This study was carried out with 26 isolates belonging to INNO-LiPa cluster CHII recovered from clinical and environmental specimens by Prof. Françoise Portaels (Institute of Tropical Medicine, Antwerp, Belgium) and Prof. Roland Schulze-Röbbecke (University of Dusseldorf, Dusseldorf, Germany). M. smegmatis mc²155, M. tuberculosis H37Rv and the type strains of M. chelonae-M. abscessus complex (M. abscessus subsp. abscessus ATCC 19977^T, M. abscessus subsp. bolletii CCUG 50184^T, M. abscessus subsp. massiliense CCUG 48898^T, M. chelonae ATCC 35752^T, M. immunogenum ATCC 700505^T, M. salmoniphilum ATCC 13758^T, M. franklinii DSM 45524^T, and M. saopaulense CCUG 66554^T) were included for comparison (Table 1).

Cultures were grown aerobically at 28–30°C on solid media including Löwenstein-Jensen (LJ) and Middlebrook 7H10 [Becton-Dickinson (BD), USA] supplemented with oleic acid, albumin, dextrose and catalase (OADC–BD) and in liquid media including Middlebrook 7H9 (BD), Mueller-Hinton and Lisogeny Broth with 1% Tween 80.

Phenotypic analyses were performed as described in standard protocols for biochemical identification of mycobacteria (Tsukamura, 1984; Kent and Kubica, 1985; Leao et al., 2004). Analysis of pigment production, single-source carbon utilization (mannitol, inositol and citrate), growth at 26° and 37°C and tolerance to 5% NaCl, 0.2% picric acid, 0.2% nitrite and para-nitrobenzoic acid (PNB) were performed on 7H10-OADC and LJ. Nitrate reduction, Tween 80 hydrolysis and arylsulfatase production were also examined.

Antimicrobial drug-susceptibility testing was performed using the microdilution method in cation-supplemented Mueller–Hinton broth, according to the recommendations of the Clinical and Laboratory Standards Institute [Clinical and Laboratory Standards Institute (CLSI), 2011] for rapidly growing mycobacteria. The antimicrobials tested were amikacin, cefoxitin, ciprofloxacin, clarithromycin, doxycycline, minocycline, moxifloxacin and tobramycin.

Chromosomal DNA was extracted using QIAamp DNA mini kit (Qiagen, Germany) as previously described (Bryant et al., 2013). DNA concentration was determined using a Qubit high-sensitivity (HS) assay kit (Life Technologies, USA).

High quality DNA of the 26 isolates and of M. abscessus subsp. bolletii CCUG 50184^T, M. immunogenum ATCC 700505^T, M. salmoniphilum ATCC 13758^T, and M. franklinii DSM 45524^T were subjected to multiplexed paired end sequencing using the Illumina Miseq platform. The genome of M. saopaulense CCUG 66554^T was sequenced in a previous project from the laboratory of the Universidade Federal de São Paulo (accession number CP010271). The genomes of M. abscessus subsp. abscessus ATCC19977^T (accession number CU458896), M. abscessus subsp. massiliense CCUG48898^T (accession number NZ_AKVF01000005NZ_AKVF01000001 to NZ_AKVF01000005), and M. chelonae ATCC 35752^T (accession number CP010946) were retrieved from the GenBank database (http://www.ncbi.nlm.nih.gov/genbank/). Sequencing errors in reads were corrected with the program Quake v0.3 (Kelley et al., 2010) and reads trimmed with the program Trimmomatic v0.33 (Bolger et al., 2014). The assembly was performed with Newbler program v3.0 (20140318_1550)—version with support for reads with Illumina's Casava accession number v1.8 format) using default parameters. Raw sequencing data was deposited on the NCBI Sequence Read Archive (http://www.ncbi.nlm.nih.gov/sra) under accession number SRP075879 and the assembled genomes were deposited as BioProject PRJNA323571.

Taxonomically informative partial sequences of 16S rDNA, rpoB, hsp65 and 16S–23S ITS-1 fragments were PCR amplified and sequenced using primers listed in Supplementary Table 1. PCR products were purified using QIAquick PCR purification Kit (Qiagen, Germany). Dideoxy sequencing was performed using BigDye® 19 Terminator v3.1 Cycle Sequencing Kit (Applied Biosystems, USA) and run in ABI PRISM 3100 DNA Analyzer (Applied Biosystems).

Individual and concatenated phylogenetic trees based on the partial sequences of the previous genomic regions were constructed using PhyML (http://www.atgc-montpellier.fr/phyml/) (Guindon and Gascuel, 2003), using as input the multiple alignment of these sequences produced by MUSCLE (http://www.ebi.ac.uk/Tools/msa/muscle/help/) with default parameters (penalty for gap opening = 400, gap extension = 0). Confidence bootstrap values were calculated with 100 replicates. The corresponding sequences of the M. chelonae-M. abscessus complex type strains and outgroups (M. tuberculosis H37Rv and M. smegmatis mc²155) were retrieved from the GenBank database (http://www.ncbi.nlm.nih.gov/genbank/) (Supplementary Table 2).

The 16S rDNA, hsp65, 16S–23S ITS-1 and rpoB, partial sequences obtained in this study were deposited in the GenBank/EMBL/DDBJ under accession numbers: KT779789, KT779792-KT779795, and KT779797-KT779815 (16S rDNA), KT779818, KT779821-KT779824 and KT779826-KT779844 (hsp65), KT779847, KT779850-KT779853, and KT779855-KT779873 (16S–23S ITS-1), and KT779876, KT779879-KT779882, and KT 779884-KT779902 (rpoB).

The genomes were deposited in the GenBank/EMBL/DDBJ under accession numbers: MAEQ00000000 (M. chelonae 96-1705), MAER00000000 (M. chelonae 96-1717), MAES00000000 (M. chelonae 96-1720), MAET00000000 (M. chelonae 96-1724), MAEU00000000 (M. chelonae 96-1728), MAEV00000000 (M. sp. D16R24), MAEP00000000 (M. franklinii D16R27), MAEW00000000 (M. sp. D16Q13), MAEX00000000 (M. sp. D16Q14), MAEY00000000 (M. sp. D16Q16), MAFS00000000 (M. franklinii D16Q19), MAEZ00000000 (M. sp. D16Q20), MAFA00000000 (M. chelonae D16Q24), MAFB00000000 (M. sp. D17A2), MAFC00000000 (M. sp. D16R12), MAFD00000000 (M. sp. D16R18), MAFE00000000 (M. salmoniphilum D16Q15), MAFF00000000 (M. chelonae D16R2), MAFG00000000 (M. chelonae D16R3), MAFH00000000 (M. chelonae D16R7), MAFI00000000 (M. chelonae D16R9), MAFJ00000000 (M. chelonae D16R10), MAFK00000000 (M. chelonae D16R14), MAFL00000000 (M. chelonae D16R19), MAFM00000000 (M. chelonae D16R20), MAFN00000000 (M. sp. 96-892), MAFO00000000 (M. abscessus subsp. bolletii BD), MAFP00000000 (M. immunogenum MC 779), MAFQ00000000 (M. franklinii CV002), and MAFR00000000 (M. salmoniphilum SC).

The type strains of the eight formally named members of the M. chelonae-M. abscessus complex and 26 isolates from the INNO-LiPA cluster CHII were analyzed. Five isolates from clinical specimens (96-1705, 96-1717, 96-1720, 96-1724, and 96-892) and one from an animal (96-1728) were received from the Institute of Tropical Medicine in Antwerp, Belgium. The remaining 20 isolates were obtained from water sources in Germany and were received from the collection of the University of Dusseldorf in Dusseldorf, Germany.

All CHII isolates and type strains grew in the presence of picric acid, 5% NaCl at 30°C and PNB and generated nonchromogenic colonies on solid culture media within 7 days. They did not reduce nitrate or hydrolyze Tween 80, but exhibited arylsulfatase activity within 3 days.

Growth in the presence of nitrite and 5% NaCl at 37°C and utilization of mannitol, inositol and citrate as single-source carbon sources, generated strain specific results and were not consistently related to any of the established species within this complex (Supplementary Table 3).

All isolates tested were susceptible to clarithromycin and resistant to cefoxitin, except for M. franklinii DSM 45524^T, which was susceptible to cefoxitin (MIC = 16 μg/mL). Variable results were obtained with the other tested antimicrobials (Supplementary Table 4).

The number of assembled bases ranged from 4,768,278 to 5,548,818, with an average G+C content of 63.94%. The number of generated scaffolds ranged from 15 to 82 with an N50 size from 110,426 to 682,599 bp. Genome sizes were consistent with the expected sizes of known species within the complex (Supplementary Table 5).

Average Nucleotide Identity (ANI) compares the nucleotide sequences of conserved genes shared by two genomes. ANI comparison measures the level of identity of nucleotides after full alignment of two genomes and selection of the most conserved regions, in such a way that only highly conserved genes are compared. This characteristic made ANI very popular in whole genome sequencing (WGS) comparative studies, because it is considered to represent more accurately the evolutionary relationships among genomes.

Pairwise ANIb and ANIm values of the M. chelonae-M. abscessus complex type strains were all below 95%, except between M. abscessus subsp. abscessus ATCC 19977^T, M. abscessus subsp. bolletii CCUG 50184^T and M. abscessus subsp. massiliense CCUG 48898^T. These results confirmed that the type strains represent distinct species within the M. chelonae-M. abscessus complex and that the strains ATCC 19977^T, CCUG 50184^T, and CCUG 48898^T are appropriately classified into a single species, M. abscessus (Table 2 and Supplementary Table 6).

Based on ANIb and ANIm values, the CHII isolates could be separated into different taxonomic groups. The pairwise genome alignment of 14 isolates (96-1705, 96-1717, 96-1720, 96-1724, 96-1728, D16Q24, D16R2, D16R3, D16R7, D16R9, D16R10, D16R14, D16R19, and D16R20) yielded >95% ANIb and ANIm values, showing that they belong to the same species. The ANI values between these isolates and the type strains yielded values slightly higher than 95% with M. chelonae ATCC 35752^T, showing that they could be classified into the species M. chelonae according to their ANI. Isolate D16Q15 yielded ANIb and ANIm values above 95% with M. salmoniphilum ATCC 13758^T, showing that it belongs to the species M. salmoniphilum by this approach; and isolates D16Q19 and D16R27 yielded ANIb and ANIm values above 95% with M. franklinii DSM 45524^T, indicating that they belong to the species M. franklinii, thus confirming previously reported data (Nogueira et al., 2015a). The remaining nine isolates yielded ANI values below 95% with all type strains indicating a clear separation from the M. chelonae-M. abscessus complex at the species level. These isolates could be grouped in three genomospecies using their ANI values: D16Q14, D16Q20, D16R24, and D17A2 (Genomospecies G1), 96-892, D16Q13, and D16Q16 (Genomospecies G2); and D16R12 and D16R18 (Genomospecies G3). Pairwise ANI values of isolates within each genomospecies were above 95% (Table 2 and Supplementary Table 6). The complete ANIb data distribution was represented in a tree (Figure 1A). The tree obtained using ANIm data showed the same distribution (data not shown).

Tetranucleotide frequency correlation coefficients (TETRA) analysis determines the relative tetra-oligonucleotide invariance along the genome sequence, including coding- and non-coding regions. The procedure is based on the hypothesis that the composition of tetra nucleotide sequences in a genome is conserved within a species; moreover, the level of similarity of that composition is related to the evolutionary distance between genomes. Two highly similar genomes would have higher than 99% TETRA coefficient (Teeling et al., 2004). Values above this percentage mean that the bacteria belong to the same species.

The pairwise TETRA coefficients were all above the 0.99 threshold, even between the different type strains (Table 2). Isolates with pairwise ANI values above 95% showed TETRA values above 0.999 and pairwise ANI values below 95% corresponded to TETRA values between 0.99 and 0.999. ANI values below 95% and TETRA values above 0.999 were obtained in pairwise comparisons of D16Q16 and D16Q20, D16R18 and D16Q15, D16R18 and M. salmoniphilum ATCC 13758^T (Supplementary Table 7). These results showed a low discriminative power of TETRA analysis applied to the CHII group.

Genome-to-Genome Distance (GGD) calculation is a web-based procedure that performs in silico genome-to-genome comparison. The method is based on BLAST nucleotide comparison of entire sequences and allows calculation of digital DNA-DNA hybridization (dDDH) values, corresponding to classical wet-lab DDH.

The genome distance displayed by the genomes under study confirmed the grouping obtained with ANI, however some partially discordant data were observed (Table 2, Figure 1B and Supplementary Table 8). The 14 isolates found to belong to M. chelonae by ANI, showed data suggesting that they could belong to a different species very closely related to M. chelonae ATCC 35752^T (GGD around 0.045 and dDDH around 64%) (Supplementary Table 8). A similar result was found when comparing genomospecies G3 genomes to each other (isolates D16R12 and D16R18) (Supplementary Table 8), while GGD and dDDH data confirmed data for genomospecies G1 and G2 (Table 2).

On the other hand, isolate D16Q15 vs. M. salmoniphilum ATCC 13758^T showed GGD values slightly higher than the accepted threshold (0.0340, see Supplementary Table 8). This result could suggest that they belong to different species; yet, dDDH percentage values higher than 70% were obtained, confirming that they should be considered as a single species. The same situation was seen when comparing D16Q19 and D16R27 vs. M. franklinii DSM 45524^T, with GGD value of 0.0336 and 0.0335, respectively and calculated dDDH of 71.90% (Table 2).

As expected, the genomes from reference type-strains showed GGD and dDDH values corresponding to those of different species (GGD between 0.1272 and 0.1627; dDDH <33%; Supplementary Table 8). When comparing genomes of M. abscessus subspecies to each other, genome distance analysis showed values slightly higher than the threshold (0.0266–0.0288) corresponding to dDDH >70%, (Supplementary Table 8).

Data obtained of the genome distance, represented in a tree (Figure 1B), showed similar genome distribution to that derived from ANI data (Figure 1A).

Genomic Signature determines the relative intragenomic invariance of di- or tetra-oligonucleotide composition along the genome sequence, similarly to TETRA analysis. Similarities among genomes are represented as delta asterisk (δ^*) values (see Supplementary Table 9). There is no general threshold for species separation using this approach. The calculated δ^* value, when a genome is compared with itself, represents the threshold value that is used for species separation for the considered genome. Higher values identify genomes of different species and equal to or lower values identify genomes of the same species.

The obtained delta values fell within the range of the calculated cutoff values, between 21 and 27, indicating that all isolates and type strains are closely related. The type strains M. abscessus subsp. abscessus ATCC 19977^T, M. abscessus subsp. bolletii CCUG 50184^T and M. abscessus subsp. massiliense CCUG 48898^T showed the lowest delta values (22 to 24) and, as expected, were grouped as a single species. In accordance with the distribution found with ANI and GGD, the three M. abscessus subspecies and M. immunogenum ATCC 700505^T appeared more separated from the other strains and isolates within the group (delta values of 24 to 28) (Table 2 and Supplementary Table S9).

Individual trees obtained with 16S rDNA, rpoB, hsp65, and ITS-1 sequences grouped the CHII isolates among members of M. chelonae-M. abscessus complex (Figure 2 and Supplementary Figure 1).

The individual trees obtained with 16S rDNA, hsp65, and ITS sequences showed some discordant groupings when compared to ANI and GGD trees (Supplementary Figure 1). In the 16S rDNA tree all M. chelonae isolates clustered with M. saopaulense CCUG 66554^T. Moreover, it was not possible to discriminate M. chelonae ATCC 35752^T from the M. franklinii isolates as well isolates of genomospecies G1 and genomospecies G2. In the hsp65 tree, isolates of genomospecies G1 clustered with M. salmoniphilum ATCC 13758^T and isolates D16Q19 and D16R27 did not cluster with M. franklinii DSM 45524^T. Moreover, isolate D16Q15 did not cluster with M. salmoniphilum ATCC 13758^T. In the 16S–23S ITS tree, isolate D16Q15 clustered with isolates of genomospecies G1 and not with M. salmoniphilum ATCC 13758^T. Furthermore, isolates of genomospecies G3 were not grouped. In the rpoB and the concatenated trees all M. chelonae isolates clustered with M. chelonae ATCC 35752^T, D16R27, and D16Q19 with M. franklinii DSM 45524^T, and D16Q15 with M. salmoniphilum ATCC 13758^T. Moreover, genomospecies G1, G2 and G3 formed clusters separated from all type strains (Figure 2). Therefore, rpoB and concatenated trees were in agreement with the isolates distribution obtained using ANI and GGD procedures.

The authors declare that the research was conducted in the absence of any commercial or financial relationships that could be construed as a potential conflict of interest.