¹³C isotopologue perturbation studies of Listeria monocytogenes carbon metabolism and its modulation by the virulence regulator PrfA

Abstract

The carbon metabolism of Listeria monocytogenes (Lm) EGD and the two isogenic mutant strains LmΔprfA and LmΔprfApPRFA* (showing no or enhanced expression, respectively, of the virulence factor PrfA) was determined by ¹³C isotopologue perturbation. After growth of the bacteria in a defined medium containing a mixture of [U-¹³C₆]glucose and glucose with natural ¹³C abundance (1:25, wt/wt), 14 amino acids were isolated and analyzed by NMR spectroscopy. Multiply ¹³C-labeled isotopologues were determined quantitatively by signal deconvolution. The ¹³C enrichments and isotopologue patterns allowed the reconstruction of most amino acid biosynthesis pathways and illustrated that overproduced PrfA may strongly influence the synthesis of some amino acids, notably that of the branched amino acids (Val, Ile, and Leu). Retrobiosynthetic analysis of the isotopologue compositions showed that degradation of glucose occurs to a large extent via the pentose phosphate pathway and that the citrate cycle is incomplete because of the absence of 2-oxoglutarate dehydrogenase activity. The reconstructed labeling pattern of oxaloacetate indicated its formation by carboxylation of pyruvate. This metabolic reaction seems to have a strong impact on the growth requirement in defined minimal medium. Bioinformatical steady-state network analyses and flux distribution predictions confirmed the experimental data and predicted metabolite fluxes through the enzymes of the pathways under study.

Terrestrial carbon is a mixture of ≈98.9%¹²C and 1.1% ¹³C. In organic matter, the distribution of these isotopes is close to random. All organic compounds are, therefore, complex mixtures of different carbon isotopologues. As an example, a 5-carbon compound consists predominantly of the [U-¹²C₅] isotopologue, accounting for ≈95 mol %. Each of the five isotopologues carrying a single ¹³C atom in any position is present at ≈1 mol %. Multiply ¹³C-substituted species are progressively rare; for example, the [U-¹³C₅] isotopologue accounts for ≈10⁻⁸ mol %. Minor deviations from the random distribution are caused by geophysical and biological processes and serve as the basis for a variety of scientific studies, such as the geographic origin of materials; however, these deviations are below the sensitivity of the methods used in this article and are, therefore, not discussed further.

The quasirandom distribution of carbon isotopes can be experimentally perturbed by the introduction of a ¹³C-enriched compound or a mixture of such compounds. In cells and organisms, such a perturbation will rapidly spread through the entire metabolic network. The quantitative analysis of that relaxation process can afford qualitative as well as quantitative information on metabolic processes in the experimental system under study (1–7).

As an example for the application of this perturbation/relaxation concept, we analyzed the carbon metabolism of Listeria monocytogenes (Lm) after extracellular growth in a defined culture medium containing [U-¹³C₆]glucose. Lm is a human pathogenic microorganism that has been extensively studied, primarily with respect to its specific virulence factors, which enable these bacteria to overcome the various physical and biochemical barriers of the infected host (for recent reviews see refs. 8 and 9). Compared with this rather comprehensive knowledge, little is known of the metabolic requirements of the pathogen necessary for efficient replication in infected host cells.

The last decade has brought an unprecedented wealth of genomic information on most major human pathogens, including Lm (10). This information can serve as a basis for the bioinformatical reconstruction of the metabolic networks of this and other bacteria. Even though the large number of unassigned genes still presents considerable obstacles, mathematical and bioinformatical studies of the reconstructed metabolic networks, using the enzymes already known to be present, can provide valuable insights into the metabolome structure of pathogens. These analyses are capable of enumerating all possible routes a metabolite can take through a reaction network. Intermediary metabolites are hereby assumed to be balanced, and the sum of their producing fluxes must equal their consumption. In a mathematical sense, the system has to fulfill the steady state condition of N × v = 0, with N being the stoichiometric matrix and v being the flux vector or flux distribution of the system under study. The solution space is described by a set of pathways, the extreme pathways or elementary modes (EMs) of a system, that reflect all possible states the metabolic network can assume. For a more detailed description of the concept of elementary modes, see refs. 11 and 12.

In this study, we used the Lm wild-type strain EGD and, in addition, an isogenic mutant strain LmΔprfA, carrying an in-frame deletion in the prfA gene encoding the central regulator of listerial virulence PrfA (13) and the same strain (LmΔprfApPRFA*), which is complemented with a multicopy plasmid carrying a prfA* gene. This gene codes for a constitutively active version of PrfA (termed PrfA*) because of a G145S exchange (14). Because of the elevated prfA* gene copies, the latter strain produces elevated levels of the PrfA* protein (data not shown). The two latter strains were included in this study to determine whether and how PrfA may interfere with the listerial metabolism, as suggested by previous investigations (15–17).

Results and Discussion

NMR Analysis of the ¹³C-Labeled Amino Acids and Reconstruction of Their Biosynthetic Pathways.

Fourteen amino acids obtained from protein hydrolysates of Lm grown in a defined minimal culture medium (18) containing a mixture of unlabeled and uniformly ¹³C-labeled glucose (25:1, wt/wt) were analyzed by ¹³C NMR spectroscopy. The averaged ¹³C enrichments are summarized in Table 1, which is published as supporting information on the PNAS web site. Because of the applied acidic hydrolysis procedure, no conclusions can be drawn regarding the isotopologue composition of Cys, Trp, Asn, and Gln. Met and His were not isolated in quantities required for NMR analysis. Four of the 14 analyzed amino acids (Arg, Ile, Leu, and Val) were present in the used medium as unlabeled supplements, and, as expected, these amino acids were not synthesized de novo or were synthesized only in small amounts (e.g., Ile; see below). The ¹³C enrichments of all other amino acids, including Ser, showed the de novo synthesis from the added ¹³C-labeled glucose at rates of 15–100% (Fig. 1).

Fig. 1.

De novo biosynthetic rates of amino acids in different strains of Lm grown in Welshimer broth medium.

Whereas the signals in ¹H-decoupled ¹³C spectra of natural organic matter typically appear as singlets, many signals of the analyzed amino acids appeared as multiplets, indicating the presence of multiply labeled isotopologue species with significant abundance. For example, the signal of C-2 of Glu was a pseudotriplet, where the central signal represents the [2-¹³C₁] isotopologue (Fig. 2A Left). The two flanking lines (shown in blue) arose by ¹³C¹³C coupling between the observed C-2 of Glu and a directly bound ¹³C atom; the coupling constant of 34.5 Hz showed that the adjacent ¹³C atom causing the signal-splitting is located at position 3 (see ref. 5 for chemical shift values and ¹³C¹³C coupling constants of biogenic amino acids). The relative signal intensities of the coupled signal pair indicated an abundance of 1.0 mol % for [2,3-¹³C₂]Glu (see also Table 2, which is published as supporting information on the PNAS web site). Small signals displaying a double doublet signature (shown in green in Fig. 2) with coupling constants of 53.4 and 34.5 Hz were caused by Glu species carrying ¹³C-label at positions 1, 2, and 3. The small doublet pair shown in yellow in Fig. 2 with a coupling constant of 53.4 Hz was due to the [1,2-¹³C₂]-isotopologue. The low intensities of these signals reflected molar abundances <0.1 mol % and were, therefore, not considered in the following sections.

Fig. 2.

¹³C NMR signatures of C-2 of Glu (Left) and C-2 of Asp (Right) from Lm EGD grown with [U-¹³C₆]glucose diluted with unlabeled glucose (1:25, wt/wt). (A) ¹³C NMR signals observed under ¹H decoupling conditions. (B–E) Simulated signals of isotopologues based on the chemical shifts and coupling constants given in Table 2. The respective isotopologues are shown next to the signals with colored bars connecting adjacent ¹³C-labeled atoms. An isolated ¹³C atom (with adjacent ¹²C atoms) is indicated by the filled circle. The amplitudes of the simulated spectra were adjusted to reflect the relative abundances of each isotopologue in the observed mixture.

The multiplet observed for C-2 of Asp is more complex and comprises seven lines of significant intensity (Fig. 2A Right). The deconvolution of this complex pattern is best achieved by numerical simulation of the signal patterns for each individual isotopologue, taking into account the coupling constants as well as the chemical shift increments. For easy viewing, the coupling pattern of Asp for each individual simulation is indicated in Fig. 2 B–E Right). The deconvolution of the experimentally observed spectrum is in line with the presence of the [2-¹³C₁], [2,3-¹³C₂], and [1,2,3-¹³C₃] isotopologues with ¹³C-enrichments of 1.2, 0.5, and 3.5 mol %, respectively.

An even higher degree of signal complexity is observed for signals arising from the ring carbon atoms of Phe and Tyr. As an example, the position 6/8 ring carbon atoms of Tyr comprises more than a dozen lines that reflect the presence of three multiply ¹³C-labeled isotopologues with significant abundance (see Fig. 7, which is published as supporting information on the PNAS web site). On the basis of the known coupling constants (19), the coupling signals can be attributed to [6,7,8,9-¹³C₄]-, [7,8,9-¹³C₃]-, and [6,8-¹³C₁]Tyr at molar abundances of 1.2%, 1.4%, and 1.6%, respectively (Table 2).

NMR analyses revealed that multiply ¹³C-labeled isotopologues were also present in Gly, Ala, Ser, Thr, and Phe at abundances higher than 1 mol % (Table 2; and see Fig. 8, which is published as supporting information on the PNAS web site). The ¹³C NMR spectra of Arg (an amino acid added to the culture medium) showed no coupling satellites and obviously did not contain multiply ¹³C-labeled isotopologues in detectable amounts. The branched amino acids (Val, Ile, and Leu), although added to the culture medium as well, contained, at least under certain conditions (see below), sufficient quantities of ¹³C-labeled isotopologues to allow their analysis.

The dominant features of the labeling patterns of Ala, Thr, and Asp were blocks of three contiguous ¹³C atoms involving carbon atoms 1–3 (Fig. 3). At lower abundance, blocks of two ¹³C atoms were found at carbon atoms 2 and 3. It should be noted that the same labeling pattern was also characteristic for phosphoenolpyruvate (Fig. 4). At increased abundance, we also observed isotopologues carrying a single ¹³C atom at the γ carboxylic position of Asp and the methyl position of Thr, respectively (shown as red colored dots in Fig. 3). These findings are all consistent with a common biosynthetic origin of all three amino acids from pyruvate. More specifically, Ala can be obtained by transamination of pyruvate. The formation of Asp is easily explained by carboxylation of phosphoenolpyruvate or pyruvate followed by reductive transamination. Finally, the labeling pattern of Thr faithfully reflects its biosynthetic origin from Asp.

Fig. 3.

Reconstruction of the incomplete citric acid cycle in Lm EGD from the experimentally determined labeling patterns of Thr, Asp, Ala, and Glu (shown in boxes). The labeling patterns of citric acid cycle intermediates were predicted on the basis of standard biosynthetic routes in bacteria. Bold bars connect ¹³C-labeled atoms that were transferred from the same molecule of [U-¹³C₆]glucose. The line widths were adjusted according to the respective abundances of each isotopologue. Filled dots represent ¹³C₁ isotopologues with ¹³C enrichments well above the natural abundance contributions. The numbers indicate ¹³C enrichments in mol %.

Fig. 4.

Reconstruction of the averaged labeling patterns of phosphoenolpyrvate and erythrose 4-phosphate from the observed labeling patterns in Phe and Tyr (shown in boxes) in the experiment with Lm EGD. For other details, see legend to Fig. 3.

The isotopologue composition of Tyr and Phe directly reflects the isotopologue composition of the central metabolic pools phosphoenol pyruvate and erythrose 4-phosphate via the shikimate pathway. More specifically, the side chains indicate the isotopologue composition of phosphoenolpyruvate (Fig. 4), and the labeling pattern of erythrose 4-phosphate can be extracted from the labeling pattern of the aromatic rings. Notably, the deduced labeling pattern of phosphoenolpyruvate was dominated by the universally ¹³C-labeled [U-¹³C₃] isotopologue. Similarly, the universally labeled [U-¹³C₄] isotopologue was also the predominant species in the tetrose phosphate pool, reflecting its formation from [U-¹³C₆]glucose by the transketolase reaction of the pentose phosphate pathway (PPP). The observed [2,3,4-¹³C₃] isotopologue of erythrose 4-phosphate can be explained by the transaldolase reaction of the PPP. In silico, if Tyr- and Phe-synthesizing pathways are highly active, this activity is reflected by predicting an up-regulation of the transaldolase and transketolase I reactions in the calculated flux distributions. Additionally, to supply the synthesis pathways with enough phosphoenolpyruvate, a high enolase activity was predicted in all cases (scenarios 1 and 2), tantamount to an up-regulation of overall glycolysis.

The labeling efficiency of Ser clearly suggests that this amino acid can be biosynthesized by Lm, although the gene for the enzyme phosphoserine phosphatase is not annotated in the Lm genome sequence (10). The labeling pattern (Fig. 5) closely resembles that of Ala and so do the predicted flux distributions for Ser synthesis. This finding is well in line with the hypothesis that Ser is predominantly formed by dehydrogenation of 3-phosphoglycerate, followed by reductive transamination and dephosphorylation. The alternative pathway, i.e., hydroxymethylation of Gly, leads to [1,2-¹³C₂]Ser that was observed only in minor amounts (Fig. 5).

Fig. 5.

Reconstruction of Ser biosynthesis in Lm EGD. Experimentally determined labeling patterns are shown in boxes. For other details, see legend to Fig. 3.

In silico flux distribution ratios, in which EM activities were chosen to maximize Ser production, showed no glycine hydroxymethyltransferase activity but an equal distribution of fluxes between the serine dehydrogenase and phosphoserine phosphatase originating from glycolysis. For a schematic overview of the different Ser production routes and their computed fluxes, see Fig. 9, which is published as supporting information on the PNAS web site.

The labeling pattern of Lys was characterized by blocks of three contiguous ¹³C atoms (C-1, C-2, and C-3) and two ¹³C atoms (C-5 and C-6). Moreover, we observed an isotopologue with a single ¹³C atom in position 4. The comparison of that labeling pattern with those of Asp, pyruvate, Glu, and acetyl-CoA clearly indicates that the biosynthesis of Lys proceeds exclusively or predominantly via Asp and pyruvate as precursors (Fig. 6). Contributions from the reaction path by condensation of 2-oxoglutarate with acetyl-CoA can be excluded on the basis of the detected isotopologue composition in Lys. The apparent lack of a symmetrical labeling pattern in Lys also excludes the involvement of one or more symmetrical intermediates in free form, such as free (LL)-diaminopimelate. Flux distribution analyses of the Lys-synthesizing pathways confirmed that pyruvate and Asp are the only precursors of Lys. One-half of the pyruvate is predicted to be used for the formation of dihydrodipicolinate. The other half is carboxylated and transformed into oxaloacetate and transaminated to Asp. The necessary nitrogen is taken from the externally supplied Gln in the medium and passes through the glutamate synthase (see Table 3, which is published as supporting information on the PNAS web site).

Fig. 6.

Reconstruction of Lys biosynthesis in Lm EGD. Experimentally determined labeling patterns are shown in boxes. For other details, see legend to Fig. 3.

Incomplete Citrate Cycle.

The decarboxylation of the detected [U-¹³C₃]- and [2,3-¹³C₂]pyruvate specimens should afford [1,2-¹³C₂]acetyl-CoA. Condensation of that central intermediary metabolite with [1,2,3-¹³C₂]oxaloacetate should predominantly yield a mixture of [1,2-¹³C₂] and [3,4,3′-¹³C₃] isotopologues of citrate (Fig. 3). Moreover, the presence of [2,3-¹³C₂]oxaloacetate predicts the formation of minor amounts of [3,4-¹³C₂]citrate. On the basis of the classical stereospecifity in the aconitase reaction (20, 21), the decarboxylation of that isotopologue mixture should predominantly yield [2,3-¹³C₂]- and [4,5-¹³C₂]2-oxoglutarate, which could be transformed into the cognate isotopologue mixture of Glu by reductive transamination. In fact, that picture is qualitatively reflected in Glu isolated from the bacterial hydrolysate, albeit at about threefold lower enrichment than expected. This finding is explained by dilution of biosynthetic Glu by Glu derived from unlabeled Gln that had been provided in the culture medium in relatively large amounts (0.6 g/liter).

The conversion of [2,3-¹³C₂]- and [4,5-¹³C₂]2-oxoglutarate into oxaloacetate via 2-oxoglutarate dehydrogenase should yield a mixture of [1,2-¹³C₂]- and [3,4-¹³C₂]oxaloacetate. However, these isotopologues were apparently absent in oxaloacetate. It can be concluded that oxaloacetate is not derived from 2-oxoglutarate via the citrate cycle and that Lm is characterized by an incomplete citrate cycle, as shown in Fig. 3. On the other hand, the detected isotopologue composition in oxaloacetate is related to that of pyruvate or phosphoenol pyruvate. More specifically, the observed triple-¹³C-labeled isotopologue in pyruvate is retained in the [1,2,3-¹³C₃]oxaloacetate, providing strong evidence that oxaloacetate is formed exclusively by carboxylation of pyruvate. A gene for pyruvate carboxylase (pycA) but not for phosphoenolpyruvate carboxylase has been identified in the Lm genome sequence (10). In agreement with this hypothesis, flux distribution predictions of the overall system confirm pyruvate carboxylase as the main enzyme for oxaloacetate formation (Table 3).

Influence of PrfA on the Amino Acid Biosyntheses.

Recent evidence suggests that PrfA, the central regulator of Lm virulence, may also influence the listerial metabolism (ref. 22 and B. Joseph and A. Marr, personal communication). We therefore compared the ¹³C isotopologue distribution in the analyzed amino acids among the wild-type Lm strain, an isogenic prfA mutant, and this mutant strain containing multiple copies of the prfA* gene, which encodes a constitutively active PrfA (14). All three strains were grown under identical conditions, i.e., in medium containing [U-¹³C₆]glucose. The isotopologue profiles of amino acids are summarized graphically in Fig. 8. Many of the analyzed amino acids showed similar isotopologue patterns for the three strains. However, the fraction of [1,2-¹³C₂]Ser in the global ¹³C enrichment of Ser accounted for ≈22% in the prfA deletion mutant compared with 6% in the wild-type strain and 13% in the PrfA-overexpressing strain. The [1,2-¹³C₂] isotopologue of Ser was assigned to a biosynthetic origin through hydroxymethylation of Gly (see above). On this basis, biosynthesis of Ser is shifted in favor of the phosphoglycerate route in the presence of high PrfA concentration.

The ¹³C enrichments of Glu, Pro, Val, Leu, and Ile indicated that only minor fractions of these amino acids were made de novo in the wild-type strain and that the major fractions apparently derived from the unlabeled amino acids provided by the culture medium (Fig. 1). Interestingly, the de novo biosynthetic rates of these amino acids were significantly affected by PrfA. Notably, the rate of the branched-chain amino acids (Ile, Leu, and Val) was found to be strongly enhanced in the ΔprfApPrfA* strain, i.e., in the presence of high PrfA concentration (see modulation of ¹³C-coupled satellite signals in Fig. 10, which is published as supporting information on the PNAS web site). Also the biosynthetic rates of Gly and Pro were about twice as high as that of the prfA deletion mutant and the wild-type strain, respectively, whereas that of Glu was significantly decreased under these conditions. Again, these data clearly indicate that overexpression of PrfA interferes with the carbon metabolism.

Recently performed studies have shown that high levels of PrfA reduce the growth rate of Lm in the used minimal medium by inhibition of glucose uptake (A. Marr, personal communication). This decrease in glucose uptake may result in a reduced energy-dependent transport of external amino acids into the listerial cell. As a consequence, de novo biosyntheses of amino acids, which are repressed by the excess amino acids in the medium, may be induced in the presence of high PrfA levels. Alternatively (or in addition), the branched-chain amino acids, especially Ile, may be, in part, degraded to fill in the gap of carbon supply caused by the reduced glucose uptake in the presence of high PrfA concentration. Indeed, the partial usage of Ile as an additional carbon source (degradation to actetyl-CoA, which could be a metabolic requirement because of the interrupted citrate cycle) may explain its need as a nutrient in the used minimal medium, despite the fact that this amino acid can be synthesized de novo (see above).

The inhibition of de novo Glu synthesis by high PrfA concentration may also reflect an increased Gln uptake as well as a reduced 2-oxoglutarate synthesis because of the impaired glucose uptake in the presence of high PrfA levels.

In Silico Analyses of Amino Acid Synthesis Pathways.

As mentioned above, we performed steady-state analyses of the metabolic network of Lm. The comparison of the isotopologue data with the bioinformatical data from EM analysis and flux distribution predictions showed that Listeria metabolism is theoretically capable of synthesizing all investigated amino acids, with the exception of Ser, from only glucose as the main carbon source. As mentioned above, the gene for the last enzyme in the Ser biosynthesis is not annotated (10), but, because of the experimentally proven de novo biosynthesis, this amino acid was also calculated in the flux mode. Gln (added to the medium) was taken as the main nitrogen source. In addition, bicarbonate (HCO₃⁻) was included as substrate for the production of oxaloacetate by pyruvate carboxylase. Indeed, the latter pathways were fully supported by the resulting EMs. By using the redox-decoupled system (scenario 2), activation of EMs producing one or more of the named amino acids showed a flux distribution with a highly active glucose transporter, strongly suggesting glucose as the main carbon source for these amino acids. On the other hand, an inverted EM activity pattern (instead of all amino acid synthesizing modes activated, the program yana analyzes the inverted scenario with all these modes deactivated) showed a low glucose transporter activity. Analyses of the Arg synthesizing modes in scenario 2 showed that glucose is not necessary for the production of Arg, if reduction equivalents NADH and NADPH are supplied, and the system uses Gln from the medium as the only carbon and nitrogen source for the synthesis of the Arg precursors citrulline and ornithine. Scenario 1 resolves that, in this case, the consumed glucose is completely oxidized via the oxidative part of the PPP, additionally indicated by a high PPP activity in flux distributions by using scenario 2. These findings could explain the missing isotopologue labeling pattern for Arg, if these pathways are truly dominant.

Elementary mode analysis results also suggest that Asp is predominantly formed through oxaloacetate, which is synthesized by carboxylation of pyruvate. The predicted flux distribution for Thr production shows a flux through the Asp transaminase precisely the same as the Thr transporter and the Asp kinase, confirming the fact that all Thr is synthesized via Asp. Similar clear results are obtained for the production of Ala, which is exclusively formed by transamination of pyruvate (Table 3).

Conclusion

Incorporation experiments with [U-¹³C₆]glucose, followed by retrobiosynthetic analysis of amino acids, were successful in delineating the primary carbon metabolism of growing Lm. Clearly, the method is a powerful tool to quickly determine the metabolism of pathogenic bacteria in general, which is only poorly understood in most cases. The need to experimentally verify annotated metabolic pathways based on genome sequence data of these microorganisms was exemplified by the biosyntheses of Ser and the branched amino acids (especially Ile) in Lm. Although the Ser pathway was annotated as incomplete (missing gene for Ser-P-phosphatase), all genes for the biosynthesis of the branched amino acids were identified in the listerial genome (10). The experimental data show, however, that Ser is efficiently made de novo, whereas there is an absolute need for Ile as a supplement in the used minimal medium, despite the fact that de novo synthesis of this amino acid is possible. Furthermore, the data readily reveal of how a metabolic gap (e.g., the disruption of the citrate cycle in Lm, also suggested by the genome annotation) is overcome (synthesis of oxaloacetate by pyruvate carboxylation) and how it may influence other pathways. Most importantly, this study shows directly that the induced expression of the major virulence regulator PrfA interferes with the primary metabolism, as previously suggested by more indirect observations (ref. 22 and B. Joseph and A. Marr, personal communication). The present investigation was carried out with a facultative intracellular microorganism but, so far, shed some light on the metabolism in Lm growing under extracellular conditions only. It will be necessary to develop the ¹³C isotopologue perturbation method for Lm and other pathogenic bacteria growing in the host cells.

Experimental Procedures

Materials.

[U-¹³C₆]glucose was purchased from Campro Scientific (Berlin).

Bacterial Culture.

The strains used in this study were Lm EGD (serovar 1/2a, wild type, provided by S. H. E. Kaufmann, Berlin) and a prfA deletion mutant Lm EGD ΔprfA (13). The latter was transformed with the pPRFA* plasmid harboring the prfA* gene of Lm P14A with its own promoter (14, 23) on the multicopy vector pERL3 (24). Bacteria were grown to late logarithmic phase in 100 ml of modified Welshimer’s broth (MWB) minimum media (18) containing 0.96 g of glucose with natural ¹³C abundance, 0.04 g of [U-¹³C₆]glucose, 0.66 g of potassium dihydrogenphosphate, 1.64 g of dinatrium hydrogenphosphate, 0.04 g of magnesium sulfate, 10 mg of iron-III-citrate, 0.01 g of Leu, 0.01 g of Ile, 0.01 g of Val, 0.01 g of Met, 0.01 g of Arg, 0.01 g of Cys, 0.06 g of Gln, 0.01 g of His, 0.05 mg of riboflavine, 0.1 mg of thiamine, 0.05 mg of biotin, and 0.5 μg of thioctic acid. The 100-ml culture was inoculated into 10 liters of MWB containing 4.0 g of [U-¹³C₆]glucose (99.9% ¹³C enrichment) and 96 g of glucose with natural ¹³C abundance (1.1% ¹³C enrichment). Bacteria were incubated aerobically at 37°C until stationary phase. Sodium azide was added to a final concentration of 10 mM to kill the Listeriae. Bacterial cells were harvested by centrifugation and washed three times with PBS. The pellet was stored at 4°C.

Isolation of Amino Acids.

Bacterial cell mass (27 g) was suspended in 600 ml of distilled methanol, boiled under reflux for 1 h, and filtered. The residue was suspended in 300 ml of dichloromethane and boiled under reflux for 60 min. The mixture was filtered. The residue was suspended in 250 ml of 1 M sodium hydroxide, and the mixture was incubated for 24 h at room temperature. The solution was neutralized by the addition of 15 ml of concentrated hydrochloric acid. Protein was precipitated by the addition of 15 ml of 25% trichloroacetic acid. The mixture was centrifuged, and the residue was suspended in 200 ml of 6 M hydrochloric acid containing 8 ml of thioglycolic acid. The mixture was boiled under reflux for 24 h under an inert atmosphere and was then filtered, evaporated to a small volume under reduced pressure, and lyophilized. The residue was dissolved in 8 ml of water and was placed on top of a column of DOWEX 50W × 8 (H⁺ form, 2 × 34 cm) (Dow) that was developed consecutively with 0.2 M ammonium formate, pH 4.5; 0.2 M ammonium formate, pH 6.5; and 0.5 M ammonium formate, pH 6.5. Fractions containing Lys, His, and Arg were collected and lyophilized. Fractions containing neutral and acidic amino acids were combined and lyophilized. The residue was dissolved in 10 ml of water and was placed on top of a column of DOWEX 1 × 8 (formate form, 2 × 24 cm) that was developed with 150 ml of distilled water, followed by 250 ml of 10 mM formic acid. Fractions containing Glu and Asp were collected and lyophilized. Fractions containing neutral amino acids were combined and lyophilized. The residue was dissolved in 10 ml of water and was placed on top of a column of DOWEX 50W × 8 (H⁺ form, 3 × 33 cm). The column was washed with 100 ml of water and was then developed with a linear gradient of 0–3 M hydrochloric acid (total volume, 2 liters). Fractions were combined, evaporated to a small volume under reduced pressure, and lyophilized.

NMR Spectroscopy.

¹H and ¹³C NMR spectra were recorded at 25°C by using a DRX 500 spectrometer (Bruker, Karlsruhe, Germany) at transmitter frequencies of 500.1 and 125.6 MHz, respectively. Samples were dissolved in 0.1 M DCl. ¹³C-Enrichments were determined by quantitative NMR spectroscopy (5). For this purpose, ¹³C NMR spectra of the biolabeled specimens and of samples with natural ¹³C abundance (i.e., with 1.1% ¹³C abundance) were measured under the same experimental conditions. The ratios of the signal integrals of the biolabeled compounds and of the compounds at natural abundance were then calculated for each respective carbon atom. Absolute ¹³C abundances for certain carbon atoms (i.e., for carbon atoms with at least one attached hydrogen atom displaying a ¹H NMR signal in a noncrowded region of the spectrum) were determined from the ¹³C coupling satellites in the ¹H NMR spectra. The relative ¹³C abundances determined for all other positions were then referenced to this value, thus affording absolute ¹³C abundances for every single-carbon atom (Table 2). ¹³C-coupled satellites were integrated separately. The relative fractions of each respective satellite pair (corresponding to a given coupling pattern, Table 2) in the total signal integral of a given carbon atom were calculated. These values were then referenced to the global absolute ¹³C abundance for each carbon atom affording concentrations of multiple ¹³C-labeled isotopologue groups (mol %). Numerical simulation of signals was performed with NMRSIM (Bruker) on the basis of chemical shift values and coupling constants given in Table 2.

Bioinformatics.

Reconstruction of the metabolic network of Lm was based on information from the Kyoto Encyclopedia of Genes and Genomes (KEGG) (25) database (www.genome.ad.jp/kegg/kegg2.html) and focused on amino acid synthesis pathways and basic carbohydrate metabolism. Enzymes stated to be missing in Lm by the KEGG system were reinvestigated by similarity searches in the Lm genome both from the National Center for Biotechnology Information web repository (26) and based on a manually annotated version by using the pedant annotation system (27). Searches were carried out by using query sequences of different closely related Gram-positive bacteria owning the questionable enzyme (preferably Bacillus subtilis because of its extensive annotation data) and both psi-blast (28) and Smith–Waterman algorithms (29). The reconstructed metabolic network comprised 114 metabolites [of which 27–31 (see below) were considered external] and 104 reactions (66 assumed to be reversible). EM analysis (11) was effected by using the program yana (30). Because amino acid synthesis strongly depends on availability of reduction equivalents, the EMs were calculated in two different scenarios, scenario 1 coupled to the production of NADH and NADPH (both internal), and scenario 2, with NADH and NADPH set as external compounds, meaning that both are supplied by standard Listeria pathways not considered further and in ample amount. The latter scenario was thus introduced to reduce network complexity in the calculations, where our interest focused on nitrogen and carbon sources rather than redox coupling. Scenario 1 resulted in a total of 4,082 EMs, of which 365 formed the convex basis of the solution space, whereas scenario 2 yielded only 811 EMs, including 120 convex basis vectors. The complete system is given in Tables 4–9, which are published as supporting information on the PNAS web site and can be downloaded in SBML format from our web site at (http://wbbi005.biozentrum.uni-wuerzburg.de/∼binf012/listeria). EM flux distributions were computed by using yana. Different flux distributions were calculated for the different amino acid synthesis subsystems of interest by assigning activities to the elementary modes. yana allows the calculation of the total flux distribution of all modes for the synthesis or uptake of a specific amino acid of interest, including deactivation of EMs that do not concern the amino acid under study. The calculation yields a flux coefficient for each individual enzyme and allows the identification of enzymes in the network that are never used for the synthesis of this amino acid. These fluxes describe the active and inactive enzymes for this amino acid, and they were further compared with the experimentally received results (Table 3).

Glossary

Abbreviations:

EM

elementary mode

KEGG

Kyoto Encyclopedia of Genes and Genomes

Lm

Listeria monocytogenes

PPP

pentose phosphate pathway.