Mycobacterium abscessus Mutants with a Compromised Functional Link between the Type VII ESX-3 System and an Iron Uptake Mechanism Reliant on an Unusual Mycobactin Siderophore

The opportunistic pathogen Mycobacterium abscessus subsp. abscessus (Mab) has become an emerging public health threat due to the increasing number of Mab-associated chronic pulmonary disease cases. Treatment requires multiple drug courses and is often combined with surgical resection. Cure rates are only ~50% due to treatment failure and comorbidities. Deeper understanding of the biology of Mab is required to illuminate potential avenues for the development of better therapeutics against Mab infections. The ESX-3 type VII protein secretion system of Mab has an important role in host inflammatory and pathological responses during infection. In this work, we demonstrate a functional link between ESX-3 and an iron uptake system based on an unusual mycobactin-type siderophore (designated MBT Ab) and exploit this link to implement a large screen for transposon mutants with an impaired ESX-3. Most mutants we identified carry insertions in genes encoding predicted ESX-3 secretion machinery components or potential ESX-3 substrates. The mutants overproduce MBT Ab, a trait consistent with an iron uptake defect. Our characterization of MBT Ab revealed structural features reminiscent of nocardial mycobactin-like compounds with cytotoxicity. This finding raises the possibility that MBT Ab may play roles in pathogenesis unlinked to iron homeostasis. The mutants generated herein will facilitate research to better understand the role of ESX-3 and its interplay with the siderophore system.


Introduction
Mycobacterium abscessus subsp. abscessus (Mab), the best-known member of the M. abscessus complex, is a ubiquitous opportunistic nontuberculous mycobacterial pathogen responsible for community-acquired and healthcare-associated infections [1][2][3][4][5][6][7][8]. Mab has become an emerging public health menace due to the rising number of Mab-associated tuberculosis-like chronic pulmonary disease cases [1,[3][4][5][6]8]. Mab is considered the most pathogenic of the rapidly growing Mycobacterium species and accounts for 65-80% of the chronic pulmonary disease cases associated with this group [1,[3][4][5][6]. Patients with Mab-associated chronic pulmonary disease often have underlying risk factors such as cystic fibrosis, chronic obstructive pulmonary disease, previous mycobacterial lung infections, lipoid pneumonia, lung transplantation, or cancer [9][10][11][12]. The pulmonary disease caused most laboratory studies and herein. We have recently highlighted the presence of orthologues of the genes encoding the nonribosomal peptide synthetases, polyketide synthases, and other proteins involved in the synthesis of MBT/cMBT siderophores of Mtb and Mycobacterium smegmatis (Msm) [41,42] in Mab ATCC 19977 T and other species [43]. The presence of these orthologues in Mab suggests that the bacterium has the capacity to produce an MBT-type siderophore. Overall, it is clear that additional studies on the ESX-3 and siderophore systems of Mab are warranted.
In this work, we demonstrate a functional link between the ESX-3 and siderophore systems of Mab and exploit this link to identify mutants with an impaired ESX-3. We also show that Mab produces an unusual MBT-type variant with features reminiscent of nocardial MBT-like compounds with cytotoxicity or antiproliferative activity. Our findings provide further insight into the ESX-3 and siderophore systems of Mab and underline the differences between the iron-acquisition capabilities of the opportunistic pathogen and Mtb. The collection of novel Mab mutants generated in this study will facilitate future research to better understand the functional dimensions of ESX-3 and its interplay with the siderophore system in Mab.

Principle and Validation of the Screen for ESX-3-Impaired Mutants and Siderophore Production Probe
As noted above, Mtb ESX-3 mutants cultured under growth-permissive conditions that bypass the essentiality of ESX-3 display a distinct orange pigmentation (i.e., OP) phenotype caused by excessive accumulation of siderophore-Fe 3+ complexes. This observation led us to speculate whether ESX-3 mutants of Mab, in which the secretion system is not essential [36], would display OP on s7H11 plates, and whether, if so, such a phenotype could be used to screen libraries of Tn mutants for strains with a dysfunctional ESX-3 or other defects leading to OP. Before exploring these ideas, we probed the capacity of Mab to produce a presumptive MBT-type siderophore. We previously reported an orthology analysis of MBT/cMBT biosynthesis genes that highlighted the potential of Mab to produce an MBT-like siderophore [43]. These genes include mbtA, which encodes the predicted salicylic acid-specific adenylation enzyme (MbtA) that catalyzes the first committed step in MBT/cMBT backbone biosynthesis [42,43]. We have also shown that exogenous [ 14 C]salicylic acid can be utilized for biosynthesis of radiolabeled MBTs/cMBTs in Mtb and Msm cultures [42][43][44]. We used this labeling approach herein to probe the ability of Mab cultures to produce presumptive salicylic acid-derived siderophores by radio-TLC analysis. The results of this experiment revealed the production of a salicylic acid-derived compound (a presumptive MBT-type siderophore) with a retention factor (R f ) different from the R f of the MBT from Msm (Figure 1a). When the results of the TLC analysis are interpreted in the context of the published Mtb and Msm work, they suggest that Mab may indeed produce an MBT variant, an inference confirmed by the MS analysis described below.
We have previously shown that the rationally designed Mtb MbtA inhibitor salicyl-AMS suppresses MBT/cMBT biosynthesis and has potent antimicrobial activity against Mtb and Msm (strain producing only MBT/cMBT siderophores) conditional to culturing in ironlimiting medium, where the siderophores are needed for growth [44,45]. We assessed the antimicrobial activity of the inhibitor against Mab. The analysis revealed robust antimicrobial activity conditional to culturing in iron-limiting medium (iron-rich IC 50 to iron-limiting IC 50 ratio~2000; iron-rich MIC to iron-limiting MIC ratio~100) (Figure 1b). The activity of salicyl-AMS against Mab paralleled the effect of the inhibitor against Mtb and Msm [44,45]. The findings of the antimicrobial testing against Mab suggest that a siderophore is likely to be particularly critical for the bacterium's growth in iron-limiting medium. Having validated the production of a salicylic acid-derived compound (a presumptive MBT-type siderophore) in Mab, we proceeded to explore the screen concept. To this end, we generated a small pilot library of Tn mutants (~2500) and screened the library for colonies with OP on s7H11 plates. The pilot screen rendered one colony with OP (isolate M1). The orange coloration of M1 was clearly distinguishable from the off-white color of the rest of the colonies on the plates and resembled the color of an MBT J-Fe 3+ standard [46] (Figure 2). Tn insertion site analysis indicated that the mutant had a single insertion ( Figure S1), and it mapped to eccC3 (MAB_2232c), the gene encoding the predicted core component EccC3 of the ESX-3 secretion machinery ( Figure 3). By analogy with the Mtb ESX-3 mutant, we hypothesized that the OP of M1 (hereafter referred to as M1 eccC3 ) is caused by accumulation of a siderophore-Fe 3+ complex secondary to an ESX-3 malfunction.  Genes are depicted as arrows labeled according to the single letter or number key shown below the gene cluster diagram. The number of insertion mutants isolated for each gene is given in parentheses. The 5 and 3 chromosomal coordinates of the genomic segment depicted are indicated. Tn insertions with the zeo R gene in the same and opposite orientations relative to the Mab genes are marked above and below the locus, respectively. The Tn mutants marked with an asterisk (*) are those displayed in Figure 4. The location of a predicted iron box (IB) sequence (shown in Figure 5) for IdeR binding upstream eccA3 is represented. Gene names noted for the predicted canonical ESX-3 substrates (pe5, ppe4, esxG, and esxH) and the conserved system components involved in substrate secretion (eccA3-eccD3, espG3, and mycP) are as per standard esx gene nomenclature [21]. Locus tags for the Mab genes (MAB_) and their orthologues in M. tuberculosis (Mtb, Rv_) and M. smegmatis (Msm, MSMEG_) are shown. Mab esx-3 locus genes hit by the Tn in our collection of mutants are marked with a filled diamond symbol ( ). Mtb and Msm genes for which individual knockouts have been reported are marked with an open diamond symbol (♦) [26,33]. The gene pairs boxed represent reported double-gene deletions [26,33]. § Mtb genes listed are essential for growth on iron-rich 7H10 agar (except for esxG and esxH, for which mutation leads to a growth defect) by analysis of Himar1 Tn libraries [47]. (b) Assemblage model of ESX-3 components and substrates encoded in the archetypal esx-3 locus. Two independent cryo-electron microscopy investigations of the Msm ESX-3 secretion complex have determined similar structures, composed of two dimerized core complexes each composed of one EccE, EccB, EccC, and two EccD components, in an organization simplified in the model depicted here. This ESX-3 dimer further trimerizes into a final hexamer structure similar to that seen for ESX-1 and ESX-5 complexes, forming the channel across which ESX-3 substrate components can be translocated. The fifth conserved ESX-3 membrane component, MycP, is not tightly associated with this described core secretion complex [48,49]. The yellow rounded rectangles (labeled with question marks) in the mycobacterial outer membrane layer (OM) signify the potential existence of protein(s) involved in facilitating the passage of ESX-3 substrates through the mycomembrane (adapted from [21,48,49]). AG, arabinogalactan layer; PG, peptidoglycan layer; CM, cytoplasmic membrane.
Unexpectedly, during our initial experiments, we noticed a sporadic weakening or loss of the OP mutant phenotype on s7H11 plates. We eventually linked this puzzling phenomenon to changes in the lots of the commercial bovine serum albumin (BSA) used to prepare the standard ADN supplement added to 7H11. We found that BSA was needed in the agar to observe the OP phenotype ( Figure S2a), and that some BSA lots supported the development of OP poorly or not at all. We also found that increasing the concentration of the BSA from such lots in the medium supplement from the standard 0.5% to 1.25-2.5% afforded reproducible OP development ( Figure S2b). In view of these early findings, we subsequently prepared the ADN supplement with BSA at higher concentrations (typically 1.5%) depending on each BSA lot's pre-assessed performance in supporting OP development by previously collected mutant strains. In all, the findings of our pilot experiments provided the first indication of a functional link between the ESX-3 and siderophore systems of Mab, validated the screening approach, and set the methodological stage for a large-scale screen.

Large-Scale Screen and Genetic Characterization of Isolates with OP Phenotype
The results described above encouraged us to scale up the screen. We screened a total of~196,000 mutants, a library reaching an average of one Tn insertion per 26 bp of genomic DNA. This frequency corresponds to a theoretical probability of the Tn missing the reference~1000-bp average size gene of Mab by chance of~10 −16 ( Figure S3). Moreover, the library could be considered to be near-saturation (i.e., having mutants of nearly all nonessential genes represented) based on the observation that reaching near-saturation in bacterial Tn insertion libraries requires 6000-7000 mutants per megabase of genome [50][51][52]. This threshold was surpassed by~6-fold in our library.
The screen led to the identification of 54 isolates with OP phenotype (including M1 eccC3 ). The results of the Southern blot analysis confirmed a single insertion in 52 of the isolates ( Figure S1). The remaining two isolates (M7 and M63) showed two hybridization bands that could have originated from two insertions or incomplete DNA digestion (not shown). However, both M7 and M63 had at least one insertion in an esx-3 locus gene. The results of the insertion site determinations for all isolates are summarized in Figure 3a and Table 1. Of the 52 isolates with a single insertion, 47 carried an insertion in one of the eleven genes in the esx-3 locus, and two had an insertion in a putative promoter region at the 5 -end of the locus ( Figure 3). The remaining five isolates had insertions outside either the esx-3 locus or its predicted promoter region ( Table 1). Four of these insertions mapped to four different genes, i.e., MAB_1912c, MAB_2276c, MAB_4275c, and MAB_4783. The fifth insertion mapped to the putative promoter region of MAB_4537c. These five genes encode proteins of unknown function and different degrees of conservation across Mycobacterium species (Table S4). To our knowledge, there are no reports of functional links between any of these genes and ESX or siderophore systems, or iron or zinc homeostasis. Notably, however, MAB_4783 encodes a paralogue of the predicted ESX-3 substrate Ppe4 (MAB_2230c; 37% amino acid identity) with close orthologues only present in the M. abscessus complex, M. chelonae, and four species closely related to the M. chelonae-M. abscessus group (Table S4).  2 The position and orientation of the transposon in each esx-3 locus mutant are displayed in Figure 3. 3 The direct repeats resulting from the most common 9-bp or the less frequent 10-bp duplication [53] at the transposon insertion point are depicted. 4 The 5 and 3 chromosomal coordinates of the duplicated segment at the insertion site are noted on the right column. 5 The genes below the gray-colored row are outside the esx-3 locus. Figure 4 shows the OP phenotype of single colonies and macrocolonies (arising from spot-inoculation of liquid cultures onto agar [54,55]) of nine representative esx-3 locus mutants (M65 eccA3 , M59 eccB3 , M23 eccC3 , M75 eccD3 , M5 eccE3 , M57 mycP3 , M72 esxH , M45 ppe4 , and M64 pe5 ; highlighted in Figure 3), two mutants with an insertion outside the esx-3 locus (M83 MAB_4783 and M50 MAB_2276c ), and the respective genetic complementation control strain for each of the eleven mutants. Robust complementation was achieved for M65 eccA3 , M59 eccB3 , M75 eccD3 , M5 eccE3 , M57 mycP3 , and M45 ppe4 , as demonstrated by a drastic reduction in pigmentation intensity in single colonies ( Figure 4a) and macrocolonies (Figure 4b) of their respective complementation control strains. Weak partial complementation was observed for the remaining five mutants (i.e., M23 eccC3 , M72 esxH , M64 pe5 , M50 MAB_2276c , and M83 MAB_4783 ). The complementation controls of these mutants showed only a slight reduction in pigmentation intensity, which was more evident in single colonies ( Figure 4a) than in macrocolonies ( Figure 4b). Interestingly, the esxH mutant could be fully complemented by an esxG-esxH fragment ( Figure S4). The remaining three mutants with insertions outside the esx-3 locus (M56 MAB_1912c , P5 MAB_4275c , and M55 MAB_4537c ) did not show signs of complementation (not shown) and were not further investigated.  Top and bottom rows in (a,b) depict the Tn mutants and their corresponding genetic complementation control strains, respectively. The wild-type strain (WT) and the mutants in the top rows carried pML1335 (empty), the vector used in the genetic complementation experiments, so that all strains could be grown in the same antibiotic-containing medium. Images of single colonies (7 days old) and spot inoculation-derived macrocolonies (5 days old; 1.5 µL inoculum, culture OD 600 = 1.0) were digitally captured using an Olympus SZX7 stereo microscope (Olympus Corp., Center Valley, PA, USA) and a T2i DSLR camera (Canon Inc., Melville, NY, USA), respectively. Scale bars are shown.

Sequence Bioinformatics Suggests the esx-3 Locus of M. abscessus Is Regulated in Response to Iron by IdeR
As noted above, M74 peccA3 and M68 peccA3 have an insertion in a putative promoter region upstream of eccA3 ( Figure 3, Table 1). A search for promoter elements in this region revealed potential -10 and -35 sequence motifs upstream of the Tn insertions ( Figure 5). Notably, these sequence motifs differed by only two mismatches (-10 motif) from those in the promoters upstream of the Mtb and Msm eccA3 orthologues. These findings are in line with the notion that disruption of the transcription of esx-3 locus genes downstream of the Tn insertion in M74 peccA3 and M68 peccA3 causes the OP phenotype of the mutants. Future studies to probe the hypothesized basis of the phenotype observed in M74 peccA3 and M68 peccA3 are warranted. The promoter of the Mab eccA3 orthologue in Msm (MSMEG_0615) is controlled by the iron-dependent regulator IdeR [22,56,58]. In contrast, the promoter of the eccA3 counterpart in Mtb (Rv0282) is controlled by both IdeR and the zinc-dependent regulator Zur [22,27,56,58]. We searched the putative promoter region of Mab eccA3 for potential binding sites for IdeR (iron box) and for Zur (Zur box). The searches revealed a potential iron box with a location analogous to that of the iron boxes present in the esx-3 loci of Msm and Mtb ( Figure 5). The analysis, however, did not reveal a putative Zur box. We extended the iron box search to the rest of the Mab genome to identify other potential IdeR-regulated genes. The search revealed several iron boxes in the chromosome, many of which have counterparts in orthologous loci of other mycobacteria ( Figure 6). Notably, a putative iron box was found in the predicted promoter regions of MAB_4783 (the ppe4 paralogue disrupted in M83 MAB_4783 ) and the MBT biosynthesis genes mbtT (MAB_2121c), mbtE (MAB_2122), mbtA (MAB_2247c), mbtE (MAB_2248), and mbtI (MAB_2245) [42,43] ( Figure 6). These results support the idea that transcription of the Mab esx-3 locus genes, MAB_4783, and the MBT biosynthesis genes might be co-regulated in response to intracellular iron levels by Mab IdeR (MAB_3029; 86% identity with Mtb IdeR). Overall, these findings strengthen the hypothesized physiological interplay between the ESX-3 and MBT siderophore systems of Mab noted above and underscore a role of Mab ESX-3 in iron homeostasis. Our findings also highlight directions for future studies on the expression levels of esx-3 and MBT biosynthesis genes and their hypothesized iron/IdeR-dependent regulation.

The Orange Pigmentation Phenotype of M. abscessus Cultures Is Influenced by Iron Availability
Based on our analysis of putative promoter regions, we hypothesize that growth in iron-limiting conditions will cause de-repression of IdeR-regulated esx-3 and MBT biosynthesis genes. In addition, we predict that the upregulation of MBT biosynthesis genes will lead to an increase in MBT accumulation in both WT and mutant cultures. To test this view, we probed for OP in cultures of selected esx-3 mutants (M45 ppe4 , M57 mycP3 , M65 eccA3 , and M75 eccD3 ) grown in GAST (iron-limiting, BSA free) and GAST+Fe (GAST supplemented with 100 µM FeCl 3 ), two media where the mutants and the WT strain display comparable growth (see below). In line with our expectation, an OP analogous to that seen in the colonies and macrocolonies of the mutants was observed in the GAST cultures. While these cultures had no visible OP initially (presumably because restricted iron in the medium does not support substantial formation of the orange MBT-Fe 3+ complex), post-culturing addition of excess FeCl 3 (5 mM) to the cultures for maximal conversion of colorless MBT into MBT-Fe 3+ led to an instantaneous development of OP in the WT and the four mutants tested; i.e., M45 ppe4 and M57 mycP3 (Figures 7a and S5a), and M65 eccA3 and M75 eccD3 (not shown). The pigmentation was chiefly associated with the cells, as determined by visual inspection of both the cell pellets and spent culture supernatants (Figures 7a and S5b). Nonetheless, the supernatants retained some pigmentation detectable by the naked eye and quantifiable by spectrophotometric analysis (Figures 7b and S5b). The pigmentation of the esx-3 mutants was markedly more intense than that observed in the WT strain (M45 ppe4 and M57 mycP3 , Figures 7a and S5a) or essentially indistinguishable from it (M65 eccA3 and M75 eccD3 , not shown). In the former case, genetic complementation reduced the pigmentation intensity to WT levels. Growth of the WT and mutant strains in GAST+Fe, however, did not lead to visible pigmentation in any of the cultures, with or without post-culturing addition of excess FeCl 3 (Figures 7a and S5a). The findings with the GAST+Fe are in agreement with the hypothesized IdeR-dependent downregulation of MBT biosynthesis in both WT and mutant strains in the iron-rich medium.

Growth Characterization of esx-3 Locus Mutants, M83 MAB_4783 , and M50 MAB_2276c
To gain further insight into the effect of the insertions in the mutants, we compared the growth of the WT strain and the representative mutants displayed in Figure 4 in standard s7H9 (iron rich), GAST (iron limiting), and GAST+Fe (iron rich) media. Figure 8 summarizes the results of growth curves shown in Figure S6. In s7H9, only the eccC3 mutant had an appreciable growth delay compared to the WT strain. The defect was modest, and it was not present in the mutant's complementation control strain ( Figure S6a). In GAST, however, all esx-3 locus mutants with insertions in genes of ESX-3 components predicted to be involved in secretion (ecc genes and mycP3) showed some degree of growth delay relative to the WT strain ( Figure S6b). The delay was small for the eccA3, eccB3, eccD3, eccE3, and mycP3 mutants, but drastic for the eccC3 mutant. The growth of the complementation control strains of these six mutants was no different from that of the WT strain ( Figure S6b). The growth defect of the eccA3, eccB3, eccD3, eccE3, and mycP3 mutants was also absent in GAST+Fe ( Figure S6c). In contrast, the growth defect of the eccC3 mutant was not fully suppressed in GAST+Fe, where the mutant had an appreciable growth delay analogous to that seen in the iron-rich s7H9 medium. The defect was, however, eliminated by genetic complementation ( Figure S6c). or left untreated (gray bars). A 450 , blank (sterile broth)-corrected absorbance at 450 nm. The data represent means ± SE from three cultures. The wild-type strain (WT) and M45 ppe4 carried pML1335 (empty), the vector used in the genetic complementation experiments, so that they could be grown along with the complemented strain M45 ppe4 -C in the same antibiotic-containing medium.
Unlike the secretion machinery mutants noted above, the mutants with insertions in predicted ESX-3 substrate genes (i.e., pe5, ppe4, and esxH mutants) did not show a growth abnormality in GAST ( Figure S6b). These mutants also had normal growth in GAST+Fe ( Figure S6c). The medium-specific growth pattern of the ppe4 mutant was not recapitulated in the MAB_4783 mutant, which carried the insertion in a ppe4 paralogue (Figure S6a-c). Both the MAB_4783 mutant and its complementation control strain showed a slight growth delay in GAST ( Figure S6b), a defect absent in GAST+Fe ( Figure S6c). Lastly, the mutant with the insertion in MAB_2276c (encoding a putative regulatory protein, Table S4) had no growth delay in any media ( Figure S6b,c). Overall, the growth characterization data indicate that insertions disrupting the ESX-3 secretion apparatus lead to a fitness cost conditional to (or in the case of eccC3 exacerbated by) culturing in the iron-limiting medium. In contrast, insertions disrupting the predicted canonical ESX-3 substrate genes do not impact fitness in iron-limiting or iron-rich media.

Mass Spectrometry Analysis of M. abscessus Mycobactin
To gain insight into the structural features of the presumptive MBT-type siderophore of Mab (hereafter called MBT Ab), we undertook a siderophore-targeted, LC-MS/MS-based metabolomics approach. To this end, we explored the structure of the siderophore produced by the WT strain and the mycP3 Tn mutant, which has a disruption of the gene encoding the predicted MycP (mycosin protease) component conserved in type VII secretion systems [18]. As noted above, the mutant displayed siderophore overproduction, and a robust OP phenotype on iron-rich s7H11 plates and iron-limiting broth readily complemented by a WT copy of the gene (Figures 4 and S5). We grew both strains in iron-limiting broth to upregulate siderophore production, and processed cell pellets and spent culture supernatants for isolation of MBT-type siderophores. The samples were then subjected to high-resolution LC-MS/MS analysis using an iron isotope-assisted screen for iron-containing metabolites in the mass range expected for MBT/cMBT siderophores. LC-MS conditions were initially guided by protocols we previously developed for the siderophore of Msm [43], and then iteratively optimized for presumptive siderophores in Mab samples. Commercially available MBT J was used as a structural reference in the LC-MS/MS studies. Our explorative pilot experiments (not shown) revealed similar presumptive siderophore ions and MS/MS signatures in the WT and mycP3::Tn strains, and across samples from cell pellets and culture supernatants. The analysis showed that MBT J and the presumptive Mab siderophore exhibit similar MS/MS fragmentation patterns, while also sharing a few of the same daughter ions. The pilot experiments also demonstrated a higher total (summed) siderophore abundance in the mutant than in the WT strain, a finding consistent with the more intense OP of the esx-3 mutant (Figure 7). In view of these results, we carried out our analysis to obtain a detailed structural characterization of the Mab siderophore with the more readily available siderophore obtained from the cells of the mutant.
Based on the findings of LC-MS analysis ( Figure S8), we propose the structure of the scaffold of MBT Ab shown in Figure 9. The figure also shows structural information for other mycobacterial siderophores and related compounds from Nocardia species reported in the literature [43,46,62,63]. Surprisingly, two unusual structural features differentiate MBT Ab from most previously characterized mycobacterial siderophore variants (e.g., Mtb and Msm MBTs, and MBT J) (Figure 9). First, MBT Ab does not have an acyl substituent on the Nε of the internal hydroxylysine residue. In other mycobacteria, the biosynthesis and addition of this substituent have been shown to require four genes (mbtK, mbtL, mbtM, mbtN) located in the so-called mbt-2 gene cluster [64][65][66]. Our orthology analysis (not shown) and analysis by others [66] did not reveal an mbt-2 locus orthologue in Mab. This finding is consistent with the lack of an acyl substituent on the internal hydroxylysine residue of MBT Ab. Second, MBT Ab has a long alkyl chain of variable length in the moiety predicted to be assembled by the MbtC (MAB_2120c)-MbtD (MAB_2119c) polyketide synthase system of the pathway [42,43]. MBT Ab shares structural features with a recently proposed structure of the MBT of M. marinum (MBT M [62]; Figure 9), which also lacks an acyl substituent on the internal hydroxylysine and has an alkyl chain with or without a terminal carboxylic acid functionality at the equivalent position of the variable alkyl chain of MBT Ab ( Figure 9). Notably, the two salient structural features of MBT Ab noted above are common in several nocardial compounds with core scaffolds remarkably similar to those of MBTs [43] (Figure 9).
Following the structure elucidation of MBT Ab, we used the LC-MS platform to investigate siderophore abundance in cultures of the WT and mycP3::Tn strains in ironlimiting and iron-rich growth media. To this end, we analyzed aggregated (summed) data of siderophore abundance determined by ion peak integrations from extracted ion chromatograms collected for seven different MBT Ab structural variants in samples of cell pellet-associated and supernatant-associated siderophore extracts (Table S5). The results of this analysis, which provides a more reliable picture than an analysis based on any single MBT Ab variant could afford, are shown in Figure 10. In the iron-rich condition, the ESX-3 mutant displayed a 17-fold increase in total MBT abundance relative to the WT strain reference. In contrast, the ratio of culture supernatant-associated MBT to cell pellet-associated MBT did not diverge drastically between the strains (1.6-fold change). In the iron-limiting condition, both strains had increased MBT abundance relative to their respective iron-rich condition reference. The increase was more drastic for the WT strain (32-fold) than for the mutant (5-fold), which had already higher MBT abundance in the ironrich condition than the WT reference (17-fold). Notably, while there was a relatively modest 3-fold difference in total MBT abundance between the strains in the iron-limiting medium, there was a drastic 11-fold change in the ratio of culture supernatant-associated MBT to cell pellet-associated MBT between the strains, with the mutant having a significantly larger proportion of cell-associated MBT. Overall, the increased abundance of MBT in the mutant is consistent with its OP phenotype. All together, these findings support the hypothesis that the mutant overproduces MBT due to an inability to secure a suitable iron supply in the presence of a malfunctioning ESX-3 with an impaired capacity for utilization of MBT-bound iron in both iron-limiting and iron-rich media.  Cell pellet-associated and culture supernatant-associated siderophores were extracted and analyzed by LC-MS. The pie charts show relative siderophore abundance obtained by ion peak integrations from extracted ion chromatograms. The areas of charts for the wild-type strain in iron-limiting medium (lower left pie; 32 cm 2 ), the mutant strain in iron-limiting medium (lower right pie; 92 cm 2 ), and the mutant strain in iron-rich medium (upper right pie; 17 cm 2 ) are proportional to siderophore abundance relative to the total siderophore abundance of the wild-type in iron-rich medium (upper left pie; 1 cm 2 ). The results shown represent aggregated data of abundance collected in duplicate for the same seven different MBT Ab structural variants for each sample.

Discussion
ESX-3 is essential for Mtb and M. bovis BCG growth in standard MB growth media, a property of ESX-3 thought to be due to its requirement for utilization of MBT/cMBTbound iron by a yet unknown mechanism. Notably, the ESX-3 essentiality seen in these slowgrowing mycobacteria is not recapitulated in Mab, despite the involvement of Mab ESX-3 in iron homeostasis, as determined by this study. Our findings demonstrate a functional link between the ESX-3 and siderophore systems of Mab. We found that Mab Tn mutants with insertions in the esx-3 locus have an OP phenotype reminiscent of the pigmentation observed in the Mtb ESX-3 mutant. Based on our results and by analogy to the observation made for Mtb, we attribute the pigmentation of the Mab mutants to an abnormally high accumulation of MBT-Fe 3+ complex. We hypothesize that a dysfunctional ESX-3 in the Mab mutants impairs the utilization of MBT-bound iron, a defect that, in turn, leads to a constant intracellular iron-deficiency state accompanied by de-repression of IdeR-regulated MBT biosynthesis genes and excessive accumulation of the MBT-Fe 3+ complex. Our laboratory hopes to explore the experimental validation of this hypothesized model in the future.
Our results show that the impact of a compromised ESX-3 on the iron uptake capability of the Mab mutants is not severe enough to prevent growth under iron-limiting or iron-rich conditions. This finding underlines a difference between the iron-acquisition capabilities of Mab and the slow-growing mycobacteria noted above. Paralleling results seen with Mtb [45], however, we found that the MBT/cMBT biosynthesis inhibitor salicyl-AMS (which targets MbtA [44,45]) produces a drastic Mab growth inhibition conditional to culturing in iron-limiting medium. This finding suggests a critical need of MBT-based iron scavenging and acquisition for Mab growth under iron-limiting conditions. Our findings also suggest that the ability of the Mab ESX-3 mutants to utilize MBT-bound iron is diminished rather than abolished. There are two possible non-mutually exclusive explanations for the iron acquisition ability of the mutants. They could acquire iron using a partially functional ESX-3 or via a secondary mechanism that allows for utilization of MBT-bound iron in an ESX-3 independent manner. Future Mab studies are warranted to probe the hypothesized essentiality of mbtA and siderophore production for growth in iron-limiting media, explore the salicyl-AMS' mechanism of antimicrobial activity, and further investigate bacterial iron acquisition.
Our serendipitous finding that the OP phenotype of the colonies of Mab ESX-3 mutants is conditional to the presence of BSA on s7H11 plates and is BSA lot-dependent was unexpected. The mechanism underlying this phenomenon remains unknown. Nevertheless, the observation that the properties of the BSA added to the growth medium can influence the behavior of mycobacterial cultures is not unprecedented [67]. Moreover, a study reporting that addition of BSA to cultures of Pseudomonas putida leads to an enhancement of siderophore accumulation in the medium provides another example of a puzzling link between BSA in the medium and siderophore production [68]. The finding that our Mab mutants do not display the OP phenotype and have WT-like growth on 7H11 plates lacking BSA suggests that the mutants are able to secure a suitable (non-limiting) iron supply (conceivably via MBT-dependent and/or MBT-independent mechanisms) in the absence of BSA. The addition of BSA to the 7H11 medium appears to compromise the ability of Mab to procure a suitable iron supply in the absence of a WT ESX-3. A tantalizing mechanistic possibility behind this phenomenon is that binding of MBT and/or the MBT-Fe 3+ complex to BSA may exacerbate the malfunctioning of an already weakened iron acquisition system in the mutants. The documented binding of the enterobactin siderophore to BSA and the earlier proposal that serum albumin may act in conjunction with other proteins in serum to restrict the iron supply needed for pathogen growth provide a conceptual framework for this possibility [69].
We expected our screen of a near-saturation library with~196,000 Tn mutants to provide the first collection of Mab mutants with gene knockouts of individual genes in the 14-kb esx-3 locus and, possibly, illuminate any genetic determinants outside the locus that are involved in ESX-3 function or transport of ESX substrates through the mycolate layer [21]. The screen rendered a collection of 54 mutants. Of the 54 mutants, 47 had an insertion in one of nine (out of eleven) genes in the esx-3 locus, two had an insertion in a hypothesized promoter region at the 5 -end of the locus, and five had an insertion distant from the locus. Interestingly, the screen did not render mutants of esxG or espG3, two genes for which, given their size, the theoretical probability of not having their Tn mutants in our library by chance is in the order of 1 × 10 −5 and 1 × 10 −15 , respectively ( Figure S3). Thus, our results suggest that disruption of esxG or espG3 does not lead to the OP phenotype. Functional redundancy might be a reason for the lack of esxG and espG3 mutant isolates. Mab has two esxG paralogues encoding proteins with 98% (MAB_0048) and 99% (MAB_0665) sequence identity to EsxG (Figure S7a,b). One or both of these paralogues might complement an esxG knockout. Mab appears to have only one espG3 paralogue (MAB_0147c) encoding an EspG family protein with~22% amino acid identity to EspG3 (not shown). The low sequence identity between the two proteins is not surprising, as EspG paralogues tend to have low sequence identity (<25%) [70,71]. Perhaps MAB_0147c can complement an espG knockout to a degree sufficient to prevent the pigmentation phenotype. Interestingly, MAB_0147c is preceded by, and possibly in the same operon with, MAB_0149c and MAB_0148c, which encode a PE protein (17% identity with Pe5, Figure S7a) and a PPE protein (21% identity with Ppe4, Figure S7b), respectively. The MAB_0149c−MAB_0147c cluster is not in proximity to either the esx-3 locus or the esx-4 locus (MAB_3759c−MAB_3753c), the only two esx loci in Mab [17,72]. Of note, the MAB_0149c−MAB_0147c array is homologous with the pe-ppe-espG array of the so-called esx-4 EVOL locus present in Mab subsp. bolletii and Nocardia brasiliensis [17]. The function of the genes in the MAB_0149c−MAB_0147c cluster remains unknown.
Our mutant complementation analysis showed weak partial complementation for some mutants with insertions in esx-3 locus genes. The poor complementation seen for some of these genes could be due to polar effect and/or loss of translational coupling. The latter possibility is likely in the case of the esxH mutant, which can be fully complemented by an esxG-esxH fragment. Interestingly, complementation of the Mtb esxH mutant has been reported only with the cognate esxG-esxH pair [33]. Moreover, folding of the co-expressed Mtb EsxG and EsxH proteins is coupled to formation of a stable 1:1 heterodimeric complex [28,73,74], a process likely facilitated by translational coupling. Of the five mutants with insertions outside the esx-3 locus, only MAB_2276c and MAB_4783 could be complemented, albeit partially. Prediction of operonic arrangements suggests that the weak and lack of complementation of the MAB_4783 and MAB_1912c mutants, respectively, might be caused by polar effects resulting from the Tn insertions (Table S4). Conversely, polar effect is unlikely to be responsible for the weak complementation of the MAB_2276c mutant or the lack of complementation of the MAB_4275c and MAB_4537c mutants (Table S4). Since only one Tn mutant was isolated for each of the five loci outside the esx-3 gene cluster, it is also possible that spontaneous mutations in the genome (e.g., possibly in the esx-3 locus) are ultimately responsible for the pigmentation seen in at least some of these mutants. Future experiments to further investigate genotype-phenotype associations in the mutants with weak or no complementation will be needed.
The number of Tn mutant isolates with insertions in the 14-kb esx-3 locus is lower than expected given the size of the library screened, which corresponds to an average of one Tn insertion per 26 bp. Assuming that no insertion compromises viability and that the probability of insertion is uniform across the genome, one could have predicted approximately 540 insertions in the esx-3 locus. Thus, it is somewhat surprising that the screen rendered 49 isolates with insertions in the locus (one Tn/286 bp observed average), even considering that not all insertions in the locus could be expected to produce colonies with OP phenotype. For example, some gene knockouts might not lead to the phenotype due to functional redundancies (e.g., esxG), some intragenic insertions near a gene's 3 end might not effectively disrupt function, and some insertions might be selected against due to generation of toxic protein truncations. A trans-complementation mechanism preventing the OP phenotype on the screen plates might have also contributed to lower the number of isolates. Mutants with an impaired ESX-3 could perhaps utilize ESX-3 substrates involved in acquisition of MBT-bound iron release into the medium by nearby colonies with a functional secretion system. The observation that the growth defect of the Mtb ESX-3 mutant can be eliminated by filtered supernatant from a culture of the WT strain or by co-culturing with the WT strain [24] provides some conceptual support for the possibility of a trans-complementation phenomenon in our screen plates.
The Tn insertion distribution pattern in the pool of isolates with insertions in the esx-3 locus was also unexpected. For example, we found four insertions in esxH (291 bp; one Tn/73 bp average) and fifteen insertions in ppe4 (1572 bp; one Tn/105 bp average), but only six insertions in eccC3 (4032 bp; one Tn/672 bp average). Interestingly, an irregular insertion pattern across the length of some genes was also evident. Most notably, the seven insertions in eccA3 (1863 bp) were exclusively distributed across the first half of the gene (~1000-bp segment; one Tn/143 bp local average), and the only three insertions in eccB3 (1566 bp) mapped to the first third of the gene (545-bp segment; one Tn/182 bp local average). The unexpected distribution of insertions in the locus is unlikely to be due to pure chance. As noted above, selection against insertions leading to production of toxic protein truncations might have contributed to biases in insertion patterns. On the other hand, the insertion pattern in some of the genes might signify the presence of functionally dispensable and functionally indispensable (sub)domains in their protein products. This is a tempting possibility for eccA3. EccA3 proteins have an N-terminal domain containing tetratricopeptide repeats (postulated to mediate interactions with ESX-3 substrates) and a C-terminal ATPase domain [75]. All the eccA3 mutants identified had insertions distributed across the N-terminal domain, perhaps suggesting that disruption of the ATPase domain does not lead to an OP phenotype. Intragenic Tn insertion patterns displaying segments with and segments without insertions that correspond to dispensable and essential protein domains, respectively, have in fact been reported [76,77].
Our structural analysis of the Mab siderophore indicates that MBT Ab is an unusual MBT variant. The siderophore has considerable similarity to the recently proposed structure MBT M from M. marinum, a slow-growing mycobacterium that causes tuberculosis-like disease in fishes and opportunistic infections in humans, most commonly leading to skin and soft tissue disease. Both MBTs have structural features uncommon in mycobacterial siderophores, but evocative of nocardial MBT-like compounds with cytotoxicity or antiproliferative activity. This finding raises the possibility of roles for MBTs Ab and M in pathogenesis unlinked to iron uptake. MBTs Ab and M lack a characteristic long-chain acyl substituent found on the internal hydroxylysine residue of other MBTs characterized thus far (e.g., Mtb and Msm MBTs). The lipid tail installed on the hydroxylysine is thought to be critical for the ability of the siderophores to interact with biological membranes and capture iron within macrophages through lipid trafficking [65,78]. The long alkyl chains in the polyketide synthase-derived hydroxy acid moieties of MBT Ab and MBT M (and related nocardial compounds) are likely the functional equivalent of the lipid tail moieties in the MBTs of Mtb, Msm, and other mycobacteria. It appears that evolution might have led to two different strategies to provide the hydrophobic tail needed for the interactions of the siderophores with lipidic environments of both bacterial and host cells.

Construction of Plasmid pMOD3Zeo and Preparation of EZ-Tn5-Derived Transposome
The EZ-Tn5-carrying plasmid pMOD-3 from the EZ-Tn5 Custom Transposome Construction Kit (Epicentre Biotechnologies Corp., Madison, WI, USA) was modified by introducing a Zeo resistance (Zeo R ) selectable marker for Mab [81]. To this end, the promoter-Zeo R gene (sh ble) fragment from plasmid pMSG360Zeo [82] was amplified with primers PM57 and PM56. The latter primer introduced the rrnB-T1 transcription terminator [83] at the 3 end of the promoter-sh ble cassette. The resulting amplicon was cloned into vector pCR-2.1-TOPO (TOPO TA Cloning Kit, Thermo Fisher Scientific Inc.). The promoter-sh ble-terminator cassette insert in the pCR-2.1-TOPO construct was subse-quently recovered as a 727-bp KpnI-PstI fragment and subcloned into pMOD-3 digested with KpnI and PstI. The resulting plasmid, named pMOD-3Zeo, carried a Tn5 transposon derivative (hereafter referred to as Tn5Zeo) with a Zeo R marker. Plasmid pMOD-3Zeo was used as a source of Tn5Zeo for the generation of the transposome complex in vitro, following recommended protocols (EZ-Tn5 Transposome Kit, Epicentre Biotechnologies Technologies Corp.).

Generation and Screening of Mutant Libraries
Transposome electroporations into Mab were carried out as noted above for plasmids, except for the addition of a glycine treatment during the preparation of electrocompetent cells [84]. Prior to library screening, the number of Tn mutants per milliliter resulting from the electroporations was determined via titration on standard s7H11 plates containing Zeo. The titers obtained guided subsequent bacterial plating for library screening on the same medium or s7H11 with an optimized BSA concentration (s7H11*) as noted in the results section and at a frequency of 100-125 colonies per plate (standard Petri dish size). Plates were incubated at 37 • C and visually screened for colonies with OP phenotype on days 4, 5, and 6. Colonies with OP were recovered from the plates and subjected to colony purification (two rounds of streak plating for colony isolation) and phenotype confirmation. Glycerol stocks (s7H9 containing 25% glycerol) were prepared for each confirmed isolate and stored at −80 • C until needed.

Insertion Site Determination and Southern Blot Hybridization Analysis
Cloning of Tn-containing genomic fragments by plasmid rescue in E. coli EC100D-pir+ (Epicentre Bioechnologies Corp.) and Tn-genome junction sequencing were carried out as reported [54], except for the use of Tn5Zeo-specific sequencing primers (primers GB27 and GB28). The genomic sequences identified were mapped onto the Mab ATCC 19977 T genome (chromosome: GenBank CU458896.1; plasmid pMAB23: GenBank CU458745.1). Southern blot analysis was performed using standard methods described elsewhere [54], except for the use of a Tn5Zeo-specific DNA hybridization probe. The probe was a DIG-labeled DNA fragment (751-bp) generated by PCR amplification (primers GG3 and ME-Plus9-FWD) from pMOD-3Zeo using a PCR digoxigenin-labeled probe synthesis kit (F. Hoffmann-La Roche, Ltd., Nutley, NJ, USA) according to the manufacturer's instructions.

Sequence Bioinformatics
Potential IdeR binding sites were identified by querying the Mab genome with the IdeR binding site consensus (a.k.a. iron box, TWAGGTWAGSCTWACCTWA; where W = A/T and S = G/C) [57,58]. Potential Zur binding sites were identified by querying selected promoter regions in the chromosome of Mab or Msm (strain MC2 155; Genbank: NC_008596.1) with the Zur binding site consensus (SNTRWYGAWAAYMRTKKYCRWYADNV; where S = G/C, R = A/G, W = A/T, Y = C/T, M = A/C, K = T/G, D = A/G/T, V = A/C/G, and N = any base) [27,56]. The computational queries were done using the Virtual Footprint tool of the Prokaryotic Database of Gene Regulation (http://www.prodoric.de, accessed on 17 August 2022) [89]. The searches were set to identify sequences with up to five mismatches relative to the consensus, a binding site search criterion used with other mycobacteria [57,60]. Predicted -10 and -35 promoter sequences were identified using the BPROM promoter prediction tool of the Softberry application package (www.softberry.com, accessed on 17 August 2022; Softberry, Inc., Mount Kisco, NY, USA). Potential orthologues of Mab genes in Mtb (strain H37Rv, Genbank: NC_000962.3), Msm, and M. avium subsp. paratuberculosis (strain K-10, Genbank NC_002944.2) were identified using the orthologue database OrtholugeDB (http://www.pathogenomics.sfu.ca/ortholugedb, accessed on 17 August 2022) [90] or the reciprocal BLAST feature embedded in the GVIEW Server (https://server.gview.ca/, accessed on 17 August 2022) with expected cut-off, alignment length cut-off, and percent identity cut-off of 1 × 10 −10 , 100, and 60, respectively. Routine sequence alignments were performed with Clustal W embedded in the MegAlign module of the DNASTAR Lasergene software package (DNASTAR, Inc., Madison, WI, USA).

Growth Inhibition Assay
Dose−response experiments were carried out using a microdilution assay comparable to those we have reported [44,45,91]. Briefly, mid-log phase cultures in s7H9 medium were spun down and resuspended in the iron-limiting GAST medium or in GAST supplemented with 200 µM FeCl 3 (GAST+Fe) [44,45,91]. The resulting cell suspensions were used to start multi-well plate microcultures (200 µL/well) at an initial optical density at 595 nm (OD 595 ) of 0.001. The MBT/cMBT biosynthesis inhibitor salicyl-AMS (obtained as reported [92]) was evaluated using a 2-fold dilution series covering a 0.004−10,000 µM range. The inhibitor was added from a 10% DMSO stock solution. The final DMSO concentration in inhibitortreated cultures and DMSO controls (no inhibitor) was 0.5%. Growth was assessed as OD 595 after four days of incubation (37 • C, 170 rpm) using a DTX 880 multimode detector microplate reader (Beckman Coulter, Inc., Brea, CA, USA). Dose-response data were analyzed with Prism v6.01 (GraphPad Software, Inc., San Diego, CA, USA).

Determination of Mycobactin Production by Spectrophotometric and Radiometirc Assays
Radio-thin layer chromatography (TLC) analysis of radiolabeled MBT was carried out as reported [43][44][45]. Briefly, 14 C-labeled MBT was obtained by feeding the MBT-specific radiotracer [ 14 C]salicylic acid (sp. Act. = 55 µCi/µmol; American Radiolabeled Chemicals, Inc., St. Louis, MO, USA) to cultures in iron-limiting GAST medium, the labeled MBT was extracted from the cultures with organic solvents, and the extracts were subjected to radio-TLC analysis on silica gel plates. Developed TLC plates were exposed to phosphor screens, which were subsequently scanned using a Cyclone Plus Storage Phosphor system (Perkin-Elmer Life and Analytical Sciences, Inc., Boston, MA, USA). Presence of the MBT-Fe 3+ complex in culture supernatants was evaluated using a spectrophotometric assay measuring the characteristic absorbance of the complex at 450 nm [93]. Absorbance determinations were done using the plate reader noted above.

Mass Spectrometry Analysis of Mycobactins
Cultures of Mab WT and the mutant M22 mycP3 (5 mL, 220 rpm, 50 mL culture tubes) in iron-limiting GASTD medium or in GASTD supplemented with 100 µM FeCl 3 (GASTD+Fe) [44,45,91] were grown to saturation, normalized to an OD 600 of 1.4, and centrifuged to obtain cell pellets and spent culture supernatants. MBT siderophores associated with pellets and supernatants were extracted into organic solvents using established protocols reported for isolation of MBT and cMBT siderophores [43,45]. After extraction, the solvent was evaporated and the remaining residue was lyophilized to dryness. The dried siderophore extracts were analyzed by high-resolution liquid chromatography-mass spectrometry (LC-MS). LC-MS was performed on an Agilent 6550 iFunnel Q-TOF mass spectrometer coupled to an Agilent 1290 Infinity LC system (including a binary pump, diode array detector, and autosampler). Data were analyzed using Agilent's MassHunter Qualitative Software (version B.06.00; Agilent Technologies, Lexington, MA, USA). Chromatography was performed using an Agilent Poroshell 120 SB-C18 column (2.7 µm, 2.1 × 50 mm) at 45 • C and a gradient of solvents A (water, 0.1% formic acid) and B (acetonitrile, 0.1% formic acid) from 5-95% solvent B or 50-100% solvent B in 10 min at a flow rate of 0.4 mL/min. Stock solutions of the siderophore samples were prepared by dissolving the extracts in acetonitrile and routinely stored at −80 • C until needed. Small aliquots of these stocks were diluted 100× and 5 µL of the diluted samples were injected for analysis. The following settings were applied to the electrospray ionization source: gas temperature, 250 • C; nebulizer, 30 psig; sheath gas temperature, 250 • C; vcap, 3500 V; and nozzle voltage, 2000 V. For MS analysis, full scan mass spectra (m/z = 100-1500) were acquired in positive-ion mode. For targeted MS/MS analyses, collision energies (CE) of 10-70 were evaluated. Compounds of interest were identified by both assessing UV chromatograms at 450 nm, and mass spectra exhibiting characteristic Fe isotopic distribution. The commercially available MBT J from Mycobacterium avium subsp. paratuberculosis [46] (Allied Monitor, Inc., Fayette, MO, USA) was analyzed by MS/MS, and its fragmentation pattern was used as a template to help determine potential structures of the siderophore analogues found in the extracts.

Conclusions
Overall, our studies provide further insight into the ESX-3 and siderophore systems of Mab and expand our knowledge of the biology of this recalcitrant pulmonary pathogen. The findings of this work highlight the need for further research to better understand the functional dimensions of ESX-3 and its interplay with the MBT Ab-mediated iron acquisition system. The collection of novel mutants generated in this study will facilitate progress on these fronts. Mutants generated herein will also be useful to probe the relevance of individual ESX-3 components in cellular and animal infection models. The structural features of the MBT Ab revealed by our study suggest that the siderophore might have cytotoxic properties that contribute to pathogenesis. This possibility warrants exploration and may open new lines of experimental inquiry into the pathogenesis of Mab.
Supplementary Materials: The following supporting information can be downloaded at: https://www.mdpi.com/article/10.3390/pathogens11090953/s1, Figure S1: Southern blot analysis of M. abscessus mutants; Figure S2: Effect of BSA on the development of the orange pigmentation phenotype in M. abscessus mutants; Figure S3: Probability (p) of missing a gene with "x" base pairs (bp) in a library of "n" transposon (Tn) mutants of M. abscessus; Figure S4: Complementation of the esxH mutant; Figure S5: The orange pigmentation phenotype of M. abscessus cultures is influenced by iron availability; Figure S6: Growth of M. abscessus strains in iron-rich and iron-limiting media; Figure S7: Substrate candidates for the secretome of the ESX systems of M. abscessus; Figure S8: Representative structures and fragmentation patterns of mycobactin Ab (a) and mycobactin J (b); Table S1: Primers used in the study; Table S2: Mycobacterial gene expression plasmids used in this study; Table S3: M. abscessus strains included in phenotypic characterizations; Table S4: Genetic loci outside the esx-3 cluster with transposon insertions; Table S5: Ion peak integration data collected from extracted ion chromatograms for seven different MBT Ab structural variants in samples of supernatant-associated (SA) and cell pellet-associated (CPA) siderophore extracts from cultures in the iron-limiting or iron-rich growth media.