Escherichia coli DH5α, the strain used for cloning, was grown in LB Lennox (BD, Difco) medium at 37°C. M. abscessus subsp. massiliense (Mmas) CIP108297, a clinical isolate recovered from the sputum and bronchoalveolar fluid of a patient with hemoptoic pneumonia (Adekambi et al., 2004) was grown at 37°C in Middlebrook 7H9-OADC broth (BD, Difco) supplemented with 0.05% Tween 80, Mueller-Hinton II broth (BD, Difco), minimal Glycerol-Alanine-Salt (GAS) medium supplemented with 0.05% tyloxapol (pH 6.6), or on Middlebrook 7H11-OADC agar (BD, Difco). Kanamycin (Kan), hygromycin (Hyg) and streptomycin (Str) were added to final concentrations of 400, 1,000, and 200 μg/ml, respectively.

Recombineering was used to inactivate the MMCCUG48898_0905 (herein renamed mmpA) and MMCCUG48898_0906 (herein renamed mmpB) genes of Mmas CIP108297 by allelic replacement. To this end, the Gp60 and Gp61 recombineering proteins from mycobacteriophage Che9c were expressed from the replicative plasmid pJV53-XylE under control of an acetamide-inducible promoter as described (van Kessel and Hatfull, 2007; van Kessel et al., 2008; Calado Nogueira de Moura et al., 2014). Acetamide-induced Mmas CIP108297 harboring pJV53-XylE were electro-transformed with linear allelic exchange substrates encompassing the mmpA and mmpB loci and double-crossover mutants were isolated on Str-containing medium. The linear allelic substrate used to delete the mmpA locus was generated by bracketing the streptomycin-resistance cassette from pHP45Ω with 571–500 bp of upstream and downstream DNA sequence flanking the internal XmnI restriction site of mmpA, respectively. The linear allelic substrate used to delete the mmpB locus was generated by similarly bracketing the streptomycin-resistance cassette with 565–357 bp of upstream and downstream DNA sequence immediately flanking the internal XmnI restriction site of mmpB. The linear allelic substrate used to try and delete both porin genes simultaneously was generated by bracketing the streptomycin-resistance cassette with 571 bp of DNA sequence located upstream the XmnI site of mmpA and 357 bp of DNA sequence downstream the XmnI site of mmpB.

pOMK-mmpA, pOMK-mmpB, and pOMK-mmpAB, the replicative plasmids used for the complementation of the knock-out mutants, were constructed by cloning mmpA, mmpB or a 2.1 kb region encompassing both genes in the multicopy plasmid pOMK (Calado Nogueira de Moura et al., 2014). The mmpA and mmpB genes are expressed from their own promoter in the complementation constructs. The sequences of the primers used to generate the different constructs are provided in Supplementary Table 1.

The susceptibility of Mmas to commercial glutaraldehyde (GTA) in suspension tests, MIC determinations and [U-¹⁴C]-glucose (5 mCi/mmol; American Radiolabeled Chemicals) uptake experiments were performed as described in Calado Nogueira de Moura et al., 2014. For in vitro NO susceptibility studies, Mmas strains (approximately 2 × 10⁷ cells) grown in 7H9-OADC medium were resuspended in 1 ml phosphate buffer (pH 7.4) and exposed to 10 mM Spermine NONOate (Cayman Chemical) for 4–24 h at 25°C at which time point viable CFUs were enumerated by plating serial dilutions of the cell suspensions on 7H11-OADC agar.

Porin extraction and immunoblot analysis using anti-MspA antibodies (Stahl et al., 2001) was carried out as described previously (Calado Nogueira de Moura et al., 2014).

Mmas RNA was extracted from 5-ml cultures grown to an OD₆₀₀ of 0.2 using the Direct-zol RNA Miniprep kit (Zymo Research) per the manufacturer’s instructions. Reverse transcription reactions were carried out using the Superscript V kit (Invitrogen), and qRT-PCRs were run using the SsoFast Evagreen Supermix kit (Bio-Rad) as per the manufacturers’ protocols and analyzed on a CFX96 real-time PCR machine (Biorad). PCR conditions: 98°C (2 min; enzyme activation), followed by 35 cycles of 98°C (5 sec; denaturation) and 55°C (5 sec; annealing/extension). Mock reactions (no reverse transcription) were done on each RNA sample to rule out DNA contamination. The target cDNA was normalized internally to the sigA cDNA levels in the same sample. The primer sequences are provided in Supplementary Table 1.

A previously published global collection of MABSC isolates, consisting of 526 isolates associated with pulmonary infections and 189 isolates collected from nine separate outbreak locations in Brazil, was used in this study (Bryant et al., 2016; Everall et al., 2017). These isolates were further supplemented by an additional 665 longitudinal isolates associated with 190 patients with Cystic Fibrosis (CF) (Bryant et al., 2021). The Illumina Hiseq 2,500 paired end raw reads of all the isolates were mapped using BWA-MEM v0.7.12 to the reference genome Mabs ATCC 19977 (Li. 2013). Assemblies for all the isolates were generated and optimized using Velvet and Velvet optimizer, respectively (Zerbino and Birney, 2008). The final draft assemblies were annotated using PROKKA (Seemann, 2014).

An alignment was produced from the mapping of 526 isolates associated with pulmonary infections and 188 isolates associated with the post-surgical wound infection epidemic in Brazil to the Mabs ATCC 19977 reference genome, using SAMtools v1.2 and BCFtools v4.2, with parameters previously described (Li et al., 2009; Harris et al., 2010). Briefly, a variant was inferred if the following criteria were met: A mapping quality of 20, a base call quality of 50, eight reads supporting the alternative allele and three reads supporting the alternative allele on either strand. Heterozygous SNP sites were masked. An alignment of the variable positions was created using SNP-sites and used to produce a maximum likelihood phylogeny with RAxML (Stamatakis, 2014; Page et al., 2016). Hundred bootstrap replicates were performed.

A local tblastx search was performed to detect homologs of msp-like porin genes in the MABSC genomes. A local database was made using the nucleotide sequences of mmpA (MAB_1080), mmpB (MAB_1081) and the possible mmpC homolog (MAB_2800). The assemblies of 1,382 MABSC isolates were queried against this database with msp-like porin gene homologs predicted if a match had a percent identity greater than 50%, E-value < 0.05 and a match length greater than or equal to 50% of the length of the reference gene of mmpA (672 bps), mmpB (669 bps) and mmpC (741 bps).

The coverage over the porin locus containing the mmpA and mmpB genes was determined for each isolate using SAMtools v1.3, with a mapping quality threshold of 20 (probability mapped to correct location 99%), base call quality of 30 (base call is 99.9% accurate) (Li et al., 2009). Presence of both msp-like porin genes at this locus was determined by an average depth of coverage greater than eight in the 436 bp intergenic region between mmpA and mmpB and an average depth of coverage greater than eight over both mmpA and mmpB. Deletion of both mmp-like porin genes at this locus was determined by the absence of coverage (average of less than eight reads mapped) in the intergenic region between mmpA and mmpB and a reduction in depth of coverage greater than 20 between the 1,000 bp region upstream of the start of the porin loci and the 94 bp signal peptide region of mmpA. The presence of a single mmp-like porin gene at this locus was determined by absence of coverage in the intergenic region between mmpA and mmpB (average of less than eight reads mapped) and no reduction in depth of coverage greater than 20 between the 1,000 bp flanking region upstream of the porin loci and the 94 bp signal peptide region of mmpA. The figures were made using easyfig (Sullivan et al., 2011). A two tailed Fisher’s exact test was performed to determine if there was an association between the mmpA and mmpB porin locus configuration and isolates classified as clustered and more virulent and those classified as unclustered by Bryant et al. (2016).

Where only one mmp-like porin gene was present at the mmpA/mmpB porin locus, an alignment of the nucleotide sequence of this gene, and the reference mmpA (MAB_1080) and mmpB (MAB_1081) genes was created using muscle v3.8.31 (Edgar, 2004). The position at which the recombination event occurred was predicted based on where the mmp-like porin gene of the isolate with one mmp-like porin gene at the mmpA/mmpB porin locus changed from being more similar to mmpA to being more similar to mmpB.

The raw reads of the 1,382 isolates were further mapped to the Mabs ATCC 19977 mmpA/mmpB porin locus using BWA-MEM v0.7.12 (Li, 2013). Variants were called using SAMtools v1.2 and BCFtools v4.2 with the following parameters (Li et al., 2009): A base call quality of 30 and mapping quality of 20, a minimum of eight reads supporting the alternative base, with a minimum of three reads supporting the alternative base on each strand. Variants, including Indels, were only determined for isolates predicted to encode both mmpA and mmpB at this locus.

Specific-pathogen-free female SCID and GMCSF knock-out mice, from 6 to 8 weeks old, were purchased from the Jackson Laboratories, Bar Harbor, Maine. Mice were maintained in the Biosafety Level III animal laboratory at Colorado State University, and were given sterile water, mouse chow, bedding, and enrichment for the duration of the experiments. The specific pathogen-free nature of the mouse colonies was demonstrated by testing sentinel animals. All experimental protocols were approved by the Animal Care and Usage Committee of Colorado State University. The CSU animal assurance welfare number is A3572-01.

Bone marrow-derived macrophages (BMDM) from GMCSF and SCID knock-out mice were infected with mycobacteria at a one bacterium per one macrophage ratio and, 2 h later, the monolayers were washed to remove extracellular bacilli. The numbers of intracellular mycobacteria were measured by plating. Briefly, monolayers were washed at each time point to remove extracellular bacilli and 1 ml double-distilled H₂O containing 0.05% Tween 80 was added to monolayers and incubated for 10 min to lyse macrophages. After passing through a 26-gauge needle five times, the lysates were serially diluted and plated onto 7H11-OADC agar. Monolayers that weren’t lysed were replenished with fresh medium as previously described (Shang et al., 2011; Obregon-Henao et al., 2015). BMDM viability was assayed by trypan blue exclusion and determined by flow cytometry using BD Viability Counting Beads, as described by the manufacturer (BD PharMingen, San Jose, CA United States) cell viability-staining methods.

SCID mice were challenged with WT, mutant and complemented mutant strains using an infection calibrated to deliver 1.0 × 10⁶ bacilli per animal. At days 1 and 15 following infection, bacterial loads in the lungs, spleen and liver, and lung histology were determined. Bacterial counts were determined by plating serial dilutions of organ homogenates on 7H11-OADC agar and counting colony-forming units after 5–10 days of incubation at 30°C. A total of five animals were infected for each time point.

The whole lung from each mouse was fixed with 10% formalin in phosphate buffered saline (PBS). Tissue sections were stained using hematoxylin and eosin and acid-fast stain as previously reported (Bryant et al., 2016).

Data are presented using the mean values from five mice per group and from values from replicate samples and triplicate macrophage infection assays. The unpaired Student’s t-test test was used to assess statistical significance between infection groups.

The reference strains Mmas CIP108297 (alias CCUG 48898) and Mabs ATCC 19977 harbor two clustered Msp-like porin genes annotated as MMCCUG48898_0905/MMCCUG48898_0906 and MAB_1080/MAB_1081, respectively (Ripoll et al., 2009; Tettelin et al., 2012). We renamed the two Mmas genes mmpA and mmpB for Mycobacterium abscessus subsp. massiliense porins A and B. In addition, the genomes of Mmas CIP108297 and Mabs ATCC 19977 encode a third, more distantly related, porin gene (MAB_2800 in Mabs ATCC 19977) which we renamed mmpC and six other proteins sharing 22–53% identity with MmpC annotated as putative porins (MAB_2799; MAB_1321; MAB_3154c; MAB_2329c; MAB_0734 and MAB_1891). The mature MmpA and MmpB porin products share 97% identity between them and 72–73% identity with MspA from M. smegmatis (Supplementary Figure 1) (98.9–100% identity with the Mabs ATCC 19977 porin products, MAB_1080 and MAB_1081). MmpC shares only 22% identity (27.8% similarity) with MmpA and MmpB from Mabs ATCC 19977 and 22.3% identity (28.4% similarity) with MspA from M. smegmatis mc²155. 463 bp of nucleotide sequence separate mmpA from mmpB in Mmas CIP108297 and Mabs ATCC 19977 suggesting that the two genes are probably each monocistronic.

At least one (mmpA/mmpB- or mmpC-like) porin gene homolog was detected in all the MABSC genomes used in this study (n = 1,382). However, analysis of the depth of coverage of the mapped reads over the mmpA and mmpB porin loci of the 1,382 MABSC genomes used herein showed that there was variation in the number of porin genes encoded at this position. Overall, 1,059 isolates encoded both mmpA and mmpB porin genes at this locus, 312 isolates encoded a single mmp-like porin gene (a fusion of mmpA and mmpB; see further) and 11 isolates encoded no mmp-like porin genes at this locus.

Of the 526 isolates associated with pulmonary infections, 469 isolates encoded two mmp-like porin genes at this locus (Figure 1A), 54 encoded a single mmp-like porin fusion gene (Figure 1B) and three isolates encoded no mmp-like porin genes at this locus (Figure 1C). A single mmp-like porin gene, a fusion of mmpA and mmpB, was observed in 180 of 189 isolates associated with the post-surgical wound infection epidemic in Brazil (Duarte et al., 2009; Leao et al., 2010; Everall et al., 2017), while the remaining 9 isolates encoded two mmp-like porin genes at this locus.

Short indels, leading to the loss of function of the protein, were observed in mmpA in seven isolates associated with pulmonary infections. One isolate incurred a non-synonymous mutation resulting in a premature stop codon, W220^∗.

Changes in genome organization at the mmpA/mmpB locus were also observed in isolates collected from the same patient over time. Isolates from 15 of the 190 CF patients for which longitudinal samples were available were found to vary in the number of mmp-like porin genes encoded by this locus. Isolates from 12 of the 15 patients either encoded one or two mmp-like porins, one patient’s isolates encoded either two mmp-like porin genes or no mmp-like porin genes, one patient’s isolates encoded either one mmp-like porin gene or no mmp-like porin genes and the final patient’s isolates varied between encoding either two, one or no mmp-like porin genes at the mmpA/mmpB locus. A further four patients’ isolates were shown to have accumulated indels in mmpA over time. No non-synonymous mutations were found to have occurred within patients over time.

As noted above, where isolates were found to have only one mmp-like porin gene at the mmpA/mmpB porin locus, a recombination event had occurred leading to the fusion of mmpA and mmpB. We examined the sequences of the 312 predicted mmpA/mmpB fusion genes to determine where in the sequence the fusion occurred. Of the 312 predicted mmpA/mmpB fusion genes, it was only possible to extract the complete sequence of the gene from the draft assembly of 297 isolates. It was possible to predict the fusion position for 38 of these isolates, with six unique fusion positions predicted which spanned close to the entire length of mmpA (Figure 2). All six fusion positions were also seen in isolates where changes in the number of mmp-like porin genes encoded at the mmpA/mmpB porin locus were observed over time within a patient. All the predicted fusion positions fall after the signal peptide sequence at the N-terminal of MmpA and, therefore, all the fusion genes use the signal peptide sequence of this porin.

Because the changes occurring at mmpA/mmpB loci of longitudinally isolated strains from CF patients was suggestive of the involvement of MmpA and MmpB in the adaptation of MABSC to the host, subsequent studies were undertaken to determine whether these naturally occurring mutations arose as a result of the selective pressure of antibiotic treatment or as part of a more general adaptation of the bacterium to stresses imposed by the host environment.

To investigate the role of MmpA and MmpB in the physiology and virulence of MABSC, we first sought to knock-out each of these two porin genes of Mmas CIP108297 by allelic replacement. To this end, the recombineering method was used wherein Mmas CIP108297 expressing the Gp60 and Gp61 recombineering proteins from mycobacteriophage Che9c from pJV53-XylE (van Kessel and Hatfull, 2007; van Kessel et al., 2008; Calado Nogueira de Moura et al., 2014) was electrotransformed with linear allelic exchange substrates designed to replace the entire mmpA or mmpB loci with a streptomycin-resistance cassette. Twelve to 55 double crossover candidates were selected for each gene and analyzed by PCR. PCR analysis confirmed that allelic replacement had occurred in 18% of the mmpA candidate mutants and 50% of the mmpB candidate mutants (Figure 3A). Two passages in liquid broth devoid of Kan followed by plating on Str-containing agar yielded single mmpA and mmpB mutant clones devoid of pJV53-XylE plasmid. In contrast, three independent attempts to inactivate both porin genes simultaneously yielded no double crossover mutant suggesting that, similar to the situation in M. smegmatis (Stephan et al., 2005), the presence of at least one functional porin gene is required for Mmas CIP108297 growth under the experimental conditions used herein. Given that our genomic analysis identified 11 isolates devoid of mmpA and mmpB genes, one may only speculate that the mmpC gene and homologs present in these isolates can compensate for the lack of mmpA and mmpB in a way that is not achievable in Mmas CIP108297.

Porin production in the mutant strains was analyzed by Western blot using polyclonal antibodies to the MspA porin of M. smegmatis (Stahl et al., 2001). While porin products of the expected size (19.6 kDa) were detected in the wild-type (WT) Mmas CIP108297 strain, they were barely detectable in the mmpA knock-out; the intensity of the porin signal in this mutant was only about 6–10% of that detected in the WT parent. In comparison, the intensity of the band was decreased by less than 40–50% in the mmpB knock-out (Figure 3B). Complementation of the mutants with their respective porin gene or with both porin genes expressed from the replicative multicopy plasmid pOMK increased porin production in the mmpA knock-out relative to the WT parent 3.4- and 5-fold, respectively, and that in the mmpB knock-out mutant 1- and 10-fold, respectively.

Porin expression in the two mutants was further analyzed by qRT-PCR to check for a potential polar effect of the inactivation of mmpA on the expression of mmpB and for the existence of compensatory mechanisms regulating the level of expression of the two porin genes. The result of two independent experiments indicated that while the level of expression of mmpA was only about 1.3–2-fold less in the mmpB mutant relative to the WT strain, the expression of mmpB increased 4–7-fold in the mmpA knock-out (Figure 3C). qRT-PCR analysis further revealed that the level of expression of mmpA relative to sigA was > 14-fold greater than that of mmpB in WT Mmas CIP108297, thereby confirming MmpA as the major porin product of this isolate and explaining the more dramatic effect of knocking-out mmpA over mmpB on the total porin content of this strain (Figure 3B).

As expected, the disruption of mmpA impacted more dramatically the growth rate of Mmas CIP108297 than that of mmpB, the latter mutant growing at a similar rate as its WT parent in minimal GAS (Glycerol-Alanine-Salts) medium (Figure 4A) and at an intermediate rate between the mmpA knock-out and the WT strain in glucose-based medium (Figure 4B). Complementation with the individual or combined mmpA and mmpB genes restored WT growth in the mutants (Figures 4A,B).

Consistent with these observations, the mmpA knock-out mutant presented a significantly reduced [¹⁴C]-glucose uptake rate relative to the WT parent (8.75 ± 0.03 vs. 17.11 ± 0.47 pmoles of glucose/mg cells/min) which was restored to WT levels upon complementation with mmpA (17.86 ± 0.41 pmoles of glucose/mg cells/min), and beyond WT levels upon complementation with mmpA and mmpB (23.35 ± 1.67 pmoles of glucose/mg cells/min) (Figure 5). The mmpB knock-out mutant, in contrast, was as proficient at taking up [¹⁴C]-glucose as WT Mmas CIP108297 in this assay (16.23 ± 0.58 and 17.11 ± 0.47 pmoles of glucose/mg cells/min, respectively).

Because of the reported impact of porins on the susceptibility of M. smegmatis and M. chelonae to biocides and nitrosative stress (Fabrino et al., 2009; Svetlikova et al., 2009; Frenzel et al., 2011), we next set out to test the resistance of the mmpA and mmpB knock-out mutants to a number of stresses including various classes of antibiotics used in the clinical treatment of fast-growing mycobacterial infections, the high-level disinfectant glutaraldehyde (GTA), and nitric oxide (NO). Apart from a modest 2–4-fold increase in the resistance of the mmpA knock-out mutant to ciprofloxacin and imipenem, the two mutants presented susceptibility patterns to antibiotics (Table 1) and GTA (Supplementary Figure 2) comparable to those of their WT parent. In particular, both mutants displayed WT susceptibilities to drugs used in the clinical treatment of MABSC infections such as clarithromycin, erythromycin, azithromycin, amikacin, cefoxitin, linezolid and tigecycline. The mmpA and mmpB knock-outs further displayed WT susceptibility to NO in vitro (Supplementary Figure 3).

We investigated the capacity of Mmas reference strain CIP108297 and corresponding porin knock-out mutants to replicate and persist in BMDM to evaluate their virulence. The growth of each strain was determined in BMDM prepared from SCID and GMCSF knock-out mice by the method of plating to measure colony-forming units (CFU) over a period of 6 days. While no significant difference in bacterial uptake was noted between strains, both mutants demonstrated significantly increased replication in SCID (Figure 6A) and GMCSF knock-out (Figure 6B) BMDM compared to their WT parent, with the mmpB mutant consistently displaying the most acute virulence phenotype of the two. Throughout the experiment, BMDM cellular viability (as assessed by the percentage of propidium iodide positive dead cells) was above 95% and similar for all strains indicating that the porin mutants were not more cytotoxic than their WT parent in these macrophage models (data not shown). The replication of the mutants in macrophages was restored to or below WT levels upon genetic complementation with both porin genes (Figure 6B), and between WT and knock-out levels upon complementation with the individual WT copy of mmpA or mmpB (Figures 6A,B).

We used a SCID mouse intravenous infection model to compare the virulence of Mmas WT, mutant and complemented mutant strains in vivo. The rationale for using this immunocompromised mouse model is related to the difficulty of establishing a high-level pulmonary infection with rapidly growing mycobacteria in immunocompetent mice (Obregon-Henao et al., 2015). Moreover, this model more closely reflects an infection which would occur in immunocompromised patients, a population particularly at risk for NTM infections (Chan et al., 2010; Prevots and Marras, 2015). Following intravenous infection, the bacterial loads in the lung, spleen and liver were quantified on days 1 and 15 post-infection. Consistent with the observations made in infected BMDM, infection with the mmpA and mmpB knock-out mutants resulted in increased bacterial loads in the lungs, spleen and liver at 15 days relative to the infection with the WT parent strain, with the mmpB mutant displaying the most dramatic increase in virulence (Figure 7). Virulence was restored to or below WT levels upon genetic complementation of the mmpB mutant with both porin genes, and that of the mmpA mutant with a WT copy of mmpA. The mmpB mutant complemented with both porin genes was used in this in vivo study because of the limited impact of complementing this mutant with mmpB only in murine macrophages (i.e., replication rate significantly above that of the WT parent strain) (Figures 6A,B).

Lung histopathology mirrored these observations in that more severe peribronchiolar inflammatory infiltrates and increased cellular foci with early granuloma formation were observed in the lung tissues of the animals infected with the mmpB knock-out strain (Figure 8C). In comparison, mice infected with the WT parent and the complemented mutant strains displayed similar lung histopathology with less inflammatory foci and pulmonary involvement, and more healthy lung tissue (Figures 8A,D,E). The mmpA mutant displayed an intermediate phenotype (Figure 8B).