Blocks in Tricarboxylic Acid Cycle of Salmonella enterica Cause Global Perturbation of Carbon Storage, Motility, and Host-Pathogen Interaction

We performed perturbation analyses of the tricarboxylic acid cycle of the gastrointestinal pathogen Salmonella enterica serovar Typhimurium. The defect of fumarase activity led to accumulation of fumarate but also resulted in a global alteration of carbon fluxes, leading to increased storage of glycogen. Gross alterations were observed in proteome and metabolome compositions of fumarase-deficient Salmonella. In turn, these changes were linked to aberrant motility patterns of the mutant strain and resulted in highly increased phagocytic uptake by macrophages. Our findings indicate that basic cellular functions and specific virulence functions in Salmonella critically depend on the proper function of the primary metabolism.

environments (1,2). Several endogenous factors, such as the energy status of the cell, influence TCA cycle activity. For example, the activity of the isocitrate dehydrogenase is allosterically stimulated by ADP (3), whereas ␣-ketoglutarate dehydrogenase is inhibited by its products succinyl coenzyme A (CoA) and NADH (4). In addition, bacterial citrate synthesis is controlled by allosteric inhibition of citrate synthase by ATP and NADH (5). However, TCA cycle activity is also influenced by exogenous factors, such as exposure to antibiotics and reactive oxygen species (ROS), which target sensitive enzymes harboring Fe-S clusters (6,7).
Salmonella enterica serovar Typhimurium (S. Typhimurium) is an invasive facultative intracellular pathogen, the causative agent of human gastroenteritis, and serves as a model organism for systemic Salmonella infections. The divergent host niches colonized during infection require S. Typhimurium to adapt its metabolism from the intestinal lumen, which is a nutrient-rich environment with a competing microbiome (8), to severe nutritional restrictions and ROS attacks inside the so-called Salmonellacontaining vacuole (SCV) during intracellular life within host cells (9,10). Its versatile and robust metabolism (11) makes S. Typhimurium an ideal model organism to study the interconnection of metabolism and virulence functions.
To address the role of the TCA cycle in pathometabolism of S. Typhimurium, we analyzed the effect of perturbations of the TCA cycle using a set of mutant strains each defective in one enzymatic step. Our previous study indicated that TCA cycle perturbations induced in S. Typhimurium by oxidative stress result from damage of Fe-S cluster-containing enzymes (12). Accordingly, a mutant strain deficient in all three fumarase isoforms (ΔfumABC) accumulated large amounts of TCA intermediate fumarate but also showed the remarkable phenotype of increased phagocytosis by murine macrophages. These observations pointed toward a link between TCA cycle metabolite fumarate and cellular functions of S. Typhimurium.
The C 4 -dicarboxylate fumarate recently gained increasing interest due to various links between metabolism and bacterial pathogenesis. In enterohemorrhagic Escherichia coli (EHEC), fumarate is essential for full virulence in a Caenorhabditis elegans infection model where it regulates the expression of a tryptophanase by the transcription factor Cra (13). In Mycobacterium tuberculosis, fumarase deficiency was shown to be fatal due to protein and metabolite succination (14). Other studies demonstrated fumarate as a factor that increases the frequency of persister formation or modulates motility and chemotaxis in E. coli (15)(16)(17).
In this work, we conducted metabolomics and proteomics studies to characterize the metabolic landscape of S. Typhimurium ΔfumABC. By this dual-omics approach, we elucidated a new example for the interplay between metabolism and cellular functions and virulence in S. Typhimurium.

RESULTS
Effects of TCA cycle enzyme deletion on the carbon metabolism of S. Typhimurium. For a global analysis of the effects of perturbations of the TCA cycle on pathometabolism of S. Typhimurium, we generated a set of isogenic S. Typhimurium mutant strains, each defective in one reaction of the TCA cycle. Using this set of strains in comparison to S. Typhimurium wild type (WT), we performed metabolomics analyses of stationary cultures, grown for 18.5 h in rich medium (LB broth), and analyzed samples as described before (12). Metabolomics revealed that the ΔfumABC strain, deficient in all fumarase isoforms, had a highly aberrant metabolic profile distinct from that of other mutant strains. Besides a strong accumulation of fumarate (115-fold compared to WT), S. Typhimurium ΔfumABC contained significantly increased amounts of glycolysis and pentose phosphate pathway (PPP) intermediates.
Moreover, the ΔfumABC strain exhibited increased levels of glucose-6-phosphate (G6P), fructose-6-phosphate (F6P) and sedoheptulose-7-phosphate (S7P), whereas all other mutant strains exhibited decreased or unchanged levels compared to WT ( Fig. 1; see also Table S1 in the supplemental material). This observation indicates distinct and unique impacts of the fumarase deletions on carbon flux.
Only a mutant strain deficient in succinate dehydrogenase also showed a larger level of F6P, but not to the same extent as observed for the ΔfumABC mutant. Furthermore, there was a strong accumulation of aspartate, likely arising from the large pool of fumarate by the action of aspartate ammonia-lyase AspA (Table S2).
In our previous analyses of ROS-induced damage of TCA cycle enzymes in S. Typhimurium pathometabolism, we found that a mutant strain unable to detoxify endogenously generated ROS was attenuated in intracellular proliferation. Surprisingly, this mutant strain was internalized by macrophages at higher rates than S. Typhimurium WT (12). Endogenous ROS cause damage of Fe-S cluster-containing TCA cycle enzymes, and also a ΔfumABC strain was internalized by macrophages at a 15-foldhigher rate than WT S. Typhimurium, without defects in intracellular proliferation. These observations point toward a link between the function of the TCA cycle and the virulence properties of S. Typhimurium, which prompted us to characterize the S. Typhimurium ΔfumABC strain in detail.
To test for increased glycogen storage, bacterial cultures grown on LB agar were treated with potassium iodine for glycogen staining (18). While S. Typhimurium WT was only lightly stained, the intense brown color of S. Typhimurium ΔfumABC colonies indicated high accumulation of glycogen (Fig. 3C). We next applied transmission electron microscopy (TEM) of ultrathin sections of S. Typhimurium WT (Fig. 3A) and ΔfumABC cells (Fig. 3B). Granular aggregates of low electron density were observed in the polar regions of S. Typhimurium ΔfumABC, but to a far lesser extent in WT cells. Accordingly, enzymatic quantification revealed 12-fold-increased glycogen content in S. Typhimurium ΔfumABC compared to WT (Fig. 3D). Complementation of S. Typhimurium ΔfumABC by plasmids harboring fumAC or fumB genes restored WT levels of glycogen (see Fig. S1A in the supplemental material). These data indicate that fumarate FIG 2 Deletion of fumarases leads to changes in carbon fluxes and amounts of metabolic enzymes. S. Typhimurium WT and ΔfumABC strains were grown aerobically in LB broth for 18.5 h at 37°C. For the proteomic approach, harvested cells were lysed and proteins were precipitated with 10% TCA. After trypsin digestion, the peptides were analyzed by quantitative LC-MS E (where E stands for elevated offset). For metabolomics analyses, harvested cells were disrupted and metabolites were extracted for GC-MS analysis. Heat map colors of oval symbols indicate relative changes in amounts of enzymes detected for ΔfumABC mutant compared to WT. Heat map colors of square symbols indicate relative changes in amounts of metabolites determined in ΔfumABC mutant compared to WT. Gray symbols indicate less than 2-fold or nonsignificant differences in enzyme or metabolite amounts. Quantitative data are shown for TCA cycle (A), glycolysis (B), pentose phosphate pathway (C), and glycogen synthesis (D). Data represent means from at least four or three biological replicates for the metabolomics or proteomics analyses, respectively. Statistical analyses were performed by Student's t test, and all data shown have significance differences between the two strains of P Ͻ 0.05 or lower. accumulation in S. Typhimurium ΔfumABC is a key factor for biasing the glycogen metabolism toward altered carbon fluxes and increased glycogen storage.
Deletion of glycogen synthase GlgA decreases amounts of G6P, F6P, and S7P in Salmonella WT and ⌬fumABC strains. To further investigate the connection of glycogen biosynthesis and fumarate accumulation, we blocked glycogen synthesis by deletion of glgA, which encodes the glycogen synthase, in the ΔfumABC mutant, resulting in the S. Typhimurium ΔfumABC ΔglgA double mutant. We verified the loss of glycogen production in the glgA-deficient strain with potassium iodine staining (Fig. 3C) and TEM analyses (Fig. S2) as before and were able to restore the original phenotype by complementation with a plasmid harboring glgA (Fig. S1B).
Subsequently, we performed quantitative comparative proteomics and metabolomics of S. Typhimurium ΔfumABC ΔglgA and compared the obtained profiles with those of S. Typhimurium ΔfumABC ( Fig. 4 and Tables S1 and S3). Deletion of glycogen synthase did not affect amounts of metabolic enzymes in glycolysis, PPP, and TCA cycle ΔfumABC (B) strains were grown aerobically for 18.5 h at 37°C in LB broth. Cells were fixed, dehydrated, and resin embedded, and ultrathin sections were prepared for TEM. Massive accumulations of polymers in the polar regions of ΔfumABC cells were observed frequently. Bars, 1 m (overview), 500 nm (detail). (C) S. Typhimurium WT, ΔglgA, ΔfumABC, and ΔfumABC ΔglgA strains were grown on LB agar plates for 18.5 h at 37°C. Potassium iodine staining was performed, and brownish color indicates intercalation of iodine with glycogen. (D) Quantification of glycogen contents of S. Typhimurium strains grown aerobically for 18.5 h in LB broth. Glycogen was degraded to glucose monomers using amyloglucosidase, and resulting glucose was phosphorylated to G6P. G6P was oxidized by G6P dehydrogenase in the presence of NAD, being reduced to NADH. Glucose concentrations were proportional to OD 340 . By subtraction of free glucose concentrations (sample without amyloglucosidase) from total glucose concentrations, glycogen amounts were quantified. Glycogen concentrations were normalized to WT (ϭ1), and error bars represent standard deviations from four biological replicates. n.d., not detected. Statistical analyses were performed by Student's t test, and significances are indicated as follows: ***, P Ͻ 0.001. but decreased the abundance of glucose-1-phosphate adenylyltransferase GlgC, an enzyme catalyzing the synthesis of ADP-D-glucose (ADPG). Metabolite analyses by gas chromatography-mass spectrometry (GC-MS) revealed strong decrease of G6P, F6P, and S7P if glgA is deleted (Fig. 4E). Furthermore, the amount of trehalose was increased by 30%, while amounts of maltose were 100-fold reduced in S. Typhimurium ΔfumABC ΔglgA.
We conclude that altered fluxes through glycolysis and PPP in a fumarase-deficient strain are induced by increased glycogen synthesis. Abrogation of storage compound synthesis by glgA knockout normalized metabolite levels, due to modified enzyme activities and regulative mechanisms, rather than altered protein amounts.
Fumarate-induced stringent response influences Salmonella physiology. The amount of stored glycogen is dependent on the abundance of synthesis enzymes (19), and glycogen synthesis in S. Typhimurium is mainly mediated by enzymes GlgA and GlgC (20). In E. coli, the main regulators for glgA and glgC transcription are the alarmones ppGpp and pppGpp [here referred to as (p)ppGpp] (21), which are induced during nutrient starvation by stringent response mediators RelA and SpoT. To elucidate whether S. Typhimurium ΔfumABC has an enhanced stringent response compared to S. Typhimurium WT, we made use of a dual-color reporter plasmid for relative quantification of wraB (ϭwrbA in E. coli) expression, which was recently used to determine the (p)ppGpp levels in E. coli (22). WrbA is known as a stationary-phase protein, whose . Data represent means from at least four or three biological replicates for the metabolomics or proteomics analyses, respectively. (E) The concentrations of metabolites glucose-6-phosphate (G6P), fructose-6-phosphate (F6P), and sedoheptulose-7-phosphate (S7P) were determined and normalized to WT (ϭ1). Statistical analyses were performed by Student's t test, and all data shown have significance differences between the two strains of P Ͻ 0.05 or lower: *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001. expression is dependent on (p)ppGpp levels (23). Whereas initial studies identified WrbA as tryptophan-repressor-binding protein (24), other groups characterized it as a flavodoxin-like protein (25). We introduced the P wraB ::sfGFP (superfolder green fluorescent protein) reporter plasmid into S. Typhimurium WT, ΔfumABC, and ΔfumABC ΔglgA strains and as negative control into S. Typhimurium ΔrelA ΔspoT, a mutant strain deficient in (p)ppGpp synthesis (26), and analyzed the expression by flow cytometry (Fig. 5). To test reporter performance, stationary LB broth cultures of S. Typhimurium WT were subcultured in defined PCN (phosphate, carbon, nitrogen) minimal medium (MOPS-buffered minimal medium without limitation of phosphate, carbon, and nitrogen) with or without supplementation by Casamino Acids (Fig. 5A). Indeed, WT grown without an additional source of amino acids showed a higher sfGFP signal intensity than S. Typhimurium WT grown with amino acid supplementation, indicating higher (p)ppGpp levels.
Next, we determined sfGFP signal intensities of S. Typhimurium WT, ΔfumABC, and ΔfumABC ΔglgA strains harboring the respective reporter plasmid cultured in LB broth ΔglgA strains were cultured o/n in LB broth, and RNA was extracted and used for cDNA synthesis and consecutive qPCR experiments. 16S rRNA expression levels were used for normalization. Depicted are the expression levels of glgA and glgC normalized to WT (ϭ1). Shown is one representative assay of three independent biological replicates, consisting each of three technical replicates. Statistical analyses were performed by Student's t test, and significances are indicated as follows: *, P Ͻ 0.05; **, P Ͻ 0.01; ***, P Ͻ 0.001. for 18.5 h as described before. Quantification of sfGFP intensity revealed higher values for S. Typhimurium ΔfumABC and ΔfumABC ΔglgA strains than for S. Typhimurium WT, whereas the negative-control S. Typhimurium ΔrelA ΔspoT exhibited the lowest signal intensities ( Fig. 5B and C). Additionally, transcript levels of glgA and glgC were determined (Fig. 5D). Strongly enhanced expression of glgA and glgC was detected for the ΔfumABC mutant compared to WT. For S. Typhimurium ΔfumABC ΔglgA, we detected only background signals for glgA but still highly increased expression levels of glgC compared to WT. In addition, glycogen accumulation in S. Typhimurium ΔfumABC was eliminated by further deletion of relA and spoT (Fig. S3). Thus, we propose that ΔfumABC enforces glycogen synthesis as a consequence of an early and strong stringent response, leading to high (p)ppGpp levels, which in turn raise the transcript and protein levels of GlgA and GlgC.
Altered amounts of chemotaxis proteins in fumarase-deficient S. Typhimurium lead to increased counterclockwise (CCW) flagellar rotation. Accumulation of (p)ppGpp can negatively affect motility, as recently described for E. coli (27). To explore this potential link, we analyzed proteomic data for modulation of chemotaxis and motility-related proteins (Fig. 6A). Decreased amounts of methyl-accepting chemotaxis proteins (MCP) and decreased abundance of CheY, CheZ, and CheW (2.14-to 3.86-fold) were detected in S. Typhimurium ΔfumABC compared to WT. In addition, CheB was found only in S. Typhimurium ΔfumABC. For S. Typhimurium ΔfumABC ΔglgA, a restoration of chemotaxis protein levels was detected for CheY. However, CheY abundance was still lower than in S. Typhimurium WT (Fig. 6B). The amount of CheY influences the number of switching events of flagellar rotation direction (28). Thus, S. Typhimurium ΔfumABC might show an altered swimming behavior, and we analyzed swim patterns of bacteria grown overnight (o/n) in rich medium (Fig. 7A). Counterclockwise (CCW) flagellar rotation bundles flagella and results in straight swimming, while clockwise (CW) rotation leads to tumbling (29). S. Typhimurium WT showed short swimming paths alternating with tumbling, whereas S. Typhimurium ΔfumABC exhibited highly prolonged swimming paths and reduced tumbling events. Furthermore, the number of motile bacteria was higher than for WT. The motility patterns of ΔfumABC and ΔfumABC ΔglgA strains were similar.
To further analyze flagellar switching from CCW to CW rotation, we performed flagellar rotation analyses of S. Typhimurium WT, ΔfumABC, and ΔfumABC ΔglgA strains grown in rich medium by microscopic inspection of single bacterial cells fixed by one flagellum to a polystyrene-coated coverslip (30) (Fig. 7B and C). We observed a statistically significant increase of CCW flagellar rotation for S. Typhimurium ΔfumABC. Whereas S. Typhimurium WT had an average proportion of CCW rotation of 33%, the ΔfumABC strain spent 78% of time in CCW flagellar rotation. Although S. Typhimurium ΔfumABC ΔglgA exhibited partly normalized amounts of the chemotaxis protein CheY, there was still an increased proportion of CCW flagellar rotation comparable to that of S. Typhimurium ΔfumABC. Furthermore, the swimming behavior was not altered by glgA deletion, indicating that the amount of CheY necessary for normalization of switching events was not achieved in S. Typhimurium ΔfumABC ΔglgA.
Thus, we conclude that fumarase deletion in S. Typhimurium leads to a downregulation of chemotaxis proteins and by this to enhanced CCW flagellar rotation.
The increased phagocytic uptake of fumarase-deficient S. Typhimurium is due to enhanced CCW flagellar rotation and partially depends on glycogen synthesis. Since bacterial motility can increase uptake of pathogenic bacteria by host cells (31)(32)(33)(34), we hypothesized that the observed enhanced uptake of fumABC mutant strains by RAW 264.7 macrophages (12) could be caused by increased CCW flagellar rotation. To test this hypothesis, we introduced additional deletions of chemotaxis gene cheY or cheZ in the mutant strain. Whereas cheY deletion strains are locked in CCW flagellar rotation, ΔcheZ mutant strains are mainly locked in the CW state (35). The combination of ΔcheY and ΔfumABC did not alter phagocytic uptake, while the combination of ΔcheZ and ΔfumABC showed uptake only 3.15-fold higher than WT (Fig. 8).
Deletion of glycogen synthase partially normalized CheY levels but not the duration of CCW flagellar rotation in S. Typhimurium ΔfumABC. Thus, we expected an increased phagocytic uptake of the ΔfumABC ΔglgA double mutant as well. However, phagocytosis of S. Typhimurium ΔfumABC ΔglgA was 6.9-fold increased over WT, but significantly lower than uptake of S. Typhimurium ΔfumABC (Fig. 8). Complementation by plasmid-borne glgA again increased levels of phagocytosis (Fig. S4B). The cheY deletion did not change phagocytic uptake of S. Typhimurium ΔfumABC ΔglgA, while phago-

FIG 8
Increased phagocytic uptake of fumarase deletion strains is dependent on CCW flagellar rotation and glycogen synthesis. S. Typhimurium WT and various mutant strains as indicated were grown for 18.5 h in LB broth and used to infect RAW 264.7 macrophages at an MOI of 1. Infection was synchronized by centrifugation for 5 min, followed by incubation for 25 min at 37°C. Next, noninternalized bacteria were removed by washing and treatment with gentamicin at 100 g/ml for 1 h and 10 g/ml for the remaining time. Cells were lysed 2 h after infection by addition of 0.1% Triton X-100, and lysates were plated onto Mueller-Hinton agar plates to determine the CFU of internalized bacteria. Phagocytosis rates were determined as percentage of internalized bacteria relative to the used inoculum. Values were normalized to WT (ϭ1), and means and standard deviations from three technical replicates are shown. Statistical analyses were performed by Student's t test, and significances are indicated as follows: ***, P Ͻ 0.001; n.s., not significant. cytosis of S. Typhimurium ΔfumABC ΔglgA ΔcheZ was reduced (Fig. 8). These results demonstrate that high phagocytosis of fumarase deletion strains is due to CCW bias of flagellar rotation and is partially dependent on glycogen synthesis.
In order to elucidate which factors reduce the phagocytic uptake of S. Typhimurium ΔfumABC ΔglgA compared to ΔfumABC, we analyzed further characteristics of swimming behavior of both mutant strains. The frequency of switching events within 1,000 frames (17.71 s) was determined, and a switching event occurred if the flagellar rotation direction changed from CW to CCW or vice versa (Fig. S5). Compared to WT (median ϭ 31 events), the switching rate was reduced in S. Typhimurium ΔfumABC (median ϭ 20 events) but not in a significant manner (Fig. 9A). Even stronger reduction of switching events was determined for S. Typhimurium ΔfumABC ΔglgA (median ϭ 10 events). Additionally the number of pauses, defined as rotation of the bacterial body of less than 5°/frame, was analyzed (Fig. 9B). Comparable to the number of switching events, WT had the highest number of pauses (median ϭ 170.5), followed by S. Typhimurium ΔfumABC (151.5), but again there is no statistically significant difference between these two strains. A stronger reduction was observed for S. Typhimurium ΔfumABC ΔglgA; here, the number of pauses within 1,000 frames was reduced to 89.
In conclusion, S. Typhimurium ΔfumABC showed strongly increased CCW bias and fewer switching events than S. Typhimurium WT. These factors influence the interaction with host cells, such as increasing phagocytic uptake by macrophages. Further deletion of glgA in the ΔfumABC strain did not reduce time spent in CCW flagellar rotation but decreased the number of switching events, resulting in reduced phagocytic uptake.

DISCUSSION
Our work investigated the effect of perturbation of the TCA cycle of S. Typhimurium on basic cellular functions and pathometabolism. Enzymes harboring iron-sulfur clusters, i.e., fumarases and aconitases, are of particular sensitivity toward ROS attacks, which are a consequence of antibiotic treatment or immune responses in phagocytic host cells. We were interested in physiological changes and aberrant virulence properties upon ROS-dependent inactivation of metabolic enzymes and focused on a mutant strain defective in all fumarase isoforms. By deploying proteomics and metabolomics, we determined that defects in fumarases biased carbon fluxes toward enhanced glycogen synthesis, likely due to elevated (p)ppGpp levels in the mutant strain. Furthermore, proteomics revealed reduced abundances of chemotaxis proteins in S. Typhimurium ΔfumABC. Analysis of flagellar rotation and swim patterns showed increased CCW bias, raising the contact frequency of S. Typhimurium and host cells, thus leading to enhanced phagocytic uptake by macrophages. Deletion of glycogen synthase GlgA relieved the metabolic perturbations but not the aberrant motility phenotype. However, phagocytic uptake was decreased.
Our metabolomics data demonstrated higher accumulation of G6P, F6P, and S7P for S. Typhimurium ΔfumABC than for WT, and deletion of glycogen synthase again normalized the metabolic flux through glycolysis and PPP (Fig. 4). Thus, the increased concentrations of these metabolites were caused by enhanced glycogen synthesis in S. Typhimurium ΔfumABC due to changes in carbon fluxes. Accumulation of the respective metabolites was also observed for E. coli with truncated CsrA, the main component of the carbon storage system (36,37). As csrA deletion strains accumulate large amounts of glycogen as well, our results indicate that the observations obtained for E. coli ΔcsrA are also consequences of the massive remodeling of the carbon metabolism due to enhanced glycogen synthesis. However, a role of CsrA was reported not only in the context of posttranscriptional regulation of carbon metabolism, and in particular glycogen metabolism, but also for chemotaxis proteins, flagellar subunits, and proteins involved in virulence functions (38,39). Thus, the involvement of CsrA as an inducer of phenotypes of S. Typhimurium ΔfumABC is conceivable. While glycogen accumulation indicates very low levels of CsrA, mutant strains with truncated CsrA showed increased levels of Pgm and reduced levels of especially PfkA in E. coli (36), observations which are contradictory to our results. However, most studies on csrA mutant strains were performed with bacteria grown in minimal medium or at early growth phases (36). Thus, we cannot exclude a role of CsrA in the enhancement of glycogen synthesis for S. Typhimurium ΔfumABC, yet we do not expect CsrA to be the sole regulating factor.
In contrast, (p)ppGpp was shown to be the most important factor influencing glycogen synthesis, at least in E. coli (40). (p)ppGpp is known to enhance the expression of glgA and glgC but not glgB during stringent response (19). Indeed, we detected GlgA and GlgC only in S. Typhimurium ΔfumABC (Fig. 2D). Using a dual-color reporter system with P wraB controlling sfGFP expression, we detected increased fluorescence intensities for S. Typhimurium ΔfumABC and ΔfumABC ΔglgA compared to WT. The promoter of wrbA was used in several studies for the indirect quantification of (p)ppGpp (22,41). Furthermore, by proteomic analyses we detected increased abundances of WrbA in the ΔfumABC strain (3.78-fold; see Table S2 in the supplemental material), supporting our results obtained by flow cytometry. Taken together, we hypothesize that a fumarase deletion strain increases glgA and glgC expression in a (p)ppGpp-dependent manner.
The main inducing factors for (p)ppGpp synthesis by RelA and SpoT are amino acid and carbon source limitations (23). Using LB broth, amino acid limitations are unlikely at early growth phase. Several studies showed that increase of (p)ppGpp levels can be induced by diauxic shifts, for example, from glucose to succinate (42). Considering the high accumulation of fumarate, the use of the TCA cycle intermediate as carbon source is conceivable. An indicator for this model is the slightly increased abundance of aspartase AspA in S. Typhimurium ΔfumABC (1.5-fold), catalyzing the reversible reaction from fumarate and ammonia to aspartate (43). Indeed, metabolomic data showed a 10-fold-larger amount of aspartate in the mutant strain, which serves as the substrate for a range of metabolic pathways (44). Furthermore, two studies indicated that high fumarate accumulation led to use of fumarate as an alternative electron acceptor, despite the presence of oxygen (15,45). However, our proteomic data gave no hints for fumarate respiration (i.e., fumarate reductase FrdABCD) in the mutant strain but rather indicated utilization of fumarate as a carbon source. Fumarate metabolism possibly leads to a physiological situation similar to exponential-to stationary-phase transition and therefore increased (p)ppGpp levels, as discussed for E. coli (15,22).
Absence of fumarases led to enhanced CCW flagellar rotation and a prolonged phase of running movement, resulting in increased uptake by RAW 264.7 macrophages (Fig. 8). The impact of CCW flagellar rotation during the infection process was discussed in several prior publications (31)(32)(33). In these studies, CCW flagellar rotation and the resulting smooth swimming phenotype were linked to enhanced frequencies of bacterial contact with host cells, prolonged duration of adhesion, and increased numbers of phagocytic uptake events. Further deletion of glgA in S. Typhimurium ΔfumABC partly restored CheY levels, and we observed reduced uptake of S. Typhimurium ΔfumABC ΔglgA by macrophages. As we determined a strongly decreased number of switching events for the glgA-deficient strain but high frequency of phases of CCW flagellar rotation, the logical consequence is that duration of phases of CW flagellar rotation after switching events is longer for S. Typhimurium ΔfumABC ΔglgA than for the ΔfumABC strain. This effect might be accompanied by the reduced number of pause events observed for S. Typhimurium ΔfumABC ΔglgA in comparison to ΔfumABC and WT strains and could lead to changes in frequency or duration of contacts between S. Typhimurium and host cells.
To conclude, our results demonstrate that accumulation of fumarate due to fumarase deletion leads to induction of glycogen synthesis by enhanced (p)ppGpp concentrations (Fig. 10). This might be triggered by utilization of fumarate as carbon source, causing an exponential-to stationary-phase transition-like physiological state during early stationary growth phase. Additionally, we revealed that the increased phagocytic uptake of the fumarase deletion strain is caused by enhanced CCW flagellar rotation, which is the consequence of reduced CheY abundance. Further deletion of glgA normalized metabolic fluxes and restored abundance of the chemotaxis protein in part but did not change CCW bias of flagellar rotation. However, glgA deletion led to reduced phagocytic uptake by RAW 264.7 macrophages, possibly due to prolonged periods of CW flagellar rotation. Our work demonstrates that perturbations of the carbon fluxes in the TCA cycle lead to dramatic changes in S. Typhimurium physiology and affect the interaction of this pathogen with host cells.

MATERIALS AND METHODS
Bacterial strains. Salmonella enterica serovar Typhimurium NCTC 12023 was used as the wild-type strain (WT), and isogenic mutant strains were constructed by Red-mediated mutagenesis (Table 1) (46). Primers and plasmids required for mutagenesis, removal of resistance cassettes, and checking for the correct insertion are listed in Table 2 and in Table S4 in the supplemental material. Transfer of mutant alleles into a fresh strain background or for combination with other mutations occurred via P22 transduction. Both methods are described in the work of Popp et al. (47).
Construction of plasmids. For generation of p3752 and p3756, wild-type promoters and coding sequences of fumAC and fumB were amplified with primers listed in Table S4. After digestion with NotI and XhoI or ApaI and XhoI, respectively, the gene products were ligated into the low-copy-number plasmid pWSK30 and transformed in E. coli DH5␣. Positive clones were confirmed with primers listed in Table S4. The plasmids were isolated and transformed in the ΔfumAC or ΔfumB deletion strain.
For construction of p4763, the promoter and sequence of glgBXCAP as well as the vector pWSK29 were amplified by PCR using primers listed in Table S4. The obtained PCR fragments were assembled by Gibson assembly according to the manufacturer's protocol (New England BioLabs [NEB]). Sequenceconfirmed plasmids were transformed in the ΔfumABC ΔglgA deletion strain.
Generation of the reporter plasmid p5330 was performed as described previously (48). Briefly, plasmid p4889 (P EM7 ::DsRed P uhpT ::sfGFP) was used as vector. The uhpT promoter was replaced by the promoter fragment of wraB by Gibson assembly of fragments generated by PCR. Primers for fragment generation are listed in Table S4. Sequence-confirmed plasmids were transformed in S. Typhimurium WT, ΔfumABC, ΔfumABC ΔglgA, and ΔrelA ΔspoT strains.
GC-MS sample preparation and measurement. Culture of strains and cell harvest occurred as described in the work of Noster et al. (12). In short, each strain was cultured for 18.5 h at 37°C in 25 ml LB broth with agitation at 180 rpm. For measurements of metabolites in bacterial cells, 5 ml of cultures was transferred onto Durapore polyvinylidene difluoride (PVDF) filter membranes (Merck, Darmstadt, Germany) with a pore size of 0.45 m by suction. After washing with PBS, cells were scraped from the filter into 1 ml of fresh PBS, pelleted, and shock-frozen in liquid nitrogen. Afterward, samples were freeze-dried and their dry weights were determined. Metabolome analysis of the TCA cycle mutant strains was performed by GC-MS using protocols according to the work of Plassmeier et al. (49) and Noster et al. (12). In short, for metabolite extraction 1 ml 80% methanol containing 10 M ribitol (RI; internal standard) was added to dried samples, and for cell disruption, 500 mg acid-washed glass beads (Sigma-Aldrich, USA) and a homogenizer (Precellys; Peqlab) were used. After centrifugation, supernatants were evaporated in a nitrogen stream. For derivatization, 50 l of 20 mg/ml methoxylamine hydrochloride in pyridine and N-methyl-N-(trimethylsilyl)-trifluoroacetamide was added successively to each sample and incubated with constant stirring at 37°C for 90 min or 30 min, respectively. RI standard was added and incubated for a further 5 min. Samples were centrifuged, and supernatants were used for GC-MS measurement using a TraceGC gas chromatograph equipped with a PolarisQ ion trap and an AS1000 autosampler (Thermo Finnigan, Dreieich, Germany) according to the work of Plassmeier et al. (49). Metabolite quantities were normalized to ribitol and dry weights of used samples as described in the work of Plassmeier et al. (49). Mean relative pool size changes of the mutant strains compared to WT were calculated, and only those data with an error probability (Student's t test) of less than 0.05 were used for further interpretation. Proteome profiling by nano LC-MS measurement. Bacteria were cultured as described for the metabolite profiling. Sample preparation and liquid chromatography (LC)-MS measurement were performed according to the work of Noster et al. (12). In short, cells from 50 ml overnight (o/n) culture were pelleted, suspended, and washed twice with PBS. Pelleted bacteria were resuspended in lysis buffer (50 mM Tris, pH 8.5, 1% SDS, protease inhibitor). Cell disruption occurred with zirconia-silica beads and a cell homogenizer. Cell debris was removed by centrifugation, and proteins were precipitated with TCA. Protein pellets were washed with acetone, dried, and used for the following sample preparation, proteomic digestion, and nano LC-MS measurement as described in the work of Noster et al. (12).
Gentamicin protection assay. Culture and infection of RAW 264.7 macrophages were performed as described in the work of Popp et al. (47). Briefly, RAW 264.7 macrophages were infected with S.   Qualitative and quantitative determination of glycogen content. Qualitative determination of glycogen contents of bacterial cultures occurred as described in the work of Govons et al. (18). Bacterial cultures were streaked on LB agar plates and incubated o/n at 37°C. Ten milliliters of Lugol's iodine solution (Roth) was added to the plate and incubated 1 min at room temperature (RT). The iodine solution was discarded, and the plates were photographed immediately.
Quantification of glycogen contents occurred according to the protocol of Fung et al. (50). Of each strain, cells of 5 ml o/n culture were pelleted by centrifugation (13,000 ϫ g, 10 min, 4°C), resuspended in 50 mM Tris-acetate-EDTA (TAE) buffer, and pelleted again. Cells were resuspended in 1.25 ml sodium acetate buffer (200 mM, pH 4.6), and the suspension was added to 500 mg glass beads and disrupted by three cycles, each of 1 min with maximal speed, using a Vortex Genie 2, equipped with an attachment for microtubes (Scientific Industries). After centrifugation, supernatants were incubated for 20 min at 80°C for denaturation of endogenous enzymes. For each strain, 60 l lysate was incubated with 6 l amyloglucosidase (200 U/ml; Sigma-Aldrich) (quantification of glucose stored as glycogen and free glucose) or 6 l water (quantification of free glucose), respectively. After incubation for 30 min at 50°C, 50 l of each sample was transferred into a 96-well plate in technical duplicates. Two hundred fifty microliters HK reagent (Sigma-Aldrich) was added to each sample, and OD 340 was determined in 10-min intervals for 1 h. A standard curve with different dilutions of a glucose solution was used for extrapolation of the determined data. For relative quantification of the glycogen amount, maximal values obtained for free glucose were subtracted from maximal values obtained from free glucose and glycogen and normalized to the OD 600 of the bacterial culture.
TEM analysis. For TEM analyses of bacteria, S. Typhimurium was grown o/n at 37°C in LB broth with aeration. Cells were harvested for 2 min at 1,250 ϫ g. The pellet was resuspended in buffer (0.2 M HEPES, pH 7.4, 5 mM CaCl 2 ), and bacteria were fixed by addition of glutaraldehyde (Electron Microscopy Sciences) in buffer to a final concentration of 2.5% for 1 h at 37°C. After fixation, bacteria were washed several times in buffer and harvested for 5 min at 625 ϫ g. The pellet was gently resuspended in liquid 2% low-melting-point (LMP) agarose prewarmed to 37°C in buffer and incubated for 10 min at 37°C. Bacteria in agarose were repelleted for 1 min at 1,250 ϫ g and cooled down to 4°C until agarose was solid. The agarose block containing the bacteria was cut into small cubes (maximum, 1 mm 3 ), and cubes were postfixed with 2% osmium tetroxide (Electron Microscopy Sciences) in buffer containing 1.5% potassium ferricyanide (Sigma) and 0.1% ruthenium red (AppliChem) for 1.5 h at 4°C in the dark. After several washing steps, bacteria were dehydrated in a cold graded ethanol series and finally rinsed in anhydrous ethanol at RT twice. The agarose cubes were flat-embedded in Epon 812 (Serva). Serial 70-nm sections were generated with an ultramicrotome (Leica EM UC6) and collected on Formvar-coated EM copper grids. After staining with uranyl acetate and lead citrate, bacteria were observed with a TEM (Zeiss EFTEM 902 A), operated at 80 keV and equipped with a 2K wide-angle slow-scan charge-coupled device (CCD) camera (TRS, Moorenweis, Germany). Images were taken with the software ImageSP (TRS Image SysProg, Moorenweis, Germany). Flagellar rotation analysis. Flagellar rotation was determined as illustrated in Fig. S5. Bacteria were cultured for 18.5 h in LB, diluted 1:100 in PBS, and subjected to shear force by pressing the suspension eight times through a syringe equipped with a 24-gauge cannula. Fifteen microliters of sample was placed onto a microscope slide and covered with a coverslip spin-coated with polystyrene, on which three small drops of vacuum grease were spotted to achieve an optimal distance allowing free movement of S. Typhimurium. Sealing the cover slip with a 1:1:1 mixture of Vaseline, lanolin, and paraffin avoided suction. Rotating cells, bound with their flagellar filaments to the coverslip, were selected, and rotation direction was recorded using the Axio Observer microscope with an AxioCam CCD camera (Zeiss) for periods of 18 s (using a 100ϫ alpha-Plan-Apochromat objective, 1.6ϫ Optovar, 10 ms exposure, frame rate of 57/s). After image processing with Fiji (conversion to 8-bit format; background subtraction: rolling ball radius 20 pixels; cropping down to rotation disk; contrast enhancement: saturated pixels 0.0%, normalized; downsizing: width 10 pixels, constraint of aspect ratio, averaged, interpolated; Gaussian Blur: sigma 0.5; contrast enhancement), rotation analyses were performed using the tool SpinningBug Tracker (user-written software available from the authors upon request, Matlab 7.17 [R2012a]). By detection of the angle between the rotating bacteria and a reference axis, the rotation direction was calculated. CCW rotation of the bacterial body has to be interpreted as CW rotation of the flagellum and vice versa. Rotations of less than 5°/frame were defined as pause. Bacteria rotating with speeds of Ͼ180°/frame were excluded, due to limited time-resolution. Switching events were defined as changes from CW to CCW rotation and vice versa.
Swimming path analysis. Bacteria were cultured for 18.5 h with aeration in LB and diluted 1:20 in PBS. The assembly of microcopy slide, sample, and coverslip was similar to that described for the flagellar rotation analysis, but without prior coating of the coverslip with polystyrene. The swimming bacteria were recorded for 100 frames (14 frames/s) using a 40ϫ LD Plan-Neofluar objective, 1.6 Optovar, 10 ms exposure. Visualization of swimming paths was performed with ImageJ, using the plug-in MTrackJ with time step size 5 frames, snap range 25 ϫ 25 pixels (51).
qPCR. For RNA preparation by the hot-phenol method, bacteria were cultured for 18.5 h in LB with aeration. Bacteria at a number of 1.2 ϫ 10 9 were pelleted, treated with stop solution (95% ethyl alcohol [EtOH], 5% phenol saturated with 0.1 M citrate buffer, pH 4.3) (Sigma-Aldrich), and snap-frozen in liquid nitrogen. All subsequent steps were conducted as described in detail in the work of Noster et al. (48) according to protocols from Mattatall and Sanderson (52) and Sittka et al. (53). For cDNA synthesis, the RevertAid first-strand cDNA synthesis kit (ThermoFisher) was used, applying 1 g RNA and random hexamer primers. qPCR was performed using the Maxima SYBR green/fluorescein qPCR master mix (ThermoFisher) and an iCycler equipped with the MyiQ module (Bio-Rad). Data were normalized to expression levels of a housekeeping gene (16S rRNA) and calculated relative to primer efficiencies, which were determined before using serial dilutions of cDNA. Used oligonucleotides are listed in Table S4.
Flow cytometry analyses. S. Typhimurium strains harboring p5330 were grown in LB broth at 37°C with aeration for 18.5 h, diluted 1:1,000 in FACS buffer (1% BSA in PBS, 1 mM EDTA, 20 mM HEPES, pH 7.2, 50 mM NH 4 Cl), and subjected to flow cytometry on an Attune NxT instrument (Thermo Fisher Scientific). The intensity of the sfGFP fluorescence per gated S. Typhimurium cell of 10,000 bacteria with constitutive red fluorescence was recorded, and x medians for sfGFP intensities were calculated.