Robustness and Plasticity of Metabolic Pathway Flux among Uropathogenic Isolates of Pseudomonas aeruginosa

Pseudomonas aeruginosa is a human pathogen that frequently causes urinary tract and catheter-associated urinary tract infections. Here, using 13C-metabolic flux analysis, we conducted quantitative analysis of metabolic fluxes in the model strain P. aeruginosa PAO1 and 17 clinical isolates. All P. aeruginosa strains catabolized glucose through the Entner-Doudoroff pathway with fully respiratory metabolism and no overflow. Together with other NADPH supplying reactions, this high-flux pathway provided by far more NADPH than needed for anabolism: a benefit for the pathogen to counteract oxidative stress imposed by the host. P. aeruginosa recruited the pentose phosphate pathway exclusively for biosynthesis. In contrast to glycolytic metabolism, which was conserved among all isolates, the flux through pyruvate metabolism, the tricarboxylic acid cycle, and the glyoxylate shunt was highly variable, likely caused by adaptive processes in individual strains during infection. This aspect of metabolism was niche-specific with respect to the corresponding flux because strains isolated from the urinary tract clustered separately from those originating from catheter-associated infections. Interestingly, most glucose-grown strains exhibited significant flux through the glyoxylate shunt. Projection into the theoretical flux space, which was computed using elementary flux-mode analysis, indicated that P. aeruginosa metabolism is optimized for efficient growth and exhibits significant potential for increasing NADPH supply to drive oxidative stress response.


Introduction
Pseudomonas aeruginosa is a metabolically versatile bacterium that resides in a wide range of biotic and abiotic habitats and is a human pathogen that causes numerous acute and opportunistic infections [1]. The clinical spectrum of P. aeruginosa infections includes wound and urinary tract infections, meningitis, and necrotizing pneumonia [2]. In particular, urinary tract infections and catheter-associated urinary tract infections are the most common bacterial infections in clinical practice [3,4] and pose a severe health threat with more than one million hospitalizations annually [5].
Research on P. aeruginosa has focused its virulence [1], resistance [6], and adaptation [7] as well as therapeutic strategies [8]. Microevolution through resistance-mediating mutations in the bacterium's resistome involves a large subset of its genetic repertoire and a complex network of metabolic pathways that mediate adaptive resistance and adaptive metabolism [9][10][11]. Therefore, a systems-level understanding of the network that drives the pathogenesis of P. aeruginosa is important for devising specific control strategies [1]. In particular, 13 C-metabolic flux analysis (fluxomics) detects common and specific pathways employed by pathogens and identifies candidate pathways as targets for therapy [12,13].
This network-wide approach provides information on the activities of central enzymes and pathways most directly linked to phenotype [14]. However, to our knowledge, such analyses of P. aeruginosa have not been published. Here, we investigated the laboratory strain P. aeruginosa PAO1 at the level of carbon fluxes by using 13 C-metabolic flux analysis that combined isotopic tracer experiments with mass spectrometric labeling analysis and stoichiometric and isotopomer balancing for flux calculation [15]. This was extended to a collection of 17 P. aeruginosa clinical isolates from patients with urinary tract infections and catheterassociated urinary tract infections. These strains are genetically diverse, differ from strains that cause chronic lung infections in patients with cystic fibrosis, and exhibit heterogeneous production of virulence factors in vitro [16].

Culture media
Starter cultures were grown in complex LB medium containing 5 g?L 21 yeast extract (Becton, Dickinson and Company, Sparks, MD, USA), 10 g?L 21 peptone (Becton, Dickinson and Company) and 10 g?L 21 NaCl. For the second and main cultures, a minimal medium developed during the study ensured balanced growth of all strains, an important prerequisite for the 13 C-flux approach [15]. The osmolality of this medium was 500 mosmol?kg 21 , which reflects that of human urine [18]. The medium contained the following (per liter): 2.5 g glucose, 13.0 g KH 2 PO 4

Culture conditions
Cultures were shaken at 1,050 rpm and 37uC in 10 mL deepwell plates (HJ Bioanalytik, Mönchengladbach, Germany) with a working volume of 1.4 mL (Titramax 1000, Heidolph Instruments, Schwabach, Germany). Starter cultures were prepared by inoculating LB medium with 20 mL of a glycerol stock. After 6 h of incubation, 50 mL of cell suspension was transferred to a second culture in minimal medium. Subsequently, exponentially growing cells (50 mL) were used as inoculum for main cultures, and strains were grown in 30 mL of minimal medium in 300 mL baffled shake flasks. These were incubated at 37uC with shaking at 200 rpm on an orbital shaker (Aquatron, Infors AG, Switzerland). Dissolved oxygen was monitored online using shake flasks with integrated sensor spots (PreSens SFR, PreSens Precision Sensing GmbH, Regensburg, Germany). The sensor device was installed in the orbital shaker as described previously [19]. In cultures incubated with 13 C-tracer, the inoculum level was always kept below 1% of the final sampled cell concentration to exclude potential interference of a differently labeled inoculum on subsequent calculation of flux [15].

Quantification of cell concentration
Cell growth was monitored spectrophotometrically at 578 nm (OD 578 ) (Libra S11, Biochrome, Cambridge, UK). Cell dry weight (CDW) was determined as follows: Dried plastic tubes were filled with 20 mL cell suspension, and cells were harvested by centrifugation (25,8006 g, 4uC, 10 min); subsequently, pellets were washed twice with deionized water, and then dried to a constant weight.

Quantification of substrates and products
The concentration of glucose was determined from filtered supernatants (Minisart, 0.2 mm, Sartorius, Göttingen, Germany) using a YSI 2700 Select biochemical analyzer (Kreienbaum, Langenfeld, Germany). The analysis of organic acids (oxalic acid, citric acid, a-ketoglutaric acid, gluconic acid, glyoxylic acid, pyruvic acid, succinic acid, glycolic acid, lactic acid, fumaric acid, acetic acid, propanoic acid, and butyric acid) and amino acids in the culture supernatant was conducted using high-pressure liquid chromatography [20]. The limit of detection was below 10 mM for organic acids and below 1 mM for amino acids, respectively.
Mass isotopomer labeling analysis of proteinogenic amino acids Cells (0.5 mg CDW) were harvested in the mid-exponential growth phase by centrifugation (3 min, 16,0006 g, 4uC, Biofuge Fresco, Heraeus, Hanau, Germany), washed twice with deionized water, and incubated with 30 mL 6 M HCl for 24 h at 100uC. The obtained hydrolysate was filtered (0.2 mm, Ultrafree MC, Millipore, Bedford, MA, USA) and then lyophilized. The proteinogenic amino acids in the lyophilisate were dissolved in 50 mL N, N-dimethylformamide containing 1% (v/v) pyridine and then derivatized at 80uC for 30 min with 50 mL N-methyl-tbutyldimethylsilyl-trifluoroacetamide (Macherey-Nagel, Düren, Germany). Mass isotopomer distributions of selected ion clusters of the analytes were then quantified using gas chromatographymass spectrometry (GC-MS) (HP6890, M 5973, Agilent Technologies, Waldbronn, Germany) as described earlier [23]. All samples were first measured in scan mode to exclude isobaric overlay, and were analyzed in triplicate by selective ion monitoring to determine relative fractions of the mass isotopomers of interest [15].

Metabolic reaction network and flux calculation
A large-scale model of the central metabolism of P. aeruginosa was assembled according to the genome-scale model of P. aeruginosa PAO1 [24], KEGG database [25], and the Pseudomonas genome database [26]. The network compromised the Entner-Doudoroff pathway (EDP), incomplete Embden-Meyerhof-Parnas pathway lacking phosphofructokinase (EMPP), pentose phosphate pathway (PPP), gluconeogenesis, reactions of pyruvate carboxylase, PEP carboxylase, PEP carboxykinase, malic enzyme, tricarboxylic acid (TCA) cycle, glyoxylate pathway, and anabolic pathways generating biomass. The chemical composition of P. aeruginosa cells was adapted from that of closely related Pseudomonas species [27,28]. Because P. aeruginosa uniquely synthesizes the capsular polysaccharide alginate [29], the cellular content of this polymer was measured (Table S1, [30]), and the corresponding precursor demand was considered for all strains. The calculation of metabolic flux through the network was performed using OpenFLUX software [31]. For each strain,  [32]. Multiple flux estimations using statistically varied starting values for free flux parameters confirmed the identification of a global minimum. For all flux data, 95% confidence intervals were calculated using a Monte Carlo approach [33].

Elementary flux mode analysis
The metabolic network of P. aeruginosa was investigated using in silico pathway analysis involving the computation of elementary flux modes. The network topology agreed with that of the related species P. putida [34]. The energy demand for polymerization of building blocks and biomass assembly was based on the genome model of P. aeruginosa PAO1 [24]. Elementary flux modes were calculated as described previously [35]. Evaluation of the modes was carried out using Excel (MS Office, Windows, 2007) and provided relative pathway flux and yields for each of the modes [34,35].

Statistical analysis
Hierarchical clustering analysis (HCA) and principal component analysis (PCA) were conducted using MATLAB (Version R2011b, The MathWorks). PCA was used to convert the set of 13 C-labeling data to a set of linearly uncorrelated variables, i.e., the principal components [36]. HCA was carried out using the complete linkage, the Euclidean distance, as a measure of proximity of experimental data to inspect strain similarity [37].

Results
Quantitative metabolism of P. aeruginosa PAO1 and the clinical isolates P. aeruginosa PAO1 consumed glucose from an early time point ( Figure S1), and its growth rate was approximately 0.91/h, corresponding to a doubling time of 45 min (Table 1). Biomass and CO 2 were considered to be the only products formed, because secreted by-products were not detected. Note that the strain grew fully aerobically as verified by on-line monitoring of the dissolved oxygen level ( Figure S2). Next, we analyzed P. aeruginosa isolates from urinary tract infections (8 strains) and catheter-associated urinary tract infections (9 strains). All individual growth profiles are presented in the Supporting Information ( Figure S3). The strains showed high variability in growth ( Table 1).
The efficiency for recruiting glucose for anabolism, i.e., the biomass yield, differed by more than a factor of 1.5. Strain MH15c produced the lowest biomass yield, which was approximately 65% of the value of the two most efficient strains, MH34c and MH29u, which yielded a biomass 0.53 g?g 21 . This approached the theoretical optimum (0.54 g?g 21 ), a value derived by in silico pathway simulation. Briefly, the corresponding simulation for the carbon core network of P. aeruginosa provided 13,138 elementary modes, each representing a theoretical metabolic state of the cell. Together, all modes span the feasible flux space of the organism and contain the optimum growth mode with maximum possible yield [35]. Differences between growth kinetics of the strains were even larger. Whereas certain strains grew rather slowly and The corresponding cultivation profiles for all strains are provided in Supporting Information ( Figure S3). Values indicate means and standard deviations of three biological replicates. By-products in the supernatants were not detected; concentrations were below the detection limit (1 mM for amino acids and 10 mM for organic acids). doi:10.1371/journal.pone.0088368.t001 exhibited doubling times of about 120 min, others duplicated within only 40 min. Similarly, specific uptake rates for glucose differed significantly.
Glucose catabolism by P. aeruginosa PAO1 P. aeruginosa PAO1 was grown on [1-13 C] glucose to estimate metabolic flux and exhibited metabolic and isotopic steady state growth (see Supporting Information). This fulfilled an important prerequisite for the 13 C-flux approach [15]. The simulated labeling data, corresponding to the optimal fit, agreed closely with experimental values (Table S2) and indicated high consistency of the flux values. Glucose was mainly consumed through the EDP (Figure 1). The flux through 6-phosphogluconate dehydrogenase, which catalyzes the first reaction of the oxidative Figure 1. In vivo carbon flux distribution in the central metabolism of P. aeruginosa PAO1. Flux is expressed as a molar percentage of the specific glucose uptake rate of 9.5 mmol g 21 h 21 . Open arrows indicate the flux toward biomass. For reversible reactions, the direction of the net flux is indicated by a dashed arrow. The errors given for each flux reflect the corresponding 90% confidence intervals. The full flux data set is presented in Supporting Information. Metabolic and isotopic steady states were ensured by constant stoichiometry, kinetics, and the constant 13 C-labeling patterns of recruited metabolites during cultivation (see Figure S1, Figure S5). The abbreviations are as follows: Entner-Doudoroff pathway (EDP), Embden-Meyerhof-Parnas pathway (EMPP), pentose phosphate pathway (PPP), glyoxylate (Gly) shunt, tricarboxylic acid (TCA) cycle, and pyruvate metabolism. doi:10.1371/journal.pone.0088368.g001 PPP, was low. The PPP exclusively served for anabolic purposes. Both, oxidative and non-oxidative PPP, contributed to supply of ribose 5-phosphate, erythrose 4-phosphate, and fructose 6phosphate for anabolism. EMPP activity was undetectable, likely caused by lack of cytoplasmic phosphofructokinase activity (,0.01 mU?mg 21 ). Accordingly, P. aeruginosa PAO1 catabolized glucose only through the EDP to subsequent glycolytic steps and further to pyruvate.
Pyruvate metabolism, anaplerosis, and the TCA cycle in P. aeruginosa PAO1 The activity of a functional PEP carboxylase (anaplerotic reaction from phosphoenolpyruvate to oxaloacetate) could not be confirmed. During growth on [1-13 C] glucose, the combination of 13 C enrichment in pyruvate and oxaloacetate can be used to discriminative between PEP and pyruvate carboxylase ( Figure 2). The weak signal generated by the singly labeled pyruvate mass isotopomer (M1), i.e., the low 13 C-enrichment in the molecule, together with a relatively high value for the corresponding M1 Figure 2. Resolution of anaplerotic flux in [1-13 C]-glucose-grown uropathogenic P. aeruginosa, P. aeruginosa PAO1, and P. putida. For exclusive use of the ED pathway for glucose catabolism, the contribution of PEP carboxylase resulted in low 13 C-enrichment of oxaloacetate (deduced from aspartate), whereas the 13 C accumulated in pyruvate (alanine), which creates an excess of its single-labeled (M1) mass isotopomer (green region). In contrast, the action of pyruvate carboxylase reduced and increased the levels of pyruvate and oxaloacetate (orange region), respectively. This allowed for discrimination between these enzymes according to the combined labeling of the two molecules. Prior to inspection, the 13 Clabelling data were corrected for the contribution of natural isotopes [32]. doi:10.1371/journal.pone.0088368.g002 isotopomer of oxaloacetate, matched the expected pattern for pyruvate carboxylase. In contrast, it differed significantly from possible combinations for pyruvate and oxaloacetate, which would result from PEP carboxylase. Thus, pyruvate was the source of anaplerotic oxaloacetate.
P. putida expresses pyruvate carboxylase [27] and shows the same pattern ( Figure 2). On flux level, decarboxylating phosphoenolpyruvate carboxykinase and malic enzyme returned carbon to the glycolytic pools ( Figure 1). Accordingly, malic enzyme powered the pyruvate shunt (malate conversion to pyruvate by malic enzyme and further to oxaloacetate by pyruvate carboxylase) [27]. The overall net flux toward the TCA cycle generated by the concerted action of the carboxylating and decarboxylating enzymes was rather low (14.8%). Therefore, this anaplerotic route alone was not sufficient to replenish the TCA cycle, which was continuously depleted by the anabolic requirement for its Flux is given as average flux of all strains and is expressed as a molar percentage of the average glucose uptake rate of all strains (8.6 mmol g 21 h 21 , calculated from the individual rates in Table 1). Open arrows indicate flux toward biomass. For reversible reactions, the direction of net flux is indicated by a dashed arrow. The errors given for each flux reflect the corresponding 90% confidence intervals. The full flux data sets are presented in Supporting Information. Metabolic and isotopic steady states are ensured by constant stoichiometry, kinetics, and the constant 13 C-labeling patterns of recruited metabolites during cultivation (see Figure S1, Figure S5). Abbreviations are as follows: Entner-Doudoroff pathway (EDP), Embden-Meyerhof-Parnas pathway (EMPP), pentose phosphate pathway (PPP), glyoxylate (Gly) shunt, and tricarboxylic acid (TCA) cycle. doi:10.1371/journal.pone.0088368.g004 Figure 5. In vivo carbon fluxes through central metabolic pathways of uropathogenic P. aeruginosa during growth on glucose. The data reflect the individual flux for each isolate. All fluxes are expressed as a molar percentage of the corresponding specific glucose uptake rate (Table 1). Statistical differences of flux between strains from urinary tract infections and from catheter-associated infections were assessed using a t test. Significant differences (**P,0.05) are indicated. Abbreviations are as follows: glucose 6-phosphate dehydrogenase (G6PDH), 6phosphogluconate dehydrogenase (6PGDH), 6-phosphogluconate dehydratase (EDD), isocitrate dehydrogenase (ICDH), isocitrate lyase (ICL), pyruvate kinase (PK), pyruvate dehydrogenase (PDH), malic enzyme (ME), pyruvate carboxylase (PC), and phosphoenolpyruvate carboxykinase (PEPCK). The full flux data sets for all strains are presented in the Supporting Information. doi:10.1371/journal.pone.0088368.g005 intermediates oxaloacetate (17%) and 2-oxoglutarate (10%). Note that the cells recruited the glyoxylate shunt as an anaplerotic pathway. At the level of isocitrate, approximately 20% of carbon was channeled into the shunt, matched by a relative flux of 12%. The activity of this pathway seemed surprising at first, because it is typically not required in glucose-grown cells. We therefore performed in-vitro measurements to detect isocitrate lyase, the key enzyme of this pathway. Indeed, isocitrate lyase was present in the cytoplasm (12461 mU? (mg protein) 21 ). Further studies showed that acetate as sole carbon source activated isocitrate lyase by approximately a factor of four ( Figure S4).

Metabolic analysis of clinical isolates using [1-13 C] glucose
The P. aeruginosa isolates grown on [1-13 C] glucose exhibited balanced growth and reached a metabolic steady state ( Figure S5). The 13 C-labeling patterns of proteinogenic amino acids (Table S2) were first explored by unsupervised statistical analysis because they were informative for directly discriminating between different types of metabolism and pathway use [38] and provided an initial qualitative overview of the strains. Briefly, the 13 C-data were analyzed using PCA ( Figure 3A) and HCA ( Figure 3B). The 18 strains clustered into three subgroups, designated as clusters a, b and c in the Euclidian tree. This revealed site-specific metabolism. Clusters b and c comprised only isolates from urinary tract infections. In contrast, all strains from catheter-related infections grouped in cluster a. The only strains assigned to cluster a from the urinary tract infections were MH26u and MH29u. The clustering resulted from differences in distinct amino acids and were therefore of metabolic origin. A prominent example is the high enrichment of label in threonine, aspartate, and glutamate, which was specific for the strains of cluster c ( Figure 3B). The three clusters were also identified using PCA ( Figure 3A), which described 98% of the labeling information.

Metabolic flux of P. aeruginosa isolates
The full set of flux analysis as well as all experimental and simulated labeling patterns, which reflect optimal fit, are presented in Table S2. Figure 4 provides an integrated view of pathway use. Briefly, flux through each reaction displays the average value of all strains, whereas the deviation indicates the corresponding variability. Flux through the initial metabolic pathways, i.e., EDP, PPP and EMPP, respectively, was conserved. This was inferred from the flux data of the single isolates ( Figure 5) where the corresponding pathways showed similar activity.
Only three catheter-associated strains (MH15c, MH37c, and MH57c) did not use 6-phosphogluconate dehydrogenase, whereas all other strains channeled carbon through this enzyme into the PPP. Using an in-vitro assay, phosphofructokinase activity was not detected in any isolate (,0.01 mU (mg protein) 21 ). All isolates, as well as PAO1, recruited pyruvate carboxylase but not PEP carboxylase ( Figure 2). Interestingly, flux varied downstream from the pyruvate node. This involved the TCA cycle and the reactions interconnecting the cycle with the glycolytic intermediates, i.e., the anaplerotic and gluconeogenetic reactions.
The most variable metabolic reactions were localized around the isocitrate node, where the isolates differed significantly in the flux partitioning between the glyoxylate shunt and the TCA cycle ( Figure 4). This is even more apparent when inspecting individual strains ( Figure 5). In certain isolates, the glyoxylate shunt was completely inactive, whereas others showed high flux (44%). As exemplified by MH34c, strains with an active shunt showed isocitrate lyase activity (see Figure S4). Similarly, the TCA cycle (17-112%) and other enzymes, positioned near the pyruvate node varied strongly in flux. Note that flux through pyruvate kinase and phosphoenolpyruvate carboxykinase was significantly different between the two clinical subgroups ( Figure 5). The catheterassociated isolates carried higher flux through these reactions (P,0.05) compared with strains isolated from patients with urinary tract infections.

Discussion
The present study describes the analysis of metabolic fluxes in P. aeruginosa PAO1 as well as uropathogenic isolates and provides novel insights into function and regulation of carbon core metabolism of this important pathogen. We show here that P. aeruginosa isolates catabolize glucose through the EDP with fully respiratory metabolism and without overflow ( Figure 2, Figure 4, Figure 5). They further recruit the oxidative and non-oxidative PPP exclusively for biosynthesis, but do not exhibit a functional EMPP. All strains utilize pyruvate carboxylase but not PEP carboxylase (Figure 2), and the glyoxylate shunt operates as anaplerotic pathway. Hierarchical clustering of the strains according to their flux reveals a site-specific metabolism among the isolates, which might indicate that P. aeruginosa differently adapts to its environment: the urinary tract and the surface of catheters infections, respectively. The isolates differ strongly in glucose uptake rate and growth efficiency. Faster uptake of nutrients and their more efficient conversion into biomass might provide an advantage to compete and persist during infection. However, one should be cautious with this particular interpretation due to potential metabolic differences, caused by different levels of oxygen in our aerobic experimental setup and the oxygen limited infection environments, in which P. aeruginosa thrives [39].
The exclusive use of the EDP as a glycolytic strategy is conserved among P. aeruginosa and other pseudomonads The predominant use of the EDP is identical to that of other glucose-grown pseudomonads, including P. putida, P. fluorescens, and P. denitrificans previously studied at the flux level (Table 2). This finding is attributed to the lack of phosphofructokinase, an essential enzyme of the alternative glycolytic EMPP [27]. However, this glycolytic strategy is uncommon among prokaryotes, and only 12% of bacteria rely solely on the EDP [40], whereas the more energy-efficient EMPP is nearly ubiquitous [41].  However, the EDP mediates superior resistance to oxidative stress [42]. In this study, the expression of a functional EMPP in P. putida, which forced the organism to redirect flux from the natural ED route, significantly decreased its robust response to oxidative stress [42], because the EMP pathway does not yield NADPH required by various antioxidant defense mechanisms [43][44][45][46]. In contrast, the ED pathway provides NADPH [40]. Accordingly, bacteria that reside in the environment, such as P. putida, favor the ED pathway to better cope with naturally occurring oxidative stress [42]. We conclude, therefore, that the high ED flux of P. aeruginosa provides a major benefit for this pathogen, which must counteract oxidative stress imposed by the host as a response to infection. Moreover, P. aeruginosa senses the redox state and switches its metabolism as a reaction to the host immune defense and other oxidative stress conditions, which involves the activation of catalase, thiol reductase, and superoxide dismutase, all of which use NADPH as a cofactor [47].
Anaplerotic and TCA cycle flux of uropathogenic P. aeruginosa reveal high plasticity Microevolution of P. aeruginosa during chronic pulmonary infection creates a metabolic shift to support adaptation to the host environment [7,48]. Transcriptome profiling has shown that this predominantly involves its carbon core machinery [11]. Briefly, adaptive evolution affected only specific pathways such as pyruvate metabolism, the TCA cycle, and the linked pathways of anaplerosis and gluconeogenesis. In contrast, the reactions that metabolize glucose to pyruvate via the EDP remained unchanged.
The integration of observations of distinct metabolic changes in P. aeruginosa during chronic lung infection determined at the transcript level [11] taken together with the metabolic flux changes in uropathogenic isolates presented here, reveal a striking picture ( Figure 6). It seems that the pathogen undergoes a similar metabolic adaptation in two very different infection environments: the lung and the urinary tract. The conserved upstream steps of metabolism, i.e. conversion of glucose to pyruvate, are combined with variations in metabolic flux through the downstream steps of metabolic pathways among all strains. For example, the isolates varied in the flux-partitioning ratio at isocitrate, i.e., toward the glyoxylate shunt and TCA cycle (Figure 7 A). Further, flux at the PEP pool showed significant differences ( Figure 7B). Similarly, these reactions were also affected during chronic lung infection [11] (Figure 6). This may reflect a more general underlying strategy that allows P. aeruginosa to adapt to the host environment. The validity of this conclusion is limited by our study's use of isolates from different patients, which did not allow a time-resolved reconstruction of the evolving flux changes during infection. Nevertheless, the alterations in metabolism detected among these isolates are likely involved in adaptation and supports their survival in the host. Further flux studies on strains, experimentally evolved in the laboratory [49,50], and on evolutive lineages of clinical isolates [10,11] seem straightforward to confirm this assumption.
The isocitrate node displays a flexible switch-point in the metabolism of P. aeruginosa The glyoxylate shunt, comprising isocitrate lyase and malate synthase, is widespread [51]. First discovered in P. aeruginosa [52], it has been commonly considered as a by-pass of the TCA cycle to assimilate carbon from C2 compounds [22] and to overcome limitations in the junction between glycolysis and TCA cycle under stress [53]. The use of the shunt in non-pathogenic bacteria grown on glucose was not significant (Table 2). In contrast, most of the P. aeruginosa clinical isolates exhibited significant flux through this pathway ( Figure 5). The presence of the glyoxylate shunt, a bit surprising at first, was confirmed by enzyme assays ( Figure S4). Its use in the presence of glucose differed from other bacteria, which repress the shunt (Figure S4), as shown previously [54].
The enhancement of isocitrate lyase activity by a factor of 4 for acetate-grown P. aeruginosa indicates transcriptional regulation of the pathway, which is known for C. glutamicum [54]. At the level of isocitrate dehydrogenase, the glyoxylate shunt competes with the TCA cycle, i.e., cells can direct carbon toward either pathway. Consistent with this, flux through both pathways is coupled Figure 7. Niche-specific metabolic flux in uropathogenic P. aeruginosa, including flux through isocitrate dehydrogenase and isocitrate lyase (A) and through pyruvate kinase and phosphoenolpyruvate carboxykinase (B). Each data point refers to a distinct isolate. The clustering represents statistical analysis using principal components (see Figure 4) that discriminated between isolates from catheterassociated infections (cluster a) and strains from urinary tract infections (clusters b and c). The full flux data sets for all strains are presented in Supporting Information. All flux values are normalized to the average glucose uptake rate for all strains (100%). doi:10.1371/journal.pone.0088368.g007 ( Figure 7A). Here, the isolates differed significantly in the relative activation of the glyoxylate shunt. Note that the shunt plays a major role in bacterial pathogenesis [55]. For example, isocitrate lyase is up-regulated in P. aeruginosa during infection [56]. Mutants lacking a functional glyoxylate pathway show reduced virulence in plants and mammals [57]. However, the shunt is essential for the survival of other pathogens inside the host, including Mycobacterium tuberculosis [58], Salmonella enterica [59], and Rhodococcus equi [60]. P. aeruginosa isolates show an enhanced turnover of fatty acids and lipids that are abundantly found in the host's environment [11]. The metabolism of such substrates ultimately requires the glyoxylate shunt to metabolize the C2 intermediate acetyl CoA formed during their degradation. Therefore, the changed flux might reflect an adaptation to nutrients available in the host.

Uropathogenic P. aeruginosa is optimized for growth efficiency
For the metabolic network of P. aeruginosa, computer-based analysis yielded 12,955 elementary flux modes, each a unique, minimal combination of reactions with fluxes that support steady state operation of cellular metabolism [34]. Together, all modes spanned the theoretical flux space [35]. Here, we focused on growth efficiency and redox metabolism of P. aeruginosa. The model allowed for NADPH overproduction. Particularly, we were interested in scenarios, which produce more NADPH than needed for anabolism, i.e. reveal apparent NADPH excess. We assumed that this extra amount of reducing power approximately reflects tolerance against oxidative stress, because NADPH drives many  Table 1) and of NADPH metabolism (considering the NADPH balance as described below). In order to evaluate individual flux phenotypes, experimental values for biomass yield and apparent NADPH excess, respectively, (black squares) are integrated into the flux space created by elementary flux mode analysis (grey squares). Most strains localize close to the growth optimum. (A), Correlation of biomass formation and NADPH metabolism (B), contribution of isocitrate dehydrogenase and 6phosphogluconate dehydrogenase to supply of reducing power (C), correlation of biomass formation and flux through the oxidative pentose phosphate pathway (D). NADPH metabolism was inspected by balancing of the redox cofactor. For this purpose, the NADPH formation flux by concerted action of glucose 6-phosphate dehydrogenase, 6-phosphogluconate dehydrogenase, isocitrate dehydrogenase, and malic enzyme was balanced with the NADPH consumption flux for anabolism. For all strains, NADPH formation was higher than consumption. The resulting apparent excess flux of NADPH supply is presented here. The full flux data sets for all strains are presented in Supporting Information. The flux values are normalized to the corresponding glucose uptake rate for each strain (set to 100%). doi:10.1371/journal.pone.0088368.g008 antioxidant mechanisms [43][44][45].Two-dimensional visualization grouped the obtained elementary modes according to their efficiency for growth and NADPH production (Figure 8 A). Therefore, we defined the yield coefficient for biomass (Y X/S ) associated with a particular mode as measure of growth efficiency. As measure of NADPH metabolism, we defined the surplus NADPH that resulted from balancing of this cofactor: For each mode, the NADPH formation flux through glucose 6-phosphate dehydrogenase, 6-phosphogluconate dehydrogenase, isocitrate dehydrogenase, and malic enzyme was balanced with the NADPH consumption flux for anabolism.
In order to evaluate the isolates, we placed their flux phenotypes, i.e. the experimental biomass yield coefficient (Figure 8 A) and the flux of apparent NADPH overproduction (Figure 8 B) into this space of feasible physiological states and looked for their distance to certain points of optimality.
The elementary modes (grey squares) span a triangle with the extremes at the corners, reflecting the corresponding theoretical maximum yield for biomass (Y X/S = 0.54 g?g 21 ) and the maximum flux for NADPH overproduction (6,000%), respectively. The experimentally determined yields and the NADPH overproduction flux of all isolates (black squares) clearly localize close to optimum growth (Figure 8 A). We conclude that the metabolism of these strains is optimized for growth. However, the metabolic network exhibits a significant potential for enhancing NADPH overproduction, if needed. Growth efficiency occurred at the expense of the NADPH supply, because flux closely coupled in the isolates (Figure 8 B). Therefore, the amount of extra NADPH, not needed for anabolism, was decreased in efficient growers. However, they all exhibited an apparent excess, which was not required for anabolism.
We conclude that, under the conditions studied, P. aeruginosa recruits a transhydrogenase to direct the extra NADPH to NADH and the respiratory chain for energy generation. Under stress, the extra amount of NADPH, which is not required for anabolism, could immediately serve for protection against oxidative stress [46].Two nucleotide transhydrogenases, the soluble Sth and the membrane-bound Pnt, respectively, are annotated in the genome of P. aeruginosa [24] and might be involved in this inter conversion process. It is unlikely that the transhydrogenases operate in the NADH-to-NADPH direction, although the flux data do not allow a definite conclusion.
Improved growth efficiency relates to preferential use of the PPP but reduced the production of excess NADPH The P. aeruginosa isolates differed in their relative use of the oxidative PPP (Figures 8 C and D), and this is directly related to their growth efficiency. Strains with an increased biomass yield recruited the PPP to a higher extent. We conclude that the activated PPP represented by its entry enzyme, 6-phosphogluconate dehydrogenase, served for biosynthesis. First, it supplied anabolic precursors such as erythrose 4-phosphate and ribose 5phosphate. Second, it compensated for reduced flux through isocitrate dehydrogenase that catalyzes the synthesis of NADPH in the TCA cycle, which was reduced as a function of increased biomass formation. Thus, the PPP may provide a significant source of redox potential. Similarly, these enzymes complement each other, supplying NADPH as in other bacteria [32]. P. aeruginosa shows niche-specific traits in carbon core metabolism related to the type of infection The flux data indicate that the metabolism of strains isolated from urinary tract infections differed from those isolated from catheter-associated infections. This represents potential adaptations to the specific conditions during infection. The strains were grouped into three clusters according to flux level. In particular, the differences involved the 13 C-labeling of amino acids generated by the TCA cycle (Figure 3 B) and the flux at the junction between the cycles, the pyruvate node, and the glyoxylate shunt ( Figures 5,  6, and 7A and B). Other studies demonstrate site-specific phenotypes of P. aeruginosa isolates [16,61,62]. These differences include extracellular enzymes such as phospholipase [16] or elastase [63], reflecting specific adaptation to nutrient status in the corresponding host environment. Changes in biofilm formation, cell adhesion, and polymer formation indicate adaptions to sessile life styles [16]. Note that the clustering of the P. aeruginosa isolates at the phenotypic flux level (Figure 3 B) was completely different from a previously reported clustering at the genetic level [16]. At seems clear that a different number of changes in the genome could have the same effect on cell activity or even no effect [64]. Consequently, characterization of function, i.e., metabolic level, promises to provide a more direct understanding of the adaption processes. In particular, central metabolic pathways, activated during the infection, appear relevant. Future metabolic flux analysis of direct evolutionary lineages of P. aeruginosa [10,11] and under conditions of oxygen limitation, typically present in the afflicted tissues in which these isolates thrive [39], should be straightforward and promises to shed more light on this interesting question. Figure S1 Growth characteristics and validation of the experimental approach exemplified for the reference strain P. aeruginosa PAO1. Cultivation profile on minimal glucose medium (A). Metabolic and isotopic steady-state are visualized by constant yield for biomass, derived as slope from the profiles of consumed glucose concentration and formed cell dry weight (B). Isotopic steady state is indicated by constant labeling patterns over time, as shown for single-labeled mass isotopomers (M1) of [M-57] amino acid fragments (C). The growth behavior in shake flasks and deep-well plates was identical regarding growth stoichiometry (B) and 13 C labeling data (D) justifying the joint use of data. Accordingly, cell dry weight measurements from shake flask cultures could be integrated with growth and labeling data from deep-well plates for flux calculations, when needed. Experiments were performed in three replicates. (PDF) Figure S2 Dissolved oxygen during cultivation of P. aeruginosa on minimal medium in shake flask culture. To ensure sufficient aeration during cultivation, the level of dissolved oxygen was monitored on-line. As exemplified for P. aeruginosa PAO1, the oxygen level was above 80% of saturation so that fully aerobic conditions were given (one of three replicates shown). The excellent agreement of growth kinetics and stoichiometry and of the 13 C labeling fingerprint ( Figure S1), confirmed that this was obviously also the case for the deep-well plate cultures. (PDF) Figure S3 Growth characteristics of uropathogenic P. aeruginosa isolates. Cultivation profiles on minimal glucose medium (three biological replicates each). For none of the strains, extracellular by-products were detected. (PDF) Figure S4 Enzymatic analysis of isocitrate lyase as key enzyme of the glyoxylate shunt. C. glutamicum and P. aeruginosa were cultivated in minimal medium, supplemented either with 40 mM acetate or with 14 mM glucose as sole carbon source, respectively. (PDF) Figure S5 Metabolic profiles of consumed glucose concentration and formed cell dry weight of clinical P. aeruginosa isolates on minimal glucose medium (three biological replicates each). Metabolic steady-state is inferred from the constant yield for biomass, derived as slope from the profiles. (PDF)

Supporting Information
Table S1 Refinement of the cellular composition of P. aeruginosa by quantification of the alginate capsule and the corresponding anabolic demand for F6P for its biosynthesis. For quantification, alginate was detached from the cells, and then analyzed by the sulfamate-biphenyl method [1]. Shortly, cells were harvested, washed once with deionized water. Subsequently, the alginate was detached by shaking for 5 h (300 mL 0.14 M NaCl suspension, 1400 min 21 , 25uC). The supernatant (200 mL), obtained by centrifugation (15 min, 16,0006 g, 4uC), was then amended with 20 mL 4 M sulfamate and with 1.2 mL 0.075 M tetraborate, dissolved in concentrated H 2 SO 4 , and then incubated for 20 min at 99uC. Afterwards, the suspension was transferred to an ice bath (5 min), followed by addition of 40 mL 0.15% 3-hydroxybiphenyl, dissolved in 0.5% NaOH. The mixture was incubated for 10 min at room temperature. Alginate was quantified by photometry at 525 nm, using isolated alginate from mucoid P. aeruginosa FRD1 as external standard [2]. The data reflect the F6P demand for alginate biosynthesis and are given as relative flux, normalized to the specific glucose uptake rate (Table 1). These values were additionally considered in the anabolic requirement for flux analysis by correcting the demand for F6P (Table S2). Generally, the requirement was low. (PDF)

Table S2
Supporting information on metabolic flux analysis. The data include the 13 C labelling analysis of proteinogenic amino acids by GC-MS for all strains, the entire set of estimated metabolic fluxes, and information on the goodness of fit, i.e. the comparison of experimental and simulated labelling data, corresponding to the optimized fit.