Phylogenomics reveal that Mycobacterium kansasii subtypes are species-level lineages. Description of Mycobacterium pseudokansasii sp. nov., Mycobacterium innocens sp. nov. and Mycobacterium attenuatum sp. nov.

Among the species Mycobacterium kansasii , seven subtypes have been previously reported based on the PCR and the restriction fragment length polymorphism of the gene hsp65 . Here, we used whole-genome sequencing to refine M. kansasii taxonomy and correct multiple inconsistencies. Average nucleotide identity (ANI) values between M. kansasii subtypes ranged from 88.4 to 94.2 %, lower than the accepted 95 – 96% cut-off for species delineation. In addition, Mycobacterium gastri was closer to the M. kansasii subtypes 1, 2, 3, 4 and 5 than M. kansasii subtype 6. The recently described species Mycobacterium persicum shared 99.77 % ANI with M. kansasii subtype 2. Consistent with the ANI results, the digital DNA – DNA hybridization value was below the 70% threshold for species delineation between subtypes and above it within subtypes as well as between subtype 2 and M. persicum . Furthermore, core-genome phylogeny confirmed the current M. kansasii species to be polyphyletic. Hence, we propose (i) Mycobacterium pseudokansasii sp. nov., replacing subtype 3, with the type strain MK142 T (=CCUG 72128 T =DSM 107152 T ), (ii) Mycobacterium innocens sp. nov., replacing subtype 5, with the type strain MK13 T (=CCUG 72126 T =DSM 107161 T ), and (iii) Mycobacterium attenuatum sp. nov., replacing subtype 6, with the type strain MK41 T (=CCUG 72127 T =DSM 107153 T ). Subtype 4 represents a new species-level lineage based on the genomic data but no strain was available. No genome sequence or strain was available for subtype 7. The proposed nomenclature will facilitate the identification of the most pathogenic subtype 1 as M. kansasii by clinicians while the new species names suggest the attenuated pathogenicity of the other subtypes.

The species Mycobacterium kansasii, a member of slowgrowing non-tuberculous mycobacteria, is an environmental mycobacterium causing opportunistic infections in humans. Mycobacterium kansasii was first described in 1953 [1] and is one of the most frequent non-tuberculous mycobacteria isolated from patients [2][3][4]. Seven subtypes have been previously described based on the restriction fragment length polymorphism (RFLP) of the hsp65 gene [5][6][7][8]. Furthermore, the rpoB and the tuf genes were also shown to successfully discriminate between subtypes [9,10]. Subtype 1 is the most frequently isolated and most pathogenic subtype [6,11,12]. Subtype 2 is the second most common subtype recovered from patients, most of them with immunosuppression [62.5 % had a co-infection with human immunodeficiency virus (HIV) and 21 % were treated with corticosteroids], whereas subtype 3 is most often associated with colonization [6]. Subtypes 4-6 are very rarely isolated from patients and generally non-pathogenic [6]. Subtype 7 wasto our knowledgeonly described by Taillard et al. and its pathogenicity remains unclear [6]. Mycobacterium gastri, a non-pathogenic and closely related species to M. kansasii, described in 1966  [13], is phenotypically distinguishable from M. kansasii because it is not photochromogenic. Despite M. gastri sharing the same 16S rRNA gene sequence as M. kansasii ATCC 12498 T (subtype 1), it was shown to differ in a phylogeny based on average nucleotide identity (ANI)-divergent values [14]. In 2017, Mycobacterium persicum was described as a new closely related species of M. kansasii and M. gastri, altogether forming the M. kansasii complex [15].
To assess the genomic differences between M. kansasii subtypes and its closely related species, M. persicum and M. gastri, we performed whole-genome sequencing of 13 strains belonging to five different M. kansasii subtypes (1, 2, 3, 5 and 6). We compared these genomes with publicly available whole-genome sequences of the M. kansasii complex (n=9), of widespread slow-growing mycobacterial species (n=9) and of the rapid-growing Mycobacteroides abscessus (n=1) ( The 16S rRNA gene alone and a concatenated nucleotide sequence of the 16S rRNA, rpoB and hsp65 genes were aligned using MAFFT version 7.310 [19] and used for phylogenetic reconstruction with FastTree version 2.1.8 with double precision and parameters '-nt -gamma -spr 4 -mlacc 2 -slownni' [20]. Phylogenetic trees were rooted on M. abscessus ATCC 19977 T , a rapid-growing mycobacterium, using Archaeopteryx 0.9921 [21] and visualized using Figtree version 1.4.2 [22]. As expected, M. kansasii subtypes 1 and 4 presented the same 16S rRNA sequence as M. gastri. However, subtypes 2, 3, 5 and 6 presented distinct unique 16S rRNA gene sequences (Fig. 2), sharing 99.61, 99.61, 99.87 and 99.54 % nucleotide identity with the 16S rRNA gene sequence of subtype 1, respectively (BLAST analysis). The phylogeny based on the concatenated 16S rRNA-rpoB-hsp65 genes could distinctly separate each subtype as well as M. gastri (Fig. S1, available in the online version of this article). Both the 16S rRNA and the concatenated 16S rRNA-rpoB-hsp65 genes also clustered M. persicum very tightly with M. kansasii subtype 2.
Groups of orthologous sequences were defined using Ortho-Finder version 2.1.2 [23]. A total of 1351 single-copy orthologous groups were identified and aligned using MAFFT version 7.310. Then, each alignment was concatenated into a core-genome alignment of 489 835 amino acids. A maximum-likelihood core-genome phylogeny was reconstructed using FastTree (as mentioned above but without '-nt'  Fig. 3. RFLP analysis of the hypervariable fragment of the gene hsp65 after in silico amplification (with Tb11, 5¢-ACCAAC-GATGGTGTGTCCAT; Tb12, 5¢-CTTGTCGAACCGCA TACCCT) and digestion with BstEII and HaeIII [5] was performed using Geneious version 9.1.8 [24]. Results (Table S1)   MK13 T (subtype 5) and MK41 T (subtype 6) were generated after growth on Middlebrook 7H10 agar. Cells were saponified, extracted and derivatised following the recommendations of the Sherlock Mycobacteria Identification System (SMIS, MIDI). Mycolic acids were separated with a gradient of methanol and 2-propanol on an Agilent ChemStation 1100/ 1200 HPLC system and analysed with the MIDI Sherlock Software version 4.0. All strains produced similar profiles with a single late-eluting peak cluster (Fig. 4) M. kansasii. Despite their genetic distance, strain MK15 (subtype 2) and strain MK41 T (subtype 6) were also identified as M. kansasii with high similarity indexes (>8.0). However, the identification of strains MK142 T (subtype 3) and MK13 T (subtype 5) was unsuccessful using the Sherlock criteria; the profiles showed similarities not only to M. kansasii, but also to M. szuglai and M. asiaticum for strains MK142 T (subtype 3), and M. bovis (BCG) for MK13 T (subtype 5; data not shown).
Strains MK22 (subtype 1), MK15 (subtype 2), MK142 T (subtype 3), MK13 T (subtype 5) and MK41 T (subtype 6) presented mature colonies on 7H10 medium after 2 weeks of growth at 37 C in aerobic conditions. Photochromogenicity on Löwenstein-Jensen culture medium (37 C) was also confirmed. No strain of subtypes 4 and 7 was available in our laboratory. Regarding phenotypic properties, Jim enez-Pajares et al, characterized 298 M. kansasii strains (subtypes 1-6) [26]. In their study, all strains were reported to grow in more than 1 week at an optimal temperature of 37 C on Löwenstein-Jensen medium. Growth was inhibited at 25 and 45 C. Furthermore, no growth was detected on Löwenstrein-Jensen medium with 10 µg ml À1 thiosemicarbazone or 5 % NaCl or on MacConKey agar without violet crystal, whereas they all grew on Löwenstein-Jensen medium supplemented with 5 µg ml À1 thiophen-2-carboxylic acid hydrazide. All strains presented strong catalase activity at 68 C but none was able to reduce potassium tellurite or exhibited an arylsulfatase activity after 3 days. However, a high degree of variability of several phenotypic testsniacin production, nitrate reduction, Tween 80 hydrolysis and urease activitywas reported within and between each former subtype of M. kansasii (described in the species description) [26]. Therefore, phenotypic testing should not be recommended to achieve reliable identifications of the species of the M. kansasii complex.
Existing species-level lineages include M. kansasii (subtype 1), M. persicum (subtype 2) as well as M. gastri. In this study, we propose to define three new species-level lineages of the M. kansasii complex, corresponding to subtypes 3, 5 and 6: Mycobacterium pseudokansasii sp. nov., Mycobacterium innocens sp. nov. and Mycobacterium attenuatum sp. nov., respectively. This new taxonomical classification is necessary to conserve the monophyly of each species (Figs S1 and 3) and corroborates common cut-offs for species using genetic distances. Furthermore, our results are congruent with the Genome Taxonomy Database (GTDB), in which the M. kansasii species was split into six specieslevel lineages [27], as well as with three recently published WGS phylogenies by Tortoli et al. [14], Gupta et al. [28] and Nouioui et al. [29]. No type strain for the former subtype 4 was available but the clear-cut genomic findings of Mycobacterium sp. 1010001458 suggest that a new species name should be defined as soon as a type strain is available.
No strain or genome of M. kansasii subtype 7 was available and this subtype was described only in one study [6]. The gel electrophoresis technique used at that time lacks precision and this subtype might have been misidentified with a subtype 3 that share a very similar restriction profile. Given the absence of available genomic sequence for this subtype, we cannot infer any recommendation on its taxonomic classification. Defining new species names may help clinicians discriminating between all members of the M. kansasii complex which present drastic differences in their pathogenicity. This species corresponds to the former M. kansasii subtype 3. The name was chosen because it is rarely pathogenic  Formerly M. kansasii subtype 6, this species is non-pathogenic, as suggested by its name, and very rarely isolated from patients. M. attenuatum is a slow-grower and displays photochromogenicity on Löwenstein-Jensen media (37 C). Beige colonies can also be observed on 7H10 medium after 2 weeks' growth at 37 C. M. attenuatum phenotypically exhibits variable niacin production, nitrate reducase and urease activities. However, Tween 80 hydrolysis is observed after 1-3 days. Other general phenotypic features are shared with the other former M. kansasii subtypes. M. attenuatum exhibits the same HPLC profile as the other members of the M. kansasii complex. Reliable molecular identification can be done using PCR-RFLP of the hsp65 or the tuf gene, or using PCRs and sequencing of various genes including hsp65, 16S rRNA and rpoB genes. The maximum-likelihood core-genome phylogeny shows M. attenuatum to be the most distant deep-branching species of the M. kansasii complex.

DESCRIPTION OF
The type strain is MK41 T (=CCUG 72127 T =DSM 107153 T ) and was isolated from bronchial secretions (aspiration of bronchial secretions) of a patient known for Still's disease and reported as a non-pathogenic colonizer. 16S rRNA gene and whole-genome sequence data are available under the accession numbers LS999934 and GCA_900566085.1, respectively.

Funding information
The authors received no specific grant from any funding agency.