We examined all known Mycobacterium species listed in List of Prokaryotic Names with Standing in Nomenclature (LPSN; https://lpsn.dsmz.de; n = 287) and NCBI (https://www.ncbi.nlm.nih.gov; n = 262) (accessed on 30 June 2023 for both). LPSN was used exclusively as a reference for valid species names (to compile the list of Mycobacterium species), not for genome retrieval. Genomes were sourced from NCBI. After manual deduplication when merging those in LPSN and NCBI together, there are a total of 298 Mycobacterium species (Supplementary Dataset S1). We conducted thorough searches for type strain genomes in NCBI (including both assembly and Sequence Read Archive [SRA] datasets). Species without a designated type strain or without an available genome sequence of the type strain were excluded. For Mycobacterium mucogenicum, as no assembled genome of the type strain (JCM 13575) was available, we identified SRA data for this strain. We downloaded and assembled these data de novo into contigs using SPAdes-based assembly pipeline Shovill v1.1.0¹ to include this type strain in our comparative genomic analysis. We excluded 75 species from further analysis due to: unavailability of the type strain genome in NCBI (n = 10), unavailability of genome assemblies in NCBI (n = 40), being actual subspecies rather than species (n = 6), being members of Mycobacterium tuberculosis (n = 5), having misspelled species names (n = 5), being heterotypic (n = 3) or homotypic (n = 2) synonyms, having been transferred to another genus (n = 2), and uncultured candidatus species (n = 2). Then, we retrieved the type strain genome sequence of all included Mycobacterium species from NCBI (n = 223, as of June 30, 2023; Supplementary Dataset S1). Notably, the only available genome of Mycobacterium aurantiacum in NCBI was contaminated. As such, we constructed a dataset comprising the remaining 222 Mycobacterium species.

Our analysis began with phylogenomic inference, which provided the evolutionary context necessary to generate species hypotheses. We first reconstructed three core-protein phylogenomic trees encompassing all 222 Mycobacterium species, which revealed the overall phylogenetic structure of the genus. These trees also served as the primary guide for identifying candidate species clusters: species that grouped closely in the phylogeny were flagged as potential conspecifics or close relatives, warranting further investigation of their genomic similarity. The ANI and digital DNA–DNA hybridization (dDDH) thresholds were used to validate the species hypotheses generated by phylogenomics. These methods were used complementarily, with phylogenomics forming the foundational framework and ANI/dDDH serving as quantitative validation of species boundaries.

We checked genome completeness and contamination using CheckM v1.0.18 (Parks et al., 2015) and then annotated genomes using Prokka v1.14.5 (Seemann, 2014). We discarded genome assemblies of low quality defined as consisting of >500 contigs, having <90% genome completeness, or >10% genome contamination (detailed in Supplementary Dataset S1), ensuring high-quality genomic data.

We calculated pairwise ANI using fastANI v1.32 (Jain et al., 2018) and used a ≥95% ANI cutoff (Dahl et al., 2021; Pan et al., 2022a) to compare each genome with type strains of known Mycobacterium species: (1) Genomes with ≥95% ANI to one single type strain were tentatively assigned to that species. (2) Genomes with ≥95% ANI to two or more type strains were flagged as ambiguous, triggering a dDDH validation step. (3) Genomes with <95% ANI to all known type strains were classified as “unassigned to known species.

For genomes with conflicting ANI results (assigned to more than one species using ≥95% ANI alone), we applied a more stringent criterion of ≥95% ANI plus ≥70% dDDH (Richter and Rossello-Mora, 2009; Meier-Kolthoff et al., 2013) (1) genomes met the dual threshold for one single type strain and were definitively assigned to that species; (2) genomes showed ≥95% ANI but <70% dDDH with two or more type species. This conflict was resolved by classifying it as “species undetermined,” which may represent a new species or part of a broader species complex need to be determined.

Genomes that could not be assigned to any known species (i.e., they had an ANI of <95% to all type strains) were considered potential novel taxa. To confirm their phylogenetic position within the FVC, we inferred a phylogenomic tree including these genomes and all type strains. A group of genomes forming a distinct, well-supported clade and sharing ≥95% ANI among themselves was defined as a new genomospecies.

Species synonyms were defined if they clustered tightly in the phylogenomic tree and exceeded the ANI/dDDH thresholds: ≥95% ANI (Dahl et al., 2021; Pan et al., 2022) plus ≥70% dDDH (Richter and Rossello-Mora, 2009; Meier-Kolthoff et al., 2013).

Phylogenomic trees were constructed for 222 Mycobacterium species using three distinct sets of conserved genetic markers (Figure 1). The first phylogenetic tree was based on concatenated amino acid sequences encoded by conserved genes-these genes were identified and aligned using GTDB-Tk v2.3.2 (Chaumeil et al., 2022) with default settings, and constructed using IQ-TREE v2.3.0 (Minh et al., 2020) under LG model allowing for sites heterogeneity with 1,000 ultrafast bootstraps. For the second tree, 1,862 core protein families were identified using the CD-HIT program as described in Gupta et al. (2018); leveraging these core proteins, the tree was also built with IQ-TREE v2.3.0 (Minh et al., 2020). The third comprehensive phylogenomic tree was constructed from concatenated sequences for 136 proteins, as detailed in Gupta et al. (2018), which form the established marker set for the phylum Actinobacteria.

Based on the phylogenomic tree of type strains of all Mycobacterium species (Figure 1), we further selected those (n = 106) belonging to the FVC to infer three FVC-specific phylogenomic trees with abovementioned methods (Figure 2 and Supplementary Figure S1). All trees were visualized and annotated using iTOL v6.9 (Letunic and Bork, 2021).

We used txid1762 [Organism:exp] AND “latest” [filter] to search NCBI GenBank, and then retrieved all available assemblies (n = 11,534, accessed on 30 June 2023) of Mycobacterium. We discarded genomes labelled ‘atypical’ in NCBI for the following reasons: (i) derived from metagenome, (ii) contaminated, (iii) with many frameshifted proteins, (iv) fragmented assembly, (v) too large or too small genome length, (vi) low quality sequence, or (vii) missing rRNA or tRNA genes. Detailed explanations of atypical genomes are available in the NCBI’s Genome Notes.² We then evaluated genomes for the quality of assemblies using QUAST v5.0.2 (Mikheenko et al., 2018) and checked for genome completeness and contamination using CheckM v1.0.18 (Parks et al., 2015). We further discarded genome assemblies of low quality defined as consisting of >500 contigs, having <90% genome completeness, or >10% genome contamination. We calculated ANI and dDDH values between each of the genomes and type strains of Mycobacterium genomes as described above.

All software was used with default parameters unless otherwise specified. Key software versions include: GTDB-Tk v2.3.2 for phylogenomics, IQ-TREE v2.3.0 for tree inference, CheckM v1.0.18 for quality assessment, fastANI v1.32 for ANI calculations, and the Genome-to-Genome Distance Calculator (GGDC) for dDDH estimates. All genomes were retrieved from NCBI database. The complete list of the 222 type strain genomes with NCBI accession numbers is provided in Supplementary Dataset S1. The list of 11,534 public Mycobacterium genomes analyzed, their quality metrics, and their final species assignments are provided in Supplementary Dataset S2. The list of 86 new genomospecies and their representative genomes is provided in Supplementary Table S2.

We constructed a dataset comprising 222 Mycobacterium species and the genome sequence of their type strains sourced from NCBI. Unlike previous investigations which have had a narrower spectrum of 110–150 species (Tortoli et al., 2017; Gupta et al., 2018; Nouioui et al., 2018b; Bachmann et al., 2020), our dataset encompasses 222 Mycobacterium species. From the retrieved genomes we inferred three independent core-protein phylogenomic trees comprising all 222 Mycobacterium species (Figure 1), to ensure the robustness and consistency of our phylogenetic conclusions, as using different marker sets helps mitigate potential biases inherent to any single method.

The phylogenomic trees segregates Mycobacterium species into three super clades, namely the “Abscessus-Chelonae” clade, the FVC, and the slow-growing Mycobacterium (Figure 1). The “Abscessus-Chelonae” clade forms a separate monophyletic clade, distinguishing itself from all other Mycobacterium species with a deep branch (Figure 1), indicative of it being the ancestral Mycobacterium species, as previously reported (Tortoli et al., 2017). The FVC shares a common ancestor with slow-growing Mycobacterium. Based on the phylogenomic trees, we were able to define the FVC. We identified 106 species within the FVC, 16 (such as M. syngnathidarum and M. yunnanensis) of which have not been previously classified in the FVC (Nouioui et al., 2018b) (Table 1). Among the 107 species, 106 are published (Table 1). We found three slow-growing species, namely Mycobacterium doricum (Tortoli et al., 2001), Mycobacterium salfingeri (Musser et al., 2022), and Mycobacterium tuscia (Tortoli et al., 1999), within the predominantly fast-growing FVC. The unexpected coexistence of slow-growing species within the predominantly fast-growing clade may suggest complex evolutionary adaptations that require further mechanistic investigation (Bachmann et al., 2020; Zhu et al., 2023). Unlike the monophyletic “Abscessus-Chelonae” clade, FVC comprised eight well-supported subclades (assigned Group 1 to 8 here) with between 6 and 23 species in each group (Figure 2, Supplementary Figure S1, and Supplementary Table S3).

We examined all 107 species within the FVC to identify possible synonyms that may not have been previously recognized to perform species curation. Notably, the genome labeled M. farcinogenes type strain DSM 43637^T (accession no. CCAY000000000) is actually sequenced from a strain of M. senegalense (Turenne, 2019). The genuine genome sequence of M. farcinogenes type strain is not available yet; hence, we did not include this species for identifying synonyms. We employed a combined approach of phylogenomic tree inference and ANI/dDDH. We applied a stringent criterion of a ≥ 95% ANI plus a ≥ 70% dDDH value to define synonyms. As such, we detected a pair of synonyms.

Mycobacterium murale (Vuorio et al., 1999) is a heterotypic synonym of Mycobacterium tokaiense (Tsukamura et al., 1981). After detailed analysis of M. tokaiense and M. murale from their original articles (Tsukamura et al., 1981; Vuorio et al., 1999), the phenotypic and genotypic features of the emended M. tokaiense are as follows. Phenotypically, as for cell morphology and staining, it consists of Gram-stain-positive, acid-fast bacilli. These bacilli exhibit both strong acid-fastness and weak or partial acid-fastness. The cells can take the form of long rods, often exceeding 7 μm in length. Regarding colony characteristics, colonies are typically smooth and may be cream-colored and non-pigmented as previously described for M. tokaiense. However, it has been observed that in certain cases, similar to what is seen in M. murale, the colonies may be pigmented. For growth characteristics, the organism grows within a temperature range of 10–37 °C. At 10 °C, growth is variable, with the majority showing positive results within 10 days, while growth at 45 °C is generally weak within 5 days. In the phylogenomic tree, M. murale JCM 13392^T and M. tokaiense NCTC 10821^T cluster together. Genotypically, the draft genome sequences of M. murale JCM 13392^T (accession no. BLKT00000000) and M. tokaiense NCTC 10821^T (accession no. UGQT00000000) have an dDDH value of 83.8% and an ANI value of 97.98%. Both the ANI and dDDH analyses indicate that the two species are the same species. Based on principles in the International Code of Nomenclature of Prokaryotes (ICNP) (2022 Revision) (Oren et al., 2023), specifically Rule 24b, M. tokaiense has the priority of the species name over M. murale.

We emended description of Mycobacterium tokaiense Tsukamura et al. 1981. Mycobacterium tokaiense was originally described by Tsukamura et al. 1981. Based on the evidence presented here, this species description is emended by Wang et al. in 2025 to include the phenotypic and genotypic characteristics of the previously described Mycobacterium murale Vuorio et al. 1999, which is now considered a later heterotypic synonym of M. tokaiense. The emended species should be cited as: Mycobacterium tokaiense Tsukamura et al. 1981 emended by Wang et al. in 2025.

Mycobacterium tokaiense Tsukamura et al. 1981

= Mycobacterium murale Vuorio et al. 1999.

The type strain is 47503 (previously, strain 5553) = ATCC 27282 = CIP 106807 = DSM 44635 = JCM 6373 = NCTC 10821.

Hence, we refined the total count of 107 initially identified species to 106 species, encompassing 105 published and one unpublished (Table 1).

We also found 10 species of five pairs with close evolutionary relatedness-clustering together within the phylogenomic tree of the FVC (Figure 2), but conflicting genomic similarity metrics: Mycobacterium obuense/Mycobacterium kyogaense, Mycobacterium fluoranthenivorans/Mycobacterium hackensackense, Mycobacterium chubuense/Mycobacterium chlorophenolicum, Mycobacterium neumannii/Mycobacterium lehmannii, and Mycobacterium septicum/Mycobacterium nivoides (Supplementary Table S1). These pairs shared a ≥ 95% ANI (range: 95.1–96.3%) but a < 70% dDDH value (<70%; range: 60.9–68.1%), failing to meet the stringent criterion that we used for defining synonyms but are nonetheless closely related and warrant further studies to clarify their taxonomic positions.

We applied our updated FVC taxonomy to curate publicly available genomes of the FVC in GenBank (accessed on 30 June 2023). Considering that strains labelled “Mycolicibacterium” in NCBI do not fully represent all FVC, we retrieved all Mycobacterium assemblies (n = 11,534, accessed by 30 June 2023) from GenBank. We excluded genomes (n = 559) labelled ‘atypical’ in NCBI for the following reasons: derived from metagenome (n = 169); contaminated (n = 146); containing many frameshifted proteins (n = 113); fragmented assembly (n = 111); too large (n = 8) or too small (n = 4) genome length; low quality sequence (n = 3); missing rRNA genes (n = 1) or tRNA genes (n = 1); and not of Mycobacterium (n = 3). We further discarded an additional 180 assemblies due to low quality defined by >500 contigs (n = 169), <90% genome completeness (n = 8), or >10% genome contamination (n = 3).

We determined the precise species for the remaining 10,795 Mycobacterium genomes (Figure 3; Supplementary Dataset S2). Using a ≥ 95% ANI cutoff, 10,553 strains were assigned to at least one known species, with 438 belonging to FVC, either to a single known species (n = 399) or to more than one species (n = 39). For the 39 genomes assigned to more than one species, we further determined their dDDH values and applied a ≥ 70% cutoff for species identification. Of these 39 genomes, 38 could be assigned to a single known species. However, the remaining one (accession no. CP070349.1) had a ≥ 95% ANI with both M. septicum (96.6%) and M. nivoides (95.8%) type strains but shared a < 70% dDDH with M. septicum (69.6%) and M. nivoides (65.4%). Therefore, its species cannot be determined and may represent a new species or alternatively, this strain together with M. septicum and M. nivoides could represent a common species with pairwise ANI ≥ 95% (Supplementary Table S1). A total of 242 genomes had a < 95% ANI with all known species and could not be assigned to any known species. To determine their taxonomic position, we inferred a new phylogenomic tree comprising of these 242 genomes and the type or reference strains of all known mycobacterial species (Supplementary Figure S2). We found that 119 were located within the FVC and could be assigned to 86 new taxa (genomospecies) using a ≥ 95% ANI cutoff, denoted as taxon 1 to 86 (Figure 2; Supplementary Table S2). These 86 new taxa also belong to the eight abovementioned well-supported major groups (Figure 2). These new taxa could be considered naming as Candidatus Mycobacterium fortuitum-vaccae clade taxon 1 to 86. Further phenotypic characterization of the new taxa is required to establish their species status with proper names according to ICNP (2022 Revision) (Oren et al., 2023), specifically Rule 27 and Recommendation 30.