Transcriptional Regulation Contributes to Prioritized Detoxification of Hydrogen Peroxide over Nitric Oxide

Bacteria can be exposed to H₂O₂ and NO· concurrently within phagosomes. In such multistress situations, bacteria could have evolved to simultaneously degrade both toxic metabolites or preferentially detoxify one over the other. Here, we found that simultaneous exposure to H₂O₂ and NO· leads to prioritized detoxification, where detoxification of NO· is hampered until H₂O₂ has been eliminated. This phenomenon resembles CCR, where bacteria consume one substrate over others in carbon source mixtures. Further experimentation revealed a central role for transcriptional regulation in the prioritization of H₂O₂ over NO·, which is also important to CCR. This study suggests that regulatory scenarios observed in bacterial consumption of growth-promoting compound mixtures can be conserved in bacterial detoxification of toxic metabolite mixtures.

ABSTRACT

Hydrogen peroxide (H₂O₂) and nitric oxide (NO·) are toxic metabolites that immune cells use to attack pathogens. These antimicrobials can be present at the same time in phagosomes, and it remains unclear how bacteria deal with these insults when simultaneously present. Here, using Escherichia coli, we observed that simultaneous exposure to H₂O₂ and NO· leads to prioritized detoxification, where enzymatic removal of NO· is impeded until H₂O₂ has been eliminated. This phenomenon is reminiscent of carbon catabolite repression (CCR), where preferred carbon sources are catabolized prior to less desirable substrates; however, H₂O₂ and NO· are toxic, growth-inhibitory compounds rather than growth-promoting nutrients. To understand how NO· detoxification is delayed by H₂O₂ whereas H₂O₂ detoxification proceeds unimpeded, we confirmed that the effect depended on Hmp, which is the main NO· detoxification enzyme, and used an approach that integrated computational modeling and experimentation to delineate and test potential mechanisms. Plausible interactions included H₂O₂-dependent inhibition of hmp transcription and translation, direct inhibition of Hmp catalysis, and competition for reducing equivalents between Hmp and H₂O₂-degrading enzymes. Experiments illustrated that Hmp catalysis and NAD(P)H supply were not impaired by H₂O₂, whereas hmp transcription and translation were diminished. A dependence of this phenomenon on transcriptional regulation parallels CCR, and we found it to involve the transcriptional repressor NsrR. Collectively, these data suggest that bacterial regulation of growth inhibitor detoxification has similarities to the regulation of growth substrate consumption, which could have ramifications for infectious disease, bioremediation, and biocatalysis from inhibitor-containing feedstocks.

IMPORTANCE Bacteria can be exposed to H₂O₂ and NO· concurrently within phagosomes. In such multistress situations, bacteria could have evolved to simultaneously degrade both toxic metabolites or preferentially detoxify one over the other. Here, we found that simultaneous exposure to H₂O₂ and NO· leads to prioritized detoxification, where detoxification of NO· is hampered until H₂O₂ has been eliminated. This phenomenon resembles CCR, where bacteria consume one substrate over others in carbon source mixtures. Further experimentation revealed a central role for transcriptional regulation in the prioritization of H₂O₂ over NO·, which is also important to CCR. This study suggests that regulatory scenarios observed in bacterial consumption of growth-promoting compound mixtures can be conserved in bacterial detoxification of toxic metabolite mixtures.

INTRODUCTION

Phagocytic cells of the innate immune response deploy a number of weapons to neutralize invading pathogens, including reactive oxygen and nitrogen species (ROS and RNS) (1–3). Of these, nitric oxide (NO·) and hydrogen peroxide (H₂O₂) are of particular interest due to their rapid ability to diffuse across cell membranes (4, 5). Accordingly, microbes have evolved detoxification and repair systems to withstand these stresses, the importance of which has been highlighted by the numerous pathogens that suffer attenuated virulence when one or more of their NO· or H₂O₂ detoxification systems are impaired (6–13). The physiological interactions of concurrent NO· and H₂O₂ stresses, which can be encountered within phagosomes (3, 14–16), are not well understood. In vitro studies have suggested that NO· can both reduce and enhance killing by high concentrations of H₂O₂ depending on the organism and treatment conditions (17–20). For example, Pacelli and colleagues observed that simultaneous exposure to 1 mM H₂O₂ and 1 mM diethylamine (DEA) NONOate, a NO· donor, led to increased killing of different strains of Escherichia coli (18). Similarly, Yadav and coworkers found that millimolar concentrations of H₂O₂ and NO· donors enhanced killing of E. coli as well as Actinobacillus actinomycetemcomitans, Actinomyces viscosus, Streptococcus mutans, and Streptococcus sobrinus (17). On the other hand, Gusarov and Nudler found that 30 μM boluses of NO· delivered right before 10 mM boluses of H₂O₂ protected Bacillus subtilis from cell death through reactivation of a catalase and inhibition of thioredoxin and thioredoxin reductase, which limited Fenton chemistry (19). Additionally, H₂O₂ has been shown to inhibit expression of the E. coli NO· detoxification enzyme flavorubredoxin under anaerobic conditions (21). With few exceptions, the concentrations used in previous studies were beyond those found physiologically, which are in the micromolar range (22, 23), and reaction fluxes through the major detoxification systems (for H₂O₂, alkyl hydroperoxidase [Ahp] and catalases [KatE, KatG] [4, 24]; for NO·, flavorubredoxin [NorV], flavohemoglobin [Hmp], and periplasmic formate-dependent nitrite reductase [NrfA] [25–27]) were not quantified. Since the physiological impacts of NO· and H₂O₂ exposure are concentration dependent (22, 28), it is important to understand the functioning of and interactions between these defense networks at phagosomal concentrations.

The H₂O₂ and NO· biochemical reaction networks of E. coli are complex, and dynamic models have proven useful in quantifying the distributions of these toxic metabolites and exploring system behaviors (29–35). These previously developed models included detoxification by antioxidants and enzymes, transcriptional regulation and inactivation of enzymes, damage and repair of DNA and Fe-S clusters, and destruction of amino acids by the hydroxyl radical, ·OH, and they were compartmentalized (intracellular, media, and gaseous) to account for the cell-dependent and cell-independent reactions. The models have correctly predicted major genetic (29) and environmental (34) perturbations and been used to dissect network behavior, such as NO· detoxification under microaerobic conditions (32) and impaired NO· dioxygenase activity in a ΔclpP mutant (30). To date, these models have been used to analyze the complex networks of single-agent stresses, and given the considerably more complex nature of multistress conditions, multistress models have the potential to be even more enlightening.

Here, we examined the response of E. coli to concurrent exposure to physiologically relevant concentrations of both NO· and H₂O₂. We observed that H₂O₂ detoxification was not affected by the presence of NO·, whereas NO· clearance was delayed in an H₂O₂ concentration-dependent manner. Interestingly, computational analyses revealed that metabolic detoxification of these stressors is prioritized, with H₂O₂ preceding NO·. Carbon catabolite repression (CCR), which produces prioritized consumption of nutrients, has been widely observed (36), and the data presented here suggest that processing of toxic metabolites has notable parallels. Possible mechanisms behind this phenomenon were explored, and we found that transcriptional regulation, which also plays a significant role in CCR (36), was a major driver of this multistress physiology.

RESULTS

NO· and H₂O₂ detoxification under simultaneous nitrosative and oxidative stress.

In this work, we sought to explore interactions within the NO· and H₂O₂ biochemical networks using concentrations that resemble those found within phagosomes (22, 23). Exponentially growing E. coli cells were washed and used to inoculate a bioreactor to an optical density at 600 nm (OD₆₀₀) of 0.025 before being treated with the NO· donor DPTA NONOate ((Z)-1-[N-(3-aminopropyl)-N-(3-ammoniopropyl)amino]diazen-1-ium-1,2-diolate; here referred to as DPTA; 50 μM; resulting in a peak NO· concentration of ∼1 to 2 μM) and/or various bolus concentrations of H₂O₂ (10 to 30 μM). NO· had no effect on detoxification of H₂O₂ (Fig. 1A); however, we observed that H₂O₂ delayed NO· detoxification in a dose-dependent manner (Fig. 1B). There was no effect on NO· or H₂O₂ levels in cell-free controls due to the addition of the other chemical (see Fig. S1 in the supplemental material). Also, we note that no loss in viability (Fig. S2) was observed during treatment, and cultures resumed growth after clearance of both H₂O₂ and NO· (Fig. S3).

FIG 1.

H₂O₂ and NO· clearance. Exponentially growing wild-type cells were inoculated into a bioreactor containing 0, 10, 20, or 30 μM H₂O₂ in M9 medium with 10 mM glucose. The culture was then immediately treated with 50 μM DPTA or an equivalent amount of the DPTA solvent. H₂O₂ concentrations (A) were quantified every 5 min, whereas NO· (B) was monitored continuously. (C) Cumulative consumption of NO· by cells was quantified by modeling with a black-box cellular compartment. Error bars show the standard errors of the means from three biological replicates.

Further analysis of the cotreatment results compared to those for cell-free controls revealed that E. coli exhibits biphasic detoxification of NO· (Fig. 1C). In the first phase, NO· is consumed at a lower rate, ∼1.5 nmol of NO· per min, whereas the rate in the second phase is >3-fold higher. Interestingly, the initial detoxification rate deviates little between cultures, and cultures do not enter the second NO· detoxification phase until after they had detoxified their respective boluses of H₂O₂. These data suggest that detoxification of H₂O₂ is prioritized over that of NO· when both are present in E. coli cultures. Such a phenomenon is reminiscent of specific nutrient consumption patterns exhibited by E. coli, where simultaneous exposure to two carbon sources leads to prioritized consumption of substrates (36). For example, when presented with glucose and fructose in the media, E. coli will prioritize glucose consumption (37). Media with glucose and gluconate provide an example of consumption in parallel, where glucose and gluconate are consumed simultaneously (38).

Possible causes of prioritized detoxification of H₂O₂ over NO· center on Hmp.

Parallels exist between the detoxifications of H₂O₂ and NO· by E. coli, such as that both are governed by enzymes, the expression of those enzymes is generally induced by their toxic substrates, and both depend to some extent on reducing power (39). Previous studies with either H₂O₂ or NO· (but not both) in the same growth phase, comparable media, and concentrations similar to those used here identified KatG and AhpCF as the main determinants of H₂O₂ detoxification and Hmp as the major NO· detoxification enzyme (30, 34). Distributions calculated with a model that includes both H₂O₂ and NO· stress networks, which was constructed in this investigation, also indicate that KatG, AhpCF, and Hmp are the major detoxification systems (Fig. 2 and Fig. S4) (see Materials and Methods).

FIG 2.

Simulation results under various H₂O₂ and NO· stress conditions. Response of the wild type to individual stress was simulated using the ensemble of models that have an ER of <10 after MCMC. Predicted H₂O₂ fluxes through Ahp, KatE, and KatG were calculated when 10 μM H₂O₂ (A), 20 μM H₂O₂ (B), or 30 μM H₂O₂ (C) was delivered. (D) Predicted NO· flux through Hmp and NorV after a bolus treatment of 50 μM DPTA was also quantified.

Since all three of those defense systems are induced by their respective stresses, we assayed whether induced responses contributed to the detoxification pattern observed. We found that inducible responses to H₂O₂ played a minor role in detoxification under cotreatment, whereas they were critical for NO· detoxification (Fig. 3). We expect that this feature largely resulted from the trace levels of Hmp that are present under normal growth conditions compared to the measurable abundances of H₂O₂-degrading enzymes (4, 40).

FIG 3.

Importance of stress-induced response to NO· and H₂O₂ detoxification. Before treatment with 50 μM DPTA and 30 μM H₂O₂, 100 μg/ml of CAM or ethanol (−CAM), its solvent, was added to the growing culture to prevent further translation. H₂O₂ (A) and NO· (B) consumption assays were then performed to assess the importance of protein synthesis during simultaneous NO· and H₂O₂ stress. The dots in panel A and dashed lines in panel B represent the means from three biological replicates, and the error bars in panel A and shadows around the lines in panel B represent the standard errors of the means.

Both AhpCF and Hmp use NADH; however, H₂O₂ detoxification depends less on the reductant because KatG exhibits catalase activity, which does not require NADH, and AhpCF has a lower apparent K_M for NADH (approximately 1 μM compared to 5 μM for Hmp) (29, 34, 41, 42). These data and knowledge suggested that H₂O₂ detoxification would be more robust than NO· detoxification under the conditions studied here and that Hmp availability or reducing equivalent supply underlies the detoxification pattern observed. In addition, it is possible that H₂O₂ or a cellular component that is modulated by it directly inhibits Hmp. Further, it is worth noting that the NO· dioxygenase activity of Hmp requires O₂ and that the O₂ level stayed far higher than the apparent K_M of Hmp for O₂ (Fig. S5, 190 μM ≫ 100 μM) (41).

The scenarios delineated above center on perturbations to Hmp activity, which led us to hypothesize that Hmp was the target for H₂O₂-dependent inhibition of NO· detoxification. We confirmed that hypothesis to be true with a Δhmp mutant (Fig. 4A). Further, complementation of the Δhmp mutant with hmp expressed from its native promoter in a low-copy-number plasmid restored the H₂O₂-dependent delay of NO· detoxification, whereas an empty plasmid control could not (Fig. 4B and C and Fig. S6).

FIG 4.

Hmp is required for H₂O₂ to delay NO· clearance. (A) Without Hmp (Δhmp), NO· detoxification was equivalently slow with or without H₂O₂. (B and C) When hmp was complemented by expressing it from its native promoter (hmpp) on the pUA66 plasmid, delay of NO· detoxification by H₂O₂ was restored. (B) Due to copy number or leakiness effects from the plasmid-borne hmp, cells could more readily consume NO· when 50 μM DPTA was delivered, resulting in a comparatively low [NO·] peak and fast NO· clearance. (C) When the concentration of DPTA was increased 2.5-fold to 125 μM, the NO· concentration was closer to what was observed in wild-type cultures (Fig. 1) and the inhibition of NO· detoxification by H₂O₂ became more evident. An empty plasmid control has been provided in Fig. S6.

To assess whether altered Hmp availability, catalytic activity, and/or NADH supply were feasible mechanisms for how NO· detoxification was stalled by H₂O₂, we performed simulations with the combined H₂O₂-NO· stress model (Fig. S7). Notably, the model did not predict that H₂O₂ would impair NO· detoxification, which suggested that one or more missing interactions between the stress networks was not in the current knowledge base (Fig. S8A). Decreasing k_max for hmp transcription or the first-order kinetic constant for translation by 35-fold was sufficient to delay NO· detoxification to levels observed in the cultures cotreated with 30 μM H₂O₂. In addition, impairing Hmp catalysis by decreasing the rate of the catalytic step for NO₃⁻ generation or initial NAD(P)H levels by 150- and 4,000-fold, respectively, was also able to recapitulate NO· dynamics in the presence of 30 μM H₂O₂. Notably, the lower availability of reducing equivalents simulated in Fig. S7D led to slower detoxification of H₂O₂ (Fig. S8B), which was not observed experimentally, and a decrease of 4,000-fold seemed unreasonable given previous measurements of NAD(P)H pools (43, 44). Collectively, these simulations revealed that while impaired transcription, translation, and catalysis might underlie the delay in NO· detoxification that we observed, it was unlikely that insufficient NAD(P)H levels contributed to the phenomenon.

Inhibition of Hmp catalysis does not underlie delayed NO· detoxification, and NAD(P)H levels are confirmed to be sufficient.

We sought to experimentally assess contributions of the scenarios delineated in the previous section to delayed NO· detoxification in the presence of 30 μM H₂O₂. One experiment we devised had the capacity to assess whether inhibition of Hmp catalysis participated in delaying NO· detoxification, as well as to assess the computational prediction that NAD(P)H supply was nonlimiting (Fig. 5A). In this experiment, a Φ(Hmp-GFP) translational fusion would be preexpressed prior to treatment, and a translational inhibitor would be included during treatment to prevent further protein production. Introduction of cells into an environment that already contained NO· with and without H₂O₂ would produce identical NO· clearance profiles if Hmp catalysis were unimpeded and NAD(P)H supply were sufficient. However, if either one contributed, the NO clearance profiles would differ significantly (Fig. 5A). When this experiment was performed, clearance of NO· by preexpressed Φ(Hmp-GFP) was indistinguishable in the presence and absence of 30 μM H₂O₂ (Fig. 5B), and initial Φ(Hmp-GFP) concentrations were confirmed to be equivalent in the NO·-only and NO·-with-H₂O₂ experiments. Uninduced controls showed that cultures could not clear NO· in the presence or absence of H₂O₂ without the induced Φ(Hmp-GFP) (Fig. S9), and controls with only green fluorescent protein (GFP) showed that the presence of the fluorophore does not impact NO· clearance (Fig. S10). These data eliminated inhibition of Hmp catalytic activity as a potential mechanism and confirmed the model prediction that NAD(P)H supply is sufficient for NO· detoxification in the presence of H₂O₂.

FIG 5.

Catalytic activity of Hmp in presence and absence of H₂O₂. A translational fusion of Φ(Hmp-GFP) was expressed from pUA66 T5p-Φ(hmp-gfp_sf) in exponentially growing Δhmp cells. After 15 min, translation was inhibited with 100 μg/ml CAM, and cells were washed and used to inoculate a bioreactor containing 50 μM DPTA and 100 μg/ml CAM. DPTA was added to the bioreactor 10 min prior to inoculation, allowing a stable NO· concentration to be reached before addition of Φ(Hmp-GFP)-containing cells. (Inset) Comparable levels of Φ(Hmp-GFP) under both treatment conditions were confirmed based on fluorescence measurements. Simulated effects of decreased Hmp catalytic activity and reduced availability of NAD(P)H are shown in panel A, and experimental observations are shown in panel B. NO· measurements and fluorescence readings show the averages from three biological replicates, with error bars representing the standard errors of the means.

H₂O₂ reduces the NO·-induced protein and transcript output from the hmp promoter.

With Hmp present before H₂O₂ and NO· treatment, H₂O₂ no longer impaired NO· detoxification. To assess whether insufficient Hmp availability underscored the phenomenon, we quantified transcript and protein output from the hmp promoter with a previously constructed reporter plasmid (pUA66 hmpp-gfp_mut2 [30]). To maintain consistent NO· dynamics between conditions, experiments were performed in a Δhmp background. GFP fluorescence levels were significantly lower in the presence of H₂O₂ (Fig. 6A). In fact, no detectable GFP was produced in the first 15 min. However, after H₂O₂ had cleared from the system (∼25 min), GFP production initiated (Fig. S11) and resembled the sample without H₂O₂ with a time delay. The lower GFP levels in cells experiencing simultaneous NO· and H₂O₂ treatment is at least partially explained by a significant defect in transcript abundance (Fig. 6B). These data suggest that the delay in NO· detoxification observed arises from difficulties associated with Hmp synthesis under concurrent NO· and H₂O₂ stress.

FIG 6.

H₂O₂ reduces protein and transcript output from the hmp promoter. Exponentially growing Δhmp cells transformed with an hmp transcriptional reporter (pUA66 hmpp-gfp_mut2) were inoculated into a bioreactor containing 0 or 30 μM H₂O₂ in M9 medium with 10 mM glucose and then immediately treated with 50 μM DPTA or an equivalent amount of the DPTA solvent. A Δhmp background was used to maintain identical NO· environments in the two samples. GFP production (A) was significantly inhibited, which is at least partially explained by a reduction in gfp_mut2 transcript levels (B). Here, the relative [mRNA] represents the [mRNA] normalized by the maximum concentration observed in the experiment, which happens at t = 15 min for NO· (blue). Samples collected at t = 0 were fixed (A) or treated with RNAprotect (B) before exposure to stress, as described in Materials and Methods.

NO· reduces the H₂O₂-induced protein output from the katG promoter.

To complement results from the hmp promoter, we performed similar experiments with a katG transcriptional reporter. We observed that H₂O₂ by itself induced protein output from the katG promoter, which is consistent with previous reports (34, 45), whereas NO· by itself failed to induce protein production (Fig. 7). Interestingly, simultaneous stress eliminated the ability of H₂O₂ to induce GFP production from the katG promoter under the conditions tested here. These data from the hmp and katG transcriptional reporters suggest that either a shared regulator of these genes is perturbed by simultaneous exposure or a more global effect is present. Also, it is worth noting that these results complement those shown in Fig. 3 and further illustrate that H₂O₂ detoxification is more robust than NO· detoxification because it can function almost equivalently with or without its induced response under the conditions considered here.

FIG 7.

Expression from the katG promoter is also inhibited by simultaneous NO· and H₂O₂ stress. Exponentially growing Δhmp cells transformed with pUA66 katGp-gfp_mut2 were inoculated into a bioreactor containing 0 or 30 μM H₂O₂ in M9 medium with 10 mM glucose and then immediately treated with 50 μM DPTA NONOate or an equivalent amount of the DPTA solvent. Measurable GFP (P < 0.05) is observed in the H₂O₂-only case within 15 min, but expression is not significant in the presence of NO· after 15 or 30 min.

Involvement of known regulators of hmp and katG expression.

Given the defective output from both the hmpp and katGp transcriptional reporters, we performed experiments to assess the involvement of known regulators of hmp or katG expression. FNR and Fur have been identified as regulators of both hmp and katG, whereas NsrR, RpoS, SdsN, and MetR have been implicated in the regulation of hmp and OxyR in the regulation of katG (39, 46–55). In addition, we assessed the involvement of DksA, the stringent response regulator, which has recently emerged as a regulator of oxidative and nitrosative stresses in Salmonella enterica (56–59) and that we have identified as important for expression of Hmp in E. coli (60).

The roles of FNR, Fur, RpoS, SdsN, and OxyR were addressed by examining NO· detoxification with and without H₂O₂ in Δfnr, Δfur, ΔrpoS, ΔsdsN, and ΔoxyR mutants (Fig. S12). There was not a notable difference in detoxification dynamics between wild-type and Δfnr, ΔrpoS, and ΔoxyR strains. A minor difference in NO· dynamics was observed in the ΔsdsN mutant; however, it was transient and NO· clearance times with and without H₂O₂ were indistinguishable from those of the wild type (Fig. S12A). A slight delay in NO· clearance in the presence of H₂O₂ was observed in the case of the Δfur mutant (Fig. S12E), but it was accompanied by a delay in H₂O₂ clearance (Fig. S12K). The delay in H₂O₂ clearance is not surprising, given that Fur is known to be a transcriptional activator of katG expression (53).

NsrR is the main transcriptional regulator responsible for repression of hmp expression in the absence of NO·; thus, when nsrR is deleted, Hmp is no longer at trace levels under normal growth (47). Indeed, when the ΔnsrR mutant was stressed with NO· in the presence and absence of 30 μM H₂O₂, simultaneous detoxification was observed and the NO· clearance profiles with and without H₂O₂ were indistinguishable (Fig. 8A). When nsrR was complemented by a plasmid-borne copy expressed from its native promoter, prioritized detoxification of H₂O₂ over NO· was restored, which was not observed with an empty vector control (Fig. 8B and Fig. S13). These data demonstrate that the prioritized detoxification observed depends on both NsrR and Hmp (Fig. 4).

FIG 8.

NsrR is required for compromised NO· detoxification in the presence of H₂O₂. (A) Exponentially growing ΔnsrR cells were transferred to a bioreactor that consisted of M9 medium with 10 mM glucose with or without 30 μM H₂O₂ and immediately treated with 50 μM DPTA. The △nsrR mutant was able to detoxify NO· in the presence or absence of H₂O₂. (B) When the △nsrR mutant was complemented by expressing nsrR under the control of its native promoter (nsrRp) on a pUA66 plasmid, the ability of H₂O₂ to inhibit NO· detoxification was restored.

As described above, delayed NO· detoxification in the presence of H₂O₂ is associated with low Hmp levels at the onset of stress and insufficient induction during simultaneous stress. The ΔnsrR mutant increases the abundance of Hmp prior to stress, but data shown in Fig. 8 do not assess whether reduced hmp expression during simultaneous stress was present in the ΔnsrR mutant. To assess this aspect of transcriptional regulation in the ΔnsrR mutant, we assessed hmpp transcript output via a gfp mRNA transcript expressed from hmpp. These experiments were performed in the Δhmp ΔnsrR mutant to ensure identical NO· environments, and we observed that the transcript level of gfp in NO· with H₂O₂ was significantly less than that in the NO·-only sample after 15 min (Fig. 9). Treatment with rifampin at the onset of the assay led to negligible transcript levels under both conditions, suggesting that the transcripts measured in the absence of rifampin were synthesized during NO·-only and cotreatment exposures. These data indicate that simultaneous H₂O₂ and NO· stress suppresses the transcript output of hmpp, even in the absence of NsrR. Collectively, these data suggest that NsrR is required for the observed prioritization of detoxification due to its impact on basal Hmp levels, whereas it is not required to observe suppressed expression of hmp under simultaneous stress compared to NO·-only stress.

FIG 9.

NsrR is not required for the suppression of hmp expression under simultaneous stress. Exponentially growing Δhmp ΔnsrR cells transformed with an hmp transcriptional reporter (pUA66 hmpp-gfp_mut2) were inoculated into a bioreactor containing 0 or 30 μM H₂O₂ in M9 medium with 10 mM glucose and then immediately treated with 50 μM DPTA NONOate. In the Δhmp ΔnsrR mutant, gfp transcripts are constitutively expressed, so transcript levels after 15 min were compared to those at time zero (prior to treatment). gfp transcript levels drop after 15 min in the presence of NO·. They drop significantly more so (P < 0.05) in the H₂O₂-containing sample, indicating that the transcriptional defect is present even in the absence of NsrR. Rifampin-treated samples (rif⁺) were included to confirm that transcripts would degrade significantly over 15 min if expression was inhibited. Error bars show the standard errors of the means for five biological replicates in rif⁻ samples and three biological replicates in rif⁺ samples.

MetR is also known to regulate hmp expression (46), although unlike the other regulators its knockout mutant cannot grow in minimal medium (61, 62), which was the medium of this study. Therefore, to test the involvement of MetR, we complemented the Δhmp mutant with hmp expressed from its native promoter and a truncated version that was missing the MetR regulatory site (Fig. S14). Removal of the MetR binding site impaired NO· detoxification, which is consistent with previous reports identifying MetR as an activator of hmp expression (46); however, H₂O₂ delayed NO· detoxification in the presence and absence of the MetR binding site. Further, we constructed hmp reporter plasmids both with and without the MetR binding site, and GFP production in the presence of NO· with and without H₂O₂ was consistent with results from the NO· detoxification assays. The initial GFP level was lower in the absence of the MetR binding site, the rate at which GFP accumulated without the MetR site was lower than that in the construct with the site, and H₂O₂ inhibited the rate at which GFP was synthesized from both reporter constructs (Fig. S15).

Recently, we found DksA to be a critical regulator of Hmp expression (60). A dksA deletion mutant is an auxotroph for multiple amino acids; therefore, to test its involvement in the phenomenon studied here, we had to assess whether H₂O₂ delayed NO· detoxification in media that allowed ΔdksA growth (63, 64). In addition, we had to confirm that the addition of those amino acids did not alter H₂O₂ and NO· concentrations in a cell-free system. As depicted in Fig. S16, the addition of those amino acids did not alter H₂O₂ or NO· concentrations appreciably in a cell-free system, and H₂O₂ still impaired NO· detoxification in wild-type cultures (Fig. S12F). When treated with NO· only, the ΔdksA mutant exhibited a significant delay in NO· detoxification compared to that of the wild type (Fig. S12F). Cotreatment with H₂O₂ delayed NO· detoxification in the ΔdksA mutant even further (Fig. S12F). For the ΔdksA mutant, the presence of NO· inhibited H₂O₂ detoxification, unlike H₂O₂-only treatments (Fig. S17), which was not an effect seen with the wild type (Fig. 1). When ΔdksA mutant culturability was measured, H₂O₂ by itself did not have much of an impact, NO·-only treatments produced an ∼3-fold loss in culturability, and the cotreatment decreased culturability by ∼1,000-fold (Fig. S18A). Due to this excessive cell death, which was not observed for the wild type (Fig. S18B), we were unable to exclude or implicate DksA in the ability of H₂O₂ to delay NO· detoxification.

DISCUSSION

For infecting pathogens, the phagosomal environment is a deadly nutrient-depriving compartment of ROS, RNS, acidic pH, antimicrobial peptides, and proteases (1–3). While many studies have examined the mechanisms of these antimicrobial defenses, most have focused on isolated stresses. Under conditions with a single insult, resources can be focused toward relieving the stress imposed; however, in multistress environments, resource allocation can be more complicated. Bacteria could have evolved to process the stresses simultaneously or prioritize defense against one or more of the insults over the others. How bacterial resources are distributed to cope with two or more phagosomal stresses remains largely an open question. In general, multistress environments are not rare for microbes; organisms used for bioremediation often encounter environments with solvents, heavy metals, oxidants, and other toxic species (65), whereas biocatalysts are exposed to furans, weak acids, and phenolic compounds in lignocellulosic hydrolysates (66, 67). We speculate that bacterial regulatory structures that have evolved to weather specific multistress environments are conserved under conditions with different stressors.

Here, we explored the response of E. coli to simultaneous exposure to H₂O₂, an ROS, and NO·, an RNS. We observed that H₂O₂ detoxification was prioritized over that of NO· and used a computational model and experimentation to enumerate and test different mechanistic possibilities. Hmp was the main enzyme used by E. coli to detoxify NO· under the conditions used here, and H₂O₂ had no effect on its catalytic activity or the requirement of the enzyme for reducing equivalents (Fig. 5B). Rather, H₂O₂ was shown to delay NO·-induced Hmp synthesis with significant repression of transcript and protein output from the hmp promoter (Fig. 6). H₂O₂ detoxification was more robust than that of NO·, which was due to the nonnegligible basal levels of defense enzymes that could cope with micromolar concentrations of H₂O₂ (4). When basal levels of Hmp were increased by synthetic preexpression or knockout of the transcriptional repressor NsrR, NO· and H₂O₂ detoxification proceeded simultaneously (Fig. 5B and 8A). These data suggest that transcriptional regulation plays a pivotal role in the prioritized detoxification of H₂O₂ over NO·.

CCR is used by bacteria to prioritize consumption of nutrients. In E. coli and many other bacteria, glucose is the preferred substrate and is consumed first when multiple carbon sources are present (36). Transcriptional regulation is a major driver of such prioritized catabolism (36). In E. coli, the major transcription factor is cyclic AMP (cAMP) receptor protein (CRP), which activates the expression of machinery to consume less preferred carbon sources when cAMP accumulates in the absence of glucose and other preferred substrates (36). In Bacillus subtilis, a different regulatory scheme, namely, transcriptional repression by CcpA in the presence of glucose, is used to achieve CCR (36). Prioritized consumption and the central role of transcriptional regulation is also observed for other growth substrates, such as with exogenous terminal electron acceptors from which O₂ is typically respired first (68). The widespread use of CCR by bacteria suggests that it has been a favorable evolutionary solution for organisms presented with multiple growth-promoting compounds. Data presented here suggest that regulation of the detoxification of growth-inhibitory compounds has similar features. However, it is worth noting that there are significant differences between the prioritized detoxification described here and CCR. For example, inducer exclusion, which is where the presence of glucose interferes with the import of other carbon sources, contributes to CCR (36), whereas it cannot exist for H₂O₂ and NO· detoxification because they passively diffuse across membranes (4, 5).

We rationalize that there are several reasons as to how regulation had evolved to yield prioritized detoxification of H₂O₂ over that of NO. For one, E. coli is exposed to H₂O₂ from its own metabolism during normal aerobic growth (4). Inadvertent loss of electrons to O₂ will produce superoxide, O₂⁻, which in turn will be dismutated to H₂O₂. Strains devoid of all catalases and hydroperoxidases, ΔkatE ΔkatG ΔahpCF mutants, have been shown to accumulate inhibitory levels of H₂O₂ in culture (4, 69). Such a scenario would explain the measureable hydroperoxidase activity that E. coli cultures exhibit in the absence of exogenous H₂O₂. On the other hand, NO· is a less common stress that E. coli encounters as a product of immune cells or metabolic intermediate released from organisms conducting denitrification. It would seem beneficial for resource conservation to not express NO· detoxification machinery at high basal levels if the stressor is not encountered frequently, and that becomes even more advantageous considering that Hmp generates ROS when NO· is not present (70, 71). We reason that E. coli regulation in this case has evolved to readily deal with oxidative stress from aerobic metabolism while limiting the use of Hmp to conditions where its benefits outweigh the costs. The consequences of such regulation would be prioritized detoxification of H₂O₂ over NO· during cotreatment if induced responses were difficult to mount in the multistress environment, which appears to be the case here (Fig. 6 and 7). In all, several factors that center on the lifestyle and physiology of E. coli could explain how its regulation has evolved to detoxify these metabolites.

Questions still remain regarding the molecular mechanism of the interaction described here. Specifically, data from the katG transcriptional reporter (Fig. 7) suggested that the effect was not confined to the hmp promoter. Further, transcript measurements in the ΔnsrR mutant showed that lower levels are found under dual stress conditions whether or not the main regulator of hmp expression is present (Fig. 9). It is possible that the H₂O₂ and NO· response networks interact more globally, for instance, as a result of their impacts on cellular metabolism (72, 73) and/or through a major transcription or translation regulator. Investigating those avenues of the phenomenon will be the focus of future studies.

MATERIALS AND METHODS

Chemicals.

The NO· donor DPTA NONOate {(Z)-1-[N-(3-aminopropyl)-N-(3-ammoniopropyl)amino]diazen-1-ium-1,2-diolate} (Cayman Chemical), which we refer to as DPTA and which releases 2 molecules of NO· per parent compound, was dissolved in 10 mM NaOH to a working concentration of 72 mM DPTA and stored on ice prior to addition to the bioreactor. H₂O₂ (35 wt% solution in water) was purchased from Fisher Scientific. M9 minimal salts (5×) were purchased from BD Difco and prepared according to the manufacturer’s instructions or were made from components. Potassium phosphate buffer (PBS) was sterile filtered with a 0.22-μm filter, and LB was sterilized by autoclaving. Bovine liver catalase at 2,000 to 5,000 U/mg protein was purchased from Sigma-Aldrich, dissolved in 50 mM potassium phosphate buffer, pH 7.0, and passed through a 0.22-μm filter before addition to media.

Bacterial strains.

All assays were performed with E. coli MG1655. For assays with mutants, mutations were moved into MG1655 via P1 phage transduction and assessed for proper chromosomal integration and lack of gene duplication using the primers listed in Table S1 in the supplemental material. When necessary, mutants were cured using pCP20 and confirmed for proper scar sequence size using the primers described in Table S1.

NO· and H₂O₂ measurements.

NO· was measured continuously (20 reads/s) using an ISO-NOP 2-mm electrode (World Precision Instruments, Inc.). The sensor was calibrated daily according to the manufacturer’s instructions in order to convert the raw output signal (pA) to NO· concentration (μM). For calibrations performed with S-nitroso-N-acetyl-d,l-penicillamine (SNAP) (Cayman Chemical), a conversion factor of 0.457 NO· per SNAP molecule was used (60). H₂O₂ from the sterile filtered samples was measured using the Amplex red hydrogen peroxide/peroxidase kit (Life Technologies) according to the manufacturer’s instructions, following dilution to below 10 μM H₂O₂. A standard curve from 0 to 10 μM H₂O₂ was used to convert raw fluorescence values to H₂O₂ concentration.

NO· and H₂O₂ consumption assays.

E. coli from a −80°C stock were grown aerobically in 1 ml LB plus 30 to 75 U/ml catalase for 4 h at 37°C with shaking at 250 rpm, and 10 μl was used to inoculate 1 ml M9 minimal medium with 10 mM glucose plus 30 to 75 U/ml catalase. This overnight culture was grown for 16 h at 37°C with shaking at 250 rpm and was used to inoculate a 250-ml baffled flask containing 20 ml M9 10 mM glucose plus 30 to 75 U/ml catalase to an OD₆₀₀ of 0.01. OD₆₀₀ measurements were performed on a Synergy H1 hybrid microplate reader (BioTek Instruments). Catalase was added in these steps to prevent H₂O₂ accumulation in strains lacking major detoxifying enzymes and added to wild-type cultures to maintain consistency across strains. The catalase concentration was determined based on the amount required for maximal growth rate in M9 minimal medium with 10 mM glucose for a mutant lacking all major H₂O₂ detoxification enzymes (ΔkatE ΔkatG ΔahpCF) and has been used in previous studies under similar conditions (34, 74). To support growth of the ΔdksA mutant, which is auxotrophic, M9 glucose medium used for growth and stress assays was supplemented with 400 μg/ml serine and 40 μg/ml each of aspartate, glutamate, histidine, isoleucine, leucine, phenylalanine, threonine, and valine (63, 64, 75). M9 glucose medium supplemented with these amino acids is referred to as M9 minAA.

When the culture reached an OD₆₀₀ of ∼0.2, 12 ml was removed to a prewarmed 15-ml Falcon tube and centrifuged at 37°C and 4,000 rpm for 10 min. A 10.8-ml volume of spent medium was removed, the cell pellet was resuspended, and 1 ml was transferred to a warm 1.5-ml microcentrifuge tube. Cells were washed a total of four times to remove all catalase. For each wash, the sample was centrifuged at 14,000 rpm for 2 min, 980 μl of medium was removed, and the cell pellet was resuspended with 980 μl fresh warm media. Washed cells were used to inoculate a bioreactor to an OD₆₀₀ of 0.025. The bioreactor consisted of 10 ml medium in a 50-ml Falcon tube with a sterile magnetic stirring bar (0.5 in.) and immersed in a 37°C stirred water bath (∼400-rpm stirring), and it was used in previous studies (29–33).

Where applicable, H₂O₂ was added just prior to inoculation with cells by adding 10 μl of the appropriate stock to reach the desired final concentration (e.g., 10 μl of 10 mM H₂O₂ to reach 10 μM H₂O₂). Time 0 H₂O₂ concentration measurements were taken before the addition of cells to allow for an initial measurement. Immediately after inoculation, 6.95 μl of 72 mM DPTA in 10 mM NaOH was added to reach the desired concentration. For assays without NO·, 6.95 μl of 10 mM NaOH was added. At each time point, 150 μl was sampled and sterile filtered (0.22-μm syringe filter) for H₂O₂ measurement. For samples containing only NO·, 150 μl was discarded at each time point to maintain consistent volumes across samples.

CFU assays.

Cultures were grown and washed identically to the method for the NO· and H₂O₂ consumption assays and used to inoculate a bioreactor. For CFU assays, DPTA and/or H₂O₂ were added following inoculation, so a time 0− (untreated) point could be sampled prior to stress. Time 0− samples, as well as samples from different times following treatment, were washed and plated using a serial dilution method. Washing consisted of removing 200 μl to a 1.5-ml microcentrifuge tube followed by centrifuging for 3 min at 15,000 rpm. The cell pellet was washed by discarding 180 μl supernatant and resuspending the pellet with 980 μl PBS. The sample was centrifuged again, and 980 μl supernatant was discarded. The pellet was resuspended a final time with 80 μl PBS, serially diluted in PBS, and plated on LB-agar plates containing 25 mM pyruvate to scavenge any residual H₂O₂. Plates were incubated for 16 h at 37°C before colonies were counted.

GFP fluorescence measurements.

Green fluorescence was quantified on a per-cell basis using an LSR II flow cytometer (BD Biosciences, San Jose, CA) using 488-nm excitation and a 525/20 bandpass filter. Data were acquired using FACSDiva software (BD Biosciences, San Jose, CA). Prior to measurement, cells were fixed using paraformaldehyde (PFA). The fixation protocol consisted of removing 200 μl of cell culture to a 1.5-ml microcentrifuge tube, centrifuging at 15,000 rpm for 3 min, and replacing 180 μl of supernatant with 180 μl of 4% PFA in PBS. After 25 min at room temperature, cells were centrifuged again, 180 μl of supernatant was removed, and 480 μl of PBS was used to resuspend the cell pellet. Fixed cells were stored at 4°C until analysis by flow cytometry. For data presentation and statistical analyses, the median fluorescence values of the entire population were used.

Φ(Hmp-GFP) preexpression assay.

A low-copy-number plasmid bearing an isopropyl-β-d-thiogalactopyranoside (IPTG)-inducible translationally fused Φ(Hmp-GFP), pUA66 T5p-Φ(hmp-gfp_sf) lacI^q (pWCKA1) (30), was transformed into a cured Δhmp mutant to eliminate NO·-induced hmp expression. Strains were grown identically to the procedures of the NO· and H₂O₂ consumption assays, except that 30 μg/ml kanamycin (KAN) was included in all steps for plasmid retention. When the OD₆₀₀ reached 0.2, the culture was induced with 20 μM IPTG. After 15 min of induction, 100 μg/ml of chloramphenicol (CAM) was added to prevent further translation. Washing was performed identically to the method described for the NO· and H₂O₂ consumption assays, except that 30 μg/ml KAN and 100 μg/ml CAM were included in all wash steps. Meanwhile, the bioreactor containing 10 ml of M9 medium with 10 mM glucose was dosed with 50 μM DPTA, such that the cells were washed and inoculated to the bioreactor 10 min after dosing. The washed cells used to inoculate the bioreactor were fixed and used to quantify Φ(Hmp-GFP) levels via fluorescence.

GFP reporter assays and gfp transcript measurement.

Transcriptional reporter plasmids (pUA66 hmpp-gfp_mut2 [pWCKA05] [30], pUA66 hmpp_trunc-gfp_mut2 [pWCKA06; this work], and pUA66 katGp-gfp_mut2 [pWCKA07] [76]) were transformed into cured Δhmp cells. Cells lacking hmp were used to obtain identical NO· dynamics throughout the assay. Cultures were grown and washed by a method identical to the method for the NO· and H₂O₂ consumption assays, except for the inclusion of 30 μg/ml KAN for plasmid retention, and used to inoculate a bioreactor containing 0 or 30 μM H₂O₂. Samples were then quickly treated with 50 μM DPTA or 6.95 μM NaOH (solvent that was added with DPTA). GFP levels were quantified as described above, with the time zero point being fixed from the washed cells prior to inoculation into the bioreactor. Samples were treated with RNAprotect bacterial reagent as described below in the case of the accompanying transcript measurements for the hmp reporter assay, with time zero being sampled from the 0 μM H₂O₂ bioreactor prior to dosing with DPTA and an equivalent volume being discarded from the 30 μM H₂O₂ bioreactor for consistency. For both GFP and transcript level measurements, the time zero point is therefore identical across samples in a replicate, since identical cells were used to inoculate multiple bioreactors.

qPCR.

Real-time quantitative PCR (qPCR) was performed as described previously (30). Briefly, 1 ml of RNAprotect bacterial reagent (Qiagen) was used to stabilize 500 μl of culture according to the manufacturer’s instructions. Stabilized cell pellets were stored at −80°C until all samples could be purified simultaneously. RNA was purified using the RNeasy minikit (Qiagen) according to the manufacturer’s instructions with on-column DNase digestion using RNase-free DNase I (Qiagen). To account for variations in RNA loss during processing, 50 μg of an external mRNA standard was spiked into each of the samples prior to purification. Preparation of the external standard was described previously (30, 77, 78). Briefly, phzM from Pseudomonas aeruginosa PAO1 was synthetically transcribed off linearized pET-11a-phzM using the T7 high-yield synthesis kit (NEB) and purified using the RNeasy minikit (Qiagen) with on-column DNase digestion.

Purified mRNA was converted to cDNA using TaqMan reverse transcription reagents (ThermoFisher Scientific) according to the manufacturer’s instructions, using a 10-μl reaction volume (3.85 μl purified RNA) and random hexamers. Controls lacking reverse transcriptase were included to determine if there was any DNA contamination. After conversion to cDNA, samples were stored at −20°C until the real-time PCR could be performed. Real-time PCR was performed using the Power SYBR green PCR master mix (Applied Biosystems) in an optical 96-well plate, using 20-μl reaction volumes that included 10 μl Power SYBR green PCR master mix, 1 μl cDNA, and 250 μM (each) forward and reverse primers (Table S2), balanced with RNase-free water. Each sample was prepared with phzM and gfp primers in separate wells to quantify individual phzM and gfp concentrations. Standard curves were included that consisted of serial dilutions of pET-11a-phzM (with phzM primers) or the plasmid encoding gfp (with gfp primers) to allow for absolute quantification of mRNA levels. The plate was run for 40 PCR cycles in an ABI 7900 HT real-time thermocycler, and threshold cycle (C_T) values were obtained from amplification curves using SDS 2.3 software. All samples were normalized by phzM concentration.

Model construction and parameter optimization.

Preexisting NO· (30) and H₂O₂ (34) models were combined, with a few minor changes: the superoxide dismutase (SOD) reaction, as well as reactions of tryptophan and cysteine with hydroxyl radical described in reference 30, were replaced with those from reference 34. The periplasmic SOD reaction (30) was removed because the model developed here does not include the periplasmic compartment, and it was replaced with the two cytoplasmic SOD enzymes (34). Reaction of tryptophan and cysteine with ·OH was changed from generating dead-end metabolites (30) to be consistent with the formalism in reference 34, which modeled the reaction as destruction of amino acids. Additionally, KatG and Ahp production from reference 34 was split into transcription (H₂O₂ concentration-dependent Hill function with n = 1), translation, and mRNA degradation, as described in reference 30. For all transcription reactions, a basal transcription rate was added, as described in reference 32. The concentrations of all amino acids were held constant because the model did not include amino acid biosynthesis to replenish the amino acid pools. Even without any decrease in amino acid concentrations, this pathway did not represent a significant sink. Ferric and ferrous iron concentrations were held constant because their reactions did not represent major sinks for NO· or H₂O₂.

All uncertain parameters were trained on data by minimizing the variance-weighted sum of squared residuals (SSR) between model simulations and data, as described previously (32, 34). Parameter sets were compared using evidence ratios (ER), which represent model likelihoods compared to the best model identified, as described previously (32, 34). Models with an ER of >10 were discarded, which is a commonly used cutoff (32, 34, 79). Medium-specific model parameters related to NO· (NO· loss to the gas phase, k_La_NO·; NO· autoxidation, k_NO·-O2; rate of NO· dissociation from DPTA, k_NONOate,DPTA) were trained on cell-free data of NO· for 2 h following addition of 50 μM DPTA. The spontaneous degradation rate of H₂O₂ was trained on H₂O₂ measurements for 30 min (10-min time points) following addition of 10, 20, and 30 μM H₂O₂ to cell-free M9 medium with 10 mM glucose. The gas phase transfer rate of O₂ between media and the atmosphere (k_La_O2) was trained on experimental measurements of [O₂] within medium that was deficient in O₂ due to prior pumping of N₂ after being exposed to the atmosphere. Uncertain cellular parameters (29–32, 34) were trained on clearance of 50 μM DPTA by wild-type cells (NO· data) and clearance of 10, 20, and 30 μM H₂O₂ by wild-type, ΔkatE, and ΔkatG strains (H₂O₂ data) (Fig. S4). Previous work suggested that clearance by single ΔkatE and ΔkatG catalase mutants is required while fitting uncertain cellular parameters to generate models with consistent detoxification distributions among the major H₂O₂ detox pathways (34).

Uncertain cell-free parameters were categorized into groups based on the chemical species (e.g., O₂, H₂O₂, and NO·) they affected. Each group of parameters was optimized from 100 random initial values. ER of the optimized values were calculated, and those with an ER of <10 were used as initial points for 10,000 iterations of a Markov chain Monte Carlo (MCMC) process to further explore the parameter space (80). Of note, after 100 optimizations, all the optimized values of parameters responsible for O₂ and H₂O₂ cell-free kinetics had an ER close to 1. Therefore, 5 sets of optimized parameter values were chosen randomly from the O₂ and H₂O₂ groups, and the sole optimized parameter set for the NO· group with an ER of <10 was used as initial conditions for the subsequent MCMC optimization. After the completion of the MCMC algorithm, ER of the accepted values was calculated, and those with an ER of <10 were retained.

Uncertain cellular parameters were optimized from 50 random initial parameter sets to account for the nonlinear nature of the optimization. During each of these optimizations, the values of cell-free parameters were randomly assigned from the ensembles of possible parameter values obtained from MCMC. Among the 50 optimized parameter sets, 4 had an ER of less than 10, and these models were individually used as the starting point for 10,000 iterations of MCMC exploration. After the MCMC process, 2,208 accepted parameter sets had an ER of <10 from the total of 40,000 iterations. All of these models accurately captured cellular responses to the NO·- and H₂O₂-only stress conditions (Fig. S4) and predicted similar flux predictions through the different NO· and H₂O₂ detoxification systems (Fig. 2). Simulations of NO· and H₂O₂ cotreatment did not agree with experimental data (Fig. S8), which suggested that interactions were missing from the combined model and knowledge base that it was built from. To evaluate the possible effects of H₂O₂ on different facets of the NO· detoxification pathway, namely, hmp transcription and translation, Hmp catalysis, and reducing agent availability, parameter values governing these cellular processes were manually changed in the best-fit model such that the resulting model was able to approximate NO· data under simultaneous NO· and H₂O₂ treatment (Fig. S7).

Black-box modeling.

Using the best model obtained from the previous section, simulation of cell-free NO· release, resulting from addition of 50 μM DPTA to fresh media, was performed by setting the cell fraction parameter to 0. Between each time point of the experimental data, cell-free simulation was performed to calculate the amount of NO· released by DPTA while accounting for the loss of NO· through gas-phase transfer and autoxidation. The difference between the experimental data and simulation results was treated as black-box consumption by the cells. This procedure was carried out for every time point of the experimental data to calculate the cumulative consumption of NO· by cells. This was done because the combined model was missing interactions that were needed to predict the impact of H₂O₂ on NO· detoxification. The resulting cumulative NO· consumption curves (Fig. 1C) depict 2 distinct regimes of NO· consumption under cotreatment conditions. The rate of NO· clearance in each of these regimes was quantified as the slope of the fitted line through the linear portion of data (Fig. S19). The number of data points used for linear fitting was determined by maximizing the number of points used while ensuring that the addition of data points did not significantly increase the SSR.

Statistical analysis.

Statistical significance was assessed using two-tailed t tests with unequal variances. Statistically different values were defined as having a P value less than 0.05.