Pseudomonas aeruginosa lasR mutant fitness in microoxia is supported by an Anr-regulated oxygen-binding hemerythrin

We previously reported that ∆lasR mutants in laboratory strains PA14 and PAO1 as well as in diverse clinical isolates had higher activity of the transcription factor Anr than their LasR+ counterparts under microoxic conditions (17, 20). To determine if Anr activity contributes to the higher fitness of lasR mutant strains, we developed an assay for the assessment of fitness in colony biofilms grown under microoxic conditions (0.2%; 2.5 μM O₂). In competitive fitness assays, wherein different strains were competed against a PA14 strain tagged with a constitutively expressed lacZ gene, the two strains had access to a common pool of secreted products including enzymes (e.g., LasA and LasB proteases) and small molecules (such as redox-active phenazines), despite differences in exoproduct production between strains (Fig. 1A for experimental design). Using this assay, we observed that PA14 ∆lasR was more fit than the PA14 wild type under microoxic conditions (Fig. 1B). The increased fitness of ∆lasR could be complemented by restoring lasR to its native locus (Fig. 1C). In both PA14 and the PA14 ∆lasR derivative, deletion of anr led to a loss of fitness that was complemented by restoration of anr (Fig. 1B). The presence of anr was essential to the microoxic fitness advantage that the ∆lasR mutant had over the wild type (Fig. 1B).

The transcription factor Anr regulates dozens of genes involved in denitrification, fermentation, and microoxic metabolism. Multiple transcriptome studies comparing P. aeruginosa strains to ∆anr derivatives in laboratory strain and CF clinical isolate backgrounds revealed that expression levels of the uncharacterized gene PA14_42860 (PA1673 in strain PAO1) are severely reduced in the absence of Anr under microoxic and anoxic conditions (8, 20). Furthermore, analysis of pairs of genetically related isolates containing a lasR mutant and a lasR+ counterpart found this gene, named here mhr for microoxic hemerythrin, to be one of the most up-regulated genes in lasR mutants (20). An Anr consensus-binding motif (TTGATCGGCGTCAA) was found 81 bp from the translational start of Mhr (34). To test whether Anr was a positive regulator of mhr, we constructed a mhr promoter fusion to lacZ and integrated it at a neutral site on the chromosome in the wild type and ∆anr backgrounds. Under microoxic conditions, mhr promoter activity was 64-fold higher in the wild type than in the ∆anr mutant. Mutation of two nucleotides in the Anr consensus motif that are essential for Anr regulation at other promoters (35) led to a large and significant reduction in β-galactosidase production (Fig. 2A). Consistent with previous reports that mhr transcripts were more abundant in lasR loss-of-function mutants compared to their lasR-intact counterparts, mhr promoter activity was 1.9-fold higher in the ∆lasR mutant compared to the wild type (Fig. 2B). As in the wild type, promoter activity was Anr-dependent in the ∆lasR mutant (Fig. 2B).

To determine if Mhr protein levels were also higher in LasR− strains, we expressed and purified P. aeruginosa Mhr from Escherichia coli (SI Appendix, Fig. S1) and used it to raise an anti-Mhr polyclonal antibody. Western blot analysis detected purified Mhr at about 15 kDa, similar to its predicted molecular weight of 17.9 kDa, and a similar band was detected in P. aeruginosa whole-cell lysates from the wild type but not in the ∆anr mutants (Fig. 2C). Mhr protein levels in the ∆lasR strain were also dependent on anr. We quantified the relative amounts of Mhr using the Licor Odyssey near-infrared Western blotting system. The Mhr protein levels were, on average, 2.8-fold higher in the ∆lasR mutant than in the wild type (Fig. 2D). We also analyzed Mhr levels in a clinical isolate from a corneal eye infection (262 K) that was a natural lasR mutant (17) and a 262-K derivative in which the native allele was replaced with the PA14 allele, as well as two clinical isolates from the same subject with CF with one isolate bearing a lasR mutation. We found that, in the two clinical isolate backgrounds, the lasR mutant in each pair had significantly more Mhr than the corresponding LasR+ strain (Fig. 2 E and F).

The Mhr sequence contains the conserved di-iron–binding motif H–HxxxE–HxxxH–HxxxxD (27, 28). When the P. aeruginosa Mhr sequence was threaded onto the crystal structure of Methylococcus capsulatus hemerythrin (Hr), all seven of the residues in the conserved motif were predicted to form the metal-binding active site (Fig. 3A) (36). In addition to its presence in other Pseudomonas spp., the closest Mhr homologs by sequence were found in other gamma proteobacteria (31) (Stenotrophomonas maltophilia, Xanthomonas campestris, Acinetobacter baumannii, Dyella spp.) and the nitrogen-fixing plant symbiont Azotobacter chroococcum, and the active-site residues are conserved across all of these homologous proteins (37) (SI Appendix, Fig. S2). Consistent with presence of the di-iron–binding motif, we detected iron in Mhr protein with a stoichiometry of two metal ions per protein.

P. aeruginosa Mhr purified from E. coli had an absorbance spectrum with maxima at 328 and 371 nm, which is similar to the absorbance spectra of other hemerythrin proteins (38). As in other hemerythrins, there were spectroscopic differences between the oxygen-bound (oxy-Mhr) and oxygen-free (deoxy-Mhr) species (Fig. 3B and SI Appendix, Fig. S3A). Analysis of spectroscopic changes upon gradual addition of an oxygen-containing solution yielded an apparent K_d value of 0.74 μM (Fig. 3C and SI Appendix, Fig. S3B). Thus, Mhr was shown to bind iron and oxygen with an affinity that is relevant to microoxic conditions.

To test whether Mhr was important for growth under microoxic conditions, we constructed ∆lasR∆mhr and ∆mhr strains and performed competition assays at 0.2% oxygen. Both the ∆mhr and the ∆lasR∆mhr mutant were less fit than their parental strains under microoxic conditions (Fig. 4A). We were able to complement the fitness phenotypes of the ∆mhr strains by expressing mhr using the arabinose-inducible plasmid pMQ70 (Fig. 4 B and C). Under anoxic conditions in which nitrate is used an alternative electron acceptor the ∆mhr mutant had no fitness defect, suggesting that Mhr does not play a physiological role in the absence of oxygen (SI Appendix, Fig. S4).

Similar to its fitness phenotype under microoxic conditions, the ∆lasR mutant was more fit than the wild type in our colony biofilm competition assay under normoxic conditions (21% oxygen) (Fig. 4D). Deleting anr in the wild type did not affect fitness under normoxia, but it did result in a fitness defect in the ∆lasR background (Fig. 4D). Similarly, the ∆mhr mutant was as fit as the wild type under normoxic conditions (Fig. 4E), but loss of mhr led to reduced fitness in the ∆lasR mutant background. Together, these data suggest that Anr, in part due to its regulation of Mhr, contributes to ∆lasR fitness even in colony biofilms grown in normoxia.

The role for Mhr in PA14 ∆lasR but not wild type in normoxic colonies raised the question of whether Anr activity is more strongly induced in ∆lasR mutants as oxygen becomes depleted. Using a liquid culture assay in which increasing volumes of medium were used to decrease the surface area for oxygen transfer, we found that, in both strains, mhr promoter induction was low with high aeration (1- and 5- mL culture volumes) and increased with culture volume, consistent with Anr activity inversely correlating with oxygen availability (SI Appendix, Fig. S5A). At the 10- and 20-mL volumes, the ∆lasR mutant had significantly higher levels of mhr promoter activity than the wild type despite similar culture densities (SI Appendix, Fig. S5 A and B). Using the 10-mL culture volume, we found that, like ∆lasR, the ∆lasI strain which cannot synthesize the LasR ligand 3-oxo-C12-homoserine lactone (3OC12HSL) also had significantly higher levels of mhr promoter activity and the elevated activity could be complemented by exogenous 3OC12HSL (SI Appendix, Fig. S5C).

To gain more insight into the mechanisms by which Mhr increases microoxic fitness, we analyzed its expression pattern relative to other Anr-regulated genes known to be involved in metabolism. To do so, we used eADAGE, a tool that enables the visualization of transcriptional pattern relationships deduced from the analysis of 1,056 publicly available samples (39). Using the Anr-regulated genes shown in color in Fig. 5A as input, we found that mhr showed an expression pattern that closely mirrored that of the ccoNOPQ-2 genes that encode components of one of the two high-affinity cytochrome cbb₃ oxidases in P. aeruginosa. The deletion of ccoNOPQ-2 has little effect on the microoxic growth because of the enzymes encoded by the remaining ccoNOPQ-1 operon, which have partially redundant activity (8). A strain lacking both loci, ccoNOPQ-1 and ccoNOPQ-2 (∆cco), has a significant microoxic growth defect (8), and thus we used this strain to determine the interaction between Mhr and cbb₃ oxidases. When the ∆cco mutant was competed against the constitutively labeled wild type in microoxia, it was recovered at 0 to 0.012% of the total population. In light of this drastic fitness difference, we competed a lacZ-tagged ∆cco strain against either an untagged version of itself or the untagged ∆cco∆mhr mutant to determine if the ∆mhr phenotype required the presence of the high-affinity cytochrome cbb₃ oxidases. The ∆cco and ∆cco∆mhr mutants had similar microoxic fitness profiles (P = 0.2) (Fig. 5B), suggesting that Mhr may improve microoxic fitness through direct or indirect interaction with cbb₃ oxidases. Because the microoxic defect of the ∆cco mutant is much more severe than that of the ∆mhr mutant, we conclude that, while the cbb₃ oxidases may benefit from the cellular function of Mhr, these enzymes do not require Mhr to function.

In light of the contribution of Mhr to the fitness of ∆lasR under normoxic conditions, we assessed the effects of Mhr on the metabolism of the ∆lasR mutant by performing a metabolomics comparison of ∆lasR and ∆lasR∆mhr colony biofilms grown with atmospheric oxygen. We analyzed cells grown on either lysogeny broth (LB) or artificial sputum medium (ASM) (40) designed to mimic the sputum conditions found in the lungs of individuals with CF. To identify metabolites that were different upon the loss of mhr in both datasets, we used a linear model with two factors (genotype and medium). Our analysis found that comparison of ∆lasR to the ∆lasR∆mhr mutant across both media revealed 60 metabolites with significant (P < 0.05) differences between these two strains and a log₂-transformed fold change >0.5 in either LB or ASM (Fig. 5C and Dataset S1). Among the largest differences between ∆lasR and ∆lasR∆mhr were those metabolites involved in energy generation, including metabolites in the tricarboxylic acid (TCA) cycle. Amino acid metabolic pathways, particularly those involved in the metabolism of sulfur-containing and aromatic amino acids, were also affected by Mhr. Other pathways included those related to the metabolism of nucleotides and lipids, as well as the synthesis of LPS, cell wall, and siderophores. The differences detected in these pathways are consistent with Mhr affecting cellular metabolism and being important for oxic respiration and intracellular redox.