Support from Phylogenomic Networks and Subspecies Signatures for Separation of Mycobacterium massiliense from Mycobacterium bolletii

Abstract

Mycobacterium abscessus subspecies classification has important clinical implications. We used phylogenomic network and amino acid analyses to provide evidence for the separation of Mycobacterium bolletii and Mycobacterium massiliense into two distinct subspecies which can potentially be differentiated rapidly by their protein signatures.

TEXT

Mycobacterium abscessus has become one of the most frequently isolated nontuberculous mycobacterium (NTM) in clinical laboratories. It is associated with chronic, recurrent infections that are difficult to treat, partly because of its resistance to many of the usual medications for NTM infections. This species was previously divided into three subspecies (M. abscessus, M. massiliense, and M. bolletii) based on biological and genetic differences (1–3). Currently, however, only two subspecies are recognized; while M. abscessus is retained as Mycobacterium abscessus subsp. abscessus, M. massiliense and M. bolletii are placed in the same subspecies designated Mycobacterium abscessus subsp. bolletii (4). This tenuous merging of M. massiliense and M. bolletii is still being debated as recent publications support the previous three-subspecies classification (5). Here, we present more evidence for the retention of the former three-subspecies taxonomic division, which correlates better with the expected treatment outcomes in infected patients (6).

(This research was conducted by J. L. Tan in partial fulfillment of the requirements for a Ph.D. from University of Malaya, Kuala Lumpur, Malaysia.)

For our genomic and amino acid analyses, we used 12 genomes from strains isolated in the Diagnostic Microbiology Laboratory of the University of Malaya Medical Centre (UMMC), Kuala Lumpur, Malaysia, and 41 downloaded from the NCBI Genome database on July 2014 (see Table S1 in the supplemental material). Eleven of the UMMC strains have been previously reported to be M. abscessus (M93, M94, and M152), M. bolletii (M24), and M. massiliense (M18, M115, M152, M172, M159, M156, and M148). One strain, M139, was shown to have an ambiguous taxonomic position in a number of studies (7, 8).

The protein sequences for all strains were retrieved using the self-training structural annotation algorithm of GeneMarkS (9). To define orthologous sequences, we used the CD-HIT program (10) with the following criteria: word length of 2, local sequence identity threshold of 0.4, alignment coverage for both sequences of 0.4, and greedy algorithm off. We also used the BLASTClust program with the following parameters: reference and query sequences must cover at least 40% of the aligned sequence and reference and query sequences must have a minimum identity of 40% (11). To reduce false-positive results due to algorithmic errors, only the consensus sequences from both programs were extracted and used as the final list of orthologs. Nonduplicated conserved protein orthologs were aligned in MAFFT (12).

The protein sequence alignments were used as the reference for codon alignments in PAL2NAL (13). The aligned nucleotide sequences were concatenated into supersequences for phylogenomic analysis using the Neighbor-Net algorithm implemented in SplitsTree4 (14). This algorithm was considered the best for the resolution of complex taxonomy (15). To assess the subspecies classification derived from our network tree, we looked for subspecies-specific polymorphisms previously described for the erythromycin ribosome methyltransferase (erm41) and 16S to 23S internal transcribed spacer (ITS) genes.

Our network-based phylogenomic tree showed reticulated branches leading to three clearly distinctive monophyletic groups representing the three subspecies of M. abscessus (Fig. 1). The M139 and the 5S strains (5S-0421, 5S-0422, 5S-0708, 5S-0817, 5S-0921, 5S-1212, 5S-1215, and 5S-0304) clustered with the other M. massiliense strains. None of the branches in any of the three major clusters bifurcated to the other two clusters. The presence of 3-dimensional-like splits within the branches indicated incompatible phylogenetic signals that are likely to be the result of recombination following the horizontal transfer of genetic material among strains. Indeed, the recombination among our M. abscessus strains is statistically supported by the pairwise homoplasy index (PHI) (P = 0) (16). The incompatible signals occurred at random points in the tree, suggesting that recombination has occurred in ancestral states and within the respective subspecies. We also noticed unusual conflicting signals within the M. massiliense cluster, appearing as a major reticulation connecting the M. massiliense strains and suggesting a higher degree of genetic recombination in M. massiliense compared to that in the other two subspecies. To test the validity of this network phylogenomics approach, we used it on three members of the M. avium complex and found a clear separation of Mycobacterium avium subsp. paratuberculosis, Mycobacterium avium subsp. hominissuis, and Mycobacterium avium subsp. avium into three distinctive monophyletic groups, as observed with the M. abscessus complex (see Fig. S1 in the supplemental material).

FIG 1.

Phylogenomic split network tree obtained from the concatenation of single-copy core genes from M. abscessus subspecies. M. massiliense (right), M. bolletii (center), and M. abscessus (left) can be seen clearly as distinct groups.

M. massiliense is known to be different from the other two subspecies in having a truncated erm41 with nucleotide deletions at the 64th to 65th and 159th to 432nd positions, as well as mutations in the ITS (a A to G substitution at the 60th position and a C insertion at the 102nd position) (2). M139 and the eight 5S strains previously classified as M. massiliense and appearing as M. massiliense in our phylogenomic tree did not show the erm41 features associated with M. massiliense (Fig. 2). M139 additionally lacked the ITS mutations characteristic of M. massiliense and did not show inducible resistance to macrolides (17). Overall, however, there was good concordance (83%) between the subspecies classifications by erm41 signatures and by the network tree.

FIG 2.

Multiple sequence alignments of erm41 showing features of M. massiliense M139 and 5S strains compared to those of the type strains of M. abscessus ATCC 19977^T, M. massiliense CCUG 48898^T, and M. bolletii BD^T. The M. massiliense signatures are (i) deletions at positions 64 and 65 and (ii) a 274-bp deletion after position 159.

In the multiple sequence alignment of the orthologous proteins from our 53 strains, we noted 46 proteins with at least one amino acid that can be used to differentiate the three subspecies (see Table S2 in the supplemental material) and another two proteins (thymidylate kinase [tk] and 30S ribosomal protein S3 [S3]) that can differentiate M. massiliense from the other two subspecies (Fig. 3 and 4). We used BLAST to search the amino acid sequences of tk3 and S3 against all Mycobacterium genomes in the NCBI database and found them in 37 and 44 species, respectively. After realigning against these mycobacterial species, we confirmed the amino acid signatures of tk (RALTRRSISQGLS at position 20 to 30) and S3 (ETGGNTSAEAPAETSTES at position 260 to 277) to be specific for M. massiliense (Fig. 3 and 4). The presence of these signatures in M139 and the 5S strains supported their classification as M. massiliense, in agreement with the classification by the phylogenomic network. They will need to be experimentally verified as suitable biomarkers for the identification of M. massiliense in clinical material.

FIG 3.

Consistent protein signatures in M. massiliense identified in multiple alignments of thymidylate kinase from M. abscessus and other selected mycobacteria: *23 strains of M. abscessus subsp. abscessus; ** 2 strains of M. bolletii; ***28 strains of M. massiliense.

FIG 4.

Consistent protein signatures in M. massiliense identified in multiple alignments of 30S ribosomal protein S3 from M. abscessus and other selected mycobacteria: *23 strains of M. abscessus subsp. abscessus; ** 2 strains of M. bolletii; ***28 strains of M. massiliense.

It is well known that M. abscessus subspecies exhibit different clinical and epidemiological features (18, 19). M. massiliense is more susceptible to antibiotics but is also more often associated with clinical infections. M. bolletii, on the other hand, is rarely isolated from clinical material but is more highly antibiotic resistant. While the reasons behind these differences are still unclear, there is sufficient justification for subspecies identification in patient care. Our analyses support the division of M. abscessus into three subspecies and the reinstatement of M. massiliense as a taxon independent of M. bolletii. The specific identification of these two subspecies which show different antibiotic susceptibilities will enable the clinician to prescribe appropriate antibiotics for the effective treatment of infections.