The nomenclature of Mycobacterium abscessus is complicated by the non-uniform use in the literature of species and subspecies designations. Currently, two subspecies are recognized: M. abscessus subsp. abscessus, and M. abscessus subsp. bolletii which unites the previous subspecies massiliense and bolletii (Leao et al., 2011). However, for the sake of clarity and because recent publications support the previous classification (Cho et al., 2013; Tan et al., 2015; Bryant et al., 2016; Tortoli et al., 2016), we use the three-subspecies designations. The M. abscessus isolates from CF patients used in this study were characterized previously (Rüger et al., 2014; Kehrmann et al., 2016). The type strain of M. abscessus subsp. abscessus (DSM 44196) was obtained from DSMZ, Braunschweig, Germany.

Pseudomonas quinolone signal and HHQ were purchased from Sigma Aldrich and dissolved in methanol.

Sequencing libraries were constructed from extracted genomic DNA with the Nextera XT kit (Illumina) and sequenced on the Illumina MiSeq instrument in a 2 × 300 bp paired end run or in a HiSeq 2 × 150 bp paired end Rapid Run.

For isolation of RNA, M. abscessus^T (DSM 44196) was grown in DSM219 medium. To possibly induce the expression of aqd genes, 20 μM PQS was added 2 h before harvesting the cells by centrifugation. Cells were frozen in liquid nitrogen and stored at –80°C. Cells were then resuspended in TE buffer (10 mM Tris, 1 mM EDTA, pH 8.0) and disrupted using the Mikro-Dismembrator S (Sartorius, 3000 rpm, 2 min) after addition of glass beads (150–212 μm diameter) to the cell suspension. Subsequently, RNA purification was performed with the innuPREP RNA Mini kit (Analytik Jena) according to the manufacturer’s instructions. RNA concentration was determined using the Nanophotometer N60 (IMPLEN). Removal of DNA contamination was executed with DNAse I (Thermo Scientific) at 37°C and checked via PCR (GoTaq Polymerase, Promega) after renewed RNA purification. For analysis of operon structures, cDNA synthesis was carried out using the RevertAid H Minus First Strand cDNA Synthesis kit (Thermo Scientific) according to the manufacturer’s instructions. Primers were designed to amplify 500 bps of regions spanning aqdAB and aqdBC.

For relative mRNA quantification of single genes, RT-qPCR was used. Primers were designed to amplify 75–150 bps of the analyzed genes (sequences shown in Table 1). All measurements were performed in a LightCycler 96 System (Roche) with a SensiFast SYBR No-Rox One-Step Kit (Bioline, London, UK) and 96 well lightcycler plates (Sarstedt) as reaction vessels, sealed with qPCR seals (Sarstedt). The concentrations of all template RNA samples were adjusted to 200 ng/μL for normalization on total RNA. One microlitre of the RNA samples was used as template and mixed with 19 μL master mix containing 1 μL of specific primers (10 μM each), 0.2 μL reverse transcriptase, 0.4 μL RNase inhibitor, 10 μL reaction mix and 7.4 μL 5 M glycine betaine. All measurements were carried out with a minimum of three biological replicates in two technical duplicates each. For each primer pair two negative controls with 1 μL H₂O as template were included. Reverse transcription was carried out at 45°C for 20 min, followed by 2 min polymerase activation at 95°C, a three step amplification (95°C 5 s, 60°C 10 s, 72°C 10 s, 60 cycles) and a melting profile analysis. Evaluation of control measurements and analysis of the melting curves as well as C_q calculation was carried out with the LightCycler 96 V1.1 software. The relative transcript amount was normalized on total RNA (200 ng) and calculated as 2^-ΔCq where ΔC_q corresponds to the difference of the mean C_q values.

The aqdB and aqdC genes were amplified by PCR from a colony of M. abscessus^T (DSM 44196). Both AqdB and AqdC were produced as His₈-MBP fusion proteins in recombinant E. coli strains, obtained by restriction-free cloning of the corresponding genes into the pET28b(+) expression vector (van den Ent and Löwe, 2006). E. coli Rosetta(DE3) was transformed with pET28b(+)::his8-mbp-aqdB due to many rare codons in the aqdB sequence. E. coli BL21(DE3) was transformed with pET28b(+)::his8-mbp-aqdC.

Mycobacterium strains were cultivated in DSM219 medium, and recombinant E. coli BL21(DE3) and Rosetta(DE3) strains were grown in LB medium supplemented with 50 μg/mL kanamycin, at 37°C. For AQ biotransformation by mycobacteria, cells from pre-cultures were suspended at an optical density at 600 nm of 3.5 in fresh DSM219 medium, supplemented with 20 μM of HHQ or PQS, and incubated at 37°C. AQs were extracted at different time points as described previously (Müller et al., 2014). Samples of extracted cell suspensions were solubilized in methanol and analyzed via HPLC. For biotransformations of AQs by recombinant E. coli strains harboring pET28b(+)::his8-mbp-aqdB or pET28b(+)::his8-mbp-aqdC, cells were grown overnight at 30°C in the presence of 0.1 mM IPTG, harvested by centrifugation, and resuspended in fresh LB medium with 0.1 mM IPTG, adjusting an OD_600nm of 3.5. Biotransformation assays were performed as described above.

Cell extracts were prepared from M. abscessus^T (DSM 44196) grown in DSM219 medium. To possibly induce the expression of AQ-converting enzymes, 20 μM PQS was added 2 h before harvesting the cells by centrifugation. Cells resuspended in 50 mM potassium phosphate buffer pH 7.5 were disrupted by sonication, and cell debris was removed by centrifugation (20.000 × g, 45 min, 4°C). The supernatant (crude extract) was desalted using Zeba^TM Spin Desalting columns (10 K molecular weight cut-off, Thermo Scientific). Total protein amount in cell extract supernatants was determined using the Bradford method as modified by Zor and Selinger (1996).

The sequence of aqdC of M. abscessus^T was optimized for codon usage of E. coli using OPTIMIZER (Puigbò et al., 2007) and synthesized by MWG Eurofins. For the purification of AqdC proteins, codon-optimized synthetic genes were cloned in pET28b(+) using restriction-free cloning (van den Ent and Löwe, 2006). The sequence coding for TEV protease cleavage site was introduced between the coding sequences for his₈-tag and AqdC protein. In the following, AqdC^I refers to the protein of the type strain, and the protein which differs from AqdC^I by the two amino acid substitutions R129P and A133T is termed AqdC^II. E. coli BL21(DE3) harboring pET28b(+)::his8-aqdC^I or pET28b(+)::his8-aqdC^II were grown at 37°C in Terrific Broth. At an OD_600nm of 1.0, cultures were supplemented with 0.2 mM IPTG and incubated at 16°C overnight (for approximately 16 h). Cells harvested by centrifugation were resuspended in washing buffer (300 mM NaCl, 20 mM Tris and 10 mM imidazole, pH 8.0), disrupted by sonication, and the AqdC proteins were purified by Ni-NTA affinity chromatography and stored in buffer containing 20 mM Tris, 10% (v/v) glycerol (pH 8.0) at -80°C.

The catalytic activity of AqdC proteins was determined spectrophotometrically at 30°C by measuring PQS consumption at 337 nm. The assays contained 20 μM PQS in assay buffer (50 mM Tris, 2 mM EDTA, 10% PEG 1500, 4% (v/v) DMSO, pH 8.0). The extinction coefficient of PQS in assay buffer is 10169 M^-1 cm^-1 at 337 nm. Apparent steady-state kinetic constants of AqdC proteins (two biological replicates with three technical replicates each) were estimated by fitting the initial velocities measured at different substrate concentrations with the Michaelis Menten equation.

Overnight cultures of P. aeruginosa PAO1 and M. abscessus strains were used as inocula for co-cultivation experiments in 10% LB medium. To account for differences in growth rates (generation times of M. abscessus and P. aeruginosa in 10% LB are 38.5 and 16.5 h, respectively), P. aeruginosa was adjusted to an initial OD_600nm of 0.05, and the mycobacterial strain to an OD_600nm of 0.15, as described by Costa et al. (2015). Cells were incubated at 37°C under vigorous shaking. PQS was extracted after 8 and 24 h as described previously (Müller et al., 2014) and quantified by HPLC analysis. Colony forming units (CFUs) of P. aeruginosa PAO1 were determined by dropping 10 μL of a diluted culture onto LB agar and counting colonies after incubation at 37°C overnight. M. abscessus formed colonies only after 36 to 48 h, so selective medium was not necessary.

For the identification and quantification of PQS and other AQs, compounds were separated on a 250 × 4 mm Eurospher II RP-18 column using a Hitachi EZchrom Elite HPLC system with diode array detector model 2450, or an Agilent 1100 series system with diode array detector model G1315B. Methanol with 0.1% (w/v) citric acid and 0.1% (w/v) citric acid in water were used as solvents. Separation of PQS and HHQ was carried out via a linear gradient from 80 to 100% methanol (v/v) over 20 min at a flow rate of 0.5 mL min^-1.

The presence of an aqdRABC gene cluster in the genome of M. abscessus^T suggested that it might degrade HHQ and PQS via reactions analogous to those identified in R. erythropolis BG43 (Figure 1). The aqdB2 gene of R. erythropolis BG43 codes for an NADH-dependent HHQ monooxygenase, whereas the gene product of aqdC2 is a PQS-cleaving dioxygenase which requires O₂ as only co-substrate (Müller et al., 2015). AqdB and AqdC activity was present in cell extract supernatant suggesting cytoplasmic localization. Desalted cell extract supernatants of M. abscessus^T supplemented with NADH transformed HHQ (Figure 2A), and HHQ consumption was accompanied by transient formation of 2.0 ± 0.3 μM PQS, as identified by HPLC. Interestingly, extracts from cells pre-incubated with PQS converted HHQ and especially PQS faster than extracts from non-induced cells (Figure 2). To analyze whether transformation of HHQ and PQS is indeed catalyzed by the mycobacterial AqdB and AqdC protein, respectively, recombinant E. coli strains expressing aqdB or aqdC of M. abscessus^T were constructed and tested for AQ biotransformation. E. coli cells producing the His₈-MBP-AqdB protein converted HHQ to PQS, and E. coli cells expressing the M. abscessus^T His₈-MBP-AqdC fusion protein were able to cleave PQS to N-octanoylanthranilic acid, as identified by HPLC and comparison of UV/Vis spectra and fluorescent properties with reference compounds (Figure 3).

Taken together, the observations strongly support the hypothesis that the mycobacterial aqdRABC gene cluster (locus tags MAB_0300c to MAB_0303 in NC_010397.1) codes for an inducible AQ degradation pathway which proceeds analogous to that identified in R. erythropolis BG43 (Figure 1). N-Octanoylanthranilic acid formed by the PQS dioxygenase AqdC presumably is hydrolyzed by AqdA, a member of the carboxylesterase type B family, to anthranilic acid and octanoate, which both can be channeled into the central metabolism. The aqdR gene codes for a putative transcriptional regulator of the TetR family.

The genomes of M. abscessus strains, previously isolated from respiratory samples of CF patients (Rüger et al., 2014), were analyzed for the presence and sequence of the aqd gene cluster with BLASTN analyses (Altschul et al., 1997). Among these 50 strains, 22 (44%) lack the aqd gene cluster. Interestingly, absence of the aqd genes correlates with the assignment of the strains to the subspecies bolletii (3 strains) and massiliense (19 strains), whereas all 28 strains of the subspecies abscessus harbor the aqd gene cluster. In all cases where two isolates were obtained from the same patient, the nucleotide sequences of the aqd gene clusters did not differ between the first and second isolate.

The nucleotide sequences of the aqdRABC genes of strains P5a, P5n, P23a, P23n, P28a, and P28n are identical to those of M. abscessus subsp. abscessus^T (DSM 44196, ATCC 19977). Compared to the aqd gene region of the type strain, those of strains P13, P24 and another 16 isolates contain the same five single nucleotide variations (SNPs). Two of these SNPs within the protein coding regions of AqdA (triplet encoding L173) and AqdC (triplet encoding I187) are silent, and one is a transition within the predicted promoter region of aqdR. The two other SNPs lead to differences in the amino acids at position 129 and 133 of the predicted PQS dioxygenase AqdC (Table 2 and Supplementary Figure 1). Two other aqd gene modifications observed in the genome of individual strains lead to single amino acid deviations in AqdA (G394S) and AqdB (G15S), respectively (Table 2 and Supplementary Figure 1). However, with respect to AqdC, the key enzyme in PQS degradation, the strains can be divided into two groups: Nine isolates produce the same protein as the type strain (AqdC^I), and 18 strains form a protein (AqdC^II) which differs in two amino acids (Table 2). Supplementary Figure 1 shows the translated nucleotide sequence of the aqdRABC region and indicates the changes in nucleotide and amino acid sequences.

Ten of the clinical M. abscessus isolates (Table 2) were analyzed for their ability to degrade PQS and HHQ. Interestingly, members of the subset of 18 strains carrying the aqd gene cluster with the five SNPs (as compared to that of M. abscessus^T) appeared to be the more potent PQS degraders. Surprisingly, even those strains that do not harbor aqd genes reduced the amount of PQS to different extents within 24 h (Figure 4).

We selected a representative of each of the groups of strains – strain P5a with an aqd gene cluster identical to that of M. a. subsp. abscessus^T, strain P13 with the five SNPs in the aqd region, and strain P4a lacking aqd genes – to determine the time course of HHQ and PQS conversion by cell suspensions. HHQ consumption by strains P5a and P13 followed similar kinetics. However, as already suggested by the preliminary data shown in Figure 4, strain P13 converted PQS faster than strain P5a, and strain P4a lacking the gene cluster very slowly converted HHQ as well as PQS (Figures 5A,B).

Strains without the aqd-cluster which slowly consume PQS and HHQ must use alternative enzymes to modify or even degrade these AQs. HPLC analyses of culture extracts revealed transient formation of 4.9 ± 3.9 μM PQS after 2 h of incubation with 20 μM HHQ, besides an arsenal of other metabolites. Thus, it appears that other monooxygenases besides AqdB can mediate HHQ hydroxylation to PQS. Due to the complexity of metabolites formed, incomplete peak separations, and the low concentrations of intermediates (expected to be produced by the cultures at the nM range), elucidation of their structures will require the establishment of improved extraction and separation protocols, as well as considerable upscaling, to enable NMR analyses for structural identification. However, because the UV spectra of many of the peaks of the HPLC elution profile resemble that of PQS, we assume that the corresponding metabolites are quinolones, with modifications introduced mainly to the alkyl chain.

To analyze whether the aqdABC genes are co-transcribed, RT-PCR was performed with primer pairs addressing adjacent gene transcripts (for primer sequences see Table 1). Formation of PCR products, as verified by gel electrophoresis (not shown), indicate that aqdAB and aqdBC are co-transcribed, suggesting organization of the aqdABC genes in an operon.

Cells induced with 20 μM PQS have higher transcript levels of aqdR (fold change between 4.0 and 6.8), indicating autoregulation of aqdR transcription. Transcript levels of aqdC were also increased when the cells were induced with PQS (fold change between 5.4 and 5.7) (Figure 6A). In strains harboring the aqd genes with five SNPs compared to the aqd region of the type strain, the transition within the predicted promoter region of aqdR changes the inverted repeat sequence TTGTCGCATCGACAA to TCGTCGCATCGACAA. To determine whether the SNP results in different expression levels, transcript levels of aqdR and aqdC were compared for strains P5a and P13. RT-qPCR analyses revealed only slight reduction of transcription amounts in strain P13 (Figure 6B). Thus, it seems unlikely that differential expression of the aqdC gene coding for the key enzyme, a PQS dioxygenase, accounts for the observed differences in PQS degradation rates (Figure 5).

To analyze the possibility that the AqdC^I and AqdC^II proteins differ in their catalytic efficiency, recombinant proteins were purified to electrophoretic homogeneity, and their steady-state kinetic parameters were determined. Under the conditions of the assay, a specific activity of 53.2 U mg^-1, an apparent k_cat of 97.8 ± 19.1 s^-1 and an apparent K_m value for PQS of 37.2 ± 9.8 μM were observed for AqdC^I, the PQS dioxygenase form of the M. a. subsp abscessus type strain and the group of strains represented by the isolate P5a. For AqdC^II, the enzyme of the group of isolates represented by strain P13,a specific activity of 50.5 U mg^-1, an apparent k_cat of 61.0 ± 4.3 s^-1 and apparent K_m value of 13.6 ± 1.6 μM were determined.

Representatives of each M. abscessus group were co-cultivated with P. aeruginosa PAO1 to test whether PQS produced by P. aeruginosa, which is packaged into membrane vesicles for trafficking between cells (Mashburn-Warren et al., 2008), is amenable to degradation by the mycobacteria. Indeed, the clinical isolates P5a and P13, which degraded synthetic PQS, also were able to reduce the PQS concentration in co-culture with P. aeruginosa PAO1 (Figure 7). Most interestingly, however, the PQS concentration increased significantly when P. aeruginosa PAO1 was cultivated with strain P4a, a representative of the group lacking the aqd genes. Compared with the P. aeruginosa PAO1 cultures, the amount of PQS in P. aeruginosa PAO1 – M. abscessus P4a co-cultures was fivefold higher after 24 h of cultivation. P. aeruginosa PAO1 cultivated with dead (autoclaved) M. abscessus strain P13 cells produced increased amounts of PQS as well. After 24 h of incubation, the PQS concentration was almost eightfold higher than in the PAO1 solo cultures. CFUs of P. aeruginosa PAO1 cultures were similar to those of PAO1 in co-culture with M. abscessus strain P13, or PAO1 cultivated in the presence of autoclaved M. abscessus P13 cells. Therefore, the increase in PQS production especially in presence of autoclaved cells is not due to increased population density of P. aeruginosa. The boosting of PQS production may rely on the recognition of cell components by PAO1. It makes the reduced levels of PQS in PAO1-P5a and PAO1-P13 co-cultures even more remarkable.