplementary Table 2]. 206 sequences were detected in total; 47 were
classified as intact (with zero to three sequences per strain), 38 as questionable (zero to six sequences per
strain) and 121 (zero to eleven sequences per strain) as incomplete. All strains harboured prophages of some