CO2/HCO3 - Accelerates Iron Reduction through Phenolic Compounds.

In an oxygenic environment, poorly soluble Fe3+ must be reduced to meet the cellular Fe2+ demand. This study demonstrates that elevated CO2/HCO3− levels accelerate chemical Fe3+ reduction through phenolic compounds, thus increasing intracellular Fe2+ availability. A number of biological environments are characterized by the presence of phenolic compounds and elevated HCO3− levels and include soil habitats and the human body. Fe2+ availability is of particular interest in the latter, as it controls the infectiousness of pathogens. Since the effect postulated here is abiotic, it generally affects the Fe2+ distribution in nature.

I ron is a vital mineral for almost all living organisms and participates in inevitable electron transfer reactions since it serves a redox potential from Ϫ500 to ϩ300 mV by switching between two oxidation states (Fe 2ϩ /Fe 3ϩ ) (1). Despite its great abundance in the Earth's crust (representing the fourth most abundant element), its availability for organisms is limited, because it is mainly present as poorly soluble Fe 3ϩ (10 Ϫ18 M at pH 7.0) in an oxidative environment (2). Sophisticated strategies to increase its accessibility have evolved, including the secretion of siderophores and smaller iron chelators that enhance the solubility of Fe 3ϩ (3). Iron complexes are then taken up via designated transport systems in an energy-dependent fashion (2). Since reduced Fe 2ϩ is incorporated as a prosthetic group in a number of enzymes that belong to respiratory complexes or that participate in tricarboxylic acid (TCA) cycle reactions and stress response inside the cell, the reduction of Fe 3ϩ is of central relevance. Several studies on the characterization of ferric reductase activity are reviewed by Schröder et al. (4). It turned out that most designated ferric reductases are in fact flavin reductases that regenerate the substrate for chemical iron reduction. This is not limited to intracellular reactions, since it was shown recently that the human pathogen Listeria monocytogenes possesses an extracellular electron transfer apparatus, in which iron reduction is mediated by flavin (5).
In the Gram-positive soil bacterium Corynebacterium glutamicum, which is used in amino acid production on an industrial scale, little is known about iron uptake mechanisms because most annotated genes lack experimental validation (6,7). During the development of a suitable minimal medium for C. glutamicum, the addition of citrate (8) or small amounts of the diphenolic compounds catechol or protocatechuic acid (PCA) (3,4-dihydroxybenzoic acid) revealed a growth-promoting effect, which was attributed to the improved iron uptake of Fe 3ϩ chelates (9). To ensure sufficient initial iron uptake, PCA became a component of the widely used CgXII medium (10).
Excessive amounts of intracellular iron are deposited in specialized storage proteins (ferritin [Ftn] and Dps) and can be remobilized from there in times of iron starvation (2). A novel mode of iron remobilization was identified recently in C. glutamicum. Pupylation of a surface lysine residue of the Ftn and Dps proteins initiates the unfolding and subsequent Fe 3ϩ release in the cytosol. However, the mechanism responsible for reduction of Fe 3ϩ remains unclear (11).
Diphenolic substances (e.g., 2,3-dihydroxybenzoic acid [2, and derivatives) are secreted under iron-restricted growth conditions by a variety of organisms such as Bacillus subtilis and Paracoccus (Micrococcus) denitrificans (3,12,13). It was suggested that 2,3-DHB and derivatives are involved in iron uptake, as mutants deficient in the biosynthesis of 2,3-DHB and derivatives imported less iron, and adding these compounds enhanced iron transport (14). PCA is a structural relative of the common catecholate-type siderophore precursor 2,3-DHB and represents the basic compound of the rather unusual siderophore petrobactin (15). Bacillus anthracis and Bacillus cereus secrete great amounts of PCA in response to iron limitation (16)(17)(18). As a diphenolic Fe 3ϩ chelator (19), it was expected to serve a function analogous to 2,3-DHB enhancing iron uptake, although evidence for this had not been provided (16).
Besides chelating iron, diphenols have the potential to reduce Fe 3ϩ (20). The redox reaction in Fe 3ϩ -PCA chelates (19,21)-and generally speaking in Fe 3ϩmonocatecholate complexes-proceeds via the sequential oxidation of two adjacent hydroxyl groups to the respective semiquinone and quinone, providing two electrons per molecule for the reduction of Fe 3ϩ (20). The reaction depends on an acidic pH, as iron and PCA bind in an 1:1 stoichiometry at a pH of Ͻ4.5 (19), and the redox reaction is inhibited by the presence of (at least) the biscatecholate complex (20,21).
The phenotypical response of bacteria to different CO 2 /HCO 3 Ϫ levels has been extensively reviewed with a biotechnological, pathogenic, or environmental focus (22)(23)(24)(25). On the one hand, the effect on cell viability might be detrimental, considering that CO 2 -based sterilization operating at high pressure is used in food processing. However, on the other hand, a class of bacteria called capnophiles demonstrates that HCO 3 Ϫ can be a vital substrate as they are unable to grow unless sufficiently high CO 2 levels are established (26). Carboxylation reactions might generally benefit from the increased availability of CO 2 /HCO 3 Ϫ . C. glutamicum is especially well equipped with enzymes catalyzing these reactions, e.g., at the phosphoenolpyruvate (PEP)-pyruvateoxaloacetate node (27). As a consequence, higher product (e.g., succinate) and biomass yields per substrate were reported in C. glutamicum at elevated CO 2 levels and could be attributed particularly to an increased flux via the anaplerotic reactions (28)(29)(30). Anaerobic growth on glucose and tryptone could be enhanced by increasing concentrations of CO 2 . The authors suggest that this might enhance acetyl coenzyme A (acetyl-CoA) carboxylation reactions, yielding greater levels of fatty and mycolic acids (31). Although the exponential growth rate of C. glutamicum could not be promoted under aerobic conditions, elevated CO 2 in the inlet air of a bioreactor cultivation provoked the transcriptional response of almost the entire DtxR regulon (30). The master regulator of iron homeostasis in C. glutamicum was originally named after its functional homologue in the pathogenic C. diphtheriae, where it controls the toxin production in response to the iron availability (32). In fact, to date, many pathogens that induce the expression of toxin genes under iron starvation conditions and typically integrate this signal via the transcriptional regulators Fur, DtxR and homologues thereof are known (33)(34)(35). In analogy with the iron limitation, elevated CO 2 /HCO 3 Ϫ levels represent another suitable indicator for the host environment (22). The expression of toxin-encoding genes in Vibrio cholerae or Pseudomonas aeruginosa is mediated by the transcriptional regulators ToxT and RegA in response to elevated HCO 3 Ϫ concentrations (23). Interestingly, regA expression in P. aeruginosa is in turn repressed in an iron-dependent manner by a Fur homologue (33). Hence, understanding the interaction between CO 2 /HCO 3 Ϫ and iron availability is particularly interesting with regard to toxin production by pathogenic bacteria (23).

RESULTS
Elevated levels of CO 2 /HCO 3 ؊ increase intracellular Fe 2؉ availability. To monitor intracellular Fe 2ϩ availability, we constructed the fluorescence-based reporter strain C. glutamicum FEM3 which responds to the activation state of DtxR (see Fig. S1 in the supplemental material). Therefore, we chromosomally integrated the lacI gene under the control of the ripA promoter (P ripA ), which is controlled by DtxR. High Fe 2ϩ concentrations induce DtxR binding to P ripA , thus provoking repression of lacI and ultimately resulting in egfp expression under the control of the strong tac promoter (P tac ). Thus, the genetic circuit of C. glutamicum FEM3 constitutes a signal amplifier and a converter of the native repression mechanism.
Under all conditions tested, C. glutamicum FEM3 and the wild type (WT) showed identical growth. Cultivating C. glutamicum FEM3 under iron starvation (1 M) resulted in a marginal increase of the fluorescence per biomass [78 Ϯ 4 arbitrary units (a.u.) · (g CDW liter Ϫ1 ) Ϫ1 where g CDW is the cell weight (dry weight) in grams] over the autofluorescence of the WT [64 Ϯ 3 a.u. · (g CDW liter Ϫ1 ) Ϫ1 ] after 24 h. At iron excess (100 M), C. glutamicum FEM3 yielded a fluorescence per biomass which was about fourfold higher [313 Ϯ 20 a.u. · (g CDW liter Ϫ1 ) Ϫ1 ] at the end of the cultivation compared to conditions under iron starvation (Fig. S2).
Having proven the functionality at different iron concentrations, we cultivated C. glutamicum FEM3 aerobically in a parallel bioreactor system to evaluate the intracellular Fe 2ϩ availability in response to altering CO 2 /HCO 3 Ϫ levels. With 20% CO 2 in the inlet air, C. glutamicum FEM3 grew exponentially (growth rate [] ϭ 0.36 Ϯ 0.02 h Ϫ1 ) until the carbon source glucose was exhausted (Fig. 1A). In contrast, aeration with an ambient CO 2 concentration (0.04%) provoked biphasic growth of C. glutamicum FEM3 with a growth rate of 0.26 Ϯ 0.05 h Ϫ1 and 0.32 Ϯ 0.02 h Ϫ1 in growth phase one and two, respectively (Fig. 1A). Biomass-specific fluorescence started at similar levels in both . C. glutamicum FEM3 was cultivated aerobically in minimal medium with 20 g glucose liter Ϫ1 with synthetic air containing 20% CO 2 (21% O 2 , 59% N 2 ) or ambient air (0.04% CO 2 ). Data points (A) and bars (B) represent mean values, and error bars indicate standard deviations for three biological independent replicates. CO 2 /HCO 3 Ϫ Accelerates Chemical Iron Reduction ® conditions but after 24 h on average reached almost twofold-higher values at 20% CO 2 content, whereas it remained at a low level at the ambient CO 2 concentration (Fig. 1B). Biphasic growth of C. glutamicum FEM3 and the WT was also observed on minimal medium containing glucose in shaking flasks at an ambient CO 2 concentration ( Fig. 2A). In accordance with the bioreactor cultivations, growth was stimulated by adding 30 mM NaHCO 3 , which shortened the initial (nonexponential) growth phase from about 7 h under reference conditions to 4 h and led to a growth rate of 0.40 Ϯ 0.03 h Ϫ1 during the second growth phase. The addition of 195 M of the iron chelator PCA restored exponential growth throughout the cultivation ( Fig. 2A) and resulted in a slightly higher growth rate compared to the rate for the second growth phase of the reference condition (reference, ϭ 0.37 Ϯ 0.01 h Ϫ1 ; PCA, ϭ 0.43 Ϯ 0.01 h Ϫ1 ).
The biomass-specific fluorescence of the shaking flask approach showed a 0.5-and 3.0-fold increase in response to the supplementation of NaHCO 3 or PCA compared to the reference cultivation after 24 h, respectively (Fig. 2B). Although growth of C. glutamicum FEM3 was not further improved by the combined addition of NaHCO 3 and PCA ( Fig. 2A), there might be a synergistic effect on fluorescence level (4.2-fold compared to the reference; Fig. 2B), which was, however, not significantly higher than in the PCA-supplemented condition (by Student's t test). Supplementing with PCA or NaHCO 3 did not show a growth-stimulating effect in iron-depleted medium (Fig. S3), where biomass stagnated at a final 0.9 to 1.1 g liter Ϫ1 regardless of supplementation. In summary, these results demonstrate that in the presence of the extracellular iron source (FeSO 4 ), elevated levels of CO 2 /HCO 3 Ϫ and/or the addition of PCA to the minimal medium results in an increased intracellular Fe 2ϩ availability in C. glutamicum. CO 2 /HCO 3 ؊ does not increase thermal stability of the DtxR protein or interact with iron storage and mobilization. The DtxR protein of C. glutamicum shares 72% sequence identity with its extensively characterized homologue from Corynebacterium diphtheriae, including conserved ligand binding sites. Dimerization of the latter and subsequent DNA recognition are induced by sequential binding of two divalent metal ions per monomer. Besides the physiological effector of DtxR, Fe 2ϩ , in vitro activation was demonstrated with Ni 2ϩ , Mn 2ϩ , and Co 2ϩ at various affinities (36). As they are stable in an oxidative environment, they are commonly found in crystallographic structures of DtxR. An anion binding site is located in close proximity to the low-affinity binding site and essential for coordination of the metal ion (37)(38)(39). To address the question whether binding of HCO 3 Ϫ to the anion binding site of DtxR could cause the higher activation, we performed differential scanning fluorimetry. The thermal stability of purified DtxR protein was enhanced by increasing concentrations of divalent metal ions (Ni 2ϩ and Mn 2ϩ ) and dissociation constants (K D ) were calculated to 5.3 Ϯ 1.1 M To investigate whether elevated CO 2 /HCO 3 Ϫ levels have an impact on iron storage or mobilization, we created deletion mutations (Δftn, Δdps, Δftn Δdps, and Δpup) in C. glutamicum FEM3 and the WT. The deletion mutants lack either one (Δftn, Δdps) or both iron storage proteins (Δftn Δdps) or are unable to initiate remobilization of the stored iron by pupylation (Δpup). With the exception of C. glutamicum Δpup, none of the deletion mutants showed a difference regarding growth and fluorescence in any of the conditions tested (data not shown). C. glutamicum Δpup revealed an even stronger growth defect than the WT under reference conditions. However, supplementation of PCA or NaHCO 3 restored the respective growth phenotype with the result that C. glutamicum Δpup and the WT grew identically again (Fig. S4). CO 2 /HCO 3 ؊ accelerates Fe 3؉ reduction through phenolic compounds. Since we could not attribute the improved Fe 3ϩ reduction capacity in the presence of elevated CO 2 /HCO 3 Ϫ levels to an interaction with stored intracellular iron, we focused on the chemical reduction by PCA and HCO 3 Ϫ using the Fe 2ϩ -specific iron chelator bathophenanthroline disulfonic acid (BPS) for detection. After 5 h of incubation, the amount of Fe 2ϩ -BPS complexes was about sevenfold higher in the presence of 19.5 M PCA compared to a sample without additives (blank control; Fig. 3A). Although the addition of 50 mM NaHCO 3 alone increased the final absorbance only marginally, it promoted the iron reduction in the presence of PCA significantly. Differences caused by the presence of NaHCO 3 were most remarkable during the initial reduction phase with a Ͼ50% greater amount of Fe 2ϩ -BPS complexes up to 100 min after the start of the assay (Fig. 3A). The sample pH was maintained at 7.4 throughout the incubation in 200 mM 3-(N-morpholino)propanesulfonic acid (MOPS) buffer, thus proving that a shift in pH was not causing the increase in absorbance. We also replaced NaHCO 3 with KHCO 3 , which led to the identical improvement in iron reduction (Fig. 3A). An equimolar NaCl concentration did not trigger iron reduction (data not shown).
We monitored the degradation of PCA by liquid chromatography-quadrupole time of flight mass spectrometry (LC-MS-QTOF) analysis and found only about half of the residual amount of PCA when samples were incubated for 5 h in the presence of HCO 3 Ϫ compared to samples incubated without HCO 3 Ϫ (Fig. 3B). Furthermore, we identified oxidation products with the chemical formula C 7 H 6 O 6 and C 7 H 8 O 6 .
Concluding that PCA serves as electron donor in the reduction of Fe 3ϩ , we tested other aromatic compounds at identical concentrations that could be associated with iron reduction.
Catechol and 2-amino-3-hydroxybenzoic acid (3-hydroxyanthranilic acid [3-HAA]) carrying two adjacent hydroxyl groups or a mix of amino and hydroxyl groups, respectively, exhibited reduction capacities that were at least as high as PCA. Iron reduction by 4-amino-2-hydroxybenzoic acid (para-aminosalicylic acid), where the amino and hydroxyl groups are not positioned adjacent to each other, was only intermediate, whereas benzoic acid lacking hydroxyl and amino groups did not reduce iron at all (Fig. 4). The addition of 50 mM NaHCO 3 alongside with either of the above compounds (except for benzoic acid) accelerated the iron reduction in the same way as observed with PCA.
To further elucidate the role of HCO 3 Ϫ in the reduction of iron, we analyzed the complex formation between Fe 3ϩ and the iron chelator PCA that is required prior to reduction. In accordance with iron reduction, the initial rate of Fe 3ϩ -PCA complex formation (0 to 35 min) was increased by 46% through the addition of 50 mM NaHCO 3 (5.2 Ϯ 1.0 ΔmA 560 min Ϫ1 [change in milli absorbance units at 560 nm per minute] versus 7.6 Ϯ 1.3 ΔmA 560 min Ϫ1 ; Fig. 5A). In addition to the altered kinetics, the maximum absorbance of those complexes shifted to a shorter wavelength ( max ). The assay pH was again not affected by the addition of NaHCO 3 , and intentional pH perturbation with 50 mM KOH or HCl failed to reproduce the max shift (Fig. 5B).

DISCUSSION
Recently, Blombach et al. (30) initially disclosed a link between a high CO 2 proportion in the inlet air and a transcriptional response of almost the complete DtxR regulon in C. glutamicum. In this study, we demonstrate that elevated CO 2 /HCO 3 Ϫ levels increased the intracellular Fe 2ϩ availability and thus stimulated growth of C. glutamicum. The higher Fe 2ϩ concentration was not coupled to biological activity. Instead, we  identified an abiotic interaction of CO 2 /HCO 3 Ϫ with chemical iron reduction through phenolic compounds.
Growth retardation had not been reported previously, when C. glutamicum cultures were aerated with pressurized air and has been introduced in this study presumably by the insertion of an additional minimal medium preculture, since all other process parameters were kept the same (30). The growth defect could be transferred to a shaking flask approach and was apparently caused by a limitation of intracellular iron availability, as suggested (i) by the lower fluorescence of FEM3, (ii) by the stronger growth retardation of C. glutamicum Δpup in the reference culture, which was found to exhibit a growth retardation only in iron starvation (11) and (iii) by the fact that PCA supplementation restored exponential growth (9). Analogously, an extended lag phase of C. glutamicum in minimal medium had been observed when the main culture was inoculated from a preculture in minimal medium or when the inoculum was washed excessively and, consequently, the addition of iron chelators like catechol, PCA, or citrate to minimal medium was suggested (8,9). Interestingly, this requirement for iron chelators became apparent only when carbonate was omitted from the medium (8,9,40). However, the role of CO 2 with regard to the iron availability was not addressed further until noting a transcriptional response of the DtxR regulon (30).
The cometabolization of glucose and PCA can improve the growth rate of C. glutamicum to about 0.61 h Ϫ1 . However, this requires essentially higher PCA concentrations and lower cell densities (40). At standard PCA concentrations (as applied in this study), growth was not found to be increased over the maximum of ϭ 0.42 h Ϫ1 when glucose was utilized as the sole carbon and energy source (41). Thus, growth of PCA-supplemented and nonsupplemented shaking flask cultures recorded herein was in qualitative and quantitative accordance with the literature (40,41).
If NaHCO 3 supplementation were required for an increased flux via the anaplerotic reactions, and thus explaining the growth stimulation over a reference cultivation, the effect should persist in the simultaneous addition of PCA and NaHCO 3 , because the small amount of PCA is readily consumed and cannot provide a long-lasting surplus of necessary precursors. However, supplementing with both PCA and NaHCO 3 did not result in a further growth improvement, and we found instead that the growthstimulating effect of PCA and NaHCO 3 was related to the presence of the extracellular iron source. In iron-depleted medium, growth ceased after two to three cell divisions as reported previously (42).
Upon dissolution of the iron source FeSO 4 in CgXII medium, iron is present as reduced Fe 2ϩ . However, it will be quickly oxidized under aerobic conditions, yielding 99% of the initial iron concentration as Fe 3ϩ after 15 min (43). Considering the standard preparation times of a shaking flask experiment and estimating the initially dissolved oxygen concentration at 7.5 to 8 mg liter Ϫ1 (22), the entire amount of iron provided in the medium will be oxidized before inoculation. This highlights the need of enhanced iron reduction mechanisms to obtain increased intracellular Fe 2ϩ levels. Since the increased Fe 2ϩ availability in the presence of elevated CO 2 /HCO 3 Ϫ levels could not be attributed to a biological function, we analyzed the chemical reduction capacity of PCA and HCO 3 Ϫ . Results of iron reduction assays and the LC-MS-QTOF approach show that PCA served as an electron donor for the redox reaction. Having performed the experiments at physiological pH (7.4), our results are in contrast with the common understanding that these redox reactions are restricted to acidic pH values (19)(20)(21). At pH 7.4, PCA coordination with Fe 3ϩ represents a mix of 2:1 and 3:1 stoichiometry (20) and was expected to inhibit redox reactions of the iron monocatecholate complexes, which are favored at pH Ͻ 4.5 (21). Furthermore, our results indicate that the primary oxidation products of PCA (quinone and semiquinone) are further degraded. The chemical formula C 7 H 6 O 6 and C 7 H 8 O 6 obtained by the LC-MS-QTOF approach might represent ␤-carboxy-(cis-cis)-muconic acid or ␤-carboxy-␥-carboxymethyl-Δ ␣,␤butenolide and butene-1,2,3-tricarboxy acid that were postulated to be degradation products, when PCA was oxidized with peroxyacetic acid (44). Interestingly, these products also form part of the enzymatic degradation of PCA via the ␤-ketoadipate pathway (45).
Beyond PCA, other phenolic compounds can promote bacterial growth and have been associated with iron reduction and transport in the past. Examples are the reductive mobilization of ferritin iron through PCA and derivatives (46) as well as 2-amino-3-hydroxybenzoic acid secretion by Cryptococcus neoformans which mediated the iron reduction required for Fe 2ϩ uptake (47). Supplementing with catechol stimulated growth of C. glutamicum comparably to PCA (9), and the chemical reduction of Fe 3ϩ through catechol oxidation is well-known and has been reviewed (20). Our results provide evidence that the postulated iron reduction is not unique to PCA but characteristic of other functionalized phenolic compounds. Reduction assays with benzoic acid and 4-amino-2-hydroxybenzoic acid suggest that the carboxyl group was not required for iron reduction and that the stabilization of the semiquinone in adjacent hydroxyl (or mixed functional) groups was advantageous for the reduction of iron (drastically reduced in 4-amino-2-hydroxybenzoic acid).
The beneficial effect of HCO 3 Ϫ on the reduction of iron has not been postulated before. Remarkably, the iron reduction capacity of all aromatic compounds was enhanced in the presence of HCO 3 Ϫ , although iron reduction in the presence of HCO 3 Ϫ alone was negligible. The promoting effect of HCO 3 Ϫ on iron reduction was apparently related to the complex formation between Fe 3ϩ and chelates, as exemplarily shown for PCA. Since we could not clarify the precise mechanism of interaction, one could assume that H ϩ and OH Ϫ concentrations were the reason for the max shift in analogy with an overall alteration of the absorbance due to the formation of Fe(OH)-PCA complexes (19). However, we carefully performed pH buffering and control experiments in order to rule out this possibility. The sample pH was not varied in our experiments by the addition of NaHCO 3 , and intentionally changing the pH failed to reproduce a comparable max shift. It might be further speculated that HCO 3 Ϫ enhances the iron accessibility to PCA in the form of ferric bicarbonate, since the initial complex formation rate was accelerated by 46%.
Although the presence of HCO 3 Ϫ alone had no effect on the reduction of iron in vitro, the growth-promoting effect of CO 2 /HCO 3 Ϫ correlated with increased intracellular Fe 2ϩ levels reported in vivo. We conclude that the growth-promoting effect of CO 2 / HCO 3 Ϫ requires the natural production of suitable reductants by C. glutamicum. Surprisingly, PCA production in C. glutamicum does not appear to be regulated in response to iron availability, unlike other organisms (16). The expression of qsuB encoding the dehydroshikimate dehydratase that catalyzes the conversion of 3-dehydroshikimic acid to PCA is controlled by QsuR in C. glutamicum and provides a shunt when chorismate accumulates (48,49). The question of major iron reductants produced by C. glutamicum must be addressed in future research.
There is continuous debate about the redox reaction in Fe 3ϩ -catecholate complexes and its interpretation as antioxidant or prooxidant, since contradictory observations have been published (20,50). It must be noted that the experimental setup is essentially decisive for the outcome. Many studies emphasized the antioxidant effect of PCA and other catecholates in the past (20). Fe 2ϩ generation was monitored, e.g., by lipid oxidation experiments, which in contrast to BPS detection, do not exhibit high affinity for Fe 2ϩ . As a consequence, observations might differ from ours. The detection of Fe 2ϩ with BPS in this study might have been beneficial for two reasons. (i) Fe 2ϩ and BPS form a stable complex, thus preventing Fe 2ϩ from reoxidation. (ii) Capturing Fe 2ϩ in the complex constantly shifts the chemical equilibrium in favor of reduction. Thus, our in vitro screening system might be well suited for representing a biological context, considering the high Fe 2ϩ affinity of iron processing enzymes, e.g., Escherichia coli IscA (the first enzyme of the iron sulfur cluster assembly machinery, K D ϭ 3.3 ϫ 10 Ϫ20 to 5 ϫ 10 Ϫ20 M) (51,52). Dissociation constants between iron and human transferrin (53) have been determined in a very similar range (2.1 ϫ 10 Ϫ21 and 4.2 ϫ 10 Ϫ20 M). However, the most obvious support of these hypothetical considerations is provided by the consistency of the results of in vitro and in vivo experiments in this study.
Given the predominance of Fe 3ϩ in an oxygenic environment, the abiotic acceleration of iron reduction through HCO 3 Ϫ is of global importance. The two prerequisitesthe presence of functional aromatic compounds on one hand and high HCO 3 Ϫ concentrations on the other hand-combine in a number of habitats. Aromatic compounds are ubiquitous in cellular metabolism, and plant-derived catecholic compounds can be released, e.g., in soils during biomass degradation or reach high concentrations in human plasma shortly after ingestion (20). A number of stress hormones and neurotransmitters, including dopamine, adrenaline, and norepinephrine, contain a functional catechol group and exhibit interesting features with regard to the iron homeostasis. The stimulating role of catecholamine on growth of a number of pathogenic bacteria in iron-restricted growth media containing serum, transferrin, or lactoferrin has been reported (54)(55)(56)(57)(58). Pathogens possess an outstanding role within the inhabitants of the human body as they encounter a constant battle for iron, the majority of which is sequestered in host transferrin and lactoferrin (35). Several studies demonstrated that catecholamines or dietary catechols can liberate iron from transferrin and lactoferrin (55,59,60) and might even reduce Fe 3ϩ by promoting bacterial growth (61). The greater availability of liberated Fe 3ϩ and Fe 2ϩ resulting from this mechanism might be further increased at the high concentrations of CO 2 species typically found in calcareous soils, but also in the human body, where HCO 3 Ϫ concentrations can reach up to 140 mM (22). Considering the HCO 3 Ϫ interaction with iron reduction demonstrated in this study thus might be of particular interest in understanding pathogenicity. Of the currently most threatening human pathogens, Pseudomonas aeruginosa, Staphylococcus aureus, Klebsiella pneumoniae, Mycobacterium tuberculosis, and pathogenic Escherichia coli strains as well as several other pathogenic bacteria regulate toxin expression on the transcriptional level in response to iron availability (33,(62)(63)(64).
In summary, we show that high CO 2 /HCO 3 Ϫ concentrations accelerate the chemical reduction of Fe 3ϩ through various phenolic compounds. This increases the intracellular Fe 2ϩ availability and stimulates growth of C. glutamicum.

MATERIALS AND METHODS
Genetic manipulation, cloning, and strain construction. Standard techniques of molecular biology were applied in accordance with the literature (65). Kits for purification of plasmids and PCR products and for isolation of genomic DNA were used according to Lange et al. (66). Cloning of plasmids and construction of C. glutamicum integration and deletion mutants have been performed as described in detail previously (66). All cloned fragments were verified by Sanger sequencing (GATC, Constance, Germany).
For markerless deletions of the genes pup (cg1689), ftn (cg2782), and dps (cg3327) in C. glutamicum, 500-bp homologous sequences flanking the target gene were amplified with the respective primer pair (see Table S1 in the supplemental material) and simultaneously inserted in BamHI-and HindIII-linearized pK19mobsacB as described before.
For purification of the DtxR regulator protein, the dtxR gene was amplified from the C. glutamicum chromosomal DNA via PCR using the primer pair dtxR-1 and dtxR-2 and cloned into pJOE6089.4 by restriction and ligation using the NdeI and BamHI restriction sites. Competent E. coli DH5␣ was transformed with the resulting plasmid pJOE6089-dtxR carrying a C-terminal Strep-tag II fused to the dtxR gene by electroporation (66).
Bacterial strains and cultivation conditions. An overview of all bacterial strains and plasmids used in this work is given in Table S2. Permanent cultures were maintained at -70°C in 30% (vol/vol) glycerol. C. glutamicum strains were cultivated at 30°C on a rotary shaker in baffled shaking flasks (500 ml filled with 50 ml medium, 120 rpm). Kanamycin was added at a working concentration of 50 g ml Ϫ1 when appropriate.
Permanent cultures were streaked on 2ϫ yeast tryptone (2ϫ YT) (65) agar (18 g liter Ϫ1 ) plates and grown for 2 days. Liquid cultures (5 ml of 2ϫ YT) were inoculated from the plate, grown overnight (O/N), and used to inoculate a 50-ml 2ϫ YT culture, which was incubated for another 6 to 8 h. An appropriate volume of the complex preculture was harvested and resuspended in 2 ml of 0.9% (wt/vol) NaCl to inoculate a CgXII preculture to a starting optical density at 600 nm (OD 600 ) of 1. The main cultures were inoculated in the same way from the CgXII O/N culture and incubated for at least 25 h.
Shaking flask cultivation. C. glutamicum cultivations were performed in CgXII medium (pH 7.4) containing 20 g glucose liter Ϫ1 supplemented with protocatechuic acid (PCA) to a final concentration of 195 M (10) or with 30 mM NaHCO 3 or not supplemented (71) as indicated. Since slight variations occurred in the past, the exact composition of the CgXII medium used in this study is outlined in Lange et al. (66). The functionality of the reporter strain C. glutamicum FEM3 was validated under iron starvation (1 M FeSO 4 ) and iron excess (100 M FeSO 4 ) conditions in the presence of 195 M PCA and starting from identical precultures as described before (32). Iron-depleted growth medium was prepared with a trace element solution lacking the iron source (FeSO 4 ), and the ferrous iron chelator 2,2-dipyridyl was added at a final concentration of 250 M immediately before inoculation to remove any residual iron from the medium (42). Precultures were performed in iron excess conditions (100 M FeSO 4 , instead of the usual 59 M FeSO 4 ).
Bioreactor cultivation. Bioreactor cultivations were performed in batch mode in 1.5-liter stainless steel bioreactors (30) with a starting volume of 800 ml CgXII medium containing 40 g glucose liter Ϫ1 as the sole carbon and energy source but lacking 3-(N-morpholino)propanesulfonic acid (MOPS) buffer and urea. The cultivation pH was maintained at pH 7.4 by the addition of 25% ammonium hydroxide. Two parallel fermentations were always inoculated from identical precultures to a starting OD 600 of 2. Both reactors were operated at a total pressure of 1.5 bar and contained two six-blade Rushton-type impellers. The stirrer speed started at 300 rpm and was gradually increased to maintain the dissolved oxygen concentration above 35%. Reactor 1 was aerated with 0.5 vvm (volume of gas per volume of liquid per minute) pressurized air (0.04% CO 2 ), and reactor 2 was aerated with synthetically mixed gas containing 20% CO 2 , 21% O 2 , and 59% N 2 (20% CO 2 ).
Biomass formation. Cell density was determined by measuring the optical density of a culture at 600 nm (OD 600 ). Biomass concentration (cell weight [dry weight] [CDW] in grams liter Ϫ1 ) was calculated from it using the correlation coefficient CDW ϭ 0.21 ϫ OD 600 (72) specific for the spectrophotometer (Ultrospec 10 cell density meter; GE Healthcare, Little Chalfont, UK).
Fluorescence readings. Fluorescence (relative fluorescence units [RFU]) was detected in 100-l samples of the pure culture using 96-well microtiter plates in the Synergy 2 device (BioTek Instruments, Bad Friedrichshall, Germany) at 37°C (excitation wavelength/bandwidth, 485/20 nm; emission wavelength/bandwidth 528/20 nm) as described before (73). The culture background (after removing cells by centrifugation, 13,300 rpm, 1 min) was subtracted from the sample value and normalized with respect to the biomass concentration (RFU CDW Ϫ1 ).
LC-MS-QTOF. PCA degradation was detected by a relative decrease of the respective peak area via liquid chromatography-quadrupole time of flight mass spectrometry (LC-MS-QTOF). Samples (1-ml samples) containing 0.5 mM FeCl 3 (from a stock in 10 mM HCl) were prepared in 1.5-ml Eppendorf cups. Samples were neutralized with 50 mM NaOH, and 19.5 M PCA, 6.5 mM BPS, and 50 mM NaHCO 3 were added when appropriate. At the indicated time points, 200-l samples were transferred to fresh 1.5-ml Eppendorf cups and stored frozen at -20°C until further processing. After thawing, samples or respective standards were prepared in 60% (vol/vol) acetonitrile, 0.75 mM EDTA, and 10 mM ammonium acetate buffer (adjusted to pH 9.2) and analyzed by an Agilent 1260 bio-inert high-performance liquid chromatography (HPLC) system (Agilent Technologies, Waldbronn, Germany) coupled to an Agilent 6540 accurate-mass QTOF (Agilent Technologies, Santa Clara, CA, USA). Alkaline hydrophilic interaction liquid chromatography (HILIC) (74) and QTOF-MS (75) parameters were set as previously described, and analysis was conducted in the negative mode with a fragmentor voltage of 100 V in the MS mode. The extraction of chromatograms and integration were done in MassHunter Qualitative Analysis with the "Find by Formula" algorithm. Potential oxidation products were identified via the molecular formula calculator feature based on the accurate mass, isotope abundances, and isotope spacing (B.07.00, Agilent Technologies, Santa Clara, CA, USA).
Iron reduction assays. The kinetics of Fe 3ϩ reduction was monitored by an increase of the absorbance at 534 nm due to complex formation with the Fe 2ϩ -specific chelator bathophenanthroline disulfonic acid (BPS) (76,77) in 96-well microtiter plates containing 100 l of sample. The iron reduction assay was performed in 200 mM MOPS buffer (pH 7.4), 0.5 mM FeCl 3 was added from a stock in 10 mM HCl neutralized with 50 mM NaOH immediately before use. 19.5 M PCA, other benzoic acid derivatives, or 50 mM NaHCO 3 was applied as indicated. The reaction was started by the addition of 6.5 mM BPS and measured every 10 min.
Fe 3؉ -PCA complex formation. Complex formation between Fe 3ϩ and PCA was monitored in a setup identical to the setup used in the iron reduction assay but in the absence of Fe 2ϩ -specific BPS. The Fe 3ϩ -PCA complex formation assay was started by the addition of 1.5 mM PCA to provide it in a 3:1 stoichiometry with iron. Fe 3ϩ -catecholate complexes show characteristic max values between 561 and 586 nm (20), and for Fe 3ϩ -PCA, 575 nm was reported as max (19). Wavelength scans of the complexes were performed with 100-l samples in 96-well plates (1-nm step size) after 5 h of incubation. Since the absorbance peaked around 560 nm in the standard setup (see Fig. S5 in the supplemental material), kinetic measurements of complex formation was performed in the following experiments every 5 min at 560 nm in the same setup.

Purification of DtxR.
For the production of DtxR protein, a cryogenic culture of E. coli DH5␣(pJOE6089-dtxR) was streaked on a 2ϫ YT agar plate containing 100 g ampicillin (Amp 100 ) ml Ϫ1 and incubated for 24 h at 37°C. A single cell was used to inoculate 5 ml of 2ϫ YT (Amp 100 ) which was incubated overnight (O/N) at 37°C on a rotary shaker at 120 rpm before 1.5 ml of this culture was transferred to fresh 150 ml of 2ϫ YT (Amp 100 ) medium in a 1-liter baffled shaking flask. This culture was incubated at 37°C on a rotary shaker at 120 rpm until an OD 600 of 0.84 was reached. The expression of dtxR was induced by the addition of L-rhamnose to a final concentration of 2 g liter Ϫ1 , and incubation continued for 6 h at 30°C before the cells were harvested by centrifugation (10 min, 5,000 rpm at 4°C in an Eppendorf centrifuge 5804 R). The cell pellets were stored frozen at -20°C until further processing.
Thermal shift assay. Differential scanning fluorimetry experiments were performed in a final sample volume of 50 l in 96-well plates using the Mastercycler EP Realplex 2 (epgradient S; Eppendorf, Hamburg, Germany). Briefly, at a gradually increasing temperature, protein denaturation is monitored by an increase of the fluorescence which is caused by the binding of a dye to successively presented hydrophobic areas of the protein (78). Ligand binding might increase the protein stability. Thermal protein denaturation was monitored by an increase of the absorbance at 550 nm due to binding of SYPRO Orange (Thermo Fisher Scientific Inc., Bremen, Germany).
Protein samples (50 g ml Ϫ1 ) were prepared in 100 mM potassium phosphate buffer (pH 7.4) containing 600 mM NaCl and 5 mM dithiothreitol (DTT). Divalent metal ions were applied at various concentrations (0.1 M to 1 mM NiCl 2 or 0.1 M to 5 mM MnCl 2 ). When 10 to 200 mM NaHCO 3 was added to the sample, osmolarity was maintained at an identical level by adjusting the NaCl concentration. One microliter of SYPRO Orange was added to each sample immediately before starting the assay. Then, the 96-well microtiter plate was sealed with adhesive foil and incubated in the mastercycler (20°C for 15 s, 20°C to 90°C in 30 min, 90°C for 15 s). The range of the greatest absorbance increase over temperature was fitted by a fourth order polynomial (best fit), and the melting temperature (T m ) was determined at the inflection point. The T m at a given ligand concentration eventually represents the mean value Ϯ standard deviation for at least four replicates. Dissociation constants (K D ) were calculated for the inflection point of the T m on the ligand concentration curve (79).
Statistical analysis. All growth experiments were performed as independent biological triplicates on different days. (Bio)chemical assays were replicated 4 to 6 times, and LC-MS-QTOF analysis was performed on the three biological replicate samples. All values represent mean values with error bars indicating standard deviations. When appropriate, statistically significant differences of sample means were tested with a two-sample t test.

SUPPLEMENTAL MATERIAL
Supplemental material is available online only.