Zinc-dependent substrate-level phosphorylation powers Salmonella growth under nitrosative stress of the innate host response

The metabolic processes that enable the replication of intracellular Salmonella under nitrosative stress conditions engendered in the innate response of macrophages are poorly understood. A screen of Salmonella transposon mutants identified the ABC-type high-affinity zinc uptake system ZnuABC as a critical determinant of the adaptation of Salmonella to the nitrosative stress generated by the enzymatic activity of inducible nitric oxide (NO) synthase of mononuclear phagocytic cells. NO limits the virulence of a znuB mutant in an acute murine model of salmonellosis. The ZnuABC transporter is crucial for the glycolytic function of fructose bisphosphate aldolase, thereby fueling growth of Salmonella during nitrosative stress produced in the innate response of macrophages. Our investigations demonstrate that glycolysis mediates resistance of Salmonella to the antimicrobial activity of NO produced in an acute model of infection. The ATP synthesized by substrate-level phosphorylation at the payoff phase of glycolysis and acetate fermentation powers the replication of Salmonella experiencing high levels of nitrosative stress. In contrast, despite its high potential for ATP synthesis, oxidative phosphorylation is a major target of inhibition by NO and contributes little to the antinitrosative defenses of intracellular Salmonella. Our investigations have uncovered a previously unsuspected conjunction between zinc homeostasis, glucose metabolism and cellular energetics in the adaptation of intracellular Salmonella to the reactive nitrogen species synthesized in the innate host response.

The metabolic processes that enable the replication of intracellular Salmonella under nitrosative stress conditions engendered in the innate response of macrophages are poorly understood. A screen of Salmonella transposon mutants identified the ABC-type high-affinity zinc uptake system ZnuABC as a critical determinant of the adaptation of Salmonella to the nitrosative stress generated by the enzymatic activity of inducible nitric oxide (NO) synthase of mononuclear phagocytic cells. NO limits the virulence of a znuB mutant in an acute murine model of salmonellosis. The ZnuABC transporter is crucial for the glycolytic function of fructose bisphosphate aldolase, thereby fueling growth of Salmonella during nitrosative stress produced in the innate response of macrophages. Our investigations demonstrate that glycolysis mediates resistance of Salmonella to the antimicrobial activity of NO produced in an acute model of infection. The ATP synthesized by substrate-level phosphorylation at the payoff phase of glycolysis and acetate fermentation powers the replication of Salmonella experiencing high levels of nitrosative stress. In contrast, despite its high potential for ATP synthesis, oxidative phosphorylation is a major target of inhibition by NO and contributes little to the antinitrosative defenses of intracellular Salmonella. Our investigations have uncovered a previously unsuspected conjunction between zinc homeostasis, glucose metabolism and cellular energetics in the adaptation of intracellular Salmonella to the reactive nitrogen species synthesized in the innate host response.

Author summary
Microbial pathogens are exposed to multiple antimicrobial defenses during their associations with host cells. Nitric oxide generated in the innate response exerts widespread

Introduction
Many of the more than 2,500 serovars of Salmonella enterica cause gastrointestinal or disseminated infections in millions of people and livestock every year [1,2]. Reactive nitrogen species synthesized abiotically in the gastric lumen and the extreme acidity of the stomach constitute a formidable barrier to most microorganisms. However, Salmonella and other enteropathogens can endure these innate host defenses [3]. In the gastrointestinal tract, Salmonella competes for nutrients and space with members of the resident microbiota and, aided by the cytoskeletal remodeling induced by effectors of the Salmonella pathogenicity island-1 (SPI-1) type-III secretion system, forces its way into enterocytes and M cells of Peyer's patches. The Salmonella SPI-1 effector SopB activates the transcription of Nos2-encoded inducible nitric oxide synthase (iNOS) long after invasion [4,5]. Transcription of Nos2 is independently activated in mononuclear phagocytic cells in response to lipopolysaccharide, fimbriae or porins imbedded in Salmonella's cell envelope [6]. The iNOS flavohemoprotein synthesizes nitric oxide (NO) from the guanidino group of L-arginine and molecular oxygen (O 2 ) [6][7][8]. The diatomic gas NO combines with O 2 , superoxide anion, iron and low-molecular weight thiols, generating a plethora of reactive nitrogen species that are endowed with vigorous antimicrobial activity [6]. By modifying thiol groups in redox active cysteines, the Feα of iron-sulfur clusters, and ferrous or ferric ions in heme cofactors, NO and its oxidative and nitrosative congeners exert cytostasis against Salmonella and various other microbial pathogens [6]. Reactive nitrogen species inhibit quinol oxidases, aconitase, pyruvate dehydrogenase, α-ketoglutarate dehydrogenase and dihydroxy acid dehydratase, thereby restricting oxidative phosphorylation, the tricarboxylic acid (TCA) cycle, and the biosynthesis of methionine, lysine, leucine, isoleucine and valine [9][10][11]. Nitrosative stress is often accompanied by demetallation and loss of protein function [6,12,13]. The released iron, manganese or zinc ions can further disrupt cellular functions by mismetallating metabolic enzymes [14]. The liberation of ferrous iron can also cause genotoxicity via the Fenton-catalyzed synthesis of hydroxyl and ferryl radicals [15]. Despite the many molecular targets poisoned by NO and its congeners, the antimicrobial activity of reactive nitrogen species varies widely against diverse intracellular and extracellular bacterial pathogens. Whereas Pseudomonas aeruginosa, Mycobacterium tuberculosis, and Burkholderia spp. are readily killed by NO [16][17][18][19][20], Staphylococcus aereus continuously replicates in the presence of nitrosative stress levels that are inhibitory to other microorganisms [21].
S. enterica shows intermediate phenotypes. Chemically-generated NO is bacteriostatic against this facultative intracellular enteropathogen [22], but intracellular Salmonella grow remarkably well in the presence of high fluxes of NO synthesized enzymatically by iNOS in the innate response of macrophages [23]. The flavohemoglobin Hmp, cytochrome bd, and lowmolecular weight thiols are the main effectors of the antinitrosative toolbox of Salmonella [6,9,21,24,25]. In addition to these detoxification systems, the Salmonella pathogenicity island-2 type III secretion system helps Salmonella evade contact with iNOS-containing vacuoles [26]. Little is known about the metabolic adaptations that protect intracellular Salmonella against nitrosative stress produced by the host. Herein, we have exploited an unbiased Tn-seq approach to identify hitherto unknown antinitrosative defenses of Salmonella. Our investigations indicate that the high-affinity zinc transporter ZnuABC enables ATP synthesis via substrate-level phosphorylation in glycolysis and acetate fermentation, thereby sustaining Salmonella growth during the nitrosative stress that is generated in the innate host response of macrophages.

Diversification of carbon sources minimizes the anti-Salmonella activity of NO
NO exerts potent bacteriostasis against Salmonella in culture media, but this diatomic radical is well tolerated by intracellular Salmonella. In contrast to specialized bacterial pathogens that have undergone genomic and metabolic reduction [27], nontyphoidal Salmonella such as serovar Typhimurium can catabolize numerous carbon sources during their association with mammalian hosts [28]. Given the pressure NO exerts on central metabolism [9,10,12,29], we investigated whether the availability of diverse carbon sources influences the tolerance of Salmonella to NO. The addition of the NO donor spermine NONOate induced a similarly prolonged lag phase in Salmonella grown in glucose or casamino acids (Fig 1A). The combination of casamino acids and glucose prevented most of the bacteriostasis associated with NO treatment (Fig 1A, S1 Fig). Similar trends were noted when Salmonella were challenged with the NO donors S-nitrosoglutathione (GSNO) or diethylenetriamine-NONOate (DETA NONOate) (S1 Fig). These findings indicate that Salmonella recovers rapidly from NO-induced cytotoxicity as long as it can gain access to diverse carbon sources. As suggested earlier [10,11,29], lysine, methionine and branch chain amino acids may allow Salmonella to circumvent NOinduced amino acid functional auxotrophies associated with the inactivation of lipoamidedependent lipoamide dehydrogenase and dihydroxy acid dehydratase. However, the mechanisms by which glucose strengthens the antinitrosative defenses of Salmonella have not been determined yet.
To gain insights into the metabolic pathways by which glucose and casamino acids power the antinitrosative defenses of Salmonella, we examined the growth of mutants in glycolysis, acetate fermentation and TCA in glucose or casamino acids media (Fig 1B). Electron transport chain mutants were also tested. The inability of Δpgi, ΔpfkAB, ΔgltA, ΔacnAB and ΔicdA mutants to replicate in glucose indicate that glycolysis and the oxidative branch of the TCA cycle support Salmonella growth in this hexose (Fig 1C and 1D). A ΔpfkAB Salmonella strain lacking both isoforms of the glycolytic enzyme phosphofructokinase grew in glycerol (S1 Fig), demonstrating that this mutant can grow in carbon entering glycolysis below fructose-1,6-bisphosphate. ΔatpB and ΔackA Δpta deletion mutants thrived in glucose, suggesting that ATP formed via oxidative phosphorylation or acetate fermentation is largely dispensable for growth of S. Typhimurium in glucose (Fig 1E). Aside from the minor growth defect of ΔsdhAB Salmonella, glycolytic and TCA cycle mutants grew robustly in casamino acids (Fig 1C and  1D). Casamino acids, a mixture of small peptides and amino acids, enter central metabolism at different steps in glycolysis and the TCA cycle (Fig 1B), likely explaining why single glycolytic or TCA cycle mutants flourished in this carbon source. Salmonella, however, required oxidative phosphorylation and acetate fermentation for optimal growth on casamino acids (Fig 1E). Moreover, casamino acids supported Salmonella growth as long as the terminal electron acceptors O 2 or NO 3 were available. Conversely, glucose fueled Salmonella growth in the absence of O 2 or NO 3 - (Fig 1F). Cumulatively, this research indicates that glucose and casamino acids energize Salmonella growth by engaging substrate-level and oxidative phosphorylation, respectively. Our data also indicate that the metabolic diversification associated with glycolysis, acetate fermentation, TCA cycle and the electron transport chain yields a population of Salmonella that is highly immune to NO.

Importance of metabolism in resistance of Salmonella to nitrosative stress
To gain further insights into the mechanisms by which nutritional diversity promotes antinitrosative defenses in Salmonella, we chose an unbiased transposon-based approach. A library of barcoded transposon mutants was separately challenged with the NO donors spermine NONOate, GSNO, or DETA NONOate in MOPS minimal media supplemented with glucose, casamino acids, or a combination of glucose and casamino acids. When the results were corrected to the media alone and the false discovery rate was set to < 10%, spermine NONOate, GSNO, and DETA NONOate exerted selectable advantages or disadvantages to 1132, 1729, and 299 genes, respectively. Differentially selected loci (S1 Table) were sorted into Venn diagrams (Fig 2A, 2C and 2E), and the annotated genes were converted into Go-terms. Panther Pathway analysis indicated that NO exerts the greatest negative and positive selection on loci related to metabolic processes (Fig 2B, 2D and 2F).
Genes were also grouped by medium (S3 Table). The NO donors tested selected for or against 11 (e.g., znuABC, thiCF, nadC) or 126 genes in Salmonella grown in glucose or casamino acids, respectively. Many of the positively selected transposon mutants in casamino acidcontaining medium were deficient in glyoxylate shunt (aceABK), or biosynthesis of amino acids (argABCEGI, thrBC), thiamine (thiCDFGH), and lipopolysaccharide (rfABGIJKLPZ, rfbABCDIKMNP, rfc, arnAT, basRS, pmrDFJLM), suggesting that expression of these genes imposes a considerable metabolic burden when diverse nutrient sources are available to Salmonella undergoing nitrosative stress. As expected, NO negatively selected against mutations in hmpA and gshAB encoding the known antinitrosative defenses flavohemoglobin and glutathione synthase, respectively (S4 Table) [24,25]. In addition, mutants in efflux and acquisition of Zn 2+ (zntA, znuC, respectively), and modification of tRNAs (cmoB) were at a disadvantage when exposed to NO in casamino acids.
Overall, there were very few mutations that were consistently deleterious in most experimental conditions tested. Mutations in genes encoding the high-affinity, ABC-type, Zn 2+ uptake system ZnuABC were the strongest exception. ZnuA is the zinc-binding, periplasmic cassette; ZnuB comprises the membrane-spanning permease; and ZnuC is the cytosolic ATPase that fuels uptake of this divalent cation [30]. Genetically, znuA is divergently transcribed from the znuCB operon [31]. Henceforth, this locus will be referred to as znuABC. Although transposon interruptions of either znuA, znuB, or znuC were under negative selection in most of the screen conditions tested, the disadvantage of the znuA, znuB and znuC mutants was most evident in glucose (Fig 2G, S5 Table), suggesting that Zn 2+ uptake is particularly important in the antinitrosative defenses of glycolytic bacteria. However, not all components of zinc uptake contributed equally to the antinitrosative defenses of Salmonella, because NO did not exert significant pressures on zntB, zitB, or zupT genes encoding low-affinity Zn 2+ transporters (S6 Table). NO did not seem to exert negative selection against sodCI or sodCII mutants deficient in copper-zinc superoxide dismutases (Cu-Zn SODs), which have been involved in resistance to phagocyte NADPH oxidase and iNOS [32].

The znuABC-encoded Zn 2+ uptake system protects Salmonella against the nitrosative stress engendered in the innate host response
We evaluated in more detail the role that the high-affinity Zn 2+ uptake system ZnuABC plays in resistance to NO generated chemically in vitro or enzymatically in the innate response of macrophages and mice. Compared to wild-type controls, ΔznuB Salmonella required longer times to enter exponential growth in all media examined (Fig 3A, S2 Fig). The ΔznuB mutant also had longer doubling times in glucose than wild-type controls (S2 Fig), further reinforcing the idea that zinc plays an essential role in glycolysis. The addition of 5 μM ZnCl 2 restored normal growth to ΔznuB Salmonella, whereas the NO donor DETA NONOate exacerbated the growth defects of ΔznuB Salmonella in all carbon sources tested (Fig 3A, S2 Fig). Expression of a znuB allele in trans restored growth of ΔznuB Salmonella in EG minimal media to wildtype levels and it prevented the hypersusceptibility of this mutant to nitrosative stress (S2 Fig).
Isogenic ΔznuA and ΔznuC Salmonella mutants also displayed hypersusceptibility to DETA NONOate (S2 Fig), indicating that mutations in any of the subunits that encode the high affinity ZnuABC zinc uptake system predispose Salmonella to nitrosative stress. In contrast, Δzur Salmonella was as resistant to DETA NONOate as wild-type controls (S2 Fig), perhaps reflecting derepression of znuABC transcription [30,31]. Collectively, these investigations suggest that glycolysis is particularly reliant on Zn 2+ and that zinc-starved Salmonella are particularly sensitive to the bacteriostatic actions of NO.
We examined whether the antinitrosative defenses associated with the ZnuABC system protect Salmonella against NO engendered in the innate response of mononuclear phagocytic cells. Compared to wild-type controls, ΔznuB Salmonella replicated poorly in NO-producing J774 cells (Fig 3B, S2 Fig) suggesting that the improved growth of ΔznuB Salmonella in J774 cells treated with aminoguanidine or L-NIL cannot be explained by direct effects of these drugs on the bacteria but rather on the inhibition of host iNOS enzymatic activity. Accordingly, ΔznuB Salmonella were also more susceptible than wild-type controls to the antimicrobial activity derived from NO congeners synthesized in the innate response of periodateelicited macrophages (Fig 3C, S2 Fig). We also examined whether the ZnuABC system is a key component of the antinitrosative arsenal of Salmonella in a model of acute infection. Wildtype Salmonella killed C57BL/6 and iNOS -/mice 9 and 7.5 days post infection, respectively (Fig 3D). Consistent with previous investigations [33], ΔznuB Salmonella was attenuated in C57BL/6 mice (p < 0.0001). Interestingly, ΔznuB Salmonella became virulent in iNOS -/mice, indicating that this Zn 2+ acquisition system is an integral aspect of the antinitrosative defenses of Salmonella. iNOS -/mice infected with ΔznuB Salmonella live on average 4 days longer (p < 0.001) than those infected with wild-type controls, suggesting that zinc acquisition also participates in Salmonella virulence in ways that are independent of antinitrosative defense. The Zn 2+ uptake system ZnuABC is needed for maximal fructose bisphosphate aldolase activity In the preparative phase of glycolysis, fructose bisphosphate aldolase splits fructose-1,6-bisphosphate into dihydroxyacetone-phosphate and glyceraldehyde-3-phosphate. As is the case for E. coli, the Salmonella genome encodes for two isoforms of fructose-bisphosphate aldolase. The constitutively expressed class II fructose bisphosphate aldolase (encoded by fbaA) is zinc-dependent, whereas the expression of the zinc-independent class I isoform FbaB is induced in response to gluconeogenic substrates. Hence, in E. coli grown in glucose, 95-100% of the total fructose bisphosphate aldolase activity is contributed by the zinc-dependent class II fructose bisphosphate aldolase [34]. We noticed that ΔznuB Salmonella harbored about two thirds of the fructose bisphosphate aldolase enzymatic activity of wild-type controls ( Fig  4A). Fructose bisphosphate activity was inhibited by the Zn 2+ -chelator N,N,N',N'-tetrakis (2-pryidinylmethyl)-1,2-ethanediamine (Fig 4A). The addition of 5 μM ZnCl 2 to the EG minimal media eliminated the differences in fructose bisphosphate aldolase activity between wild- type and ΔznuB Salmonella (Fig 4B). Together, these findings suggest that most differences in fructose bisphosphate aldolase content between wild-type and ΔznuB Salmonella are due to poor zinc uptake. NO inhibited (p < 0.05) by 10% the enzymatic activity of fructose bisphosphate aldolase in wild-type Salmonella (Fig 4C). NO appears to inhibit fructose bisphosphate aldolase to a greater extent in ΔznuB Salmonella (p < 0.01) than wild-type controls (p < 0.05).
We wondered if a 33% drop in fructose bisphosphate aldolase activity has a significant impact in the intracellular growth defect of ΔznuB Salmonella. To test this idea, ΔznuB Salmonella were complemented with the fbaA or fbaB genes, encoding zinc-dependent or-independent fructose bisphosphate aldolase isoforms, respectively (Fig 4D). Although intracellular growth was dramatically improved (p < 0.001), expression of fbaA gene did not fully restore growth of ΔznuB Salmonella to wild-type levels (p < 0.001 when ΔznuB pFBAA is compared to wild-type Salmonella). However, expression of the fbaB gene encoding the zinc-independent fructose bisphosphate aldolase isoform fully restored the intracellular growth of ΔznuB Salmonella. Together, these findings strongly indicate that the decreased enzymatic activity of fructose bisphosphate aldolase is a major contributor to the poor intracellular growth exhibited by ΔznuB Salmonella.
Zinc serves as cofactor for multiple proteins, including Cu-Zn SODs that detoxify superoxide anion in the periplasm. By limiting peroxynitrite formation from superoxide and NO, Cu-Zn SODs contribute to the antinitrosative defenses of Salmonella [32]. To assess if ΔznuB Salmonella has defects in detoxifying superoxide anion formed in the plasma membrane from the adventitious reduction of oxygen by NADH dehydrogenases [35], we monitored the formation of nitrotyrosine residues as a proxy of peroxynitrite. Wild-type and ΔznuB Salmonella harbored similar levels of nitrotyrosine after treatment with 500 μM spermine NONOate (S3 Fig). As predicted by the critical role NADH dehydrogenases play in the production of endogenous superoxide in Salmonella, a Δnuo Δndh mutant showed a dramatic decrease in nitrotyrosine formation in response to spermine NONOate. Together, these findings suggest that ΔznuB Salmonella has normal SodC function, perhaps because, as it has been suggested earlier [36], metallation of apo-SodC is likely to occur in the periplasm.

Glycolysis contributes to the antinitrosative defenses of Salmonella
Given the role the ZnuABC uptake system plays in glycolysis and antinitrosative defense, we tested the contribution of glycolysis in the resistance of Salmonella to NO. The gene fba, encoding the zinc-dependent, fructose bisphosphate aldolase is an essential gene in E. coli. The lack of transposons in the gene indicate that it is also essential in Salmonella. Therefore, we tested the phenotypes of a glycolytic mutant deficient in the pfkAB-encoded phosphofructose kinase, and found that ΔpfkAB Salmonella suffered a prolonged lag phase when challenged with DETA NONOate (Fig 5A, S4 Fig). Consistent with previous reports [37,38], ΔpfkAB Salmonella failed to replicate within macrophages (Fig 5B). Blockage of NO synthesis with the iNOS inhibitor N6-(1-iminoethyl)-L-lysine (L-NIL) partially rescued the intracellular growth defect of ΔpfkAB Salmonella in J774 macrophage cells (Fig 5B, S4 Fig). The intracellular replication defect and hypersuceptibility of ΔpfkAB Salmonella to DETA NONOate could be complemented by expression of pfkA or pfkB genes in trans (S4 Fig). The ΔpfkAB Salmonella strain was extremely attenuated in C57BL/6 mice but became virulent in iNOS -/mice (Fig 5C), strongly suggesting that glycolysis helps Salmonella to overcome the cytotoxicity of NO synthesized in the innate host response. As noted with ΔznuB Salmonella, iNOS -/mice succumbed more slowly to ΔpfkAB Salmonella infection than controls infected with wild-type Salmonella (p < 0.01), suggesting that glycolysis supports Salmonella virulence in NO-dependent and -independent ways.
In addition to glycolysis, oxidative phosphorylation can be a sizable source of ATP. Therefore, we also tested whether oxidative phosphorylation contributes to the antinitrosative defenses of Salmonella. We noted that a Salmonella strain deficient in the atpB-encoded subunit of ATP synthase recovered from nitrosative stress as efficiently as wild-type controls ( Fig  5D). Furthermore, wild-type and ΔatpB Salmonella replicated to similar densities within NOproducing J774 macrophages (Fig 5E, S4 Fig). Given its excellent growth in J774 cells, we were surprised to find that ΔatpB Salmonella were not only severily attenuated in C57BL/6 mice but remained avirulent in iNOS -/mice. Similar results were obtained with a Δnuo Δndh strain lacking both isoforms of the NADH dehydrogenase of the electron transport chain. Thus, the ATP synthase seems to be dispensable for antinitrosative defenses of intracellular Salmonella, but it is critical for Salmonella pathogenesis.

Substrate-level phosphorylation sustains growth of intracellular Salmonella undergoing nitrosative stress
Because substrate-level phosphorylation in glycolysis can be a sizable source of ATP, we examined the degree to which glucose utilization helps Salmonella maintain the ATP pool during nitrosative stress. Thin layer chromatography and independent firefly luciferase determinations indicated that spermine NONOate reduces the ATP pool in Salmonella growing in glucose or casamino acids (reductions of 90 vs. 99%, respectively) (Fig 6A and 6B). Salmonella growing in both glucose and casamino acids retained about 60% of the ATP pool after NO treatment (Fig 6A and 6B), indicating that access to diverse carbon sources that feed into glycolysis and the electron transport chain protects most of the ATP pool from the toxic actions of NO. The ATP pools were also measured in glycolytically-deficient ΔznuB and ΔpfkAB Salmonella and oxidative phosphorylation-deficient ΔatpB Salmonella. Because of its extreme growth defect in glucose, the ΔpfkAB mutant was tested in LB broth. Wild-type, ΔatpB, ΔpfkAB and ΔznuB Salmonella suffered about 3-, 5-, 8-, and 21-fold reductions, respectively, in the ATP pool upon NO challenge (Fig 6C and 6D, S5 Fig). Cumulatively, these findings suggest that zinc-dependent glycolysis is more effective than oxidative phosphorylation at preserving cellular energetics during periods of nitrosative stress.
The pgk-encoded phosphoglycerate kinase and pykAF-encoded pyruvate kinases generate ATP by substrate-level phosphorylation in the payoff phase of glycolysis. In addition, an ATP molecule is synthesized by the fermentation of acetyl-CoA to acetate by the ackA-encoded acetate kinase. Because pgk is an essential gene in Salmonella, we directed our attention to the pykAF and pta-ackA pathways. In increasing order, spermine NONOate depleted ATP from wild-type, ΔackA Δpta, ΔpykAF and ΔpykAF ΔackA Δpta Salmonella (Fig 6E), suggesting that glycolysis contributes most of the ATP synthesized by substrate-level phosphorylation in NOtreated Salmonella but that acetate fermentation is also an important source of ATP. Spermine NONOate inflicted more severe bacteriostasis to a ΔpykAF ΔackA Δpta mutant than ΔpykAF or ΔackA Δpta isogenic strains (Fig 6F), and NO produced in the innate response of J774 cells stunted the growth of ΔpykAF ΔackA Δpta Salmonella more severily than that of ΔpykAF or ΔackA Δpta controls (Fig 6G). The growth defects of ΔpykAF ΔackA Δpta Salmonella could be complemented by expression of the ackA pta operon in trans (S5 Fig). ΔpykAF ΔackA Δpta Salmonella were also more susceptible to the antimicrobial activity of iNOS expressed in the innate response of primary macrophages (Fig 6H, S5 Fig). Moreover, in contrast to the ΔpykAF or ΔackA Δpta parent strains, the ΔpykAF ΔackA Δpta mutant was attenuated in C57BL/6 mice (p < 0.05 as compared to wild-type Salmonella). Interestingly, ΔpykAF ΔackA Δpta Salmonella became as virulent as wild-type bacteria in iNOS -/mice (p = 0.5) (Fig 6I). Collectively, these investigations demonstrate that ATP synthesized by substrate-level phosphorylation in both glycolysis and acetate fermentation protects Salmonella against the cytotoxicity of NO produced in the innate response.

Discussion
Zn 2+ , the second most abundant metal cofactor, provides structural, regulatory, antioxidant, and catalytic properties to diverse metalloproteins [39,40]. Given its critical importance in bacterial cell physiology, mammalian hosts actively limit the bioavailability of Zn 2+ to bacterial pathogens, thereby contributing to what is now known as nutritional immunity [41]. As a countermeasure, bacterial pathogens use high affinity Zn 2+ transporters, such as ZnuABC or the Gram-positive AdcABC orthologue, to compete with the host for zinc [41]. Interestingly, Salmonella utilizes ZnuABC to compete for zinc with indigenous microbiota of the gut, and exploits this high-affinity uptake system to gain advantages in systemic sites and macrophages  [33,42]. The mechanisms underlying poor growth of ΔznuB Salmonella within macrophages have not been identified yet. Herein, we show that ZnuABC potentiates Salmonella pathogenesis by in part antagonizing the nitrosative stress generated in the innate response of macrophages and mice. Despite its widespread utilization in metabolism and multiple regulatory processes, the scarcity of zinc in ΔznuB Salmonella is particularly detrimental to zinc-dependent fructose bisphosphate aldolase in glycolysis as suggested by the complementation of the intracellular growth defects of ΔznuB Salmonella with the fbaB gene encoding zinc-independent fructose bisphosphate aldolase. Thus, not only does high affinity zinc uptake defend microbes against the metabolic stress associated with either nutritional immunity or competition with microbiota, but it also arms bacteria with the glycolytic flexibility needed to overcome the deficiencies in energetics that are triggered by the inhibition of cytochromes by NO of the innate response of professional phagocytes.
The widespread utilization of zinc in multiple metabolic pathways may explain why ΔznuB Salmonella exhibits growth delays in all carbon sources tested. Our investigations have demonstrated that the ZnuABC uptake system plays a particularly salient role in the antinitrosative defenses of Salmonella, by perhaps promoting DksA-mediated regulation of transcription and the enzymatic activity of RecBCD DNA repair proteins [13,40,43]. Salmonella infection results in elevated levels of free zinc in macrophages, which is then readily available for the needs of the pathogen, but also undermines production of reactive oxygen and nitrogen species [44]. Despite the widespread utilization of zinc in the cell, a sizable component of the antinitrosative defenses associated with zinc uptake appears to be dependent on glycolysis. We noticed that ΔznuB Salmonella is particularly sensitive to the antimicrobial actions of NO when grown in glucose. The high demand for zinc in glycolysis likely reflects usage of zinc by class II fructose bisphosphate aldolase. Poor glycolytic activity might predispose ΔznuABC mutants to NO, an idea that is independently supported by the attenuation of ΔpfkAB Salmonella in mice expressing a functional iNOS hemoprotein. Glycolysis allows intracellular Salmonella to use the glucose available in macrophages [28], satisfying the requirements for carbon. In addition, the attenuation of ΔpykAF ΔackA Δpta Salmonella indicates that glycolysis protects against nitrosative stress by promoting ATP synthesis in substrate-level phosphorylation (Fig 7). Glycolysis could also boost antinitrosative defenses by balancing redox away from the electron transport chain, as has been demonstrated in S. aureus [45].
ATP generated by oxidative phosphorylation contributes very little to the antinitrosative defenses of Salmonella (Fig 7). Cytochrome bd and cytochrome bo are among the most sensitive targets of NO [46]. Nitrosylation of terminal cytochromes of the electron transport chain has a devastating effect on the ATP synthesized by oxidative phosphorylation [47]. Additionally, aconitase, pyruvate dehydrogenase, and α-ketoglutarate dehydrogenase are all sensitive to NO [10,48], thereby not only limiting the ATP synthesized in the TCA cycle but also preventing the generation of NADH reducing power that fuels oxidative phosphorylation. Although Salmonella does not seem to strongly rely on oxidative phosphorylation to overcome nitrosative stress, ΔatpB and Δnuo Δndh mutants lacking ATP synthase and NADH dehydrogenases, respectively, are highly attenuated. These phenotypes may be explained by the fact that the electron transport chain is critical for maintaining redox balance, uptaking and effluxing substrates across the cytoplasmic membrane, and driving protein folding in the cell envelope [49]. The ATP synthase working in reverse can also be a important aspect of the proton motive force [50].
As shown by others in Borrelia burgdorfori [12], fructose bisphosphate aldolase was found to be inactivated in Salmonella after NO treatment. However, fructose-bisphosphate aldolase enzymatic activity appears to be more resistant to NO toxicity than terminal cytochromes of the electron transport chain [9,46,51]. The high sensitivity of the electron transport chain and TCA cycle to nitrosative stress provides a reasonable explanation for the capital importance of substrate-level phosphorylation in maintaining cellular energetics of Salmonella undergoing nitrosative stress. The excellent recovery of Salmonella from nitrosative stress in the presence of both glucose and casamino acids could be partially explained by the retention of much of the ATP pool. Amino acids may also promote the recovery of Salmonella from NO by relieving the functional auxotrophies for branch chain amino acids, lysine and methionine that follow damage of dihydroxy-acid dehydratase and lipoamide-dependent lipoamide dehydrogenase by reactive nitrogen species [10,11].
In summary, the ability of Salmonella to simultaneously exploit different carbon sources during its intracellular life-cycle [28] and diversification of metabolic pathways used to synthesize ATP may underlie the remarkable resistance of Salmonella to NO generated by the innate response of macrophages [7,23]. Our investigations demonstrate that zinc allows for metabolic flexibility during nitrosative stress by allowing for proper fructose bisphosphate aldolase activity and glycolysis.

Ethics statement
All methods and experimental procedures were carried out in accordance to protocols approved by the University of Colorado School of Medicine (UCSOM) Institutional Biosafety Committee, authorization number 01-028. Mouse experiments were performed at Animal   Fig 7. Model for zinc and energy metabolism in the antinitrosative defenses of Salmonella. ATP can be generated by substrate-level phosphorylation in glycolysis and acetate fermentation or oxidative phosphorylation in the electron transport chain. Nitric oxide (NO) produced by iNOS preferentially targets oxidative phosphorylation, and thus most ATP used for growth under nitrosative stress is derived from substrate-level phosphorylation. Zn 2+ acquired through the ABC-type high affinity zinc transport system ZnuABC plays a critical role to the antinitrosative defenses of intracellular Salmonella by enabling fructose bisphosphate aldolase enzymatic activity in glycolysis. The utilization of zinc in multiple cellular processes also adds to the antinitrosative defenses of Salmonella. https://doi.org/10.1371/journal.ppat.1007388.g007 Care Facility of the UCSOM in accordance to the guidelines established by the UCSOM Institutional Animal Care and Use Committee (IACUC) protocol # 56413(07)1E.

Bacterial strains
Salmonella enterica serovar Typhimurium strain ATCC 14028s and its mutant derivatives were used in these studies (S7 Table). In-frame deletion mutations were constructed using the λ Red recombinase system as originally described [52][53][54]. Briefly, the kan gene from pKD4 or pKD13, or the cat gene from pKD3, were PCR amplified using primers with 40-45 bases of overhang on the 5' end that were homologous to the 40-45 bases following the ATG and the 40-45 bases preceding the stop codon of the gene of interest that was to be deleted. These linear DNA segments were gel purified, and electroporated into Salmonella expressing the λ Red recombinase from the plasmid pKD46 or pTP233. Transformants were selected on either LB (Luria Bertoni) kanamycin (50 μg/mL) or LB chloramphenicol (20 μg/mL) plates. Mutations were verified by PCR and were then phage (P22) transduced into 14028s wild-type Salmonella. Transduced colonies were selected on antibiotic plates, were confirmed by PCR, and were then cured of phage contamination. Primers used to generate and check these mutations are listed in S7 Table.

Construction of the Tn-seq Salmonella library
Construction of the barcoded Tn-seq library in S. enterica sv Typhimurium 14028s has been described in detail elsewhere [55]. In brief, EZ-Tn5 <KAN-2> (http://www.lucigen.com) was modified to introduce an N 18 barcode adjacent an Illumina Read 1 sequence. A library of over 230,000 different insertion mutants was constructed by mixing transposase and barcoded construct and subsequent electroporation into electrocompetent bacterial cells. The barcode associated with each unique Tn5 insertion position was determined by Illumina sequencing of PCR-amplified flanking regions, as described [55].

NO recovery assays
Salmonella grown in MOPS minimal media to an OD 600 of 0.1-0.2 were challenged with either 750 μM spermine NONOate, 5 mM GSNO or 2.5 or 5 mM DETA NONOate (Cayman Chemical Company, Ann Arbor, MI). Growth was recorded by measuring the OD 600 of a 200 μL sample every 30 or 60 mins in a 96-well plate. The cultures' doubling times of the initial OD 600 were calculated by exponential regression. Alternatively, 200 μL of overnight culture diluted 1:1000 into fresh media were seeded into honeycomb microplates and treated with either the polyamine diethylenetriamine (DETA) or DETA NONOate, and OD 600 was recorded every 15 min for up to 40 h in a Bioscreen C plate reader (Growthcurves USA, Piscataway, NJ). Doubling times were calculated according to the equation DT = ln(2)/r, where r is the growth rate as calculated by regression analysis. The time at which the cultures reached half of their maximum growth (T = ½ OD 600 Max), a method to calculate the growth delay [10], was calculated by exponential regression.

Screen conditions
Three different NO donors with different chemical properties were selected for our screen. Spermine NONOate spontaneously releases 2 molecules of NO per parent molecule with a half-life of 39 minutes at pH 7.4, 37˚C. The NO donor DETA NONOate decays with a half-life of 20 h at pH 7.4, 37˚C, producing a long and sustained flux of NO for the duration of the experiment [58]. GSNO can homolytically release NO, or heterolytically transfer an NO + -like species to redox active thiol groups in cysteine residues. The transposon library was challenged with these NO donors in early exponential phase in MOPS minimal media supplemented with either glucose, casamino acids, or glucose and casamino acids. The resulting populations were compared to the library grown in media alone. More specifically, the Salmonella Tn5 library was grown for 20 h in LB broth at 37˚C with vigorous shaking. An aliquot of the culture was stocked in 10% glycerol at -80˚C as the input, and another aliquot was diluted 1:100 into 10 mL of MOPS minimal media supplemented with glucose, casamino acids, or glucose and casamino acids. The cultures were grown for 2.5-4 h to an approximate OD 600 of 0.2. The cultures were split into 2 mL aliquots and were either left untreated or were challenged with either 750 μM spermine NONOate, 5 mM GSNO, or 5 mM DETA NONOate. After 20 h of culture, the resulting populations were stocked in 10% glycerol at -80˚C. The procedure was performed in biological triplicates on 3 different days. We analyzed two of those biological replicates by deep-sequencing.
Sequencing and analysis of the barcoded input and output Tn libraries was performed as described in detail in de Moraes et al [55]. Briefly, bacterial pellets representing approximately 5 x 10 7 bacteria were washed three times in water followed by proteinase K digestion, enzyme inactivation, and nested PCR to amplify the N 18 barcode region and add sample-and experiment-specific N 8 indices. Different samples, with different indexes, were pooled and QIAquick purified (Qiagen, Hilden, Germany), followed by Illumina sequencing with standard primers. The first 18 bases in each sequencing read, which represented the unique N 18 tag for each Tn5 mutant, were extracted, and the abundance of these 18-mers was calculated using custom perl scripts.

Statistical analysis of the Salmonella Tn-seq library screen
Detailed methods of how the location and sequence of N 18 barcode tag flanked by conserved priming sites for each Tn5 insertion mutant were identified are reported elsewhere [55]. Abundances of all mutants within each annotated feature were summed up which resulted into a data matrix where rows were annotated features and columns were samples. This data matrix was used to generate log 2 ratios as compared to either input library or library passaged in control media (without stress inducing agents) using edgeR [59].

Fructose bisphosphate aldolase assay
Fructose-1,6,-bisphosphate aldolase activity was assayed as previously described [12]. Briefly, wild-type and ΔznuB Salmonella were grown in EG minimal media to an OD 600 of 0.4-1.0. Where indicated, the medium was supplemented with 5 μM ZnCl 2 . Some of the cultures were treated with 750 μM spermine NONOate for 5 min. Cell pellets were stored at -80˚C until assayed. The pellets were resuspended in 500 μL of 100 mM Tris, pH 7.2 and the bacterial cells were lysed by sonication on ice. The specimens were centrifuged at 13,000 RPM for 15 min at 4˚C. Lysates were diluted to approximately 5 mg/mL of protein as measured by A 280 readings. Some of the specimens were treated with with 100 μM of the zinc chelator N,N,N',N'-tetrakis (2-pyridinylmethil)-1,2-ethanediamine (TPEN) at room temperature for 30 min. Aliquots of 100 μL of lysate were added to 700 μL of 100 mM Tris, pH 7.2, 100 μL of 4 mM NADH in 100 mM Tris, pH 7.2, 100 μL of 58 mM fructose-1,6-bisphosphate trisodium salt in water, 2.5 U of type X triosephosphate isomerase from rabbit tissue and 1.5 U of type I 3-phospho-glyceraldehyde dehydrogenase, from rabbit muscle (Sigma-Aldrich, St. Louis, MO). Specimens were mixed well and four 200 μL samples were pipetted into a 96-well plate. The consumption of NADH was measured by recording the A 340 every 20 sec for 10 min. The ΔA 340 /min was generated by linear regression and technical replicates were averaged and normalized to the protein in the sample. All activities were normalized to untreated samples gathered from wild-type Salmonella.

Intracellular replication
J774A1 murine macrophage-like cells (ATCC TIB67) were maintained at 37˚C, 5% CO 2 in RPMI + media (RPMI media supplemented with 2 mM L-glutamate, 1 mM sodium pyruvate, 15 mM HEPES buffer, 10% fetal bovine serum, and 100 U/mL Penicillin-Streptomycin). Twenty hours prior to infection, J774 cells were plated at 10 5 cells per well in a 96-well plate in 100 μL of media. Salmonella cultures were grown for 20 h at 37˚C with vigorous shaking. Overnight bacterial cultures were diluted to approximately 2x10 6 CFU/mL in macrophage media without Pen/Strep. J774 cells were washed with prewarmed RPMI + media, and were challenged with Salmonella at an MOI of 2. The plates were centrifuged at 4000 RPM for 1 min at room temperature, and then incubated at 37˚C, 5% CO 2 . After 25 min, the culture media was removed and fresh RPMI + media containing 50 μg/mL of gentamicin was added to the macrophages at 37˚C and, after 1h, the media was replaced with media containing 10 μg/ mL gentamicin. One and seventeen h later (T = 2 and 18 h of infection), the media was removed and the macrophages were lysed with 0.25% (w/v) deoxycholic acid prepared in PBS. The bacterial burdens were quantified after 10-fold serial dilutions on LB agar plates. Fold-replication was calculated by normalizing the CFU/mL at T = 18 h to the CFU/mL at T = 2 h. When applicable, nitrite in the culture supernatants was assayed with the Griess reaction as previously described [60].

Survival in primary macrophages
The intracellular survival of wild-type and mutant Salmonella was tested in periodate-elicited macrophages as described [61]. Macrophages from C57BL/6 and iNOS -/-[62] mice were infected at MOI of 2, and survival was determined after 16-20 h of challenge.

Murine infections
These following studies were approved by the Institutional Animal Care and Use Committee at the University of Colorado-Denver Anschutz Medical Campus. Eight to ten week-old C57BL/6 and congenic iNOS -/mice were infected i.p. with 100-200 CFU of the indicated Salmonella strains. Mouse survival was monitored over time.

Thin layer chromatography estimations of ATP pools
Wild-type Salmonella were grown for 20 h in MOPS supplemented with 2 mM HK 2 PO 4 and either glucose, casamino acids, or glucose and casamino acids. The overnight cultures were diluted 1:100 into MOPS supplemented with the same carbon source and 0.4 mM HK 2 PO 4 . The bacteria were grown to an OD 600 of 0.2 and then labeled with 10 μCi of 32 P orthophosphate in 1 mL aliquots. After approximately 2.5 generations, at an OD 600 of 0.5, cultures were either left untreated or were challenged for 5 min with 750 μM spermine NONOate. One mL cultures were mixed with 0.4 mL of ice-cold 50% formic acid and samples were put on ice for at least 15 min prior to centrifugation for 5 min at 13,000 RPM. 10 μL of lysates were spotted along the bottom of polyethyleneimine-cellulose TLC plates (EMD Millipore, Darmstadt, Germany). The TLC plates were air-dried and then placed into a chromatography chamber containing either 1.25 or 0.9 M KH 2 PO 4 , pH 3.4. The solvent system was allowed to migrate 15-19 cm up the 20 cm TLC plate. The plates were air-dried, placed inside plastic wrap, and placed on a phosphorscreen overnight. The following day the screens were scanned with a phosphorimager. Images were cropped and brightness and contrast were adjusted in Photoshop 11.0.

Firefly luciferase estimations of ATP pools
Intracellular pools of ATP were calculated with the luciferase-based ATP Determination Kit (Molecular Probes, Eugene, OR) as instructed by the manufacturer with a few minor adjustments. Briefly, wild-type Salmonella grown to an OD 600 of 0.2-0.5 were either left untreated or challenged with 750 μM spermine NONOate for 5 min. Samples (0.5 mL culture) were thoroughly mixed with 0.6 mL freshly prepared, ice-cold 380 mM formic acid and 17 mM EDTA, and small samples were taken to quantify bacterial density. Formic acid-EDTA samples were saved at -80˚C until assayed. Specimens were centrifuged for 1 min at 13,000 RPM, and the supernatants were diluted 25-fold into 100 mM TES buffer, pH 7.4 to neutralize the formic acid. Ten μL of samples or solutions with known concentrations of ATP were mixed with 90 μL of reaction master mix (10 mL; 8.9 mL water, 500 μL of 20X buffer, 500 μL of 10 mM Dluciferin, 100 μL of 100 mM DTT, and 2.5 μL of 5 mg/mL firefly luciferase) in white 96-well plates, and luminescence was recorded in a Lmax 1.1L machine (Molecular Devices, San Jose, CA), and the data analyzed using the SOFTmax Pro software. Concentrations of ATP in the culture were generated using linear regression of ATP standards. Intracellular ATP concentrations of the original samples were calculated using the CFU/mL counts assuming a cell volume of 1 fL.

Statistical analysis
Statistical analyses were performed using GraphPad Prism 5.0b Software. One-way and twoway ANOVA, t-tests and logrank tests were used. Results were determined to be significant when p < 0.05. Supporting information S1  Salmonella were grown in EG minimal media to OD 600 of 0.4 at 37˚C with shaking. The bacteria were lysed by sonication and the specimens were tested for the presence of nitrotyrosine residues by Western blotting as described [1]. Where indicated (+), the bacteria were treated with 500 μM spermine NON-Oate for 30 min prior to sonication. The blot is representative of 2 independent samples. � , proteins nonspecific labeled by the anti-nitrotyrosine antibodies. Arrows indicate proteins bearing nitrotyrosine residues. (DOCX)