Background

The non-tuberculous mycobacterium Mycobacterium abscessus was long confused with Mycobacterium chelonae[1]. Other closely related species include Mycobacterium salmoniphilum[2], Mycobacterium immunogenum[3], Mycobacterium massiliense[4], Mycobacterium bolletii[5] and Mycobacterium franklinii[6] altogether forming the Mycobacterium chelonae-abscessus complex. This complex is the third most frequent mycobacterial complex infecting humans in developed countries besides the Mycobacterium tuberculosis and Mycobacterium avium complexes [7, 8]. Bibliometrics retrieving over 1,700 publications in the Medline database illustrates the fact that this complex is emerging, causing both sporadic cases and outbreaks of community-acquired and health-care associated infections [9]. Not only humans but also cats [10, 11] and dolphins [12–14] are infected while fishes are uniquely infected by M. salmoniphilum[2, 15].

Current nomenclature is that the species M. abscessus comprises two subspecies named M. abscessus subsp. abscessus and M. abscessus subsp. bolletii. Later taxon accommodates isolates previously referred as M. bolletii or M. massiliense[16]. This nomenclature however may obscure the true diversity of mycobacteria in this complex. While the 16S rRNA gene yields an identical sequence for M. abscessus and M. bolletii, it shares 99% sequence identity with M. massiliense. RpoB gene sequencing founded the description of recent species [17–19] but yielded further conflicting results [20–22]. Multilocus sequencing analysis [23] and multispacer sequence typing [24] differentiated M. massiliense from M. bolletii. In this report, the previous nomenclature M. abscessus, M. bolletii and M. massiliense forming the M. abscessus complex, has been retained for clarity.

The availability of 39 M. abscessus, 13 M. massiliense and two M. bolletii genomes in the National Center for BioInformatics (NCBI) genome database provides new opportunities to assess the diversity of this species. Here, we review 14 complete published M. abscessus complex genomes and compare them with the re-annotated M. tuberculosis H37Rv genome (Table 1) in order to in-depth analyse the diversity of M. abscessus.

Table 1 List of Mycobacterium abscessus genomes here studied
Full size table

Results and discussion

M. abscessuscomplex pan- and core-genome

M. abscessus complex genomes comprise one circular chromosome. In addition, M. abscessus ATCC 19977 contains one 23-kb plasmid identical to the Mycobacterium marinum pMM23 plasmid, encoding mer operon and mercury reductase protein, which may confer resistance to organo-mercury compounds [25]. In order to normalize the predicted proteins and to minimize the differences of presence/absence of genes and length, coding sequences were predicted using prodigal software [26]. We identified a total of 70,309 protein-coding sequences which number varies from 4,651 to 5,079 in each genome (Table 2). The core-genome contains 57,172 protein sequences accounting for 64.15% of the pan-genome. This figure indicates a non-conservative genome contrary to that of Mycobacterium tuberculosis, a conservative-genome pathogen which core-genome accounts for 96.1% of the pan-genome [27]. Using orthoMCL [28] with a conservative parameter value of 50% sequence identity, we categorized these 70,309 proteins into 6,153 orthologous protein groups, including 3,947 core-genome groups and 55 strain-specific groups.

Table 2 Mycobacterium abscessus core genome and unique genes
Full size table

M. abscessuscomplex diversity

The average percentage of amino-acid sequence identity (AAI) of core proteins was determined as previously described [29]. The AAI values indicate that M. abscessus complex forms three main clusters: cluster 1 (C1) includes M. abscessus type strain and strains M93, 94, M152 and Go06; cluster 2 (C2) contains two subclusters: cluster 2A (C2A) includes M. massiliense type strain and strains M154 and M18; cluster 2B (C2B) includes strains 47 J26, M115, M172 and M139; cluster 3 (C3) includes two subclusters: cluster 3A (C3A) includes M. bolletii type strain and cluster 3B (C3B) includes M. bolletii strain M24 (Table 3).

Table 3 Average nucleodite identity and characteristics of Mycobacterium abscessus genomes
Full size table

M. abscessus complex proteomes were further aligned using Mauve software [30] to infer phylogeny using the Neighbor-Net algorithm in the package SplitsTree4 [31]. The phylogenomic network confirms the three clusters C1, C2 and C3 (Figure 1A). A phylogenomic tree based on gene content (i.e., the presence or absence of orthologs) (Figure 1B) organizes M. abscessus differently from the whole genome concatenated tree (Figure 1A) or even the phylogenetic tree based on rpoB gene sequence (Figure 1C). Phylogenomic analysis indicates that the M. abscessus gene repertoires have different evolutionary histories and suggests that differential gene loss and lateral gene acquisition are playing important roles in the evolution of some M. abscessus strains. Notably, the situation of strain Go06 is confusing, as it presents 98.4% AAI with M. abscessus type strain in C1 (Figure 1A) whereas its rpoB gene sequence and single nucleotide polymorphisms analysis are indicative of M. massiliense[8, 32]. Our analyses indicate that strain Go06 have an ambiguous classification as a chimera between M. abscessus and M. massiliense and is the only example compatible with a lateral transfer of rpoB gene.

Figure 1
figure 1

Phylogenomic analysis of M. abscessus . A. Aligned whole genomes phylogenetic network. B. Gene content phylogeny constructed from the matrix of discrete characters using the neighbor-joining method. C. RpoB gene based phylogentic tree using neighbor-joining method.

Full size image

Functional analysis using Clusters of Orthologous Groups database (COG) [33] and BLASTP indicates unique genes in C1, encoding hypothetical proteins, proteins implicated in transcription, energy production and transport, carbohydrate metabolism and transport, lipid metabolism, nucleotide metabolism and transport, amino-acid metabolism and transport, post-translational modification and inorganic ion transport and metabolism (Table 2, Figure 2). Within C2, unique genes encode hypothetical proteins, proteins implicated in transcription and lipid metabolism. Within C3, unique genes encode hypothetical proteins, proteins implicated in amino-acid metabolism and transport and translation. These unique genes could base a refined identification of the three genomospecies. However, we could not exclude that these unique genes could be due to a coding sequence, which arose de novo, to HGT or gene loss for the other subspecies. In the case of absence of a gene, this could also be due to a real loss or to an assembly artefact.

Figure 2
figure 2

Core genomes in M. abscessus clusters.

Full size image

C1 strains have been isolated from American and Malaysian patients suffering knee infection and lower respiratory infection, respectively (Table 1). C2A strains were isolated from Malaysian and French patients suffering severe, respiratory tract infections. C2B strains were isolated from Nepalese, Malaysian and English patients suffering respiratory tract infections, including cystis fibrosis and pulmonary tuberculosis patients. C3A and C3B strains were exclusively isolated from patients suffering respiratory tract infections, in France and Malaysia, respectively. Therefore, clusters specify the clinical form and geographical origin of the infection.

Altogether, genomics analyses revealed a more heterogeneous structure of M. abscessus complex than the one currently suggested by the nomenclature, which recognizes only two subspecies within M. abscessus[16]. It has been proposed that two genomes exhibiting AAI >96% belong to the same species [34, 35]. Therefore, AAI analysis indicates that M. abscessus is in fact comprising of three genomospecies, corresponding to previous nomenclature of M. abscessus (C1), M. massiliense (C2) and M. bolletii (C3). Using an AAI <97% threshold would further determine two subspecies in M. massiliense (C2A and C2B) and in M. bolletii (C3A and C3B). Recent whole genome sequencing analyses of clinical isolates in Great Britain also clearly distinguished three clusters in agreement with the three here reported [8]. All these data support revaluating the taxonomy of M. abscessus complex, to recognize three genomospecies M. abscessus (C1), M. bolletii (C2), and M. massiliense (C3); and four unnamed subspecies C2A, C2B; C3A, C3B.

M. abscessusprophagome

M. abscessus median GC% content is 64.2%, ranging from 62.7% (M. abscessus ATCC 19977) to 64.2% (strain Go 06). The GC% is not characteristic of the clusters as the median GC% content of C1, C2A and C3 is 64.2%, close to the median 64.1% GC% content in C2B.

However, there is a significant 14.7% variation in the genome length from 4.8-Mb (M. abscessus M154) to 5.51-Mb (M. abscessus M24) with a median of 5.07-Mb. The median of genome size is 5.07-Mb in C1, 4.89-Mb in C2A, 5.01-Mb in C2B and 5.28-Mb in C3. Differences in the genome size correlate with the number of prophage regions which are detected in 13/14 M. abscessus genomes (Figure 3): M. abscessus M154 (M. massiliense C2A) has the smallest genome encoding no prophage whereas M. bolletii M24 (C3) has the largest genome encoding seven prophage regions (Figure 3). Prophage regions comprise up to 5% of the genome lenght in M. abscessus M172. The number of prophage regions in other genomes is diverse, ranging from one to six regions encoding putative genes in the subsystem of phages, prophages, transposable elements and plasmids, which might contribute to species diversity [36]. The mean number of prophage regions is 2 in C1, 1.33 in C2A, 3.5 in C2B and 5 in C3. This observation confirms the particularity of C3: M. bolletii CIP108541T contains a 13-kb and a 63-kb prophage whereas M. bolletii strain M24 contains seven prophage regions including one 17-kb region homologous to the M. bolletii CIP108541T 13-kb region and a 27-kb region homologous to the M. massiliense CCUG 48898 50-kb region [37, 38] (Table 4). M. abscessus genomes encode putative phage-related genes necessary for phage replication as well as phage-tail protein, phage endolysin, capsid proteins (major protein and scaffold proteins) and phage tape measure protein. Both ends of this region encode putative phage integrases. M. abscessus genomes encode small prophage-like regions. However, only M. bolletii has been reported to produce a mycobacteriophage that we named Araucaria after we recently resolved its electron microscopy 3D structure [39]. M. abscessus M94 genome harbours one particular pseudo-tRNA spanning the region 51,150-57,394 in contig 33, which is not observed in the other M. abscessus genomes [40]. Phages have been reported to increase virulence of their host and encode antimicrobial resistance genes [41]. In M. abscessus however, no such genes were identified but phages could be targeted for the differentiation between the three M. abscessus genomospecies.

Figure 3
figure 3

Correlation between Mycobacterium abscessus genomes size (y axis) and the number of prophages (x axis).

Full size image
Table 4 Mycobacterium abscessus prophages
Full size table

M. abscessuscomplex resistome

As all mycobacteria, M. abscessus complex is embedded into a hydrophobic cell wall barrier to hydrophilic antibiotics. Accordingly, M. abscessus is multidrug resistant organisms exhibiting different drug resistance [42–44]. M. abscessus genomes encode many proteins potentially involved in drug-efflux systems, including members of the major facilitator family, ABC transporters and MmpL proteins; Small Multidrug Resistance-family, a family of lipophilic drug efflux proteins [45]; and a multidrug resistance stp protein similar to M. tuberculosis involved in spectinomycin and tetracycline resistance [46]. M. abscessus, M. bolletii and M. massiliense were reported to be in-vitro susceptible to amikacin; however, comparison with the M. tuberculosis H37Rv resistome and the antibiotic resistance databases indicate that M. abscessus encodes an aminoglycoside 29-N-acetyltransferase and aminoglycoside phosphotransferases involved in resistance to aminoglycosides. Also, genetic analyses disclosed 16S rRNA gene mutations conferring aminoglycoside resistance [4, 5, 47]. Indeed, the presence of a single rRNA operon in all of the M. abscessus genomes favours the occurrence of dominant mutations conferring resistance to aminoglycosides and macrolides. M. abscessus genomes encode a rifampin ADP-ribosyl transferase and monooxygenases potentially involved in resistance to rifampin and tetracyclines. Moreover, M. abscessus genomes encode three tet(M) genes conferring resistance to tetracyclyine and doxycycline; the number of tet(M) genes was correlated to the resistance to cyclines in Escherichia coli[48]. However, M. massiliense was reported to be susceptible and M. abscessus and M. bolletii to be resistant to doxycycline [49]. M. abscessus genomes encode resistance to fusidic acid, glycopeptides, MLS (Macrolide-Lincosamide-StreptograminB), phenicols, rifampicin, sulphonamide and trimethoprim. Also, M. abscessus genomes encode FolP homologs conferring resistance to cotrimoxazole, homolog of UDP-N- acetylglucosamine 1-carboxyvinyltransferase, a MurA protein conferring resistance to fosfomycin and homologs of 23S rRNA methylases conferring resistance to macrolides. Also, M. abscessus genome encodes an erm(41) gene which mutations were reported to confer clarithromycin resistance [50]. In-vitro tests showed that M. massiliense clinical isolates could be distinguished from M. abscessus isolates for their susceptibility to ciprofloxacin [51] whereas M. bolletii isolates were reported to be resistant to all quinolones [52]. A mutation at codon 90 in gyrA gene was reported in clinical isolates of M. abscessus exhibiting high resistance to ciprofloxacin [51]. This observation contrasts with our genome analysis, which found no such mutations, suggesting that other mechanisms of resistance may be involved in high-level resistance to quinolones [52]. Accordingly, we found that M. abscessus mycobacteria encode qepA2, a plasmidic gene conferring quinolone resistance in gram-negative bacteria [53]. M. abscessus mycobacteria were reported to be in-vitro resistant to penicillin, amoxicillin, cefoxitin, ceftriaxone, cefotaxime and imipenen [4, 5]. This contrasts with the fact that they encode Penicillin-binding proteins (PBPs), targets for β-lactam antibiotics (except for tabtoxinine-β-lactam, which inhibits glutamine synthetase), which are essential for peptidoglycan synthesis [54, 55]. M. abscessus genomes encode an Ambler class A β-lactamase homologous to β-lactamases in gram-negative bacteria and to two β-lactamases in M. tuberculosis. β-lactamases inhibitors have not been evaluated against M. abscessus sensu lato mycobacteria.

Genome-based analysis of host-interactions

M. abscessus are ubiquitous environmental organisms in soil and water [9] where they may have to cope with amoeba: M. chelonae, M. abscessus, M. massiliense and M. immunogenum were reported to survive within Acanthamoeba polyphaga tropohozoites and cysts [5]. Accordingly, our analyses indicate that M. abscessus genomes encode factors implicated in host interactions. The mean number of genes encoding proline-glutamate (PE), proline-proline glutamate (PPE), 10-kDa lipoprotein antigen precursor (LpqH), Mammalian Cell Entry (MCE), oxidoreductase (Yrbe) and type VII secretion system is of 70 in C1, 80 in C2A, 74 in C2B and 93 in C3. In M. abscessus, rough colonies lack mmpL4 (a gene required for glycopeptidolipid biosynthesis) and lost surface colonization, replication into human macrophages and stimulation of innate immune response; these observations suggested that glycopeptidolipid was a virulence factor [56–58]. Accordingly, glycopeptidolipids are required for sliding motility [59] and biofilm formation [60]. Glycopeptidolipids have also been suspected to inhibit phagocytosis of M. avium subsp. avium[61]. M. abscessus genomes encode MCE proteins similar to M. tuberculosis H37Rv. MCE operon promotes internalization of M. tuberculosis by mammalian cells [62] and initiates rapid induction of transcription of genes involved in substrate trafficking [63]. The number of mce operons which correlated with pathogenicity [64], varies from six in C2B to 13 in C3. In parallel, M. abscessus genomes encode 12 (C1) to 21 (C3A, C3B) copies of Yrbe proteins. As for secretion systems, recent evidences showed that mycobacteria evolved specialized type VII secretion systems to transport extracellular proteins across the cell wall [65]. Type VII secretion systems ESX-1 and ESX-5 are involved in cell-to-cell migration of M. tuberculosis[65, 66]. In M. abscessus, our analyses indicate that ESX-3 and ESX-4 systems are conserved (Figure 4). However, M. abscessus M139 (C2B) lacks two proteins of the ESX-3 system and M. abscessus M93 (C1) lacks ESAT-6 like and CFP-10-like proteins secreted by the ESX-4 system. Interestingly, M. abscessus M18 (C2A) encodes ESAT-6 and CFP-10 proteins secreted by ESX-1 system. In addition, there are two or three PE and six (M. massiliense, M. abscessus M115 or M. abscessus 47 J26) to 12 (M. bolletii M24) PPE proteins, which are reported to be involved in the virulence of M. tuberculosis[67]. Our analyses further indicated that proteins related to phenazine biosynthesis, homogentisate catabolism, phenylacetic acid degradation and DNA degradation might have been transferred from Actinobacteria (e.g. Rhodococcus spp., Streptomyces spp.) and pseudomonas (Pseudomonas aeruginosa and Burkholderia cepacia). Although distantly related, these bacteria share the same ecosystem as M. abscessus within cystic fibrosis microbiota.

Figure 4
figure 4

Heatmap clusterisation of Mycobacterium abscessus type VII secretion system compared to Mycobacterium tuberculosis H37Rv. M. abscessus strains are listed on the left side of the map.

Full size image

Conclusions

Our in-depth genomic analyses indicate that M. abscessus has a non-conservative genome, suggesting the possibility of on-going transfer of additional genetic material. Unsurprisingly, M. abscessus has already acquired antibiotic resistance. Also, phages have mediated diversity and horizontal gene transfer which drived the rapid evolution of this complex. Indeed, gene transfers have driven the evolution of M. abscessus towards three different genomospecies M. abscessus, M. massiliense and M. bolletii; and the evolution of four different yet unnamed subspecies. Each genomospecies has its own specificities in terms of genome size, prophagome and genome content. We identified 66 genes uniquely present in each genomospecies; these genes could be used in refined detection and identification of M. abscessus organisms. These genomic differences support differences in host interactions and the clinical presentation of infection with M. massiliense (C2A and C2B) being more virulent than the two other genomospecies. Host-interaction factors may contribute to the ability of M. abscessus to colonize mammalian hosts where its respiratory tract habitat put it in close proximity to other serious opportunist pathogens which can act as donors of additional host-interaction factors.

Here reported informations regarding differences between M. abscessus genomespecies will help understanding their pathogenesis factors and could reveal new, more specific targets for drug design and diagnosis tools.

Methods

Genome dataset

The whole genomes of 14 M. abscessus strains were downloaded from Genbank (Table 1). The genomic sequence, either contigs or finished genomes were concatenated to one pseudogenome per genome.

Prophage detection and genome annotation

Protein sequences were predicted using prodigal software [26] to generate normalized files containing the combined protein sequences of all 14 genomes. Prophage regions were detected using PHAST software (Table 4). Predicted proteins were annotated using BLASTp against the National Center for Biotechnology Information (NCBI) non-redundant (NR) database, UNIPROT (http://www.uniprot.org/), the Clusters of Orthologous Groups (COG) [33] and a home-made antibiotic resistance gene database.

Genome clustering and calculation of core genomes

Proteome sequences were compared using by BlastP and pairwise alignments using ClustalW and the ANI was determined by the mean percentage of nucleotide sequence identity of core proteins [29]. We clustered the M. abscessus homologous genes using orthoMCL [28] on the translated protein sequences of all predicted genes with a conservative parameter value of 50% sequence identity. The determination of the different unique core genomes was based on the homology clusters found by orthoMCL.

Phylogenetic analysis

M. abscessus proteomes were aligned using Mauve software [30] to infer phylogeny using the Neighbor-Net algorithm in the package SplitsTree4 [31]. The orthologous group data found by orthoMCL were used to construct a whole-genome phylogenetic tree based on gene content. We generated a matrix of binary discrete characters (“0” and “1” for absence and presence, respectively) [68]. Using this matrix, we constructed a phylogenetic tree implementing the neighbor-joining (NJ) method within SplitsTree4 [31].

Availability of supporting data

The data set of Figure 1C supporting the results of this article is available in the TreeBase (http://treebase.org/treebase-web/home.html) repository, under the accession URL http://purl.org/phylo/treebase/phylows/study/TB2:S15632.

Reviewer access URL: http://purl.org/phylo/treebase/phylows/study/TB2:S15632?x-access-code=6fa2ebc53b96e3ae412a8df19187ab41&format=html.

The data sets of Figure 1A and B supporting the results of this article are included as the Additional file 1.

The data sets of Figure 4 supporting the results of this article are included as the Additional file 2.