Electron-shuttling antibiotics structure bacterial communities by modulating cellular levels of c-di-GMP

Significance

In cells residing within communities, limitation for electron acceptors (e.g., oxygen) leads to redox imbalance and hampers metabolism. Pseudomonas aeruginosa bacteria in biofilms produce antibiotics called phenazines that facilitate redox balancing. When phenazines and oxygen are not sufficiently available, P. aeruginosa biofilms increase production of ECM, which increases surface area-to-volume ratio and access to oxygen. Here, we describe RmcA, a protein that regulates matrix production in response to phenazine exposure. RmcA contains four sensing domains and two domains that modulate levels of a small-molecule inducer of matrix secretion. These results provide molecular insight into mechanisms underpinning the adaptability of P. aeruginosa biofilms, which likely contributes to their persistence in diverse hosts and clinical settings.

Abstract

Diverse organisms secrete redox-active antibiotics, which can be used as extracellular electron shuttles by resistant microbes. Shuttle-mediated metabolism can support survival when substrates are available not locally but rather at a distance. Such conditions arise in multicellular communities, where the formation of chemical gradients leads to resource limitation for cells at depth. In the pathogenic bacterium Pseudomonas aeruginosa PA14, antibiotics called phenazines act as oxidants to balance the intracellular redox state of cells in anoxic biofilm subzones. PA14 colony biofilms show a profound morphogenic response to phenazines resulting from electron acceptor-dependent inhibition of ECM production. This effect is reminiscent of the developmental responses of some eukaryotic systems to redox control, but for bacterial systems its mechanistic basis has not been well defined. Here, we identify the regulatory protein RmcA and show that it links redox conditions to PA14 colony morphogenesis by modulating levels of bis-(3′,5′)-cyclic-dimeric-guanosine (c-di-GMP), a second messenger that stimulates matrix production, in response to phenazine availability. RmcA contains four Per-Arnt-Sim (PAS) domains and domains with the potential to catalyze the synthesis and degradation of c-di-GMP. Our results suggest that phenazine production modulates RmcA activity such that the protein degrades c-di-GMP and thereby inhibits matrix production during oxidizing conditions. RmcA thus forms a mechanistic link between cellular redox sensing and community morphogenesis analogous to the functions performed by PAS-domain–containing regulatory proteins found in complex eukaryotes.

When microbial cells cannot import or are physically separated from metabolic electron donors or acceptors, diffusible compounds can act as electron carriers and support survival on these substrates (1, 2). These conditions arise in the presence of poised-potential electrodes or insoluble minerals, such as iron oxides (3–5), and in multicellular communities (biofilms) where the formation of chemical gradients leads to oxidant limitation for cells at depth (6–9). Diverse microbes secrete redox-active compounds with the capacity to function as electron shuttles (10–12). In the pathogenic bacterium Pseudomonas aeruginosa PA14, electron-shuttling antibiotics called phenazines support survival on poised-potential electrodes and balance the intracellular redox state of cells in anoxic biofilm subzones (1, 9).

Similar to those formed by many species of microbes, colonies of PA14 develop intricate wrinkle structures on agar-solidified growth media (13). Phenazines profoundly alter PA14 colony morphogenesis, inhibiting the onset of wrinkle formation and changing the organization of wrinkles (12) (Fig. 1A). Modeling of resource availability within colonies suggests that the earlier increase in the colony surface area-to-volume ratio in phenazine-null (∆phz) mutants maximizes access to oxygen for cells that would otherwise become limited for oxidant (14). Measurements of intracellular redox state during development support this hypothesis, as ∆phz colonies exhibit a transient increase in cellular NADH/NAD⁺ just before wrinkling (9). The response of PA14 colonies to electron acceptor limitation has parallels in complex eukaryotic systems, such as embryos and tumors, where biochemical pathways enabling redox-driven developmental control have been described in detail (15). However, the mechanisms whereby electron shuttling compounds influence bacterial community behavior have not been defined.

Fig. 1.

Identification of c-di-GMP-modulating proteins that may contribute to the effect of phenazines on PA14 colony biofilm development. (A, Bottom) Schematic of the suspected relationship between the physiological effects of phenazine deficiency and associated effects on colony morphogenesis. (A, Top) Shown Left to Right: Quantification of cellular c-di-GMP levels in colonies, quantification of colony Congo red (CR) binding, and representative images of PA14 WT and ∆phz colonies grown for 3 d in the colony morphology assay. Error bars represent SD and n = 3. P values were calculated using unpaired, two-tailed t tests (**P ≤ 0.01; ***P ≤ 0.001). (B) Representative images of WT, ∆phz, and strains containing disruptions in the indicated genes (ORF numbers for these genes in P. aeruginosa strains PA14 and PAO1 are shown). Colony images are arranged according to the timing of the onset of wrinkling, with the time of inoculation represented by the start of the timeline (Left). Cartoons shown next to each colony depict the domain organization of each disrupted protein. (C) Surface coverage over time for WT and for mutants with early-onset wrinkling. Error bars represent SD and n = 3. (D) Colony development over time for the indicated strains grown in the colony morphology assay. (Scale bars, 1 cm.)

PA14 colony wrinkling requires secretion of ECM, which is stimulated by the second messenger bis-(3′,5′)-cyclic-dimeric-guanosine (c-di-GMP). Here, we describe the identification and molecular characterization of the protein RmcA and its role in the network of proteins that modulate c-di-GMP levels to influence matrix production and colony morphogenesis in PA14. Our findings suggest that RmcA degrades c-di-GMP in response to the presence of phenazines and/or their effect on the cellular redox state, thereby forming a critical regulatory link between electron acceptor availability and PA14 community structure formation.

Results and Discussion

Identification of c-di-GMP–Modulating Proteins Required for Wild-Type PA14 Colony Development.

In P. aeruginosa PA14, the production and redox-cycling of phenazine antibiotics oxidizes the intracellular redox state and affects multicellular development. Biofilm morphology assays show that WT PA14 colonies are initially smooth and develop concentric wrinkle structures over time, whereas phenazine-null (∆phz) colonies show early and enhanced wrinkle formation, with radial ridges extending from the colony center to the edge. We found that ∆phz colonies exhibit higher levels of intracellular c-di-GMP and matrix production than those formed by the WT (Fig. 1A), consistent with the established model that c-di-GMP stimulates matrix production, which then induces wrinkling (16–18). We sought to identify a sensor that could translate information about the presence of oxidized phenazines into c-di-GMP–dependent changes at the community level. To screen for such a protein, we tested mutants lacking each of the 40 PA14 proteins predicted to produce or degrade c-di-GMP for altered colony morphogenesis (19). These proteins contain one or more of three types of domains: (i) GGDEF domains, which can confer diguanylate cyclase (DGC) activity, and/or (ii) EAL or (iii) HD-GYP domains, which can both confer phosphodiesterase (PDE) activity (20). Eight mutants showed colony phenotypes that differed from that of the WT (Fig. 1B and SI Appendix, Fig. S1 and Movies S1–S3). Several of the proteins represented by these mutants have previously been implicated in P. aeruginosa social behaviors (19, 21–26). However, the early-wrinkling phenotype of ∆PA14_07500, which has not been reported previously, was particularly interesting to us because it closely resembled that of ∆phz, with a high degree of colony spreading and the organization of wrinkles in pronounced radial spokes (Fig. 1 C and D). Genetic complementation confirmed that the altered phenotype of ∆PA14_07500 can be attributed to the function of this locus (SI Appendix, Fig. S2).

∆PA14_07500 and ∆phz Mutants Exhibit Similar Phenotypes in Assays for Social Behavior.

Cumulative reports in the literature have revealed trends in the relationships between a strain’s matrix production during colony growth, ability to form a biofilm at the air–liquid interface of a static culture (pellicle), and capacity for swarming (motility over semisolid surfaces) (24, 25, 27, 28). We note that the ∆phz mutant is unusual in this context because it shows increased matrix production in the colony morphology assay but also exhibits swarming behavior comparable to that of the WT in motility assays; typically, P. aeruginosa mutants with altered c-di-GMP signaling show an inverse correlation between matrix production and swarming. We assessed matrix production, pellicle formation, and swarming for ∆phz and selected mutants identified in our initial screen, and found that only ∆PA14_07500 behaved similarly to ∆phz in all assays (Fig. 2 A–C). This finding suggests that PA14_07500 and phenazines function within the same pathway to affect PA14 social behaviors. ∆phz and ∆PA14_07500 mutants may deviate from the canonical “matrix or swarming” paradigm because the effects of phenazines and PA14_07500 are relevant specifically during the electron acceptor-limited conditions that cells encounter during colony development.

Fig. 2.

Characterization of ∆PA14_07500 (∆rmcA) in assays for social behaviors and in-colony matrix distribution. (A) Biofilm formation, assessed using the microtiter dish assay, after 24 h of growth in M63 medium at 37 °C (50). Error bars represent SD and n = 5. (B) Representative images of swarming assay plates after 16 h of incubation. (C) Matrix content, quantified using the Congo red binding assay, in colonies grown for 2 d on 1% tryptone, 1% agar medium with no added dyes. Error bars represent SD and n ≥ 3. (D) Distribution of Pel polysaccharide (a c-di-GMP regulated component of the matrix) in colony thin sections, visualized by staining with fluorescein-labeled lectin. Colonies were grown for 3 d on 1% tryptone, 1% agar medium, with no added dyes. Four biological replicates of each strain were prepared and imaged in two separate experiments; representative images are shown. Arrow indicates colony surface. DAPI staining is in blue and lectin staining is in green. (Scale bar, 10 µm.)

Colony Matrix Distribution Phenotypes Suggest That Phenazine Exposure Stimulates PA14_07500 (RmcA)-Dependent Degradation of c-di-GMP.

To examine the effects of PA14_07500 deletion on matrix production in more detail, we prepared thin sections of WT, ∆phz, and ∆PA14_07500 colonies and stained them with fluorescein-labeled lectin, which specifically binds the c-di-GMP–regulated PA14 matrix polysaccharide Pel (29). In agreement with our results for matrix quantification (Fig. 2C), we found more pronounced lectin staining in ∆phz and ∆PA14_07500 compared with WT in thin sections of 3-d-old colonies. However, we note that the distribution of lectin binding differed between the two mutant strains (Fig. 2D). ∆phz showed most staining at 15- to 40-µm depth, which corresponds to the microaerobic region of the colony (9). As we have proposed that phenazines act as electron acceptors in the hypoxic and anoxic regions below the colony surface, this observation fits with our model, in which matrix production is stimulated by electron acceptor limitation. ∆PA14_07500 colonies showed most lectin staining at the colony surface and at a greater intensity than that observed for the WT, suggesting that matrix production is increased, and decoupled from the influence of phenazines in this mutant.

PA14_07500 contains four Per-Arnt-Sim (PAS) domains, a GGDEF domain, and an EAL domain (Fig. 1B). PAS domains function to sense environmental signals, such as light, oxygen, and redox state (30, 31), and control regulatory processes in diverse organisms. Although the presence of GGDEF and EAL domains suggests that PA14_07500 could harbor both DGC and PDE activity, the early-onset and enhanced wrinkling of ∆PA14_07500 mutant colonies indicate that PA14_07500 functions as a PDE under the tested conditions. We hypothesize that PA14_07500 senses the presence of phenazines directly or indirectly and degrades c-di-GMP in response, thereby mediating the switch between ∆phz and WT colony morphotypes. We note that PA14_07500 expression levels are comparable in PA14 WT and ∆phz (SI Appendix, Fig. S4 A and B), supporting the idea that protein activity is differentially regulated in the two strains. We also performed control experiments to examine possible confounding effects of the PA14_07500 deletion. Comparing the intracellular redox states of ∆PA14_07500 and ∆phz colonies, which show similar morphogenesis, we found that ∆PA14_07500 does not show the transient increase in NADH/NAD⁺ observed for ∆phz just before wrinkling (SI Appendix, Fig. S3A). Furthermore, ∆PA14_07500 colonies produced phenazines at WT levels (SI Appendix, Fig. S3C). These results indicate that the colony morphotype of ∆PA14_07500 is not a response to electron acceptor limitation (9). ∆PA14_07500 also showed no growth defects in shaken liquid cultures (SI Appendix, Fig. S3B). Together, our observations are consistent with the notion that PA14_07500 degrades c-di-GMP and inhibits wrinkling specifically when phenazines are present and the cytoplasm is relatively oxidized, as depicted by the model shown in Fig. 1A. We chose to further characterize the physiological function of PA14_07500, which we have named RmcA for “redox modulator of c-di-GMP.”

RmcA Functions as a Phosphodiesterase.

To test the hypothesis that RmcA PDE activity modulates c-di-GMP levels to affect matrix production and colony morphogenesis, we created a series of strains with mutations predicted to alter this protein’s function. Strains in which the native rmcA promoter was replaced with a constitutive promoter, leading to a >10-fold increase in protein levels, formed colonies that remained smooth after 3 d of incubation, whereas WT colonies were wrinkled at this time point (SI Appendix, Fig. S4 A and C). Accordingly, colonies of the RmcA-overproduction (“rmcA+”) strain also showed an ∼twofold reduction in c-di-GMP levels and matrix production compared with WT (Fig. 3A). EAL domains typically contain an EAL motif that is required for PDE activity (20). We changed this motif to AAA and found that the mutant phenocopied ∆rmcA in the colony morphology assay (SI Appendix, Fig. S5A). When we mutated this motif in the rmcA+ background, generating strain rmcA+^AAA, the resultant colonies also showed enhanced wrinkling. Assessment of protein levels from strains containing these constructs showed that RmcA expression was not affected by the tested mutations (SI Appendix, Fig. S4D). These results confirm that RmcA’s PDE activity was responsible for the smooth colony phenotype of rmcA+ (Fig. 3A).

Fig. 3.

In vivo and in vitro analysis of RmcA-dependent c-di-GMP degradation. (A) Representative images of colonies grown for 3 d. (Scale bar, 1 cm.) Relative c-di-GMP content, quantified after 2 d of growth, and matrix content, quantified using the Congo red binding assay after 3 d of growth, is shown for colonies of each strain. (B, Top) DRaCALA-based detection of competition between the indicated ligands and ³²P-GTP for binding to His₆-RmcA_cyt. NC, no competitor. (Bottom) His₆-RmcA_cyt shows conversion of c-di-GMP to pGpG in a TLC-based assay for PDE activity. For statistical analysis in A and B, P values were calculated using unpaired, two-tailed t tests (*P ≤ 0.05; **P ≤ 0.01; ***P ≤ 0.001; ****P ≤ 0.0001) and error bars represent SD. (C) Timing of the onset of wrinkling for the indicated strains. Blue points represent the onset of wrinkling for the ∆rmcA parent strain, and yellow arrows indicate the shifts caused by mutating the indicated DGC-encoding gene. (D) Model depicting the relative contributions to the RmcA-sensitive c-di-GMP pool made by DGCs identified in the colony morphology screen.

Next, we tested the ability of RmcA to bind c-di-GMP and investigated the kinetics of RmcA activity. These studies were conducted using His₆-RmcA_cyt, a histidine-tagged version of RmcA that lacks domains predicted to localize to the periplasm and cytoplasmic membrane. Differential radial capillary action of ligand assays (DRaCALA), which enable quantitation of protein–ligand interactions based on ligand migration on a nitrocellulose membrane, showed that His₆-RmcA_cyt was able to bind c-di-GMP (32) (Fig. 3B). In substrate competition assays, binding to radiolabeled c-di-GMP was abolished by unlabeled c-di-GMP and moderately affected by pGpG, consistent with the pGpG binding and inhibition of PDE activity observed for other EAL domain proteins (33). GTP, GDP, and GMP caused only modest decreases in c-di-GMP binding. We also used DRaCALA to measure the affinity of His₆-RmcA_cyt for c-di-GMP and obtained a K_d of ∼8 µM (SI Appendix, Fig. S6B). Finally, incubation of His₆-RmcA_cyt with c-di-GMP led to a modest conversion to pGpG after 60 min (Fig. 3B). We suspect that the periplasmic and cytoplasmic membrane-associated components of RmcA, which were not included in the tested protein, contribute to PDE activity in vivo, as membrane localization can facilitate protein dimerization. Nevertheless, our observations of RmcA substrate binding and c-di-GMP conversion activity support the model that RmcA acts as a PDE to decrease c-di-GMP levels, and therefore matrix production, in phenazine-producing PA14 colonies.

The Diguanylate Cyclases RoeA and SadC Are the Primary Sources of c-di-GMP That Promotes Colony Wrinkling.

The observation that RmcA lowers c-di-GMP levels during PA14 colony development implies that there is at least one source of c-di-GMP that stimulates wrinkling when RmcA is not fully active. We investigated this by monitoring the onset of colony wrinkling for a series of combinatorial mutants lacking rmcA and each of the four DGCs identified in our screen (Fig. 1B). The double-mutants ∆sadC∆rmcA and ∆roeA∆rmcA both showed large delays in the onset of wrinkling relative to ∆rmcA (Fig. 3C and Movies S4 and S5), suggesting that SadC and RoeA are primary producers of the c-di-GMP that RmcA then degrades to exert its inhibitory effects on colony structure formation (Fig. 3D). These results are similar to those described for double mutants lacking each of these DGCs in combination with removal of BifA (21). SadC is also a primary contributor to the c-di-GMP pool that promotes matrix production when the Gac/Rsm regulatory cascade is active in the P. aeruginosa strain PAK (34).

Mutation of Individual RmcA PAS Domains Reveals a DGC Activity for the GGDEF Domain.

Having established that RmcA likely functions as a PDE under the tested conditions, we shifted our attention to its sensory domains. We mutated the RmcA PAS domains (Fig. 4A), which may affect protein activity as a function of cellular redox conditions or in response to phenazine binding. We note that a version of RmcA with all four PAS domains deleted showed expression levels comparable to those of WT RmcA when assessed by Western blotting, suggesting that the individual PAS domain deletions did not simply abolish protein production (SI Appendix, Fig. S4D). Deleting PASb, PASc, and PASd resulted in colonies that phenocopied ∆rmcA, indicating that these domains support RmcA PDE activity (Fig. 4B). Intriguingly, deleting PASa had the opposite effect on colony morphology: in both the WT and ∆phz backgrounds, the absence of PASa yielded colonies that were smoother than those formed by their corresponding parent strains (Fig. 4B), suggesting that the RmcA PASa domain functions to stimulate matrix production. One possible explanation for this is that RmcA promotes c-di-GMP synthesis under some conditions, perhaps through activity of its GGDEF domain. RmcA stands out among the PDEs identified in our screen as the only one that contains both an intact GG(D/E)EF motif and an intact “I-site” (an RxxD motif that is found in some GGDEF domains) (Fig. 4A). I-sites can participate in noncompetitive allosteric inhibition, and can be found in noncatalytic c-di-GMP binding proteins (35) but are often found in proteins with DGC activity (20). Indeed, we note that properties of rmcA+^AAA colonies, constitutively overexpressing a version of RmcA with an intact GGDEF motif and no EAL motif, are consistent with a DGC activity for this protein, as they exhibit: (i) a distinct phenotype, with taller structures forming emanating spokes, and (ii) higher c-di-GMP and matrix levels, relative to colonies of ∆rmcA (Fig. 3A). To test if this colony morphology phenotype depended on RmcA’s GGDEF domain, we mutated it to GGAAF in the rmcA+^AAA background. Indeed, wrinkling and spreading were less pronounced in strain rmcA+^GGAAF;AAA (Fig. 4C). In contrast, in strain backgrounds with an intact RmcA-EAL domain (WT and rmcA+), the GGAAF mutation caused enhanced wrinkling compared with the respective parent strains (Fig. 4C and SI Appendix, Fig. S5A). One explanation for this could be that, under these conditions, the GGDEF motif is required for full activity of the EAL domain, similar to observations reported for the mutational analysis of the PDE BifA (24). Alternative point mutations in the GGDEF domain recapitulated this phenotype (SI Appendix, Fig. S5B). Assessment of protein levels from strains containing RmcA with GGDEF or EAL domain mutations showed that their expression levels were comparable to corresponding strains that contain WT RmcA (SI Appendix, Fig. S4D).

Fig. 4.

Contributions of individual domains to RmcA function. (A, Top) Alignment of RmcA PAS domains with the FAD-binding PAS domain of NifL (Klebsiella pneumoniae). Residues highlighted in blue are only conserved between PASd and NifL. Importantly, one of these is a tryptophan (W745 in PASd), which is required for FAD binding (30). (Bottom) Alignment of the GGDEF domains of PDEs identified in the colony morphology screen. (B) Representative images of PA14 WT, ∆phz, and various mutants with deletions in rmcA. (Scale bar, 1 cm.) (C) Representative images of PA14 mutants overexpressing RmcA protein with the indicated deletions or point mutations. Colonies in B and C were grown on colony morphology assay medium for 3 d before imaging. (Scale bar, 1 cm.) (D) The cartoon depicts hypothesized interactions between individual domains of RmcA.

Finally, we used the in vitro assays described above to test His₆-RmcA_cyt for GTP binding and DGC activity. DRaCALAs showed that this protein bound GTP and that binding to radiolabeled GTP was abolished by unlabeled GTP and GDP but not significantly affected by other substrates (SI Appendix, Fig. S6A). This finding suggests that GTP and c-di-GMP bind to different sites on the protein. We were not able to obtain purified His₆-RmcA_cyt at concentrations high enough for precise determination of its K_d for GTP. However, we can compare the curves generated by GTP and c-di-GMP binding to their respective sites. This approach allows us to estimate that c-di-GMP binds to the His₆-RmcA_cyt protein ∼8× better than GTP, giving a K_d for GTP of ∼60–70 µM (SI Appendix, Fig. S6B). We did not observe c-di-GMP synthesis during a 60-min incubation of His₆-RmcA_cyt with GTP (SI Appendix, Fig. S6A), perhaps because of condition-dependence of RmcA DGC activity. As above, we note that the periplasmic and cytoplasmic membrane-associated components of RmcA, which were not included in the tested protein, may facilitate dimerization or otherwise support functionality and may therefore be required to see levels of activity relevant for in vivo effects. Taken together, our analyses of the RmcA GGDEF domain indicate that it can play opposing roles in the modulation of c-di-GMP levels by either directly contributing to c-di-GMP synthesis or supporting the c-di-GMP degradation activity of the EAL domain (model in Fig. 4D). The opposing DGC and PDE activities of RmcA likely function under differential growth conditions, as has been observed for the DGC/PDE MucR in P. aeruginosa strain PAO1 (36, 37).

RmcA’s PAS Domains Differentially Influence RmcA Activity and Exhibit Properties Consistent with FAD Binding.

To further investigate the opposing roles of the PAS domains in RmcA function, we made additional mutations in the rmcA+^AAA strain. Among the four PAS domains of RmcA, PASa stands out because of its promotion of colony wrinkling, whereas PASd is unique in that it shares many residues with the FAD-binding PAS domain of the redox-sensing protein NifL, including a crucial tryptophan (W745) implicated in aromatic stacking with the adenine moiety of FAD (30) (Figs. 4A and 5A). Deleting PASa in the rmcA+^AAA background decreased the extent of colony wrinkling, furthering the notion that PASa is required for DGC activity of RmcA (Fig. 4C). Because a defective GGDEF domain can inhibit PDE activity (Fig. 4C), we also tested whether PASd may act via the GGDEF domain to affect PDE activity. Indeed, deleting PASd in the rmcA+^AAA background resulted in colonies phenocopying the rmcA+^GGAAF;AAA mutant. These results raise the possibility that the PASa and PASd domains influence the DGC activity of the GGDEF domain, and suggest that PASd, but not PASa, influences the stimulating effect that the GGDEF domain has on the EAL domain (model in Fig. 4D). In accordance with the noted sequence features of PASd, purification of a truncated form of RmcA composed of the four PAS domains (RmcA_PAS-His₆) yielded a yellow preparation exhibiting an absorbance spectrum consistent with FAD binding (Fig. 5A). Mutating the tryptophan residue W745 in the PASd domain of the recombinant protein decreased FAD binding significantly (SI Appendix, Fig. S7). The W745A mutation in the endogenous protein was sufficient to recapitulate the ∆rmcA colony phenotype (Fig. 5B).

Fig. 5.

Evidence for phenazine-dependent modulation of RmcA activity. (A, Top) Purification of RmcA_PAS-His₆ yields a preparation that is bright yellow in color, consistent with the presence of an oxidized FAD cofactor. The UV-visible spectrum for purified RmcA_PAS-His₆ shows peaks corresponding to oxidized FAD. (Bottom) Structure of FAD. (B) A 745W→A point mutation in RmcA leads to enhanced colony wrinkling. Colonies were grown on colony morphology assay medium for 3 d before imaging. (C) Quenching of the intrinsic fluorescence of His₆-RmcA_cyt by the phenazine pyocyanin. Tryptophan fluorescence emission was monitored at 330 nm before and after addition of increasing concentrations of pyocyanin. Analysis of these results indicated a K_d of 3 µM (SD = 0.9; n = 4) for binding of pyocyanin to the protein. (D) Timing of the onset of wrinkling for the indicated strains. Orange and teal shading represent the shift in the onset brought about by the indicated mutations made in the ∆phz and WT parent strains, respectively. (E) Graphical representation of the effect of electron shuttle limitation and high NADH/NAD⁺ levels on signaling and morphological output in PA14 colonies. “X” represents a putative redox-sensitive DGC. AU, arbitrary units. (Scale bar, 1 cm.)

RmcA Binds Phenazines.

The eukaryotic PAS-domain–containing protein AhR (aryl hydrocarbon receptor) is required for activation of various aspects of the innate immune response during exposure to phenazines or infection with P. aeruginosa or Mycobacterium tuberculosis in cell lines or mice (38). Intrinsic fluorescence quenching has shown that AhR binds the P. aeruginosa product 1-hydroxyphenazine (38). To test the possibility that RmcA binds phenazines, we monitored the effects of phenazines on the intrinsic protein fluorescence of His₆-RmcA_cyt. We observed a decrease of intrinsic His₆-RmcA_cyt fluorescence in the presence of the endogenous methylated phenazine pyocyanin. Analysis of these results indicated a K_d of 3 µM (assuming a one-site binding model) (Fig. 5C). Another methylated phenazine, the synthetic compound phenazine methosulfate (PMS), bound the protein with a slightly lower affinity (K_d of 6 µM), whereas the endogenous, nonmethylated compound phenazine-1-carboxylate (PCA) bound the protein with markedly lower affinity (K_d of 16 µM) (SI Appendix, Fig. S8). It is interesting to note that the higher affinities of the methylated phenazines, relative to PCA, for His₆-RmcA_cyt binding are consistent with the colony morphology phenotypes of PA14 mutants defective in specific steps of the phenazine biosynthetic pathway. As we have reported previously, a mutant that produces PCA but is unable to produce methylated phenazines shows increased matrix production and wrinkle formation (39), implying that methylated phenazines are the primary signals controlling RmcA activity.

RmcA Inhibits Colony Wrinkling in a Phenazine-Dependent Manner.

Our mutational analysis suggests that the PAS domains support modulation of RmcA activity in response to cellular conditions. We propose that RmcA is activated to degrade c-di-GMP when oxidized phenazines are present and the redox state of the cytoplasm is relatively oxidized (9). According to this model, RmcA is responsible for the delayed colony wrinkling exhibited by WT PA14 relative to ∆phz, and the ∆phz early-wrinkling phenotype can be attributed to low RmcA activity. To test this, we created combinatorial mutants lacking phenazines and each of three PDEs identified in our screen, including RmcA. For PDEs that function independently of phenazines and their effect on the cellular redox state, we expected the deletion of their corresponding genes to shift the onset of wrinkling of ∆phz colonies to an earlier time point. In contrast, for a PDE with activity that responds to phenazines or their effect on cellular redox state, we expected its deletion to have little or no effect on the onset of ∆phz colony wrinkling. Among the ∆PDE-∆phz mutants tested, ∆rmcA∆phz was the only one that showed a negligible shift in the onset of wrinkling relative to its ∆phz counterpart (Fig. 5D and Movie S6). This result constitutes further evidence that RmcA mediates phenazine-dependent effects on PA14 colony morphogenesis (Fig. 5E). We note that the delay of the onset of ∆rmcA wrinkling relative to ∆rmcA∆phz suggests the involvement of an additional regulator that mediates the effects of phenazines on colony wrinkling, such as a redox-sensitive DGC or another redox-sensitive PDE (Fig. 5 D and E). A recent study by Schmidt et al. identified putative redox-sensing proteins that may act on SadC to modulate c-di-GMP production in response to oxygen availability (40).

Our results indicate that the PAS domains of RmcA may sense the presence of phenazines and influence the c-di-GMP–modulating activity of this protein. RmcA could sense phenazines indirectly via their effects on an FAD cofactor in the PASd domain, directly through phenazine binding, or through a combination of these mechanisms. PAS domains are well-known for their roles in sensing redox conditions and subsequently modulating multicellular development in diverse eukaryotes. A salient example is that of HIF-1, which affects mammalian development, metabolism, and tumor formation in response to limitation for molecular oxygen and other redox cues (41, 42). Although PAS-domain–dependent, redox-sensitive regulation of exopolysaccharide production and biofilm development have been reported for selected nonpathogenic bacteria (43–45), the observations presented herein extend this model to the morphogenetic control of the pathogen P. aeruginosa, and specifically to the effects of endogenously produced electron shuttles (Fig. 5E). The PAS-domain–containing kinase KinA has been implicated in redox-dependent control of colony development in the model spore-forming bacterium Bacillus subtilis (46), which is not known to produce electron shuttles but shares the soil habitat with producers of these compounds. Whether additional electron shuttle-producing microbes, or those that exploit the shuttles produced by others, have the capacity to control social behavior during this process is not known. However, as electron shuttle production and utilization is observed in phylogenetically disparate organisms (3, 12), and developmental regulation by the secondary messenger c-di-GMP is a common bacterial trait (20, 47), similar mechanisms could be operating in multicellular communities formed by diverse microbes.

Methods

Strains and Growth Conditions.

Strains used in this study are listed in SI Appendix, Table S1. Unless otherwise stated, liquid cultures of P. aeruginosa UCBPP-PA14 (48) were grown in LB (Miller) (49) at 37 °C, with shaking at 250 rpm. For genetic manipulation, strains were routinely plated on LB + 1.5% agar. For selection purposes, gentamicin was added to the medium at 100 µg/mL and 15 µg/mL for P. aeruginosa and Escherichia coli, respectively. Transposon insertion strains shown in SI Appendix, Fig. S1 were obtained from the PA14NR Set, described in Liberati et al. (50). Transposon insertion sites were verified by PCR.

Liquid-Culture Growth Assay.

Overnight precultures were grown in LB at 37 °C with shaking at 250 rpm. Precultures were diluted in 1% (wt/vol) tryptone to a starting optical density at 500 nm (OD₅₀₀) of 0.01 into 96-well plates (Greiner Bio-One). Cultures were shaken in a plate reader (Synergy 4; Biotek) at “medium” speed at 37 °C and absorbance at 500 nm was measured every 30 min.

Colony Morphology Assay.

Colony morphology assay medium was prepared by autoclaving a mixture of 1% tryptone (Teknova) and 1% agar (Teknova). The mixture was cooled to 60 °C and 20 µg/mL Coomassie blue (EMD) and 40 µg/mL Congo red (EMD) dyes were added. Sixty milliliters of medium was poured into 10- × 10- × 1.5-cm square plates (LDP) and allowed to solidify and dry for 16–24 h. Ten microliters of overnight precultures were spotted on dried plates and colonies were grown at 25 °C and >90% humidity. Images of colonies were taken daily using a Keyence VHX-1000 digital microscope.

Time-Lapse Imaging of Colony Development.

Ten microliters of overnight precultures were spotted on colony morphology assay plates and grown for 24 h at room temperature. Images of the developing colonies were then taken every 15 min over a 4-d period using a customized recording system (HD Webcam C525; Logitech) under LED illumination. Lighting and image capture were synchronized with LabView (National Instruments).

Microtiter Dish Biofilm Assay.

Pellicle formation was assessed using a microtiter dish assay and quantification of biofilm staining with the dye Crystal violet, as described in ref. 51. Briefly, overnight LB precultures of WT PA14 and mutants of interest were diluted 1:100 into M63 medium [3 g KH₂PO₄, 7 g K₂HPO₄, and 2 g (NH₄)₂SO₄ per liter amended with 0.2% glucose, 0.5% casamino acids, and 1 mM MgSO₄] and dispensed, 100 µL per well, into PVC nontissue-culture–treated flexible U-bottom 96-well microtiter dishes (Falcon #353911). The outer wells of the microtiter dish were filled with distilled water to prevent evaporation. Data shown are representative of biological replicates.

Swarming Assay.

Overnight LB precultures were centrifuged at 10,000 rpm for 2 min (Eppendorf Centrifuge 5418). The pellet was resuspended in 1× PBS at an OD₅₀₀ of 1. Two microliters of this suspension were spotted in the center of a polystyrene Petri dish (100 mm × 15 mm; Fisher Scientific) containing 20 mL of swarming assay medium (M9 salts medium amended with 0.2% glucose and 0.5% casamino acids, solidified with 0.5% agar) that had been dried for 45 min in a laminar flow hood. Plates were sealed with parafilm and incubated in an inverted position for 16 h at 37 °C (52) before imaging with a Canon CanoScan 5600F scanner.

Quantification of c-di-GMP from Colonies.

Ten microliters of overnight precultures were spotted in biological triplicate on colony morphology assay agar plates (prepared as described above) and grown for 2 d at 25 °C and >90% humidity. Each colony was scraped from the plate, transferred to 1 mL PBS, and homogenized using a BeadBug (Benchmark Scientific) for 3 min. Homogenized colonies were transferred to preweighed MicroTubes and pelleted by centrifugation for 1 min at 16,873 × g. The supernatant was removed and each pellet was resuspended in 250 µL extraction buffer [methanol:acetonitrile:water (40:40:20) with 0.1 N formic acid] before incubation at −20 °C for 1 h. The extract was then centrifuged at 14,549 × g for 5 min at 4 °C. Two-hundred microliters of supernatant were transferred to a fresh tube and neutralized with 8 μL of 15% NH₄HCO₃ and dried with a speed vacuum concentrator. The c-di-GMP pellet was resuspended in 200 μL of 10 mM tributylamine + 15 mM acetic acid in 3% methanol. Samples were quantified at the Mass Spectrometry Facility of Michigan State University by electrospray ionization analysis with Quattro Premier XE LC/MS/MS. The pellet from the extraction step was dried using a speed vacuum concentrator and weighed. The final concentration of cyclic di-GMP was normalized to cell dry weight.

Matrix Quantification.

Congo red binding, which has been used for quantification of exopolymeric substances (matrix) from various bacterial species (ref. 16 and references therein), was used to quantify matrix production by P. aeruginosa PA14 colonies. Ten microliters of overnight precultures were spotted in biological triplicate on 1% tryptone, 1% agar solidified medium, and spots were grown for 2 or 3 d, as indicated, at 25 °C. Each colony was scraped from the plate, resuspended in 1 mL PBS + 40 µg/mL Congo red dye, and incubated at 37 °C for 1 h. Samples were then centrifuged at 16,873 × g for 2 min, and supernatants were transferred to a clear 96-well plate (Greiner Bio-One) for measurement of absorbance at 490 nm using a plate reader (Synergy 4; Biotek ). PBS + 40 µg/mL Congo red was measured as the “no matrix” standard.

Fluorescence-Based Detection of Pel Polysaccharide in Colony Thin Sections.

Two-layer agar plates (bottom: 5 mm; top: 1 mm) were prepared using 1% agar and 1% tryptone, pouring the 5-mm layer first, allowing it to polymerize, and then pouring the 1-mm layer. LB cultures were inoculated from single colonies and incubated ∼16 h with shaking at 250 rpm at 37 °C. These were diluted 1:100 into fresh medium and grown for ∼3.5 h to reach exponential phase. The exponential-phase cultures were spotted as 5-µL droplets on the two-layer agar plates, allowed to dry at ambient temperature, and then incubated at 25 °C for up to 3 d. At specified time points, colonies were laminated with a 1-mm-thick, 1% agar overlayer and a scalpel was used to lift each laminated colony into a Histosette (Fisher Scientific). Samples were fixed by incubating first for 4 h at 4 °C in 50 mM l-lysine hydrochloride (in PBS), then for 4 h at 4 °C in 50 mM l-lysine hydrochloride and 4% (wt/vol) paraformaldehyde (in PBS), and finally for 16–24 h at 37 °C. Fixed samples were then washed 2× with PBS, dehydrated with increasing concentrations of ethanol in PBS (25%, 50%, 70%, 95%, 3 × 100%), and cleared with Histoclear II (3 × 100%) at room temperature before infiltration with paraffin wax (2 × 100%) at 62 °C. Buffer wash, dehydration, clearing, and infiltration were performed with a STP120 Tissue Processor (Thermo Fisher Scientific) at 1 h for each solution and 2 h for each paraffin wash. Colonies were then embedded in molds with paraffin wax (Paraplast Xtra, Fisher Scientific) and cut into 10-µm vertical sections at the center of the colony with a Leica RM2255 rotary microtome using low-profile blades (Sturkey, Fisher Scientific). Sections were mounted onto frosted-glass microscope slides (Fisher Scientific) by floating on a 42 °C water bath, heat-fixed for 1 h at 46 °C on a hot plate, and rehydrated with Histoclear II followed by decreasing concentrations of ethanol followed by PBS. Mounted sections were then stained for 15 min with 100 µg/mL fluorescein-labeled Wisteria floribunda lectin (Vector Laboratories) (in PBS), washed 2× in PBS, and then sealed with Fluoro-gel II with DAPI (Electron Microscopy Sciences, Fisher Scientific) under glass coverslips. Sections were imaged with an LSM-700 confocal fluorescence microscope (Zeiss). Laser strength and gain in the DAPI channel was modulated to maximize contrast in the sample. For the EGFP channel (used to visualize fluorescein), laser strength and gain were standardized for every sample at 55 and 625, respectively.

NADH/NAD⁺ Extraction Assay.

Extraction of NADH and NAD⁺ was performed as described previously (9). Ten microliters of precultures were spotted in biological triplicates on colony morphology assay agar plates (prepared as described above). At the indicated time point, each colony was scraped off the plate into 1 mL 1% tryptone using a sterile 1-mL pipette tip and homogenized using a pellet disruptor. For each colony, two 450-µL samples were transferred to separate MicroTubes (Sarstedt) for NADH and NAD⁺ extractions. NADH and NAD⁺ were quantified using an enzyme-cycling assay as described in Kern et al. (53).

Phenazine Quantification.

Ten technical replicates (10 µL each) of overnight precultures were spotted on 10-cm round Petri dishes containing 40 mL of solidified 1% agar, 1% tryptone medium. Colonies were incubated for 3 d at 25 °C and 80–90% humidity, then scraped from the plate. Half of the solidified medium containing released phenazines was transferred to a 50-mL conical tube containing 3 mL of water and nutated overnight in the dark at room temperature to extract phenazines. Extract was filtered using 0.22-µm Spin-X filter tubes (Costar) and loaded directly onto a Waters Symmetry C18 reverse-phase column (4.6 × 250 mm; 5-mm particle size) in a Beckman SystemGold HPLC with a photodiode array detector. Phenazines were separated following a previously described protocol (54). Conversion factors used for pyocyanin, PCA, and phenazine-1-carboxamide (PCN) were 8 × 10⁻⁶ mM/AU, 9.5 × 10⁻⁶ mM/AU, and 9.5 × 10⁻⁶ mM/AU, respectively.

Western Blot Analysis.

Lysis buffer [50 mM Tris⋅HCl, pH 7.5, 0.5 M NaCl, 10% (vol/vol) glycerol, 1 mM MgCl₂, 1 mg/mL lysozyme (Sigma), 13.2 µg/mL DNase (Sigma), and 1 complete EDTA-free protease inhibitor tablet (Roche) per 50 mL] was degassed and chilled to 4 °C. Two-milliliter cultures were grown overnight with shaking (250 rpm) at 37 °C, then used to inoculate 1-L cultures. The 1-L cultures were grown overnight at 37 °C with shaking (250 rpm). Cells were pelleted by centrifugation at 3,250 rpm for 30 min (Beckman J6-MI) and were resuspended in 50 mL lysis buffer.

Cells were lysed by four passages through an Avestin Emulsiflex C-3 cell disrupter at an operating pressure of 15,000 psi. Cell lysate was centrifuged at 4,500 rpm for 15 min (SX4750 rotor; Beckman Coulter) at 4 °C to remove cell debris and unlysed cells; the supernatant was then ultracentrifuged at 45,000 rpm for 1 h (L8-70M; Beckman). The pellet was rinsed twice and then resuspended in 10 mL of resuspension buffer [0.5 M NaCl, 50 mM Tris⋅HCl, pH 7.5, 0.3% (wt/vol) n-Dodecyl β-d-maltoside, 1 mM Tris(2-carboxyethyl)phosphine (TCEP) and 30% (vol/vol) glycerol]. The resuspended pellet was incubated overnight with slow rotation at 4 °C and then ultracentrifuged at 45,000 rpm for 1 h (L8-70M; Beckman). Protein concentration in supernatant was determined using the Bio-Rad protein assay with BSA as the standard.

Protein samples were separated on a 4–12% NuPage Bis-Tris gel (Life Technologies) and blotted onto a 0.45 µM-pore size nitrocellulose membrane (GE Water and Process Technologies). The membrane was stained with 0.1% Ponceau S stain (G Biosciences) and destained with 0.1 M NaOH. The membrane was incubated overnight in blocking solution [1× Tris-buffered saline with 1% (vol/vol) Tween-20 (TBST) and 5% (wt/vol) nonfat dry milk powder]. The membrane was then probed with anti-6× His tag (HIS.H8; Abcam) at a 1:2,000 dilution in 50% (vol/vol) blocking solution (diluted with 1× TBST) for 1 h at room temperature. Blots were washed and probed with goat anti-mouse HRP-conjugated secondary antibody (Sigma) and developed using Amersham ECL Plus Western Blotting Detection kit (GE Healthcare).

Construction of Deletion, Complementation, Point Mutation, His-Tag, and Constitutive Expression Plasmids and Strains.

Deletion, complementation, point mutation, His-tag, and constitutive expression plasmids were generated by modifying allelic-replacement vector pMQ30 (7.5-kb mobilizable vector, oriT, sacB, Gm^R) using yeast gap-repair cloning (55). All strains that were generated with these plasmids are listed in SI Appendix, Table S1 and the primers used to construct the plasmids are listed in SI Appendix, Table S2.

For deletion plasmids, two ∼1-kb regions flanking the genes to be deleted were amplified and joined in pMQ30 by yeast gap-repair cloning. For point mutations, the flanking regions extended from the sites to be mutated, with the desired mutations, were incorporated in the respective primers. For the complementation plasmid, a fragment including rmcA and ∼1-kb up- and downstream regions were amplified. A “promoter swap” plasmid was made to generate strains that constitutively express rmcA. For this, three fragments were amplified: two ∼1-kb regions flanking the native rmcA promoter and the constitutively expressing synthetic lac promoter PA1/04/03 (56, 57). The lac promoter was cloned into pMQ30 surrounded by the two flanking regions using yeast gap-repair cloning. The His-tag plasmid was designed to insert a N-terminal 9×-histidine tag after RmcA’s cleavable signal sequence at residue position 25. For this process, ∼1-kb flanks that extended from the appropriate position in rmcA (72 bp from the start site) and encoded the 9× histidines were amplified.

The respective PCR fragments were assembled into the linearized pMQ30 vector by gap repair cloning using the yeast strain InvSc1 (Invitrogen) (55). The resulting plasmid was transformed into E. coli BW29427 and transformed into PA14 using biparental conjugation. PA14 single recombinants were selected on LB agar containing 100 µg/mL gentamicin. Potential complemented strains, deletion and point mutants, and strains containing His-tag constructs were generated by selecting for double recombinants by identifying strains that grew in the presence of 10% (wt/vol) sucrose. Strains with properties of double recombination were further analyzed by PCR.

Construction, Expression, and Purification of Recombinant Histidine-Tagged RmcA Proteins.

Two plasmids encoding truncated versions of recombinant histidine-tagged RmcA were created: His₆-RmcA_cyt (amino acid residues 320–1,232), with the cytoplasmic portion of RmcA and an N-terminal 6×-histidine tag; and RmcA_PAS-His₆ (amino acid residues 320–810) encoding only the four PAS domains of RmcA followed by a C-terminal 6×-histidine tag. The gene fragment encoding the cytoplasmic portion of RmcA was first ligated into the pMAL-c5× MBP fusion vector (New England Biolabs) using BamHI and HindIII restriction sites, creating MBP-RmcA_cyt. The cytoplasmic portion of RmcA was then excised from MBP-RmcA_cyt and ligated into pET28b vector using NdeI and HindIII restriction sites and expression was carried out in E. coli BL21 (DE3) (Agilent). The gene fragment encoding just the PAS domains of RmcA was ligated into the pET21a vector using NdeI and XhoI restriction sites. A tryptophan residue in the PASd domain that corresponds to residue 745 in the full-length protein was changed to an alanine by site-directed mutagenesis (GenScript). Expression of RmcA_PAS-His₆ and RmcA_PAS-W745A-His₆ was carried out in E. coli BL21 (DE3) pMgK, which contains a plasmid encoding the genes for three rare tRNAs (Northeast Structural Genomics Research Network). Cultures were grown at 37 °C with shaking at 250 rpm to OD₆₀₀ = 0.5 in Terrific Broth medium with 35 μg/mL kanamycin and 100 µg/mL ampicillin. Expression was induced by the addition of 0.5 mM isopropyl β-d-1-thiogalactopyranoside (IPTG) and cells were incubated for an additional 16–18 h at 16 °C, with continued shaking at 250 rpm. Cells were harvested by spinning at 2,700 × g for 30 min and resuspended in 3× pellet volume of lysis buffer [50 mM Tris⋅HCl, pH 7.5, 0.5 M NaCl, 10% (vol/vol) glycerol, 1 mM MgCl₂, 1 mg/mL lysozyme (Sigma), 13.2 µg/mL DNase (Sigma) and one complete EDTA-free protease inhibitor tablet (Roche) per 50 mL]. Cells were sonicated and lysate was centrifuged at 24,500 × g for 60 min (RC5C, SA-600 rotor; Sorvall Instruments). Clarified lysate was applied to a gravity-flow column containing charged Ni-NTA resin and eluted with 300 mM imidazole. Eluted protein was further purified using a Superdex 200 prep grade gel-filtration column (GE Healthcare) on an AKTA FPLC (GE Healthcare) equilibrated with 10 mM Tris, pH 7.5, 100 mM NaCl, 5% glycerol (wt/vol). Gel filtration was done at 4 °C at a flow rate of 0.4 mL/min. Protein purity of collected fractions was checked by SDS/PAGE.

DRaCALA-Based Assessment of Protein–Ligand Interactions.

Binding of c-di-GMP and GTP to His₆-RmcA_cyt was determined by DRaCALA, as previously described (32). ³²P Radiolabeled GTP or c-di-GMP was added to 5 µM His₆-RmcA_cyt and 1× c-di-GMP calcium buffer (10 mM Tris, pH 8.0, 100 mM NaCl, 5 mM CaCl₂). For competition assays, 200 µM of unlabeled c-di-GMP, GTP, GDP, GMP, or pGpG (final concentration) was used in the binding reactions. Protein–ligand mixtures were spotted onto nitrocellulose paper using a 2-µL pin tool. The nitrocellulose was allowed to air-dry and then exposed on a phosphoimager screen and imaged using a Fuji FLA-7000 phosphoimager (GE). The area and intensity of spots were measured using Fujifilm Multi Gauge software (GE). The fraction bound was calculated as described previously (32).

Assays for in Vitro Phosphodiesterase and Diguanylate Cyclase Activity.

Kinetics of His₆-RmcA_cyt activity was determined using TLC to detect the accumulation of either c-di-GMP or pGpG for diguanylate cyclase and phosphodiesterase activity, respectively. For diguanylate cyclase activity, 5 µM of His₆-RmcA_cyt was added to 1× c-di-GMP magnesium buffer (10 mM Tris, pH 8.0, 100 mM NaCl, 5 mM MgCl₂), and ³²P GTP. For phosphodiesterase activity, 5 µM His₆-RmcA_cyt was added to 10× c-di-GMP buffer with MgCl₂, ³²P c-di-GMP, 1 µM cold c-di GMP, and water. Reactions were incubated at room temperature. At the indicated time-points, 3-µL aliquots were taken from the total reaction mixture and stopped by adding 1.5 µL of 0.2 M EDTA and heating samples for 10 min at 98 °C. Samples were spotted on TLC polyethyleneimine cellulose plates (EMD Millipore) and run using buffer containing (40%) saturated NH₄SO₄ and (60%) 1.5 M KH₂PO₄. Dried TLC plates were exposed on a phosphoimager screen and imaged using a Fuji FLA-7000 phosphoimager. Intensity of spots was measured using Fujifilm Multi Gauge software. Percent conversion was calculated by dividing the intensity of either the c-di-GMP or the pGpG spot by the total intensity of the entire sample.

Fluorescence-Monitored Binding of Phenazine Ligands to His₆-RmcA_cyt.

Fluorescence emission spectra of the His-tagged cytosolic portion of RmcA (His₆-RmcA_cyt) were recorded using a PTI QuantaMaster instrument fitted with a temperature-regulated sample chamber set at 25 °C. Protein samples in buffer (10 mM Tris⋅HCl, 100 mM NaCl, pH 7.5) were placed in a quartz cuvette (1 cm × 0.4 cm) and stirred continuously with a small magnetic stir bar situated in a masked portion of the cuvette. Intrinsic fluorescence of the protein was monitored by recording emission spectra (310- to 460-nm, slits at 5 nm) when the excitation wavelength was set at 295 nm (slits at 1 nM). In a typical experiment, successive 1-µL additions of concentrated ligand (pyocyanin, PCA, or PMS) were made to a 0.2 µM protein solution, and emission spectra were recorded following each addition. Under these conditions, photobleaching causes a 16% decrease in the emission signal over a time course that is fit well as a double exponential decay (k_fast 0.012 s⁻¹; k_slow 0.0046 s⁻¹). This decrease had to be taken into account in performing and analyzing the results of binding experiments. Emission signals were corrected for photobleaching using the results of “mock titrations” in which microliter additions of buffer were made to the diluted protein sample and emission scans were recorded using the same conditions and timing as used for the phenazine titrations. Analysis included corrections of emission signal (Em) at 330 nm for the effects of dilution and for inner filter effects using the equation: Em_corrected = Em_observed × (total volume/initial volume) × 10^{(A295 + A330)/2}), where A₂₉₅ and A₃₃₀ are the absorbance values for the solution at 295 nm and 330 nm (58). Corrected results were fit by DynaFit, assuming a simple, one-site binding model (59).