A. de Graaf, Institut für Biotechnologie I, Forschungszentrum Jülich GmbH, D-52425 Jülich, Germany. Fax: + 49 2461 612710, Tel.: +49 2461 613969, E-mail: firstname.lastname@example.org
The glucose-6-phosphate (Glc6P) and 6-phosphogluconate (6PG) dehydrogenases of the amino-acid-producing bacterium Corynebacterium glutamicum were purified to homogeneity and kinetically characterized. The Glc6P dehydrogenase was a heteromultimeric complex, which consists of Zwf and OpcA subunits. The product inhibition pattern of the Glc6P dehydrogenase was consistent with an ordered bi-bi mechanism. The 6PG dehydrogenase was found to operate according to a Theorell–Chance ordered bi-ter mechanism. Both enzymes were inhibited by NADPH and the 6PG dehydrogenase additionally by ATP, fructose 1,6-bisphosphate (Fru1,6P2), d-glyceraldehyde 3-phosphate (Gra3P), erythrose 4-phosphate and ribulose 5-phosphate (Rib5P). The inhibition by NADPH was considered to be most important, with inhibition constants of around 25 µm for both enzymes.
Intracellular metabolite concentrations were determined in two isogenic strains of C. glutamicum with plasmid-encoded NAD- and NADP-dependent glutamate dehydrogenases. NADP+ and NADPH levels were between 130 µm and 290 µm, which is very much higher than the respective Km and Ki values. The Glc6P concentration was around 500 µm in both strains.
The in vivo fluxes through the oxidative part of the pentose phosphate pathway calculated on the basis of intracellular metabolite concentrations and the kinetic constants of the purified enzymes determined in vitro were in agreement with the same fluxes determined by NMR after 13C-labelling. From the derived kinetic model thus validated, it is concluded that the oxidative pentose phosphate pathway in C. glutamicum is mainly regulated by the ratio of NADPH and NADP+ concentrations and the specific enzyme activities of both dehydrogenases.
Corynebacterium glutamicum has a high NADPH demand for the overproduction of amino acids such as lysine and isoleucine. We previously found that the main site of NADPH generation during amino-acid production is the oxidative part of the pentose phosphate pathway . Therefore the reaction catalysed by the glucose-6-phosphate (Glc6P) dehydrogenase (EC 184.108.40.206):
and 6-phosphogluconate (6PG) dehydrogenases (EC 220.127.116.11):
(Rib5P is ribulose 5-phosphate) were examined in detail in this study.
There is detailed information about the flux distribution at the branch point between glycolysis and the pentose phosphate pathway in C. glutamicum. NMR analysis of proteinogenic amino acids after 13C labelling of two isogenic strains with plasmid-encoded NAD- and NADP-dependent glutamate dehydrogenases has shown that the flux distribution between these pathways alters by a factor of 3 when the cofactor specificity is changed . According to similar studies the pentose phosphate pathway flux is reduced more than twofold in the case of fructose-grown cells as compared to glucose-grown cells . As yet little information is available about the detailed regulation of the respective enzymes in C. glutamicum.
In some cases the pentose phosphate pathway flux is regulated by changes of specific activities of the Glc6P and 6PG dehydrogenases, as, for example when using gluconate instead of glucose as the carbon source . In other cases, regulation by metabolite concentrations must be responsible, as with fructose- and glucose-grown C. glutamicum where no differences in specific enzyme activities, but strongly altered contributions of the pentose phosphate pathway were found . This has been explained by an increased intracellular fructose 1,6-bisphosphate (Fru1,6P2) concentration, which has been found to inhibit the Glc6P and 6PG dehydrogenases of several organisms [5–7]. Key metabolites like energy or electron donors have also been shown to play an important role in directing the carbon flux through central metabolism. For the pentose phosphate pathway of Saccharomyces cerevisiae it was found that ATP and NADPH are co-metabolites for directing the flux through the pentose phosphate pathway . Another example is glycolysis in Escherichia coli, which is mainly regulated by the intracellular ATP level .
The available data on the C. glutamicum enzymes clearly do not allow any well-founded conclusion on regulation of the pentose phosphate pathway flux in this organism. Therefore the aim of this work was to examine the Glc6P and 6PG dehydrogenases of C. glutamicum with respect to reaction mechanism and regulation by intracellular metabolites, and to correlate their in vivo activities with intracellular metabolite concentrations. In two isogenic C. glutamicum strains with plasmid-encoded NAD- and NADP-dependent glutamate dehydrogenases, the in vivo pentose phosphate pathway fluxes based on pool sizes and enzyme kinetic constants were compared with fluxes measured by 13C labelling and NMR.
Materials and methods
Organism and cultivation
For purification of the Glc6P and 6PG dehydrogenases C. glutamicum ATCC 13032 was grown aerobically on minimal medium at 30 °C in a Labfors fermentation system (Infors AG, Bottmingen, Switzerland). A preculture (320 mL bacto brain heart infusion medium, Difco Laboratories, Detroit, USA) was incubated for 15 h at 30 °C and used for the inoculation of 2.5 L minimal medium. The medium contained the following constituents (amounts per litre): 20 g (NH4)2SO4; 1 g KH2PO4; 1 g K2HPO4; 0.25 g MgSO4·7H2O; 10 mg CaCl2; 0.2 mg biotin; 30 mg protocatechuic acid; 1 mg FeSO4·7H2O; 1 mg MnSO4·H2O; 0.1 mg ZnSO4·7H2O; 0.02 mg CuSO4; 0.002 mg NiCl2·6H2O; 1.2 g HCl; 0.2 g polypropylene glycol; 75 mg tritriplex II and 100 g glucose. During fermentation, sodium hydroxide was continuously added in order to keep the pH value constant at 7.0. The cells were harvested in the late exponential growth phase. After centrifugation at 6400 g for 15 min at 4 °C and washing in 100 mm Tris/HCl, pH 7.5, containing 10 mm MgCl2 the cells were stored at −20 °C until use.
For metabolite pool determinations, two isogenic strains derived from the C. glutamicum lysine-producing strain MH20-22B were cultivated in chemostat culture. The first strain, containing plasmid-encoded NAD-dependent glutamate dehydrogenase from Peptostreptococcus asaccharolyticus (MH20-22BΔgdh pEK1.9gdh-1 ), is referred to henceforth as the heterologous glutamate dehydrogenase mutant. The second strain, containing plasmid-encoded NADP-dependent glutamate dehydrogenase from C. glutamicum (MH20-22BΔgdh pEKExpgdh ), is henceforth called the homologous glutamate dehydrogenase mutant. The two strains were cultivated continuously with a dilution rate of 0.09 h−1. While this is about 10% different from the rate of 0.082 reported in , the relative contribution of the pentose phosphate pathway was identical within measurement error, as was verified by an independent NMR experiment with the heterologous mutant (data not shown). The medium was the same as that already described, except that it contained 40 g (NH4)2SO4, 1 g KCl, 7.5 g glucose, 150 mg leucine, 2.1 µg isopropyl thio-β-d-galactoside and 20 mg kanamycin monosulphate (amounts per litre).
Disruption of cells was carried out in a disintegration system (Disintegrator S, BIOmatic, Rodgau-Hainhausen, Germany). The cells had been previously resuspended in a pH 7.5 buffer consisting of 100 mm Tris/HCl, 10 mm MgCl2, 0.75 mm dithiothreitol and the protease inhibitor Complete (Roche). The ratio of the cell wet weight to the total suspension weight was adjusted to 30 g per 100 mL. After addition of the same volume of glass beads with a diameter of 0.1–0.25 mm (Fisher Scientific), cell disruption was performed at 5000 r.p.m. for 12 min. After removal of the glass beads, an ultracentifugation step was carried out at 235 000 g for 90 min at 4 °C. The supernatant was used as crude extract for purification of the Glc6P and 6PG dehydrogenases.
In order to measure specific enzyme activities of the homologous and heterologous glutamate dehydrogenase mutants, cell disruption was performed at 24 kHz by an ultrasonic disintegrator (UP 200S, Dr Hilscher, Teltow, Germany), run for 5 min with a power of 110 W and a 50% duty cycle. During sonication the sample was cooled in ice water.
Preparation of extracts for metabolite pool determinations
Cell extraction for determination of metabolite pools was carried out in the metabolic steady state. As turnover rates may be in the subsecond range , the metabolism must be quenched instantaneously. Therefore 5 mL culture volume (30 °C) was drawn into a syringe filled with 10 mL 60% methanol and precooled to −58 °C [10,11]. The final sample temperature was −20 °C. After centrifugation at 9400 g at a temperature of −20 °C for 5 min (Avanti TM 30, Beckman), the cell sediment was extracted. To check for metabolite leakage, samples were stored at −20 °C for up to 46 min before centrifugation, but no loss of Glc6P or Fru1,6P2 was observed with C. glutamicum, in contrast to Lactococcus lactis. Extraction was performed by the addition of 2 mL perchloric acid 35% (w/v) and sonication (2200 sonifier, Branson Ultrasonics, Danbury, USA) at 0 °C for 10 min. Cell debris and proteins were removed by centrifugation at 41 000 g for 30 min. Neutralization of the extract to pH 7.0 was carried out by the slow addition of ≈ 900 µL 5 m K2CO3 solution. The extract was stored at −70 °C until analysed. Any impacts of KClO4 precipitation on metabolite loss were accounted for by internal standardization prior to extraction.
For NAD+ and NADP+ determinations, cell extraction was performed in perchloric acid (pH 1.0; 7 min 55 °C), whereas for NADH and NADPH potassium hydroxide was used (pH 12.4, 3 min 55 °C) [3,13]. The basic extraction medium additionally contained ethanol (25% v/v). Ten millilitres of culture volume were drawn into a syringe containing the preheated extraction medium. Stability of all metabolites during the different extraction procedures was tested by internal standardization with NADP(H).
All purification steps were carried out with a Biosys2000 system (Beckman). The crude extract was applied to an XK 50/30 column (Pharmacia) containing Fractogel EMD DEAE-650(S) material (Merck). The total bed volume was 500 mL. The column had been previously equilibrated with 50 mm Tris/HCl, pH 7.5, containing 30 mm MgCl2 and 0.75 mm dithiothreitol. After application of the crude extract, the column was washed with 500 mL of the same buffer containing 144 mm KCl. Elution was performed within 95 min by a linear KCl gradient from 144 to 320 mm at a flow rate of 7.4 mL·min−1. The 6PG dehydrogenase eluted at 240 mm KCl, the Glc6P dehydrogenase at 300 mm KCl. The active fractions (activity measurements described below) were pooled and concentrated five or six times in Centriprep 30 concentrators (Amicon). By dilution with 50 mm Tris/HCl, pH 7.5, containing 30 mm MgCl2 and 0.75 mm dithiothreitol the KCl concentration was reduced to 40 mm.
Thereafter, the partially purified Glc6P and 6PG dehydrogenases were applied to an XK26/20 column (Pharmacia), packed with 65 mL Red Sepharose CL6B gel (Pharmacia) [14,15]. The column was equilibrated with 50 mm Tris/HCl, pH 7.5, containing 30 mm MgCl2 and 0.75 mm dithiothreitol. Elution was carried out within 590 min by a linear 0–800 mm KCl gradient at a flow rate of 0.87 mL·min−1. The 6PG dehydrogenase eluted at 200 mm KCl, the Glc6P dehydrogenase at 450 mm KCl. After pooling of the active Glc6P and 6PG dehydrogenase fractions, the KCl concentration was reduced to 10 mm in the same way as described above. The solution was then applied to an XK16/20 column (Pharmacia) containing 20 mL of a 2′5′-ADP–Sepharose matrix (Pharmacia) . The column was equilibrated with the same buffer as the Red Sepharose CL6B column. Elution was performed during 240 min by a 0–6 mm NADP+ linear gradient at a flow rate of 0.5 mL·min−1. The 6PG dehydrogenase eluted at 3 mm NADP+, the Glc6P dehydrogenase at 0.8 mm NADP+. NADP+ was removed by repeated concentration in Centriprep 30 concentrators and dilution with equilibration buffer. The 6PG dehydrogenase fraction was frozen in liquid nitrogen and stored at −70 °C.
For further purification, the Glc6P dehydrogenase fraction was applied to a Superdex G200pg gel filtration column (Pharmacia) with a diameter of 1.6 cm and a bed volume of 114 mL. Elution at a flow rate of 1 mL·min−1 was carried out with 50 mm Tris/HCl, pH 7.5, containing 30 mm MgCl2, 200 mm KCl and 0.75 mm dithiothreitol. The active fractions were pooled and concentrated by ultrafiltration in Centriprep 30 concentrators. After the addition of 50% (v/v) glycerol to the purified Glc6P dehydrogenase solution, it was stored at −20 °C.
N-Terminal sequencing of the Glc6P dehydrogenase was performed according to the procedure of Edman  using a Procise Protein Sequencing System (Applied Biosystems).
Native gel electrophoresis of the Glc6P dehydrogenase
Native gel electrophoretis of the crude extract of C. glutamicum ATCC 13032 was performed in Hepes/imidazol-buffered gels at pH 7.4 according to the method of McLellan . Because of the fluorescent reaction product, NADPH, the Glc6P dehydrogenase became visible in the gel at a wavelength of 366 nm under the usual assay conditions.
Kinetic characterization of the Glc6P and 6PG dehydrogenases
The assay mixture was buffered with 50 mm Tes/KOH, pH 7.5, and contained different concentrations of NADP+, 6PG and Glc6P. Unless otherwise stated, all enzyme activities were measured in the presence of 10 mm MgCl2 and 200 mm potassium glutamate to simulate the intracellular environment . The investigations were carried out using a microplate reader (Spectramax 190 Plus, Applied Biosystems) measuring the extinction at 340 nm. The temperature was 30 °C in all cases. The exact concentrations of all metabolites and substrates are given in the results section.
Measurement of specific enzyme activities
The assay system for determination of the Glc6P and 6PG dehydrogenase activities contained 50 mm Tris/HCl, pH 7.5, 10 mm MgCl2, 1 mm NADP+ and 200 mm potassium glutamate. The reaction was initiated by the addition of 4 mm Glc6P or 1 mm 6PG, and the absorption at 340 nm was monitored. Determination of the glutamate dehydrogenase activity was carried out according to a modified method of Meers et al. . The assay mixture contained 200 mm Tris/HCl, 20 mm NH4Cl, 2.7 mm NAD(P)H and 10 mm ketoglutarate. All measurements were carried out at 30 °C.
Using Coomassie brilliant blue  to determine protein concentration in the crude extract a ratio of 30% protein per g dry weight was obtained. This differs from the generally found 50% which would have been obtained using the method of Lowry [22,23].
Determination of intracellular metabolite concentrations
For the intracellular volume of C. glutamicum a value of 2 µL per mg dry weight was used . Intracellular concentrations were calculated on the basis of the cell dry weight and dilution factors during extraction. The concentrations of NADP+ and NADPH were determined by the enzymatic cycling procedure of Bernofsky and Swan . AMP, ADP and ATP were quantified by ion-pair RP-HPLC .
The quantification of Glc6P, 6PG, Fru1,6P2, d-glyceraldehyde 3-phosphate (Gra3P) and Rib5P was performed with a Spectramax 190 Plus microplate reader (Applied Biosystems) using modified procedures of Bergmeyer . In the case of 6PG and Glc6P, the assay mixture contained 125 µL 0.4 m triethylamine buffer, pH 7.6; 5 µL MgCl2 250 mm; 5 µL NAD+ 25 mm; 10 µL 6PG dehydrogenase 12 U·mL−1 or 10 µL Glc6P dehydrogenase 10 U·mL−1 and 125 µL sample.
The assay mixture for quantification of Gra3P and Fru1,6P2 contained 125 µL buffer (0.5 m triethylamine, pH 7.6, 5 mm EDTA), 10 µL NADH 7 mm, 10 µL glycerol 3-phosphatedehydrogenase 3 U·mL−1, 10 µL triosephosphate dehydrogenase 200 U·mL−1 and aldolase 6 U·mL−1. The reaction was started by the addition of triosephosphate isomerase in the case of Gra3P and aldolase in the case of Fru1,6P2. Rib5P was quantified in a system containing 25 mm glycylglycine, pH 7.4; 6 mm MgCl2; 220 µm thiamine pyrophosphate; 4.5 mm sodium arsenate; 3 mm NAD+ 1 mm glycolaldehyde; 12.8 U·mL−1 Gra3P dehydrogenase; 0.32 U·mL−1 transketolase and 2 U·mL−1 Rib5P 3-epimerase.
The Glc6P and 6PG dehydrogenases were purified to homogeneity (Fig. 1) with purification factors of 650 and 1380, respectively (Tables 1 and 2). Surprisingly, the enrichment of the Glc6P dehydrogenase activity resulted in two bands in the SDS gel, suggesting that the Glc6P dehydrogenase exists as a complex of two different proteins. These proteins were N-terminally sequenced and the derived sequences compared with the zwf sequence of C. glutamicum ssp. flavum(Fig. 2). The N-terminal sequence of the 60-kDa protein agreed with the deduced polypeptide of the zwf gene. Surprisingly the N-terminal sequence of the 30-kDa protein was identical with the opcA gene directly upstream of zwf. Native gel electrophoretic separation of the crude extract resulted in a single band, suggesting that only a single multimeric form of the Glc6P dehydrogenase is present in C. glutamicum.
Table 1. Purification of the Glc6P dehydrogenase from C. glutamicum. C, crude extract; IE, DEAE ion exchange chromatography; DA, procion red dye affinity chromatography; LA, 2′5′-ADP ligand affinity chromatography; GF, gel filtration chromatography.
Table 2. Purification of the 6PG dehydrogenase from C. glutamicum. Abbreviations as in Table 1.
Kinetic mechanism of the Glc6P and 6PG dehydrogenases
In order to distinguish between a ping-pong and a sequential reaction mechanism, the dependence of the initial velocity of the Glc6P and 6PG dehydrogenases on their substrate concentrations was investigated. The Lineweaver–Burk plots of both dehydrogenases were linear with respect to each substrate and had a single intersection point (Fig. 3). Therefore it was concluded that both enzymes operate according to a sequential mechanism, i.e. Glc6P dehydrogenase according to a sequential bi-bi mechanism and 6PG dehydrogenase according to a sequential bi-ter mechanism. A general description of two substrate sequential reactions is given by Eqn (1), wherein [A] and [B] denote substrate concentrations according to the order of binding, Vmax the maximal velocitiy under conditions of substrate saturation and KmA, KmB and Kia the Michaelis–Menten and dissociation constants.
To estimate the Michaelis–Menten kinetic constants and the dissociation constant of the first substrate, Eqn (1) was regressed on the enzyme data that consisted of initial reaction velocity vs. substrate concentrations. The fitting was carried out using the program scientist (MicroMath). The results yielded estimates for KmA, KmB and Kia, as shown in Table 3. The Michaelis–Menten constants of the Glc6P dehydrogenase were twofold to fourfold higher than those of the 6PG dehydrogenase. The dissociation constants of the first substrates of both dehydrogenases were between 30 and 60 µm.
Table 3. Kinetic constants of the 6PG and Glc6P dehydrogenases from C. glutamicum. The constants were determined by nonlinear regression with equations 1 or 3, as indicated. Because in the case of the 6PG dehydrogenase the order of substrate addition is unknown two dissociation constants were calculated. All Km values are given in µm.
a If 6PG is the first substrate; b if NADP+ is the first substrate.
34 ± 13
17 ± 3
62 ± 25
34 ± 13
169 ± 23
40 ± 4
33 ± 10
150 ± 21
37 ± 3
45 ± 11
By investigating the inhibition patterns of the products Rib5P and NADPH, more information about the kinetic mechanisms of the two enzymes was obtained. To determine product inhibition constants for NADPH and Rib5P series of experiments with only one substrate varied were conducted at different inhibitor concentrations for both enzymes. A general description of mixed-type inhibition observed in these experiments is given by Eqn (2):
where [S] is concentration of the varied substrate, [I] is concentration of the inhibitor, Km is Michaelis–Menten constant, Kic is competitive inhibition constant and Kiu is uncompetitive inhibition constant. Eqn (2) was regressed on the enzyme data consisting of initial reaction velocity vs. substrate and inhibitor concentrations, using Scientist. The results yielded estimates for Kic, Kiu, for NADPH and Rib5P in the case of 6PG dehydrogenase and for NADPH in the case of Glc6P dehydrogenase (Table 4). The fits with Eqn (2) for the Glc6P dehydrogenase indicated competitive inhibition of NADPH with respect to NADP+. Therefore an ordered bi-bi-mechanism was assumed for Glc6P dehydrogenase, where NADP+ binds first and is released last from the enzyme. Thus, Glc6P dehydrogenase velocity can be described by Eqn (3), where KmGlc6P and KmNADP denote Michaelis–Menten constants, the dissociation constant of NADP+ and the competitive inhibition constant of NADPH. Eqn (3) was regressed on the data using scientist. The resulting estimates for KmGlc6P, KmNADP, KicNADPH and KiNADP are also shown in Tables3 and 4. The results were approx. the same as those obtained with Eqns (1) and (2), so that Eqn (3) is well suited to describe the activity of the Glc6P dehydrogenase. Therefore Eqn (3) was used to predict the in vivo activity of the enzyme.
Table 4. Product inhibition patterns and inhibition constants of the 6PG and Glc6P dehydrogenases.Kic, competitive inhibition constant; Kiu, uncompetitive inhibition constant. All constants have been determined by nonlinear regression with Eqns (2) or (3), as indicated.
a [NADP+]= 50 µm; b [glucose6P]= 750 µm; c [NADP+]= 30 µm; d [6PG]= 150 µm.
With respect to the substrate NADP+ the 6PG dehydrogenase is competitively inhibited by the product NADPH with an inhibition constant of 30 µm. In the case of the 6PG dehydrogenase, Rib5P competes with 6PG for the sugar phosphate binding site, however, with an inhibition constant of 350 µm, which is much larger than Km for 6PG (Tables 3 and 4). In the case of the 6PG dehydrogenase, the competitive inhibitions by NADPH and Rib5P with respect to NADP+ and 6PG were consistent with a Theorell–Chance ordered bi-ter mechanism , characterized by the short half-life of the ternary central complex, which immediately decays while forming the first product (Fig. 4). This was shown by deriving the complete velocity equation according to the rules of King, Altman and Cleland  (Appendix). As the product inhibition pattern is symmetrical with respect to NADPH and Rib5P, the order of substrate addition could not be predicted on the basis of these investigations.
Regulation by metabolites
One group of widespread 6PG and Glc6P dehydrogenase inhibitors show structural similarities to the substrate NADP+, as they all have an adenine phosphate moiety [5,6,31–33]. Therefore palmitoyl-CoA, acetyl-CoA, AMP, ADP, ATP, NAD+ and NADH were tested in this work, of which only ATP, ADP, AMP, acetyl-CoA and NADPH had significant effects (Table 5). However, although acetyl-CoA caused 50% inhibition of the 6PG dehydrogenase at a concentration of 2 mm, reported acetyl-CoA concentrations in C. glutamicum range from 15 to 150 µm, so that the inhibition by this metabolite is considered not important in vivo. No NAD+-dependent activity of either enzyme could be detected and NADH also had no inhibitory effect on any of the enzymes. ATP inhibited the 6PG dehydrogenase competitively with respect to 6PG and mixed with respect to NADP+ (Table 5). AMP and ADP also inhibited the enzyme, but less. In the case of Glc6P dehydrogenase there was no significant inhibition by ATP, ADP and AMP.
Table 5. Regulation of the 6PG and Glc6P dehydrogenases by metabolites. The competitive and uncompetitive inhibition constants were determined by nonlinear regression with Eqn (2).
a [6PG]= 250 µm; b [6PG]= 25 µm; c in the absence of MgCl2; d [NADP+]= 25 µm; e [Glc6P]= 150 µm; f[NADP+]= 45 µm.
Another group of inhibitors of both enzymes from various organisms consists of different sugar phosphates [5,6,32,33], of which Fru1,6P2, 5-phosphorylribose-1-pyrophosphate (Prib-PP), Gra3P, erythrose4-P, Rib5P, Xyl5P, phosphoenolpyruvate and Fru6P were tested (Table 5). The Glc6P dehydrogenase was inhibited by PRib-PP and phosphoenolpyruvate, with 5 mmPRib-PP causing 30% inhibition and 10 mm phosphoenolpyruvate causing 18% inhibition. Increasing Mg2+ concentrations significantly reduced the inhibition of the Glc6P dehydrogenase by Fru1,6P2: at 10 mm Fru1,6P2 and 375 µm MgCl2 the inhibition was 16%, whereas at 10 mm MgCl2 no inhibition could be detected. The inhibition of the Glc6P dehydrogenase by Gra3P (8% at 2.3 mm) might be significant in vivo in the case of fructose-grown cells, as in this case the intracellular concentration is 4 mm, but not in glucose-grown C. glutamicum where the intracellular concentration is 1 mm. For all inhibition studies the substrate concentrations were at their Km values to ensure good detectability of possible competitive inhibition effects.
In contrast to Glc6P dehydrogenase, 6PG dehydrogenase showed a much stronger inhibition by Gra3P (Kic = 1.45 mm), Fru1,6P2 (Kic = 3.6 mm) and also erythrose4P (Kic = 0.735 mm). All were found to be competitive inhibitors with respect to 6PG (Table 5). This showed that the 6PG binding site of the 6PG dehydrogenase has a rather low sugar phosphate specificity. The Kic for Gra3P and Fru1,6P2 are in the range of measured intracellular concentrations , so that these compounds might act as significant inhibitors in vivo. As Kic for erythrose4P is unphysiologically high and detection of this metabolite in cell extracts very problematic , quantification of erythrose4P was not attempted in this study. PRib-PP caused 50% inhibition of the 6PG dehydrogenase at a concentration of 5 mm.
For both enzymes no inhibitory effects of citrate, glyoxylate and isocitrate could be detected. Oxaloacetate caused only 20% inhibition of Glc6P dehydrogenase at 10 mm, whereas no inhibition of the 6PG dehydrogenase by this metabolite could be detected. In cell extracts of C. glutamicum it has not yet been possible to quantify oxaloacetate, which suggests that its intracellular concentration is far below 10 mm. Therefore an in vivo effect of oxaloacetate is unlikely.
Measurement of intracellular metabolite pools and specific Glc6P and 6PG dehydrogenase activities
The results of the metabolite pool determinations for the homologous and heterologous glutamate dehydrogenase mutants cultivated continuously under glucose limitation are shown in Table 6.
Table 6. Intracellular concentrations of relevant metabolites in the homologous and heterologous glutamate dehydrogenase mutant of the lysine-producing C. glutamicum strain MH20-22B. All metabolites except NADPH were quantified after perchloric acid extraction. In order to quantify NADPH an extraction in ethanolic potassium hydroxide solution was carried out. Metabolite concentrations were calculated using the intracellular volume of 2.0 µL·mg−1 dry cell mass as determined by Gutmann et al.. All values in mm. ND, not determined.
Homologous glutamate dehydrogenase mutant
Heterologous glutamate dehydrogenase mutant
0.50 ± 0.08
0.54 ± 0.08
0.95 ± 0.15
0.95 ± 0.10
0.82 ± 0.16
0.84 ± 0.17
1.4 ± 0.3
2.0 ± 0.3
0.48 ± 0.03
0.62 ± 0.03
0.118 ± 0.005
0.29 ± 0.03
0.13 ± 0.01
0.260 ± 0.015
0.27 ± 0.03
Although pentose phosphate pathway flux in the homologous glutamate dehydrogenase mutant was reported to be three times higher than in the heterologous glutamate dehydrogenase mutant , our data revealed the remarkable fact that Glc6P was identical in both strains, indicating that Glc6P dehydrogenase flux is not regulated by the Glc6P concentration in vivo.
6PG however, was at least six times higher in the homologous glutamate dehydrogenase mutant suggesting an increased flux through the 6PG dehydrogenase. Fru1,6P2 is a competitive inhibitor with respect to 6PG (Table 5). Therefore, although the Fru1,6P2 concentration was similar in both strains (Table 6), the inhibition by this metabolite is weaker in the case of the homologous glutamate dehydrogenase mutant because of the higher 6PG concentration. However, as the intracellular Fru1,6P2 concentration is significantly lower than its Ki values for the competitive inhibition (Tables 5 and 6) of 6PG dehydrogenase in both strains, no great activity modulation effect of this metabolite is expected in vivo under our conditions. In the case of the heterologous glutamate dehydrogenase mutant, inhibition of the 6PG dehydrogenase by Gra3P is increased because, on the one hand, its intracellular concentration increased up to the Ki range and, on the other hand, the 6PG concentration was significantly lower. In the heterologous glutamate dehydrogenase mutant no Rib5P was detectable, which suggests that this metabolite does not participate in down-regulation of the 6PG dehydrogenase. This also was true for both dehydrogenases in the case of AMP and ADP as their concentrations are well below the respective Ki values (Table 5). The ATP concentration (1.4–2 mm) was in the range of the Ki values for 6PG dehydrogenase. Therefore this compound might be expected to play an important role in regulating the activity of this enzyme in vivo. In the case of the Glc6P dehydrogenase, ATP inhibition is primarily competitive with respect to both substrates and, as the intracellular substrate concentrations were found to be greater than the respective Km values (Table 3), probably not significant in the homologous and heterologous glutamate dehydrogenase mutants. The value of 0.8 calculated for the energy charge of the homologous glutamate dehydrogenase mutant is representative of growing cells .
It is surprising that the concentration of the most potent inhibitor NADPH was identical within experimental error in both strains, and that its concentration (260–270 µm) was far above the Ki values of both dehydrogenases. The most striking difference between the two strains with regard to Glc6P and 6PG dehydrogenase effectors is the concentration of NADP+, which was reduced twofold in the heterologous glutamate dehydrogenase mutant. Nevertheless, it was still three times as great as the Km value of 40 µm (Table 3). These observations suggest that it is not the direct NADP+ or NADPH concentrations which influence the activities of the Glc6P- and 6PG dehydrogenase, but rather their concentration ratio, as will be shown below.
Specific activities during metabolic steady state for the homologous and heterologous glucose dehydrogenase mutants, respectively, of the Glc6P dehydrogenase were 69 ± 3 and 45 ± 6 nmol per min per mg dry weight, whereas for the 6PG dehydrogenase 105 ± 6 and 72 ± 6 nmol per min per mg dry weight was found. Thus, activities of both enzymes were 30% lower in the heterologous glutamate dehydrogenase mutant in agreement with results of 13C labelling experiments  that have shown that the pentose phosphate pathway flux in the heterologous glutamate dehydrogenase mutant was strongly reduced.
Prediction of the in vivo pentose phosphate pathway activity of the homologous and heterologous glutamate dehydrogenase-mutants
Based on the kinetic data presented (Tables 3 and 4) and Eqn (3), the in vivo activity of the Glc6P dehydrogenase was calculated from measured pool sizes (Table 6), whereby the specific activity of the Glc6P dehydrogenase as determined in the crude extract was used as Vmax value. The pentose phosphate pathway flux was assumed to be identical to the calculated activity of the Glc6P dehydrogenase activity. For the homologous mutant, 17.1 nmol per mg dry weight per min were calculated against 6.0 nmol per mg dry weight per min for the heterologous mutant. These results are in remarkable agreement with the fluxes through the oxidative part of the pentose phosphate pathway of 15.2 and 4.2 nmol per mg dry weight per min, respectively, as previously determined by 13C labelling . From the metabolite pool measurement imprecisions (Table 6) alone, it can be calculated that the error in the flux estimate is about 30%. Incorporation of the imprecisions in the Km and Ki estimates will further increase the possible error. While the individual calculations therefore may be of limited importance, the difference in pentose phosphate pathway fluxes found in the comparison of the two strains is still very significant. With these reservations in mind, we can use Eqn (3) to predict the influence of different metabolite concentrations on the activity of the Glc6P dehydrogenase in vivo. As the intracellular NADPH concentration in C. glutamicum is much greater than the respective and Ki value (Tables 4 and 6), the activity of the Glc6P dehydrogenase can elegantly be described as a function of the [NADP+]/[NADPH] ratio. This can be seen by transforming Eqn (3) for the limiting case that [NADPH] ≫KicNADPH:
with R = [NADP]/[NADPH]. This equation describes a familiar saturation curve (Fig. 5A) with an apparent dimensionless Km of 2 for the [NADP]/[NADPH] ratio R. Interestingly, the data for the homologous and heterologous mutant [(R = 1.11 and 0.48, respectively (Table 6)] lie in the most sensitive regulatory domain for the pentose phosphate pathway. In a second simulation, intracellular metabolite concentrations of NADP+ and NADPH as measured in the homologous glutamate dehydrogenase mutant (Table 6) were taken as a basis for an estimation of the pentose phosphate flux sensitivity towards the Glc6P concentration (Fig. 5B). The results showed that the flux is insensitive to alterations of the Glc6P concentration unless the latter falls below 500 µm.
Summarizing, the quantification of metabolite pools and specific enzyme activities as well as simulations showed that the regulation of the oxidative part of the pentose phosphate pathway of C. glutamicum mainly occurs by different specific activities of the 6PG and Glc6P dehydrogenases and by changes in the ratio of NADPH and NADP+ concentrations, rather than by Glc6P and ATP.
Glc6P dehydrogenase complex
By purification of the Glc6P dehydrogenase of C. glutamicum we have shown for the first time that the protein products of the zwf and opcA (oxidative pentose phosphate cycle) genes form a stable complex, which is not disrupted during the whole purification procedure. The opcA gene has also been found in other microbial organisms such as Mycobacterium and Cyanobacteria  and they all have homologous sequences. According to Sundaram , the OpcA protein in Cyanobacteria may be involved in the oligomerization process of zwf monomers. Whereas multiple molecular forms of the cyanobacterial Glc6P dehydrogenase have been found , only a single heteromultimeric form of the Glc6P dehydrogenase was detected in C. glutamicum after native gel electrophoretic separation of the crude extract in the present study. As mutations within the opcA gene cause loss of the cyanobacterial Glc6P dehydrogenase activity [37,38], it is expected that this protein is also essential in the case of C. glutamicum. No remarkable similarities of the OpcA proteins to other known proteins were found.
Regulation of the pentose phosphate pathway by intracellular metabolites
NADPH is a widespread inhibitor of the Glc6P and 6PG dehydrogenases from many organisms [32,33]. The observed absence of any influence of NAD+ and NADH on the Glc6P or 6PG dehydrogenase activities in this study was also reported for C. glutamicum ssp. flavum[5,6] and is consistent with investigations indicating that the 2′-phosphate on NADP+ plays a crucial role in coenzyme binding [39,40].
In the case of the 6PG dehydrogenase, the ATP inhibition constant was in the range of intracellular concentrations of the homologous and heterologous glutamate dehydrogenase mutants, suggesting that ATP is a potential regulator of the pentose phosphate pathway activity in microorganisms. This conclusion is corroborated by a study of S. cerevisiae, for which it was found that not only NADPH but also ATP is a key metabolite for directing the flux through the pentose phosphate pathway . ATP inhibition has also been found for the enzymes of Candida boidinii, Pseudomonas C and Gluconobacter suboxydans. In the case of AMP and ADP the inhibition was weaker and probably irrelevant in vivo, as the intracellular concentrations were significantly lower than the respective Ki values. As the inhibitions by AMP, ADP and ATP are all competitive with respect to 6PG and mixed with respect to NADP+, these metabolites may be dead-end inhibitors of the 6PG dehydrogenase competing with 6PG for addition to the free enzyme E.
The inhibition studies with adenine nucleotides were carried out in the absence of Mg2+, as it is known that the inhibition is significantly reduced by Mg2+[32,41]. As 80–90% of intracellular ATP is complexed with Mg2+[44,45], ATP inhibition is likely to be weaker under in vivo conditions.
Our data indicate that the Glc6P dehydrogenase of C. glutamicum is only weakly inhibited by Gra3P and Fru1,6P2 even at the high intracellular concentrations reported under conditions of glucose excess .
The 6PG dehydrogenase was also not significantly inhibited during glucose-limited cultivation of C. glutamicum, as the intracellular concentrations of Fru1,6P2 and Gra3P were lower than the respective Ki values (Tables 5 and 6). However, during exponential growth on glucose, the intracellular concentrations of Fru1,6P2 and Gra3P (23 and 1 mm, respectively ) are greater than these inhibition constants, so that in this case the inhibition by these sugar phosphates will become physiologically relevant according to our data. This is supported by the reduced flux through the pentose phosphate pathway as determined by 13C labelling  in the case of growth on fructose, where the intracellular concentrations of Gra3P and Fru1,6P2 are increased still further (46 and 4 mm, respectively ). Fru1,6P2 is often described as an inhibitor of the 6PG dehydrogenase, e.g. from C. boidinii, sheep liver , Streptococcus faecalis and G. suboxidans.
In contrast to the Glc6P and 6PG dehydrogenases of C. glutamicum ssp. flavum[5,6], no significant inhibition by oxaloacetate could be found in the present study. As Shiio and Sugimoto showed that the inhibition by oxaloacetate may disappear due to purification procedures [5,6] the inhibition was also examined with the partially purified enzymes after DEAE ion-exchange chromatography, a step that reportedly has no influence on the inhibition by oxaloacetate [5,6]; the inhibition was not increased in this case either.
Regulation of the carbon flux at the branch point glycolysis/pentose phosphate pathway
As far as the oxidative part of the pentose phosphate pathway is concerned, the Glc6P dehydrogenase must be regarded as the rate-controlling enzyme rather than the 6PG dehydrogenase, because the rapid hydrolysis of 6-phosphoglucono-δ-lactone, which in yeast has been shown to proceed spontaneously , in effect makes the reaction irreversible . Therefore the regulation of this enzyme must be considered in detail.
Changes of the intracellular Glc6P concentration may strongly influence the distribution of the carbon flux between glycolysis and the pentose phosphate pathway by control of the first glycolytic enzymes . During glucose limitation the intracellular Glc6P concentration was around 500 µm, which is 2.5 times its respective Km value (Tables 3 and 6). Nevertheless, the activity of the Glc6P dehydrogenase and thus that of the oxidative part of the pentose phosphate pathway may still be weakly sensitive towards Glc6P pool alterations (Fig. 5B). This is different during batch cultivations of C. glutamicum with excess glucose supply  and in glucose pulse experiments with the yeast S. cerevisiae, where intracellular Glc6P concentrations are very much higher. In these cases, the activity of the Glc6P dehydrogenase cannot be altered by moderate changes of the Glc6P pool (Fig. 5B). Although the pentose phosphate pathway fluxes in the homologous and heterologous glutamate dehydrogenase mutants differ by a factor of three, the Glc6P concentrations are unchanged indicating that this pathway is not regulated by Glc6P even in glucose-limited C. glutamicum.
In contrast, our results clearly indicate that the in vivo activity of the Glc6P dehydrogenase is primarily regulated by the specific enzyme activities and the concentration ratio [NADP+]/[NADPH] (Fig. 5A). It is therefore probably adequate to state that the oxidative pentose phosphate pathway in C. glutamicum is feedback-regulated by its main product NADPH (i.e. driving power for reductive biosyntheses), just as glycolysis can be regarded as being feedback-regulated by its main product ATP (energy) via the ATP/ADP ratio . This stresses the importance of these key metabolites for carbon flux regulation in the central metabolism.
The feedback regulation implies that the pentose phosphate pathway of C. glutamicum can respond very flexibly to altered NADPH demands, as was indeed demonstrated in the present work by the kinetic characterization of the two dehydrogenases in combination with the quantification of intracellular substrate and product levels. This flexibility is also well documented in our previous studies [1,2] employing 13C labelling and NMR.
The fact that the specific enzyme activities determined are four to eightfold in excess of the actual carbon flux suggests that NADPH availability is very unlikely to become a limiting factor in the metabolism of C. glutamicum. This is supported by results showing that an excess of NADPH generation by gluconate catabolism does not increase lysine yield .
The pentose phosphate pathway flux calculated using Eqn (3) developed in this work from the enzyme kinetic constants determined and the measured intracellular metabolite concentrations agreed remarkably well with the carbon fluxes obtained in previous 13C labelling experiments . While more experiments are needed to verify the general applicability of our model, the results shown demonstrate that it can be used to adequately predict at least the steady-state pentose phosphate pathway activity from metabolite pool measurements in C. glutamicum.
We thank Lothar Eggeling for critical reading of the manuscript. This research was funded by the German Federal Ministry of Education and Research.
Derivation of the complete velocity equation of the Theorell–Chance ordered bi-ter mechanism was carried out according to the rules of King & Altmann  and Cleland [50–52] following the procedure outlined in Segel . The reaction scheme is illustrated in Fig. 6.
The complete velocity equation is given by
Therein, Kmx denotes Michaelis–Menten constants, Kix denotes inhibition constants and Vf denotes maximal velocity in the forward direction. A, B, P, Q and R are defined as in Fig. 6.