Materials and methods 

Ethical statement 

Tissue sampling and fish manipulation was performed under the Guidelines for the Use of Fishes in Research (Nickum 2004) and the Care and Use of Fish in Research, Teaching and Testing (Canadian Council on Animal Care 2005). 

Strain isolation, cultivation, and characterization 

Three mycobacterial strains (myc161, myc162 and myc182) were isolated from S. salar and one (myc151) was isolated from O. kisutch, both of which were obtained from freshwater farms located in the IX and X regions of Chile between 2015 and 2018 (Table 1). The bacterial isolates were recovered following the method described by Zerihun et al. (2011a, b). Briefly, bacteria were recovered from the livers of infected fishes in which grey-white nodules were present. Pieces of infected tissue were carefully disaggregated and suspended in sterile phosphatebuffered saline (PBS). Volumes of 1 mL of serial dilutions (ranging from 10–5 to 10–7 of the initial solution concentration) were plated on tryptic soy agar medium (TSA; Becton Dickinson, USA), and plates were incubated at 25 8C for 5 to 10 days (Austin and Austin 2016). White creamy, brilliant, slightly convex colonies were subsequently transferred to modified Anacker and Ordal0 s agar (MAOA) medium as previously reported (Zerihun et al. 2011b; Aro et al. 2014). Growth sensitivity to temperature was evaluated by culturing the isolates at 16, 25, and 37 C for 5 days on solid TSA, MacConkey agar, brain and heart infusion agar (BHI), MAOA, blood agar (BA), and TSA supplemented with 5% NaCl. Biochemical characteristics of the strains were determined using the Analytical Profile Index (API) 20E system (BioMe´rieux, France) following the manufacturer’s instructions

Histopathologic analysis 

Tissue samples from the gills, heart, liver, kidney, spleen, stomach, intestines, and pyloric caeca were collected from euthanised fish and fixed in 10% buffered formalin to prepare them for analysis. The fixed tissues were processed as previously described (Prophet 1992), and paraffin-embedded samples were sectioned into 3- to 4-lm blocks. The sections were mounted on glass slides and stained with haematoxylin and eosin according to previously described protocols (Fischer et al. 2008). Microscopic examination was carried out using a Leica DM2000 microscope (Leica, Germany) and visualized utilizing LAS software V3.3–2019 (Leica).

DNA isolation and sequencing 

Bacterial DNA extraction was performed using a modified protocol consisting of phenol chloroform and tissue homogenization using glass beads (Anahtar et al. 2016). To facilitate bacterial wall rupture, enzymatic digestion was performed with lysozymes at 37 C for 1 h, followed by Proteinase K incubation at 55 C overnight prior to bead-beating homogenization (Gill et al. 2016). DNA integrity was analysed using 1% agarose gel electrophoresis, and nucleic acid quantification was performed using a NanoDrop Lite (Thermo Scientific, USA). To confirm that the isolates belong to the Mycobacterium genus, a PCR assay was performed using the 27F and 1492R primers for the amplification of the 16S rRNA gene (Weisburg et al. 1991). PCR products were analysed using Sanger sequencing with an ABI PRISM 3500 XL sequencer (Applied Biosystems, USA), and the obtained sequences were used for BLAST analysis (Altschul et al. 1990). For whole-genome sequencing, 2 lg DNA was sequenced by the Novogene Corporation Inc. (Sacramento, USA) on an Illumina Novaseq 6000 platform using 2 9 150 paired-end. The raw data reads obtained from the Illumina platform have been deposited at the ENA-EMBL database under the accession number PRJEB38737 (Table 2).

Quality control, assembly, and genome annotation 

Raw sequences generated in this study and those from a previous report (Behra et al. 2019) were preprocessed using Trimmomatic v0.39 to remove the leading and trailing bases with quality \28 and drops reads \120 bases long. Bases with average quality scores below 20 were trimmed (Bolger et al. 2014). The filtered reads were mapped against the Msal DSM43276 genome using SMALT aligner v0.7.6 available at the Sanger Institute (https://www.sanger. ac.uk/science/tools/smalt-0). SAMtools and BCFtools (http://www.htslib.org) were used to generate the consensus FASTA sequences (Li et al. 2009). The genomes were annotated using Prokka v1.12 (Seemann 2014), and the core- and pan-genomes were assessed using Roary v3.12.0 with a minimum percentage identity of 95% for blastp (Page et al. 2015). The core- and pan-genome were drawn using the script roary2svg.pl as described in the Roary pan-genome pipeline (https://sanger-pathogens.github.io/Roary/).

Comparison of Mycobacterium genomes 

Full-length 16S rRNA sequences from our four newly sequenced genomes and from 30 additional mycobacterial genomes retrieved from the NCBI database (Supplementary Table 1) were extracted using metaxa2 v2.2.1 (Bengtsson-Palme et al. 2015), and phylogenetic dendrograms were constructed using the Decipher package (Wright 2016) in R environment (R Core Team 2013). The full-length rpoB gene sequences obtained from the Prokka annotation were used for phylogenetic classification. The evolutionary distance at the species level was assessed using average nucleotide identity (ANI) analysis of the homologous genomic regions, and the 28 publicly available Mycobacterium complete genome sequences were compared using pyani module v0.2.10 (Pritchard et al. 2016). For this purpose, genomes were aligned using the MUMmer tool v3.23 (Kurtz et al. 2004).

Functional annotation of orthologous genes 

The sequences previously obtained from the Prokka analysis were annotated against NCBI’s cluster of orthologous groups (COG) database using the web tool WebMGA (Wu et al. 2011), and orthologous clustering was subsequently performed using OrthoVenn2 (Xu et al. 2019) (https://orthovenn2. bioinfotoolkits.net/). Unique elements were retrieved for orthologous classification from OrthoVenn2, and their sequences were used for protein identification in Blastp (Altschul et al. 1990). The virulence and antimicrobial resistance genes were identified using ABRIcate v0.9.8 (https://github.com/tseemann/ abricate) by means of the NCBI Bacterial Antimicrobial Resistance Reference Gene Database (Feldgarden et al. 2019), Resfinder (Zankari et al. 2012), and the Virulence Factors of Pathogenic Bacteria database (VFDB) (Chen et al. 2016).