Reference M. abscessus CIP104536^T, M. abscessus DSMZ44567 (German Collection of Microorganisms and Cell Cultures, Braunschweig, Germany), M. abscessus subsp. bolletii CIP108541^T (herein referred as “M. bolletii”) and M. abscessus subsp. bolletii CIP108297^T (herein referred as “M. massiliense” 23 ) were used in this study. In addition, a collection of 17 M. abscessus clinical isolates from the mycobacteria reference laboratory of the Méditerranée Infection Institute, Marseille, France were also studied (Table  1). All of the mycobacteria were grown in 7H9 broth (Difco, Bordeaux, France) enriched with 10% OADC (oleic acid, bovine serum albumin, dextrose and catalase) at 37°C. As for the identification, DNA extraction and rpoB partial sequence-based identification were performed using the primers MYCOF and MYCOR2 (Table  1) as previously described 24 . In addition, the rpoB gene sequence retrieved from 48 M. abscessus sequenced genomes was also analysed (Additional file 1) ( http://www.ncbi.nlm.nih.gov/).

* With reference to M. abscessus ATCC 19977 genome.

Fragments from five housekeeping genes argH (argininosuccinate lyase), cya (adenylate cyclase), murC (UDP N-acetylmuramate-L-Ala ligase, pta (phosphate acetyltransferase) and purH (phoshoribosylminoimiazolcarboxylase ATPase subunit) were amplified using the sets of primers as previously described (21). The sequences of each one of these five housekeeping genes retrieved from 48 M. abscessus sequenced genomes, were also included in the MLSA analysis (Additional file 1).

Sequences of the whole intergenic spacers were extracted from the reference M. abscessus CIP104536^T (ATCC19977) genome (GenBank accession CU458896.1) using the perl script software and a total of 8 spacers with a 200-700-bp sequence size were further used in analysis. For each of these 8 spacers, specific PCR primers were designed using Primer3 software v 0.4.0 ( http://frodo.wi.mit.edu/primer3) and tested in silico for specificity using BLAST software ( http://www.ncbi.nlm.nih.gov). The PCR conditions were first optimized using DNA extracted from the reference M. abscessus, “M. bolletii” and “M. massiliense” isolates before analysis of DNA extracted from the 17 clinical isolates (Table  1). The PCR amplifications were performed in a 50 μl PCR mixture containing 5 μl 10x buffer (Qiagen, Courtaboeuf, France), 200 mM each dNTP, 1.5 mM MgCl₂, 1.25 U HotStarTaq polymerase (Qiagen), 1 μl each primer (10 pM), 33 μl nuclease-free water and 5 μl DNA template. The amplification program consisted of an initial 15 min denaturation step at 95°C followed by 40 cycles of 30 s at 95°C, 30 s at 60°C and 1 min at 72°C; the amplification was completed by a final 5-min elongation step at 72°C. Negative controls consisting of PCR mixture without DNA template were included in each PCR run. The products were visualized by gel electrophoresis, purified using a MultiScreen PCR filter plate (Millipore, Molsheim, France) and sequenced in both directions using the BigDye Terminator sequencing kit (Applied Biosystems, Villebon-sur-Yvette, France), as previously described 27 . The sequences were edited using the ChromasPro software (version 1.42; Technelysium Pty Ltd), aligned using Clustal W (MEGA 5 software) and compared with the reference M. abscessus ATCC 19977 sequences (GenBank accession CU458896.1). For MST and MLSA discrimination power was calculated using the Hunter-Gaston Index 31 :

DI = 1 − 1 N N − 1 ∑ j = 1 8 n j n j − 1

where D is the numerical index of discrimination, N is the total number of isolates in the sample population, s is the total number of different types and nj is the number of isolates belonging to the jth type.

Phylogenetic trees were constructed based on rpoB gene, concatenated MLSA genes, concatenated spacers and MST spacer n°2 sequences using the neighbor–joining method with Kimura’s two-parameter (K2P) distance correction model with 1000 bootstrap replications in the MEGA version 5 software package 32 . The rpoB gene sequence-based tree was rooted using M. chelonae strain CIP 104535^T and M. immunogenum strain CIP 106684^T rpoB gene sequences. A heatmap was constructed using the R statistical software based on the spacer profile as a distance matrix.

The identification of M. abscessus CIP104536^T, M. abscessus DSMZ44567, M. bolletii CIP108541^T and M. massiliense CIP108297^T was confirmed by partial rpoB sequencing. The sequences were deposited in the GenBank database (GenBank accession: KC352778 - KC352795). Isolates P1, P2.1, P2.2, P2.3, P2.4, P2.5, P3.1, P3.2, P4, P5, P6, P7 and P8 exhibited 99% rpoB sequence similarity with M. abscessus ATCC19977^T and were identified as M. abscessus. Isolates P9 and P10 exhibited 99% rpoB sequence similarity with “M. bolletii” CIP108541^T and were identified as “M. bolletii” whereas isolate P11 exhibited 99% rpoB sequence similarity with “M. massiliense” CIP108297^T and was identified as “M. massiliense”. A total of 23 M. abscessus sequenced genomes were identified as M. abscessus since they exhibited 98 to 100% similarity with the M. abscessus type strain rpoB partial gene sequence. M. abscessus M24 shared 99% similarity with the M. bolletii type strain partial rpoB gene sequence. A total of 26 M. abscessus and “M. massiliense” sequenced genomes shared 99% to 100% similarity with “M. massiliense” partial rpoB gene sequence. The tree built from 69 partial rpoB gene sequences showed three distinct groups, each comprising the type strain (Figure  1a).

Fragments for the expected size were amplified and sequenced for the five MLSA genes. The sequences were deposited in the GenBank database (GenBank accession: KC352742 - KC352759, KC352760 - KC352777, KC352796 - KC352813, KC352814 - KC352831, KC352832 - KC352849). Concatenation of the five sequences yielded a total of 19 different types, including 9 types for 37 M. abscessus organisms, four types for 4 “M. bolletii” organisms and M. abscessus M139 and five types for 27 “M. massiliense” organisms. The Hunter-Gaston Index for MLSA was of 0.903. The MLSA tree based on the five gene concatened sequences showed three principal clusters, i.e. a M. abscessus cluster, a “M. bolletii” cluster and a “M. massiliense” cluster (Figure  1b). Latter cluster comprised of five sub-clusters with “M. massiliense” type strain and P11 strain sub-clustering together close to M. abscessus 5S strain. Also, MLSA-derived tree clustered M. abscessus M139 strain and P5 strain respectively identified as “M. massiliense”, close to the “M. bolletii” whereas both strains clustered with M. abscessus in the rpoB gene sequence-derived tree.

Analysis of the reference M. abscessus ATCC 19977 complete genome sequence yielded 3538 intergenic spacers with > 300 spacers were 200–700 bp in length. Successful PCR sequencing was achieved for 8 spacers in all the isolates studied; the sequences were deposited in the GenBank database (GenBank accession: KC352850 - KC352890). In M. abscessus isolates, including the 37 sequenced genomes, the spacer sequence variability was generated by one to 12 single nucleotide polymorphisms (SNPs) (spacers n°1 and n°8), one to 18 SNPs and one to two nucleotide deletions (spacer n°2), one to two SNPs (spacers n°3 and n°7) and nucleotide insertion (spacers n°2 and n°5). In “M. bolletii” isolates, the spacer sequence polymorphisms were generated by one SNP for spacer n°1, two SNPs and one deletion for spacer n°2, two SNPs for spacer n°3 and nine SNPs for spacer n°7. In “M. massiliense” isolates, including 28 sequenced genomes, the spacer sequence polymorphism were generated by nine SNPs and one insertion (spacer n°1), one insertion (spacer n°3), five SNPs and two insertions (spacer n°4), one SNP (spacer n°5) and two SNPs (spacer n°7). Concatenation of the eight spacer sequences yielded a total of 24 types, with the 37 M. abscessus organisms grouped into 12 spacer types, four formerly “M. bolletii” organisms grouped into three spacer types and 28 formerly “M. massiliense” organisms grouped into nine spacer types. This yielded a Hunger-Gaston Index of 0.912. Spacer n°5 was found to be the most variable of the eight spacers under study, exhibiting 13 different alleles (Table  2). When combining the eight spacer sequences, a unique MST profile for each reference isolate was obtained, i.e., MST1 and MST2 for M. abscessus CIP104536^T and M. abscessus DSMZ44567 respectively, MST13 for “M. bolletii” CIP108541^T and MST16 for “M. massiliense” CIP108297^T. At the sequence level, we found that MST1 and MST2 genotypes differ by at most nine SNPs, whereas MST1 differed from MST13 by up to 18 SNPs, one insertion and two deletions and from MST16 by 14 SNPs, 11 deletions and two insertions (supplementary material). The 17 clinical M. abscessus isolates were grouped into eight MST types, named MST1 to MST8, with five M. abscessus isolates exhibiting the M. abscessus CIP104536^T MST1 genotype and one isolate (P1 strain) exhibiting the M. abscessus DSMZ44567 MST2 genotype. The P9 “M. bolletii” clinical isolate yielded the MST13 genotype in common with the reference “M. bolletii” CIP108541^T, whereas the P10 “M. bolletii” clinical isolate yielded a unique MST14 genotype that differ from MST13 by two SNPs in spacer n°1. M. abscessus M24 yielded the MST15 and differed from MST13 by four polymorphic spacers. In “M. massiliense” nine different profiles were generated MST 16 to MST24. The P11 “M. massiliense” clinical isolate shared the MST16 genotype with the reference “M. massiliense” CIP108297^T. “M. massiliense” 2B isolate, “M. massiliense” 1S isolate and “M. massiliense” M18 isolate shared the same MST profile (MST17). M. abscessus 5S isolate exhibited the MST21 profile.

^a MST = Multispacer Sequence Typing. ^b isolates were listed with reference to their corresponding patient, for example P1 = isolate 1 from patient 1, P2.1 = isolate 1 from patient 2, etc. ^c DI = Discrimination index.

The MST-phylogenetic tree clustered isolates from patients P1 to P8 with M. abscessus reference strain, isolates from P9 and P10 with “M. bolletii” and isolate from P11 with “M. massiliense”, in agreement with their rpoB sequence-based identification and MLSA analysis (Figure  1c). The MST, MLSA and rpoB phylogenetic trees separated the M. abscessus isolates into three principal clusters depicted by M. abscessus, “M. bolletii” and “M. massiliense” isolates (Figure  1a, b and c). However, MST resolved “M. bolletii” cluster into two sub-clusters formed by isolate P5 and all of the other M. bolletii isolates with a 76% bootstrap value, wich is discordant with MLSA and rpoB based tree. Each cluster or sub-cluster of the M. abscessus isolates corresponded to different genotypes. The “M. massiliense” cluster was more disperse and divided into six sub-clusters with isolate P11 and “M. massiliense” type strain sub-clustering alone. The results of this analysis were consistent for 67 isolates and inconsistent for two isolates P5 and M. abscessus M139. A heatmap incorporating all spacer patterns into a matrix further demonstrated that spacer n°2 was the most discriminating spacer (Figure  2). Hence, the tree based on the spacer n°2 sequence also discriminated the three M. abscessus, “M. bolletii” and “M. massiliense” clusters (Figure  3). This discrimination potential makes spacer n°2 a useful new tool for the accurate identification of M. abscessus subspecies. Furthermore, these data indicated that it was readily possible to discriminate isolates that would have been identified as “M. bolletii” 26 or “M. massiliense” 23 using a previous taxonomy proposal and are now grouped as M. abscessus subsp. bolletii according to a recent taxonomy proposal 20 21 .