Enzymatic Degradation of Phenazines Can Generate Energy and Protect Sensitive Organisms from Toxicity

PCA is the precursor of all phenazines produced by Pseudomonas spp. (9, 10). Therefore, we reasoned that the ability to degrade PCA would be common among organisms capable of degrading Pseudomonas-derived phenazines. Soil was collected from 16 locations around the California Institute of Technology campus and the nearby San Gabriel mountains. Samples from six sites yielded isolates capable of growing in medium with 2.5 mM PCA as the sole source of carbon and energy (Fig. 1B). Among the isolates that were verified to degrade PCA, all had similar colony morphology; thus, one isolate from each site was randomly selected for follow-up studies. Partial 16S rRNA gene sequences were determined for each isolate and phylogenetic trees were constructed. PCA-degrading organisms grouped with two members of the genus Mycobacterium, Mycobacterium fortuitum (GenBank accession number NR_118883) and Mycobacterium septicum (accession number NR_042916), with >99% sequence identity (Fig. 1C). Strains CT6 (M. fortuitum-like) and DKN1213 (M. septicum-like) were chosen as representative strains for further study.

While PCA is produced by all pseudomonads capable of phenazine production, PYO is produced only by P. aeruginosa and is the best-studied phenazine due to its clinical relevance. We therefore sought to determine if our isolates were capable of degrading PYO. Because PYO is poorly soluble in aqueous solutions at circumneutral pH, we monitored PYO degradation using micromolar—but physiologically relevant (24)—concentrations with cultures grown in 10% LB-Tw (lysogeny broth with 0.05% Tween 80) medium. Both CT6 and DKN1213 were capable of degrading PCA under these conditions. Both strains were additionally capable of degrading PYO from a starting concentration of 200 µM to a final concentration of less than 20 µM (the detection limit of our assay) (Fig. 1D) with a concomitant change of the medium from blue to colorless. No change in medium color or measured PYO was observed for the close relative Mycobacterium smegmatis (see Fig. S1 in the supplemental material).

Phenazines are polycyclic, aromatic, nitrogen-containing heterocycles (Fig. 1A); therefore, we hypothesized that the genes responsible for their degradation likely encode dioxygenases, a family of enzymes known to break down aromatics. Several observations lend credence to this notion. First, because the only source of carbon in PCA that is sufficiently reduced to support growth is bound within the ring, it must be cleaved; the carboxylate moiety alone cannot support growth (Fig. 1A). Second, previous studies of the PCA degradation pathway in Sphingomonas wittichii strain DP58, the only other known PCA degrader, demonstrated that ring cleavage is essential for PCA catabolism (25). Furthermore, in our cultures, we never observed the accumulation of the unsubstituted core phenazine molecule, despite the fact that our mycobacteria can degrade this molecule (see Fig. S2 in the supplemental material); its absence strongly suggests that the core phenazine ring is cleaved. Finally, the fact that our mycobacteria can also use PCA as a sole nitrogen source (see Fig. S3 in the supplemental material) further indicates that ring cleavage must occur in order for nitrogen to be liberated from PCA (Fig. 1A). Accordingly, we focused on identifying putative dioxygenase enzymes that might catalyze PCA degradation.

The genome of S. wittichii strain DP58 has been sequenced and contains 91 predicted dioxygenase genes, though none have been identified as important for PCA degradation (19, 26). We sought to identify candidate dioxygenase genes for PCA degradation by comparative genomic analysis of S. wittichii DP58 and one of the phenazine-degrading mycobacteria described here. The genome of strain CT6 was sequenced to completion using the PacBio single-molecule, real-time (SMRT) genome sequencing technology. The strain CT6 genome is 6.25 Mbp and contains no plasmids. There are a predicted 6,051 protein-encoding genes according to the RAST (Rapid Annotation using Subsystem Technology) server, and the genome contains 35 to 47 dioxygenase genes as annotated by RAST or the Bacterial Annotation System (BASys) pipelines, respectively (27–29). Reciprocal best BLAST analysis of annotated dioxygenase genes was performed between S. wittichii strain DP58 and strain CT6 (PCA degraders) or M. smegmatis mc²-155 and strain CT6 to identify genes present in both PCA-degrading organisms but absent from M. smegmatis. This identified three predicted dioxygenase genes—XA26_16830, XA26_16860, and XA26_16890—that are co-oriented in an ~10-kbp chromosomal locus in strain CT6 (Table 1; also, see Table S1 in the supplemental material). Because the same set of genes is present in both sequenced PCA-degrading organisms, we hypothesized that these genes would be regulated specifically by phenazines. Quantitatve reverse transcription-PCR (qRT-PCR) analysis of the expression of putative dioxygenase genes in strain CT6 demonstrated that these three genes, as well as a nearby fourth gene predicted to encode a dioxygenase (XA26_16730), are highly induced in the presence of phenazines (Fig. 2A). mRNA abundance for these genes was increased >1,000-fold in the presence of both PCA and PYO after a 3-h exposure but not in the presence of 9,10-anthraquinone-2,6-disulfonate (AQDS) or methylene blue (MB), two other 3-ringed, aromatic, redox-active molecules (13). AQDS and MB have midpoint redox potentials that are lower and higher, respectively, than those of phenazines (13) suggesting that a redox switch is not responsible for increased expression. The mRNA abundance of two predicted dioxygenase genes located distantly on the chromosome was unchanged in the presence of any compounds tested.

A comparison of the genome of CT6 to that of the type strain M. fortuitum ATCC 6841 revealed the presence of the predicted PCA-degrading genes, and this organism was verified to degrade both PCA and PYO (see Fig. S1 in the supplemental material). The putative dioxygenases predicted to be involved in PCA degradation were analyzed in the Integrated Microbial Genomes/Expert Review (IMG/ER) web server (30, 31), and Rhodococcus sp. strain JVH1 was also found to contain these genes (32, 33). Based on their presence, we predicted that Rhodococcus sp. strain JVH1 would be capable of PCA degradation. When suspended at high density (optical density at 600 nm [OD₆₀₀] of ~2 to 3), JVH1 degraded 200 µM PCA to ~50% of the starting concentration in 48 h—a rate of degradation significantly lower than that observed for mycobacteria. No activity was seen in Rhodococcus sp. strain RHA1, the closest sequenced relative that lacks the genes of interest (Fig. 2B). No degradation activity was observed with PYO for either strain.

The presence of the same putative dioxygenase genes in all PCA degraders and their phenazine-specific regulation in strain CT6 strongly suggest an involvement in PCA degradation. We used mutagenesis to test this hypothesis. Attempts to genetically manipulate strain CT6 were ineffective, but a similar approach succeeded in the M. fortuitum type strain. A recombineering approach was employed to create mutations in the four putative dioxygenase genes (MFORT_16269, MFORT_16319, MFORT_16334, and MFORT_16349) (Table 1) (34), and each gene was replaced with a gentamicin resistance cassette. Gentamicin-resistant colonies with disruptions in each gene were streaked onto agar medium with 2.5 mM PCA as the sole carbon source. None of the mutants grew under these conditions, indicating that the mutated genes are essential for growth with PCA as a carbon source (Fig. 2C). High-performance liquid chromatography (HPLC) analysis of supernatants from cultures grown on LB medium with 100 µM PCA revealed the loss of PCA from Δ16319::Gm^r, Δ16334::Gm^r, and Δ16349::Gm^r strains but not from the Δ16269::Gm^r strain (see Fig. S4 in the supplemental material). This suggests that the product of MFORT_16269 is necessary to catalyze PCA degradation, whereas the other genes are involved in a downstream reaction (these mutants cannot use PCA as a carbon source [Fig. 2C] but still remove it from culture supernatants [see Fig. S4 in the supplemental material]). None of these mutants had a defect in PYO removal.

Notably, only the putative single-subunit, ring-cleavage dioxygenase genes, MFORT_16319 and MFORT_16349, could be complemented in trans (see Fig. S5 in the supplemental material). It is unclear why the putative multisubunit, ring-hydroxylating dioxygenase genes MFORT_16269 and MFORT_16334 could not be complemented, but perhaps these enzymes are nonfunctional when overexpressed. The position of the gentamicin resistance cassette on the chromosome was verified by sequencing for each mutation; however, we cannot rule out the possibility that a secondary site mutation resulted in the phenotype for the MFORT_16269 and MFORT_16334 mutants.

Because the four predicted dioxygenase genes were expressed specifically in the presence of phenazines, we took a transcriptome sequencing (RNA-Seq) approach to identify candidate genes important to PYO degradation. Strain CT6 was used for RNA-Seq because the genome of this organism has been closed, whereas the genome of ATCC 6841 exists as 82 contigs (35). As expected, the genes important for PCA-dependent growth showed increased mRNA abundance after a 20-min exposure to either PCA or PYO (Fig. 3A). In fact, the entire region of the CT6 genome that shares homology with Rhodococcus sp. strain JVH1 (an organism that can degrade only PCA) was highly expressed in the presence of either PCA or PYO. Additionally, genes flanking this region had increased mRNA abundance and were induced to a greater extent by PYO.

A mutant lacking the entire ~40-kb region (Δ40kb::Gm^r) of the genome induced by phenazines was constructed in M. fortuitum and found to completely lack the ability to degrade phenazines, including PYO (Fig. 3B). This locus contains genes for an additional predicted dioxygenase (MFORT_16204 [XA26_16600]) and a monooxygenase (MFORT_16224 [XA26_16640]) in one of the flanking regions; however, mutants with mutations in both genes were still capable of PYO degradation (Fig. 3B). Of note is that Δ16224::Gm^r degrades PYO more slowly than the wild type (WT), suggesting a possible role in a downstream reaction in the PYO degradation pathway. A mutant missing a three-gene operon in the other flanking region (MFORT_14352 to MFORT_30529 [XA26_16990 to XA26_16960]) lacked the ability to degrade PYO. A single gene mutant lacking MFORT_14352, annotated as a hypothetical protein, was deficient in PYO degradation; additionally, expression of this gene in Rhodococcus sp. strain JVH1 allowed PYO degradation by this strain (see Fig. S6 in the supplemental material). Therefore, MFORT_14352 is necessary for the first step of PYO degradation and may be sufficient for the removal of this phenazine from the medium.

M. fortuitum is ubiquitous and found in environments where phenazine-producing pseudomonads are also present (36, 37). Using a subset of the mutants described above, we sought to determine whether a mixed culture of M. fortuitum with a pseudomonad would lead to alterations in the phenazine pool. M. fortuitum and several mutants were inoculated into LB medium with P. aeruginosa PA14, and culture supernatants were analyzed by HPLC after 24 h. The Δ40kb::Gm^r mutant was used as a control to determine the amount of phenazines produced by PA14 under coculture conditions. The WT M. fortuitum strain decreased the abundance of both PYO (~50% decreased) and PCN (~90% decreased) but not PCA in coculture with PA14 (Fig. 4A). This level of phenazine degradation impacted neither the rate of P. aeruginosa growth nor its transcription of the phenazine biosynthesis genes. As expected, the PYO degradation mutant, Δ14352::Gm^r, decreased the abundance of PCN but not PYO. The mutant Δ30529::Gm^r decreased PYO but not PCN, suggesting that MFORT_30529—annotated as an amidase—may convert PCN to PCA. In fact, cocultures that are proficient for PCN degradation seem to have higher PCA concentrations in their supernatants, supporting this hypothesis (Fig. 4A; also, see Fig. S6 in the supplemental material). The PCN degradation phenotype of Δ30529::Gm^r was assayed in monoculture, and this strain was confirmed to have a specific defect in the removal of PCN from culture supernatants (see Fig. S7 in the supplemental material); additionally, expression of MFORT_30529 in Rhodococcus sp. strain JVH1 conferred the ability to degrade PCN to this organism (see Fig. S6).

Because M. fortuitum in coculture with PA14 prioritizes PYO and PCN removal from the medium, to determine whether PCA could also be degraded in coculture, we mixed M. fortuitum and Pseudomonas fluorescens 2-79, an organism that produces PCA as its only phenazine. WT M. fortuitum was capable of removing ~50% of the PCA from culture supernatants compared to the Δ40kb::Gm^r control (Fig. 4B). As expected, Δ16269::Gm^r was defective for PCA removal. For unknown reasons, Δ30529::Gm^r was also defective for PCA removal in the coculture condition.

Alterations in the abundance of phenazines in cocultures of M. fortuitum and Pseudomonas spp. demonstrate that phenazine degradation can influence the concentration of structurally diverse phenazines in the coculture environment. Because phenazines can function as antimicrobials (16, 17), it is ecologically important to consider whether degradation might impact other members of a microbial community. We hypothesized that PYO degradation would be protective to organisms that are otherwise susceptible. PYO survival was assayed in cocultures of M. fortuitum mixed with diverse bacteria that have a range of sensitivity to different phenazines: Staphylococcus aureus, Agrobacterium tumefaciens, Shewanella oneidensis, and Escherichia coli (16, 17, 38, 39). After 24 h of incubation in monoculture in the presence of 100 µM PYO, S. aureus, E. coli, and S. oneidensis were all inhibited by PYO to various degrees (Fig. 5). In all three cases, coculture with WT M. fortuitum rescued growth of these organisms, though not always to the same extent as growth in monoculture in the absence of PYO. A PYO degradation mutant provided no protection. A. tumefaciens was resistant to PYO, consistent with previous reports (38), and coculture had little effect on this organism.

Strains used in this study are listed in Table S2 in the supplemental material. PCA-degrading organisms were isolated by inoculating 5 ml of PCA medium with 100 mg of soil collected on and around the California Institute of Technology campus and growing at 30°C with shaking. PCA medium was composed of minimal medium containing (per liter) KH₂PO₄ (1.93 g), K₂HPO₄ (6.24 g), NaCl (2.5 g), MgSO₄ (0.12 g), NH₄Cl (0.5 g) and supplemented with PCA (0.56 g). Trace minerals were also included (45). Cultures were allowed to grow for 5 to 8 days before transfer to fresh PCA medium. After 3 rounds of growth in PCA liquid medium, cultures were streaked on plates containing PCA medium amended with 1.5% Bacto agar. After isolation, strains were routinely grown in lysogeny broth with 0.05% Tween 80 (LB-Tw). PCA and PCN were purchased from Princeton Biomolecular Research Inc., and PYO was purified from cultures of P. aeruginosa strain PA14 grown with 40 mM succinate as the sole carbon source using organic extraction into dichloromethane as described previously (14).

Growth curves were performed in 10% LB-Tw medium supplemented with 200 µM PCA, PYO, or PCN. Cultures were grown to mid-exponential phase and diluted to an OD of ~0.1 to 0.2 for growth analysis. Growth curves were carried out in 96-well plate format on a BioTek Synergy 4 microplate reader (BioTek, Inc.) set to 30°C and medium agitation. Ninety-six-well plates were sealed with sterile adhesive film to prevent medium evaporation. OD was monitored at 500 nm; PCA and PCN were monitored at 370 nm, and PYO was monitored at 310 nm. Standards were included to calculate the concentrations of each phenazine. PCA and PYO degradation by Rhodococcus was performed at an OD₆₀₀ of ~2 to 3. Rhodococcus spp. were suspended in 10% LB medium with 200 µM PCA or PYO in a volume of 5 ml. A 500-µl portion of culture supernatant was collected by centrifugation at various time points and analyzed on a plate reader.

Coculture experiments were carried out in LB with various M. fortuitum strains at a starting OD of 0.15. Pseudomonads were inoculated to an OD of 0.05 and other organisms to 0.001. Cocultures were grown for 24 h in all cases except for A. tumefaciens (48 h). To test the protective effects of PYO degradation, 100 µM PYO was included, and cultures were plated on LB at the end of growth to enumerate CFU (all organisms formed colonies in 1 to 2 days versus 3 to 5 days for M. fortuitum, so selective medium was not necessary). To analyze coculture supernatants, samples were diluted 1:10, sterilized by filtration on a 0.2-µm filter and analyzed by HPLC as described previously (12) using the same method but a flow rate of 950 µl min⁻¹.

16S rRNA genes were amplified using primers 9bf and 1512uR and sequenced with 9bf and 519uF (46, 47). Relatives were found using BLAST (http://blast.ncbi.nlm.nih.gov) and collected for analysis. All phylogenetic analyses were performed using the phylogeny.fr web server and further refined using Interactive Tree Of Life (iTOL) (48–51).

DNA extraction of strain CT6 was performed using a modified protocol from (52). Fifty milliliters of mid-exponential-phase culture was harvested and stored at −80°C before DNA extraction. Pellets were thawed at 80°C for 30 min and suspended in an equal volume of chloroform-methanol (2:1) for 1 h with periodic shaking at room temperature. This mix was centrifuged for 20 min at 2,500 × g, and the organic and aqueous phases were discarded without disturbing the cell material. The cell pellet was dried for 10 min at 55°C, suspended in 550 µl of 100 mM Tris (pH 9.5), and incubated overnight at 37°C with lysozyme (final concentration, 0.5 mg ml⁻¹). After lysozyme treatment, SDS (final concentration, 1%) and proteinase K (final concentration, 0.75 mg ml⁻¹) were added, and the tube was left at 60°C for 1 h with periodic mixing. During the last 15 min of incubation, 100 µl each of 5 M NaCl and cetyltrimethylammonium bromide (CTAB) solution (10% CTAB solution in 0.7 M NaCl) was added. DNA was extracted with 1 vol of chloroform-isoamyl alcohol (24:1), and the aqueous phase was treated with RNase A. RNA-free DNA was ethanol precipitated and stored at −80°C before sequencing.

SMRT sequencing was performed using the Pacific Biosciences RS II platform with 10-kb libraries and P5/C3 chemistry using the standard protocols. The de novo genome assembly, using HGAP v2 (53), resulted in a single circular contig. The genome of strain CT6 was annotated using RAST and BASys (27–29). Genomes were analyzed using BLAST (http://blast.ncbi.nlm.nih.gov) and the Integrated Microbial Genomes/Expert Review (IMG/ER) web server (https://img.jgi.doe.gov/er/) (26, 30, 31, 35).

To generate material for qRT-PCR analysis, overnight cultures were diluted 1/50 in LB-Tw and grown overnight to an OD₆₀₀ of ~0.3. At this point, PCA, PYO, AQDS, or MB was added to a final concentration of 200 µM. An equivalent volume of water was added to a set of cultures as a control. After 3 h, cell material was collected, flash frozen, and stored at −80°C until RNA extraction. Frozen cell pellets were suspended in 350 µl AES buffer (50 mM sodium acetate [pH 5.3], 10 mM EDTA, 1% SDS) and 350 µl acid phenol-chloroform (5:1; pH 4.5), and 200 µl glass beads (≤106 µm) were added. Samples were homogenized in an analog Disruptor Genie (Scientific Industries, Inc.) four times for 30-s each with a 30-s interval on ice between disruptions. Glass beads were removed by centrifugation, and liquid was transferred to a heavy phase lock tube containing 300 µl chloroform. The phase lock tube was centrifuged for 5 min at 12,000 × g. The aqueous phase was extracted with an additional 300 µl chloroform and transferred to a microcentrifuge tube containing 40 µl 3M sodium acetate (pH 5.2), and RNA was alcohol precipitated and suspended in 100 µl H₂O. This crude RNA extract was cleaned using an RNeasy kit (Qiagen, Inc.) with a modified protocol with optional on-column DNase treatment, as described previously (54). RNA was additionally treated with Turbo DNA-free using the manufacturer's directions (Life Technologies, Inc.).

Purified RNA was converted to cDNA using an iScript cDNA synthesis kit and following the manufacturer's directions (Bio-Rad, Inc.). cDNA was used as a template for qPCR using iTaq universal SYBR green Supermix (Bio-Rad, Inc.) on a 7500 fast real-time PCR system (Applied Biosystems, Inc.). Samples were analyzed in at least duplicate, and the signal from each treatment (PCA, PYO, AQDS, or MB) was first normalized to a water-only control using the following equation: relative expression = P_e ^{−[CT(sample) − CT(control)]} where P_e is the calculated efficiency for a given primer pair. Data were then standardized to rpoA. qRT-PCR primers are listed in Table S2 in the supplemental material.

For RNA-Seq, RNA was extracted after a 20-min exposure to 200 µM PCA, PYO, or water. rRNA was depleted with the magnetic RiboZero kit for Gram-negative bacteria (Epicentre). A library was prepared using the NEBNext mRNA library prep master mix set for Illumina (NEB). Sequencing was performed at the Millard and Muriel Jacobs Genetics and Genomics Laboratory at the California Institute of Technology to a depth of 12 to 15 million reads on an Illumina HiSeq2500 and processed using the Illumina HiSeq control software (HCS version 2.0). Low-quality bases were removed using Trimmomatic (LEADING:27 TRAILING:27 SLIDINGWINDOW:4:20 MINLEN:35) (55), mapped using Bowtie (56), and sorted with SAMtools (57). The number of reads per locus was calculated with easyRNAseq (58). The .gff file was generated using the RAST annotations (27, 28). Reads were normalized by number and compared to the water control. RNA-Seq results can be found in Dataset S1 in the supplemental material.

M. fortuitum ATCC 6841 was used for mutagenesis using a recombineering protocol (34, 59, 60). Plasmids and primers are listed in Table S2 in the supplemental material. M. fortuitum ATCC 6841 was electroporated with pJV53 using a protocol established for M. smegmatis to make a recombineering strain (34, 59). Cultures were plated on LB plates with 100 µg ml⁻¹ kanamycin and allowed to grow for 3 to 5 days. M. fortuitum ATCC 6841 pJV53 was grown overnight in LB-Tw to an OD₆₀₀ of 0.5 to 1, induced for 3 h with 0.2% acetamide, washed and suspended in 10% glycerol as described previously, and frozen at −80°C (34).

Linear DNA constructs for generating allele replacement mutants were generated by PCR amplifying genomic regions flanking the target gene and the gentamicin resistance cassette from pMQ30 (61). PCR products were treated with XbaI and NotI (New England Biolabs) when necessary (see Table S2 in the supplemental material) and T4 ligase (Invitrogen). Constructs were PCR amplified and electroporated into acetamide induced M. fortuitum ATCC 6841 pJV53 and outgrown overnight in LB-Tw medium at 30°C. Allele replacement mutants were selected on LB plates with 100 µg ml⁻¹ gentamicin at 30°C. The position of the gentamicin cassette on the chromosome of single-gene-deletion mutants was confirmed by PCR and sequencing.

To complement mutations in trans, cultures were cured of pJV53 by daily passage at 37°C for 5 days and streak purified. Colonies that had lost pJV53 and by extension kanamycin resistance were identified by patch plating on LB and on LB supplemented with kanamycin. Mutants cured of pJV53 were subsequently transformed with plasmid pSD5 (60) bearing the appropriate gene to complement the mutation.

The complete genome sequence of strain CT6 has been deposited in GenBank (accession number CP011269).