Universal antibiotic tolerance arising from antibiotic-triggered accumulation of redox metabolites

Pseudomonas aeruginosa is an opportunistic pathogen that often infects open wounds or patients with cystic fibrosis. Once established, P. aeruginosa infections are notoriously difficult to eradicate. This difficulty is in part due to the ability of P. aeruginosa to tolerate antibiotic treatment at the individual-cell level or through collective behaviors. Here we describe a new mechanism by which P. aeruginosa tolerates antibiotic treatment by modulating its global cellular metabolism. In particular, treatment of P. aeruginosa with sublethal concentrations of antibiotics covering all major classes promoted accumulation of the redox-sensitive phenazine - pyocyanin (PYO). PYO in turn conferred general tolerance against diverse antibiotics for both P. aeruginosa and other Gram-negative and Gram-positive bacteria. We show that PYO promotes energy generation to enhance the activity of efflux pumps, leading to enhanced antibiotic tolerance. This property is shared by other redox-active phenazines produced by P. aeruginosa. Our discovery sheds new insights into the physiological functions of phenazines and has implications for designing effective antibiotic treatment protocols. Author Summary Antibiotic tolerance can facilitate the evolution of resistance, and here we describe a previously unknown mechanism of collective antibiotic tolerance in Pseudomonas aeruginosa. In particular, P. aeruginosa treated with sublethal concentrations of antibiotics covering all major classes promotes accumulation of pyocyanin (PYO), an important virulence factor. In turn, PYO confers general tolerance against diverse antibiotics for both P. aeruginosa and other bacteria. Our discovery is a perfect example of what Nietzsche once said: That which does not kill me makes me stronger.


Introduction
The overuse and misuse of antibiotics have led to a global crisis (1): bacteria have developed resistance against every existing antibiotic and are doing so at an alarming rate, considering the timescale at which new antibiotics progress from development to clinical application (2,3). The drying antibiotic pipeline further heightens the global threat created by infectious bacteria (4,5).
A critical approach to the problem is developing ways to revitalize existing antibiotics (6). To extend the use of existing antibiotics, we need to develop a mechanistic understanding of the diverse ways by which bacteria survive antibiotics. Such an understanding is critical for designing therapeutic approaches that subvert the survival tactics by bacterial pathogens.
Past efforts on antibiotic resistance have focused on responses of individual cells, such as mutations in the antibiotic targets, enzymatic activity that inactivates antibiotics, and increased activation of efflux pumps (6). Unlike antibiotic resistance, which is due to inherited or acquired mutations (7,8), tolerance reflects the ability of individual cells (9) or cell populations (10) to survive antibiotic treatment without acquiring new mutations. However, it has been long realized that antibiotic tolerance precedes resistance (11)(12)(13). Recent in vitro experiments have shown that antibiotic tolerance can facilitate the evolution of resistance (14). In particular, antibiotic tolerance can pave the way for the rapid subsequent emergence of antibiotic resistance in bacteria; thus, preventing the evolution of tolerance may shed light on alternative strategies for antibiotic treatments. Importantly, the ability to survive antibiotic treatment is typically considered an intrinsic property of the single bacterial cells or bacterial populations, before antibiotics are applied. In contrast to these prevailing views, here we describe a previously unknown mechanism of collective antibiotic tolerance in Pseudomonas aeruginosa. 4 PYO, one of the most studied phenazines, is a redox-active metabolite giving the characteristic blue-green pigment of P. aeruginosa cultures (15). PYO is typically produced when P. aeruginosa enters the stationary phase when the cell density is high (16). Recent studies have demonstrated a number of physiological roles of PYO, such as serving as a signaling compound (17), facilitating biofilm development (18), promoting iron acquisition (19), and influencing colony formation (20). It is also known to confer a broad-spectrum antibiotic activity (21). Our work provides novel insights into the biological function of bacterial redox-active metabolites, and suggests a role of such metabolites, including PYO and other phenazines, in the survival of bacteria during acute stress.

Subinhibitory concentrations of antibiotics induce PYO accumulation
We observed that sub-inhibitory concentrations of kanamycin (Kan), a commonly used aminoglycoside antibiotic, induced a blue-greenish color change in P. aeruginosa PAO1 (ATCC 47085) cultures. PYO is a typical pigment for such color change (15). We confirmed this notion by purifying PYO using an established method (22), which also allowed us to measure PYO accumulation. The induction of PYO followed a biphasic dependence on the Kan dose. The blue-green color of the cultures became more visible with increasing Kan concentrations (Fig.   1A). This color change is consistent with direct quantification of PYO in the supernatants (Fig.   1B). Concurrently, the bacterial density (A600 nm) decreased with the Kan concentrations (Fig.   1B). This enhanced PYO accumulation related to the culture without antibiotic treatment was transient, as evident in the direct measurements of PYO purified from the supernatants of cultures treated with 20 μg/mL Kan (Fig. 1C). 5 We wondered whether the enhanced PYO accumulation could represent a general stress response to antibiotic treatment. To test this notion, we measured PAO1 strain responses to other antibiotics with different modes of action, including chloramphenicol that targets ribosome 50S subunit, norfloxacin (quinolone family) that inhibits DNA replication, polymyxin B (polypeptide family) that alters cell membrane permeability, and carbenicillin (β-lactams) that inhibits cell-wall synthesis. All tested antibiotics promoted accumulation of PYO in a similar biphasic manner (Supplementary, Fig. S1A). This response was not unique to PAO1. The PA14 strain, which is known to produce more PYO than PAO1 (17), exhibited qualitatively the same responses to these antibiotics (Supplementary, Fig. S1B). The biphasic PYO accumulation and its temporal dynamics reconcile the apparently contradictory conclusions on PYO accumulation under antibiotic stresses measured previously by single-point measurements (23)(24)(25).
Cyclic di-GMP (c-di-GMP), an intracellular second messenger, has been shown to confer tolerance to antibiotics (26,27). Moreover, c-di-GMP has been shown to promote PYO production (28). These studies suggest a potential role of c-di-GMP in antibiotic-mediated PYO accumulation. Consistent with this notion, we observed increased levels of c-di-GMP in PAO1 in response to several antibiotics (Fig. 1D). In the absence of antibiotics, the accumulation of c-di-GMP was approximately 2.5 fold higher in PA14 than in PAO1 during early stationary phase (Supplementary, Fig. S2A). This observation provides a potential explanation for the typically higher PYO accumulation in PA14 than in PAO1. Additionally, exogenous GTP -the precursor of c-di-GMP (27), leads to increased c-di-GMP over time and increased PYO 6 accumulation in a dose-dependent manner (Supplementary, Fig. S2B, S2C, and S2D). This result confirms the recent finding that PYO production is c-di-GMP dependent (29).

Accumulated PYO enhances antibiotic tolerance in bacteria
We wondered whether the enhanced accumulation of PYO could represent a survival mechanism for the population under acute antibiotic stress. Indeed, growth of P. aeruginosa PAO1 in the presence of Kan was enhanced by exogenously added PYO ( Fig. 2A). This PYO-mediated tolerance was effective against other aminoglycosides (gentamicin, streptomycin and tobramycin) and antibiotics of other classes (norfloxacin, chloramphenicol and carbenicillin) ( Fig. 2A). In the presence of PYO, PAO1 cultures exhibited a much shorter lag time before recovering in the presence of an antibiotic, in comparison to cultures without exogenously added PYO (Supplementary, Fig. S3). Endogenous PYO, if accumulated at a sufficiently high level such as in the cystic fibrosis respiratory tract (30), can also provide protection. In addition, PYO-mediated tolerance was maintained when PAO1 was cultured in other media or different oxygen conditions (Supplementary, Fig. S4). An exception to the general PYO-mediated tolerance was polymyxin B, where PYO enhanced its ability to inhibit bacterial growth ( Fig. 2A).

Oxidized PYO serves as electron acceptors to mediate antibiotic tolerance
As the PYO-mediated tolerance was general against different classes of antibiotics for diverse bacteria, it likely resulted from PYO-mediated modulation of cellular physiology that is common for all these bacteria. As a redox-active molecule (33), PYO can exist in reduced or oxidized state, depending on its chemical environment. The oxidized form of PYO can act as an electron acceptor to modulate the global energy metabolism in the cell by regulating the flux of electron shuttling (34). Similar functions have been proposed for other chemicals that can act as 7 electron acceptors. For instance, the electron-shuttling ability of oxidized tetrathionate (35) and nitrate (36) has been implicated in enhancing bacterial growth advantage by promoting respiration. Additionally, hydrogen sulfide has been shown to enhance bacterial survival during antibiotic stress (37). Consistent with this notion, several reductants, including nicotinamide adenine dinucleotide (NADH), β-mercaptoethanol, and N-acetyl-L-cysteine (NAC), all eliminated PYO-mediated antibiotic tolerance ( Fig. 3A and Supplementary, S6A). By themselves, these chemicals did not affect P. aeruginosa growth or response to antibiotics. The presence of these reductants in excess prevented significant accumulation of the oxidized PYO, thus interfering with its ability to act as an electron acceptor. As a side, our results might provide an explanation to previously observed unknown benefit of reductants as antibiotic-adjuvants in treating CF patients (38), and suggest new strategy to devise antibiotic adjuvants.

Oxidized PYO stimulates energy metabolism to pump antibiotics out
Studies have suggested a role of PYO in promoting ATP production (34), particularly when 8 cells enter stationary phase or are under stress (40). Consistent with this notion, we observed that exogenously added PYO promoted ATP accumulation during stationary phase ( Supplementary,   Fig. S7). An increase in ATP accumulation can facilitate generation of the proton-motive force (PMF) (41), which plays a critical role in promoting bacterial survival (but not growth) under anaerobic condition (34), in cell division (42) and increased efficiency of efflux pumps (EPs) (43,44). To test this notion, we measured the membrane potential with the carbocyanine dye DiOC 2 (3), which confirmed an elevated PMF in the presence of PYO (Fig. 3D). Additionally, the multidrug resistance phenotype in P. aeruginosa is in major part due to the expression of abundant EP systems (45,46). For example, the MexAB-OprM pump plays a key role in the intrinsic resistance for several types of antibiotics (44). An increasing efficiency of EPs, mediated by an elevated PMF, would provide an explanation for the general PYO-mediated tolerance to a wide variety of antibiotics. The PYO-mediated promotion of PMF also explains the lack of PYO-mediated tolerance against polymyxin B ( Fig. 2A and Supplementary, S3B): the loss of cell membrane integrity caused by polymyxin B completely eliminates the PMF. This is consistent with the previous observation that none of EPs extrude polymyxin B in P. aeruginosa (47), because EPs are anchored in membrane and most are driven by PMF (43)(44)(45). Aside from polymyxin B, an increased PMF may not always benefit the cells. In particular, as the uptake of aminoglycosides is energy dependent, an increased PMF can enhance their uptake and subsequent inhibition of bacterial growth (48) (i.e. by overcoming the increased efficiency of EPs). Thus, we reasoned that PYO-mediated tolerance would be eliminated at high levels of aminoglycosides. Indeed, 100 μg/mL Kan and 10 μg/mL gentamycin effectively suppressed P. aeruginosa growth, further demonstrating that PYO mediated tolerance was due to PMF.
Given this model (Fig. 3C), we expected that inhibition of the hydrolysis of ATP would abolish PYO-mediated antibiotic tolerance. Indeed, sodium azide, which inhibits the F 1 subunit of ATP synthase (49), abolished the tolerance (Fig.  3E). Similarly, We further tested modulation of EP efficiency by quantifying intracellular accumulation of ethidium bromide (EtBr), a well-established substrate of EPs in P. aeruginosa (44). Indeed, PYO accelerated the extrusion of EtBr by cells (Fig. 4A). If an enhanced efficiency of EPs was responsible for PYO-mediated antibiotic tolerance, we would expect chemical inhibitors of EPs to abolish this tolerance. We tested this notion by using phenylalanine-arginine β-naphthylamide (PAβN), which inhibits the resistance nodulation cell division (RND) pumps, the major contributors to antibiotic resistance in P. aeruginosa (50). Indeed, PAβN reduced PYO-mediated tolerance in a dose-dependent manner (Supplementary, Fig. S8D).
When the EPs are already expressed at a sufficiently high level, an elevated pump activity in the presence of PYO was primarily due to an enhanced energetic state by the cells, instead of up-regulation of pump genes. For example, various pump gene expression was not regulated (Supplementary, Fig. S9). P. aeruginosa features a transcriptional factor, SoxR, which is activated by phenazines (20). These results also indicate that the PYO-mediated protection is not due to its 10 regulation of SoxR (Supplementary, Fig. S10A, which can modulate expression of EPs (17).
The tolerance was not mediated by SoxR in P. aeruginosa, however, this result does not exclude its potential role in other bacteria such as E. coli. Moreover, PAO-JP2 strain has multiple quorum sensing (QS) genes knocked out (31), combining the assay with the addition of exogenous QS signals (Supplementary, Fig. S10B), and suggesting a dispensable role of QS in PYO-mediated tolerance. Altogether, we found that the PYO mediated tolerance was neither through working as a signal to regulate the hierarchical QS network (Supplementary, Fig. S6), nor through activating SoxR.

Discussion
The antibiotic-induced early PYO accumulation could result from reduced degradation or increased production of PYO. For example, antibiotics could inhibit enzymes involved in PYO degradation, though such enzymes have not yet been identified in P. aeruginosa (51). Alternatively, cells could increase PYO synthesis to counteract stress, akin to that experienced by cells entering the stationary phase. It has been well established that different antibiotics can trigger accumulation of reactive oxygen species (ROS) (52). Likewise, quorum sensing (QS) has been implicated in PYO metabolism (17). However, our experiments demonstrated that the antibiotic-mediated PYO accumulation was unlikely due to these mechanisms (Supplementary,   Fig. S11).
Collectively, our results reveal a previously unknown mechanism by which P. aeruginosa can and PYO conferred tolerance against all antibiotics with the exception of polymyxin B (Fig. 4B).
Moreover, the PYO-mediated protection is not limited to the producing cells; rather, PYO also enhances survival of other P. aeruginosa strains and other bacteria. This property highlights a particular challenge in treating infections involving P. aeruginosa: incomplete suppression of P.
aeruginosa by antibiotics would enhance growth of survivors, as well as other pathogens in the growth environment. Nevertheless, the PYO-mediated tolerance results from the cell's intrinsic ability to counteract a disruption in the cellular state, whereas PYO itself can be considered as a stress. As such, this tolerance is a double-edged sword. An elevated energy generation can enhance efficacy of certain antibiotics by promoting their uptake (48). Also, if pushed beyond cell's buffering capacity, the PYO-mediated disruption in the redox balance can inhibit growth (15,20). Thus, this tolerance mechanism also presents an opportunity for designing treatment strategies that can synergistically enhance effects of antibiotics. For instance, specific inhibitors can be developed to target the multiple enzymes involved for the synthesis of phenazines, antioxidants can be used as adjuvants to counteract the protective effect of PYO.

EXPERIMENTAL SECTION
Chemicals and strains Antibiotics used in this study include carbenicillin, chloramphenicol, gentamicin, kanamycin, norfloxacin, polymyxin B, streptomycin and tobramycin. Oxidized forms of pyocyanin (PYO) and other phenazines were prepared as previously described (1).
N,N'-dicyclohexylcarbodiimide (DCCD) and sodium azide were used to inhibit different subunits of ATP synthase. Ethidium bromide (EtBr) was used to test the function of efflux pumps, and phenylalanine-arginine β-naphthylamide (PAβN) was used to inhibit the efficiency of efflux pumps.
All chemicals were purchased from Sigma-Aldrich and Cayman Chemical.

P. aeruginosa, Escherichia coli, Salmonella typhimurium, Bacillus cereus, Bacillus subtilis and Staphylococcus
aureus strains were used in this study. More details of these strains were shown in Supporting Information Table S1.  h.

PYO accumulation in culture tubes
Assays of efflux pumps Firstly, PAβN as a specific pump inhibitor was used to investigate the efficiency of pumps in the presence of streptomycin. Furthermore, the growth curves of P.
aeruginosa PAO1 supplemented with 20 μg/mL streptomycin were measured, initially treated with 10 μmol/L CCCP and 1 mmol/L sodium azide, respectively. The growth curves were measured by a plate reader (Tecan) at the wavelength of 600 nm (A600). 50 μL mineral oil was added to prevent evaporation during long term culture.

Statistical analysis and software
The statistical significance of differences was determined by a Student's t-test. The fluorescent intensities and fold changes in mRNA levels were normalized prior to analysis. GraphPad Prism 5.0 was used for analyzing the data and generating figures.

ACKNOWLEDGMENTS
We thank D. K. Newman (Caltech) for sharing PA14 and its mutants (Δphz, ΔsoxR and   were presented for A to C (n = 3), for D (n = 6).   , Fig. S5). The tolerance was maintained for other Gram-negative (Escherichia coli, Salmonella Typhimurium) and Gram-positive (Bacillus cereus, Bacillus subtilis and Staphylococcus aureus) bacteria treated with antibiotics (Fig. 2B). Means ± s.d. were presented for A to C, and E (n = 6), for D (n = 9).