In Vivo Fluxes in the Ammonium-Assimilatory Pathways in Corynebacterium glutamicum Studied by ¹⁵N Nuclear Magnetic Resonance

Abstract

Glutamate dehydrogenase (GDH) and glutamine synthetase (GS)–glutamine 2-oxoglutarate-aminotransferase (GOGAT) represent the two main pathways of ammonium assimilation in Corynebacterium glutamicum. In this study, the ammonium assimilating fluxes in vivo in the wild-type ATCC 13032 strain and its GDH mutant were quantitated in continuous cultures. To do this, the incorporation of ¹⁵N label from [¹⁵N]ammonium in glutamate and glutamine was monitored with a time resolution of about 10 min with in vivo ¹⁵N nuclear magnetic resonance (NMR) used in combination with a recently developed high-cell-density membrane-cyclone NMR bioreactor system. The data were used to tune a standard differential equation model of ammonium assimilation that comprised ammonia transmembrane diffusion, GDH, GS, GOGAT, and glutamine amidotransferases, as well as the anabolic incorporation of glutamate and glutamine into biomass. The results provided a detailed picture of the fluxes involved in ammonium assimilation in the two different C. glutamicum strains in vivo. In both strains, transmembrane equilibration of 100 mM [¹⁵N]ammonium took less than 2 min. In the wild type, an unexpectedly high fraction of 28% of the NH₄⁺ was assimilated via the GS reaction in glutamine, while 72% were assimilated by the reversible GDH reaction via glutamate. GOGAT was inactive. The analysis identified glutamine as an important nitrogen donor in amidotransferase reactions. The experimentally determined amount of 28% of nitrogen assimilated via glutamine is close to a theoretical 21% calculated from the high peptidoglycan content of C. glutamicum. In the GDH mutant, glutamate was exclusively synthesized over the GS/GOGAT pathway. Its level was threefold reduced compared to the wild type.

Corynebacterium glutamicum is an important industrial amino-acid-producing bacterium. This organism is known to possess very high intracellular concentrations of glutamic acid, i.e., up to 200 mM (24). In microorganisms, glutamate can be synthesized by two alternative pathways (12). NADP-dependent glutamate dehydrogenase (GDH) catalyzes the reductive amination of 2-oxoglutarate to glutamate (GDH pathway). Glutamate can also be formed by the consecutive reactions of glutamine synthetase (GS) and glutamine 2-oxoglutarate-amidotransferase (GOGAT) (i.e., the GS/GOGAT pathway). GS catalyzes the ATP-dependent amination of glutamate to glutamine. GOGAT catalyzes the reductive transfer of the δ-amidogroup of glutamine to 2-oxoglutarate with concomitant oxidation of NADPH. All of these ammonium-assimilating enzymes have been described in amino-acid-producing bacteria (41, 46, 50), indicating that both pathways of ammonium assimilation may function simultaneously in C. glutamicum.

It has long been assumed that C. glutamicum requires GDH as a key enzyme for NH₄⁺ assimilation during glutamate production. However, a GDH mutant of C. glutamicum was recently constructed, and it was shown that neither growth nor glutamate production was impaired in this mutant (4). While it seems plausible to assume that the GS/GOGAT pathway was responsible for ammonium assimilation in that case, no detailed tests to verify whether alternative (e.g., alanine dehydrogenase) routes may be active were performed.

While determination of enzyme activities in crude extracts may give a first clue as to which ammonium assimilation pathways are active, they cannot be used to predict the in vivo flux distribution over competing enzyme systems such as GDH and GS/GOGAT.

In vivo nuclear magnetic resonance (NMR) spectroscopy, especially when used in combination with stable isotope labeling, does allow the characterization of metabolic activities in the living cell (34, 42). With ¹⁵N-NMR, important metabolites such as glutamate, glutamine, alanine, lysine, aspartate, and N-acetylglutamate have been detected in bacteria, fungi, and algae (13, 25, 27). The relative importance of especially the GDH and the GS/GOGAT pathways in NH₄⁺ assimilation has been studied successfully by using the ¹⁵N-NMR technique in, e.g., Bacillis macerans and B. polymyxa (16, 17), B. azotofixans (18), Aspergillus nidulans (26), Clostridium kluyverii and C. butyricum (19), and Agaricus bisporus (2). All of these studies used cell extracts taken at different time points after incubation with ¹⁵N-labeled substrates to identify the principal metabolites and pathways involved in ¹⁵N assimilation in a mainly qualitative manner. In contrast, detailed quantitative studies of nitrogen flux distribution in vivo in microorganisms have not yet been reported.

Although the analysis of carbon fluxes in the central metabolism by metabolite balancing (44, 51) and computer-aided stationary-state isotopic analysis of ¹³C-labeled intermediates (30, 54, 55) nowadays may be considered as an established technique that allows the differentiation between parallel pathways (31, 43) and even the identification of bidirectional metabolic fluxes (30, 48, 54, 55), it is not well suited for nitrogen flux analysis. This is due to the fact that nitrogen labeling of metabolic intermediates upon prolonged incubation with a one-nitrogen substrate (e.g., urea, ammonium, glutamate, or aspartate) does not show any positional effects, i.e., at isotopic steady state the nitrogen enrichment in all intermediates is equal to that of the substrate, independent of the fluxes. Although the use of extensive modelling of glutamate metabolism (29) allows determination of some nitrogen fluxes (e.g., GDH [8]) on the basis of glutamate ¹³C enrichments, a complete analysis of the ammonium assimilation flux network as envisaged for the present study requires time-resolved measurements of ¹⁵N label incorporation kinetics.

While such experiments were hitherto practically impossible due to the very low sensitivity of conventional ¹⁵N-NMR techniques, which do not allow for a useful time resolution in dynamic metabolic studies, recent developments in the field of integrated NMR and fermentation equipment fortunately have opened new perspectives. A continuous-flow NMR bioreactor enabled monitoring of the metabolism of anaerobic Zymomonas mobilis with in vivo ³¹P-NMR at fourfold-increased sensitivity compared to conventional NMR approaches (9). A further development of this system resulted in a hydrocyclone bioreactor that permitted the continuous aerobic cultivation of C. glutamicum at a cell density of 25 g (dry weight)/liter allowing the metabolism to be monitored with in vivo ¹³C-NMR at a time resolution of 10 min (14). In view of the high glutamate pool in C. glutamicum, this system seemed also to offer good perspectives for application of in vivo ¹⁵N-NMR.

This study is concerned with the detailed quantitative investigation of ammonium assimilation in C. glutamicum wild type (ATCC 13032) and its GDH mutant. The aim was to determine over which pathways these C. glutamicum strains assimilate ammonium and to quantify the primary nitrogen fluxes in both strains in vivo.

MATERIALS AND METHODS

Bacterial strains and growth conditions.

The experiments were performed with C. glutamicum wild type (ATCC 13032) and a GDH-negative mutant C. glutamicum EB1 (4). Continuous cultures were run with a fully synthetic medium as described elsewhere (20), modified for the glucose concentration (100 g/liter) and the (NH₄)₂SO₄ concentration, which was varied between 10 and 15 g/liter by separate dosage. The medium was inoculated with washed cells from an overnight culture in BHI medium (Difco) to give an initial optical density at 600 nm (OD₆₀₀) of about 5. Fermenter operation in the continuous cultivation mode was started directly after inoculation. Thus, growth to high cell densities took place during continuous culture operation and not in a batch phase. All of the continuous culture fermentations reported here were performed in vivo in the NMR bioreactor at a high cell density of 50 g (dry weight)/liter. The medium dilution rate was set at 0.1 h⁻¹, and the growth rate was set to 0.05 h⁻¹ by adjusting the bleed flow. In the fermentations, pH was controlled at 7.0 by using 2 M NaOH, and the temperature was kept at 30°C. The pO₂ was regulated at 20% of air saturation. The continuous cultures were aerated with a 50% (vol/vol) N₂-O₂ mixture to reduce gas input to the fermentation broth.

Preparation of crude extracts and enzyme assays.

Cells from 2 ml of culture broth were harvested by centrifugation (10 min, 6,000 × g), subsequently washed two times in 40 ml of washing buffer solution (for specifications, see below), and resuspended in 1 ml of sonication buffer (for specifications, see below). Sonication (UP 200 S sonifier; Dr. Hielscher GmbH, Teltow, Germany) was performed at 0°C for 5 min. Cell debris were removed by centrifugation (13,000 rpm, 30 min, 4°C), and the resulting cell extract was used for enzyme assays. The protein concentration was determined by the biuret method with bovine serum albumin (Boehringer Mannheim) as the standard.

For GDH assays, 200 mM potassium phosphate (pH 7.0) was used as a washing buffer, whereas the sonication buffer contained 50 mM potassium phosphate (pH 7.5), 1 mM EDTA, 10 mM l-cysteine, and 10 μM 2-oxoglutarate. The specific GDH activity was determined by a modified method of Meers et al. (32). The assay (1 ml) contained 100 mM Tris-HCl (pH 8.0), 20 mM NH₄Cl, 0.25 mM NADPH+H⁺, and 50 μl of crude extract. The reaction was started by the addition of 10 mM 2-oxoglutarate, and the extinction decrease at 340 nm was monitored for 2 min at 30°C.

For GOGAT assays, 50 mM potassium phosphate (pH 7.5) was used as a washing buffer, whereas the sonication buffer contained 50 mM potassium phosphate (pH 7.5), 1 mM EDTA, 10 mM l-cysteine, and 10 μM 2-oxoglutarate. To eliminate any competing GDH activity, residual ammonium was removed by first applying 1 ml of crude extract on a gel chromatography column (PD10; Pharmacia, Freiburg, Germany) previously equilibrated with 25 ml of ice-cold sonication buffer, then by elution with 3 ml of sonication buffer, and finally by concentration to a 1-ml volume by subsequent membrane filtration (Centricon 10; Amicon). The specific GOGAT activity was determined again by using a modified method of Meers et al. (32). The assay (1 ml) contained 100 mM Tris-HCl (pH 7.6), 0.1 mM dithiothreitol, 0.25 mM NADPH+H⁺ (sodium salt), 10 mM 2-oxoglutarate, and 50 μl of crude extract in appropriate dilution. The reaction was started by the addition of 10 mM l-glutamine, and the extinction decrease at 340 nm was monitored for 2 min at 30°C.

For GS assays, 100 mM imidazole (pH 7.15) was used both as washing buffer and as sonication buffer. The specific GS activity was determined by using a nonphysiological discontinuous test modified from Shapiro and Stadtman (40), where GS catalyzes the reaction of glutamine plus hydroxylamine to γ-glutamyl hydroxamate in the presence of ADP and arsenate. The assay (400 μl, 30°C) contained 135 mM imidazole-HCl (pH 7.15), 18 mM NH₂OH, 0.27 mM MnCl₂, 25 mM KH₂SO₄, 0.36 mM ADP, and 1 to 50 μl of crude extract. The reaction was started by the addition of 20 mM l-glutamine and stopped after 10 min by the addition of 1 ml of stop solution containing 55 mg of FeCl₃ · 6H₂O, 20 mg of trichloroacetic acid, and 21 μl of 37% HCl per ml. The precipitate was removed by centrifugation (2 min, 13,000 rpm), and the γ-glutamyl hydroxamate formed by the GS reaction was determined photometrically at 540 nm.

NMR spectroscopy.

All NMR experiments were performed by using an AMX-400 WB spectrometer system (Bruker, Karlsruhe, Germany).

(i) In vivo ¹⁵N-NMR.

Cells were cultivated in a continuous-flow membrane bioreactor system equipped with a hydrocyclone reactor vessel adapted for defined bypass volume flow as described before (14). NMR experiments were started after at least 50 h of continuous cultivation. The reactor vessel was introduced into the magnet approximately 1 h before the start of the ¹⁵N experiments. To reduce the amount and size of the gas bubbles in the NMR-sensitive region, the system flow was then increased to 800 liters/h, resulting in a better gas-liquid separation. ¹⁵N-NMR was carried out at 40.5 MHz by using a custom-tailored 20-mm ³¹P-¹³C-¹⁵N-NMR probehead (Bruker Spectrospin, Faellanden, Switzerland). Shimming was performed under flow conditions by using the water signal measured with the detuned ¹H decoupling coil. Line widths of 10 to 12 Hz for ¹⁵N were routinely achieved. The extremely long ¹⁵N longitudinal relaxation times (the T₁ of ammonium-¹⁵N, for example, was determined to be ca. 50 s) forced use of continuous broadband ¹H decoupling, exploiting the large (but negative) ¹⁵N nuclear Overhauser effect (NOE). Although under our continuous-flow conditions only a fraction of the theoretically maximal NOE can build up inside the NMR measurement chamber in the sensitive volume of the decoupling coil, sensitivity was still much better than without the NOE. The following ¹⁵N pulsing conditions were found to be optimal: 90° pulses (70 μs) with a recycle delay of 200 ms, 20.8-kHz sweep width, 2,048 complex data points, continuous Waltz-16 composite pulse broadband proton decoupling (90° ¹H pulse, 200 μs). A total of 4,440 and 2,220 scans per spectrum were used for wild-type and GDH mutant strains, resulting in a time resolution of 15 and 7.5 min, respectively. One initial ¹⁵N spectrum was recorded just before the addition of the labeled ammonium. The recording of a series of consecutive ¹⁵N spectra (a 3-h total NMR measurement time) was started immediately after application of a single pulse of (¹⁵NH₄)₂SO₄ (>99% ¹⁵N; Cambridge Isotope Laboratories). The spectra were processed without line broadening. Peak areas were determined by using the deconvolution package of the UXNMR spectrometer software (Bruker, Karlsruhe, Germany).

(ii) ¹⁵N-NMR of ethanolic cell extracts for calibration.

Samples of cell suspension (10 ml) were drawn from the reactor for amino acid pool size quantification and ¹⁵N-NMR analysis at the following time points after addition of the labeled (¹⁵NH₄)₂SO₄: 1, 2, 3, 4, 5, 10, 15, 20, 25, 30, 45, 60, 75, 90, 105, 120, 150, and 165 min. This loss of biomass from the system was partially compensated for by the cell growth, such that after 2.5 h only about a 10 to 15% loss of cell density was anticipated. The samples were injected and kept incubated in 30 ml of boiling absolute ethanol for 5 min and were subsequently kept on ice (0°C) for 15 min. They were then centrifuged (6,000 × g, 20 min, 4°C), and the supernatant was lyophilyzed. The lyophilysate was redissolved in 4 ml of distilled water. Then, 2.8 ml of this solution was transferred to a standard 10-mm NMR tube holding a coaxial 5-mm NMR tube with D₂O and 1 M K¹⁵NO₃ as a chemical shift reference and concentration standard. The ¹⁵N spectra of these cell extracts were run with the following NMR parameters: 70° pulses (15 μs) with a 6-s recycle delay, 20.8-kHz sweep width, 32,768 complex data points, and continuous Waltz-16 composite-pulse broadband proton decoupling (90° ¹H pulse, 200 μs). A total of 4,000 scans per spectrum were averaged. For concentration calibration, 50 mM ¹⁵N-labelled glutamic acid (99% ¹⁵N; Cambridge Isotope Laboratories) was added to a sample obtained 90 min after the addition of the (¹⁵NH₄)₂SO₄, the pH was adjusted to 7.0, and the increase in the ¹⁵N_α-NMR signal was determined. Thereafter, 25 mM ¹⁵N₂-labelled glutamine (99% ¹⁵N₂; Cambridge Isotope Laboratories) was added to the same sample at a constant pH, and the signal increase of the ¹⁵N_α and ¹⁵N_δ of the glutamine, as well as of the ¹⁵N_α of the glutamate, which was resolved from that of the ¹⁵N_α of the glutamine, was determined. It appeared that the peak area per millimolar concentration of ¹⁵N was equal for the α-amino nitrogens of glutamate and glutamine but was 1.5 times larger for the δ-amino nitrogen of glutamine due to different NOEs. Spectra were routinely processed with 1-Hz line broadening, and peak areas were determined by interactive integration by using the spectrometer software. In cases where it was needed, Gaussian resolution enhancement was used to resolve overlapping peaks. Resonance assignments were made by comparison with literature data (25). After ¹⁵N-NMR analysis, the cell extracts were filtered and the amino acid concentrations were determined by reversed-phase high-pressure liquid chromatography (HPLC) (LC 1090 HPLC; Hewlett Packard, Waldbronn, Germany) after automatic ortho-phtaldialdehyde precolumn derivation. For pool size determinations, the cytoplasmatic volume in the culture was measured by the ³H₂O–¹⁴C-taurine method (38).

(iii) Determination of ammonium ¹⁵N enrichment.

Next, 600 μl of the redissolved lyophilysate (see above) was transferred to a standard 5-mm NMR tube, and the pH was adjusted to ca. 1. A single-scan proton NMR spectrum was obtained by using a 90° pulse, an 8-kHz sweep width, and 8,000 complex data points to determine the percent ¹⁵N enrichment of ammonium as described elsewhere (36).

(iv) Calibration of ¹⁵N enrichments in glutamate and glutamine.

In order to determine the percent ¹⁵N enrichments in cytosolic glutamine and glutamate at key time points, the amino acids of several cell extracts were fractionated by cation-exchange chromatography on an Ultrapac 11-μm resin column (Pharmacia Biotech GmbH; Freiburg, Germany) (43) with triethylamine buffer (0.2 M, pH 3.2 to 10.5) for elution. Freeze-dried powders of the amino acids were dissolved in 700 μl of D₂O and passed through a 0.2-μm DynaGard filter (Microgon Inc., Laguna Hills, Calif.) to remove insoluble impurities, and the pH was carefully adjusted to 7.0. ¹⁵N enrichments in glutamine and glutamate were determined by ¹H-NMR by using a heteronuclear spin-echo difference spectroscopy protocol adapted from earlier studies (45, 53). The basic pulse sequence schematically reads: ¹H: 90°_X − T_H − (90°_Y − τ_A − 90°_−Y) − T_H − Acquire¹⁵N: − ΔT − (90°_X − τ_B − 90°_−X) − T_N −

The use of two independent spin echo delays T_H and T_N for ¹H and ¹⁵N, respectively, allows for an independent adjustment of homonuclear and heteronuclear ³J_NH scalar coupling modulation effects to achieve an optimal signal-to-noise ratio. Two spectra are acquired for each determination, with the ¹⁵N semiselective 180 pulse (90°_X − τ_B − 90°_−X) only applied for the second spectrum. Application of this pulse results in a selective inversion of the ¹⁵N-coupled proton signals compared to the first spectrum. The efficiency of this inversion was calibrated with pure amino acid standards (typically 85%). The difference spectrum yields only the ¹⁵N-coupled proton signals. The optimal timing and frequency settings used for the determinations are given in Table 1. Spectra were measured with the following parameters: 8-kHz sweep width, 16,384 complex data points, 15-s recycle delay, and 32 to 64 scans. Low-power presaturation was applied to suppress the residual water signal. Spectra were processed without line broadening. Peak areas were determined by interactive integration.

TABLE 1.

NMR heteronuclear spin-echo difference pulse sequence (see text) parameters for the determination of ¹⁵N enrichments in extracted cytosolic glutamate and glutamine

Objective	1H carrier position (ppm)	15N carrier position (ppm)	Pulse sequence parameters (ms)
			TH	τA	TN	τB
15Nα glutamate	2.48	−264	203	3.7	170	0.173
15Nα glutamine	2.44	−264	220	4.2	170	0.173
15Nδ glutamine	2.14	−335	548	4.2	544	0.173

Estimation of the principal ammonium assimilating fluxes.

The time-dependent incorporation of ¹⁵N in cytosolic glutamine and glutamate observed with in vivo NMR was described by using a model of differential equations. In this model of the principal ¹⁵N ammonium assimilation pathways in C. glutamicum, the following reactions were included: (i) reversible ammonium diffusion over the cell membrane, characterized by a rate constant K_in for the resulting import of ammonium into the cell and a rate constant K_ex for export from the cell; a constant net uptake of ammonium F₀; the GDH reaction, formulated as a bidirectional flux F_1X superimposed on a unidirectional net flux F₁; the GOGAT reaction, formulated as a bidirectional flux F_2X superimposed on a unidirectional flux F₂; the GS reaction, formulated as an irreversible flux F₃; a net flux of glutamate consumption F₄, accounting for transamination of glutamate in a variety of biosyntheses (e.g., of amino acids as well as for incorporation of glutamate in cell protein and for the build-up of its cytoplasmic pool); a net consumption F₅ of glutamine, accounting for the build-up of its intracellular pool and for incorporation in the cell protein; and a net flux of amidotransferase of glutamine to glutamate F₆, accounting for glutamine functioning as a nitrogen donor in the biosynthesis of, for example, nucleotides, cell wall components, and aromatic amino acids (12). Incorporation of N into the biomass was considered irreversible. The model is schematically shown in Fig. 1.

FIG. 1.

Model of the nitrogen fluxes involved in ammonium assimilation in C. glutamicum. This model was fitted to the experimentally determined ¹⁵N-NMR data in order to estimate the fluxes in vivo. Each rectangle represents a specific nitrogen atom pool.

For each nitrogen atom pool, the rate of change of the ¹⁵N content was formulated in a differential equation form as the difference of incoming and outgoing contributions, e.g., for the N_α and N_δ pools of glutamine:

As initial conditions, the ¹⁵N content of all intracellular nitrogen pools was set to zero, whereas that of the extracellular NH₄⁺ pool was set to the value that was measured after the addition of the isotopically enriched substrate. It was assumed that all intracellular metabolite pools remained constant during the experiment. Therefore, the network depicted in Fig. 1 dictates the following linear constraints for the nitrogen fluxes: F₁ + F₃ = F₀, the measured specific ammonium uptake rate; F₁ + 2F₂ + F₆ = F₃ + F₄; and F₆ = F₃ − F₂ − F₅. This constrained model was fitted to the experimental ¹⁵N-NMR data by using the least-squares parameter estimation routine implemented in the software program Scientist (MicroMath Scientific Software, Salt Lake City, Utah) on an IBM PC.

The model used the following as constants: (i) the percent ¹⁵N labeling of the substrate ammonium, (ii) the specific ammonium uptake rate (F₀), and (iii) the net consumption of glutamine for biosynthetic purposes (F₅). The latter was derived from the measured glutamine pool size and the codon usage of a selection of important genes from C. glutamicum. The model parameters fitted to the data were as follows: (i) one of the two net fluxes F₁ or F₃, (ii) F₂, (iii) the exchange fluxes F_1X or F_2X, (iv) the cytoplasmic pool sizes of glutamate and glutamine (constrained to remain in the experimentally determined error intervals), (v) the rate of decrease of the cell density in the reactor due to the sample taking, and (vi) proportionality factors for the NMR peak areas. The estimated biomass dilution rate due to sample taking was 8% per h for the wild-type fermentation and 12% per h for the GDH mutant fermentation. For the GDH mutant and the wild type, F_1X or F_2X, respectively, was set to zero as an additional fixed constraint. In the statistical evaluation of the final fit result only the estimated fluxes were taken into account, i.e., the determined standard deviations represent lower bound values.

RESULTS

Activities of ammonium assimilatory enzymes in C. glutamicum during N and C limitation.

In continuous fermentations of C. glutamicum the ammonium availability was varied by using the independent NH₄⁺ dosing system according to the following protocol: after an initial growth phase of 40 h, ammonium-limited continuous cultivation was run for 40 h, whereafter the ammonium limitation was relieved and gradually (20 h) changed to C limitation, after which carbon-limited continuous cultivation was continued for another 40 h. The glucose concentration during N limitation was 15 to 20 g/liter, and the NH₄⁺ concentration during carbon limitation was 30 to 50 mM. The specific activities of GS, GDH, and GOGAT were determined in cell extracts taken at several time points during the fermentation. The results are given in Table 2.

TABLE 2.

Specific activities of the ammonium-assimilating enzymes of C. glutamicum wild type and its GDH mutant^a

Strain	Sp act (U/mg of protein) of:
	GDH		GS		GOGAT
	N	C	N	C	N	C
Wild type (ATCC 13032)	2.5	1.8	11	0.3	0.05	<0.003
GDH mutant	0.00	0.00	6.5	0.3	0.15	0.07

^aCells were taken from both ammonium (N)- and carbon (C)-limited continuous cultures. 

These data show two clearly different regulatory states of the NH₄⁺-assimilating enzymes of C. glutamicum ATCC 13032 according to the type of culture limitation. Under NH₄⁺ limitation, all three enzyme systems show higher specific activities than under C limitation. Under C limitation, where the ammonium concentration was 30 to 40 mM, the GS activity was 30-fold reduced and the GOGAT activity sank below the detection limit of ca. 3 mU/mg of protein. The GDH activity essentially remained at a constant level. This suggests a shutdown of the energy-costing ammonium assimilation processes under conditions where NH₄⁺ is abundant and energy is limited. As expected, the GDH mutant lacked any GDH activity. Whereas the regulation of GS in the mutant was the same as in the wild type, a strongly different behavior was found for GOGAT. The GOGAT activity under N limitation was three times higher than in the wild type, and it remained still higher than the wild-type N limitation value under carbon limitation, indicating that the cells used GOGAT as a compensation for lacking GDH. In order to test whether an alternative to the GS/GOGAT pathway was present in the mutant, a substrate analogue of glutamine was employed. It was previously shown that dl-methionine–dl-sulfoximine (MSX) can be used as glutamine analogon to inhibit GS (28). Tests with C. glutamicum crude extracts indicated that MSX inhibited GS with a K_i of 90 μM and GOGAT with a K_i of 4.5 mM (data not shown). No inhibition of GDH even at 100 mM MSX was found. Accordingly, growth of C. glutamicum ATCC 13032 was not inhibited in cultures with 100 mM MSX, whereas the GDH mutant was unable to grow when MSX concentrations were greater than 100 μM (data not shown).

Growth to high cell density and mixing time of labeled ammonium.

For the present study, we further optimized the combined in vivo NMR-fermentation system until it was suited for the aerobic continuous cultivation of C. glutamicum at 50 g (dry weight)/liter, where in vivo ¹⁵N-NMR spectroscopy could be performed with a time resolution of down to 8 min. The in vivo ¹⁵N experiments were performed on carbon-limited high-cell-density continuous cultures. The wild type reached a final density of 50 g (dry weight)/liter already 15 h after inoculation, whereas the GDH mutant, cultivated with identical fermentation parameters, needed considerably more time, 35 h, to reach the same density (data not shown). This indicates a significant growth advantage of the wild type over the GDH mutant strain. The in vivo experiment was started by the addition of 20 g of labeled (¹⁵NH₄)₂SO₄ after another 50 or 30 h of steady-state continuous cultivation for the wild type and the GDH mutant, respectively. Ideally, mixing of the labeled and unlabeled ammonium should be much faster than the time resolution of the in vivo experiment to avoid an observable influence on the kinetics of label accumulation in the ammonium assimilatory pools. The ¹H-NMR spectra of acidified cell extracts show separate signals of ¹⁴NH₄⁺ and ¹⁵NH₄⁺ and therefore were used to check the kinetics of the mixing of the added labeled ammonium with the unlabeled NH₄ initially present. An example ¹H-NMR spectrum is shown in Fig. 2. The experiments showed that equilibrium ¹⁵N labeling of ammonium was indeed reached within 2 min after application of the (¹⁵NH₄)₂SO₄ pulse (data not shown), illustrating the short mixing time of the fermenter system due to the high recirculation flow of 800 liters/h. Thus, the total ammonium concentration rose within 2 min from 40 mM to ca. 150 mM in the experiments. The final percents ammonium ¹⁵N labeling were 78% in the wild-type experiment and 80% in the experiment with the GDH mutant.

FIG. 2.

Example ¹H-NMR spectrum of an acidified cell extract obtained 120 min after the addition of ¹⁵NH₄⁺ to the C. glutamicum ATCC 13032 culture. The doublet results from ¹⁵NH₄⁺ due to scalar coupling of the protons with ¹⁵N (spin 1/2); the triplet results from ¹⁴NH₄⁺ due to scalar coupling of the protons with ¹⁴N (spin 1). These spectra were used to follow the kinetics of the mixing of the added labeled ammonium with the unlabeled ammonium initially present.

In vivo monitoring of ¹⁵NH₄ assimilation.

To identify the primary products and follow the kinetics of ammonium assimilation in the wild type and the GDH mutant, in vivo ¹⁵N-NMR spectroscopy experiments were performed on carbon-limited high-cell-density continuous cultivations, one for each strain. The series of in vivo ¹⁵N-NMR spectra obtained with the wild-type strain are shown in Fig. 3. By employing a measurement time of 15 min per spectrum, the signals from the amino nitrogen of glutamate at −335 ppm and from ammonium at −355 ppm could readily be observed already in the first spectrum, identifying glutamate as a prime target of ammonium nitrogen assimilation. However, unexpectedly, the first spectrum also contained a signal from the amido nitrogen of glutamine at −264 ppm, indicating that this compound must play a key role in ammonium assimilation in C. glutamicum ATCC 13032. After ca. 1 h of incubation with the labeled substrate, signals from the amino nitrogen of proline (−320 ppm) and from the ɛ-amino nitrogen of lysine (−343 ppm) appeared. Proline directly derives from glutamate, whereas the lysine nitrogen is labeled via transamination reactions. The late appearance of these compounds in the spectra shows that they do not play a role in primary ammonium assimilation. The same holds true for alanine, of which even in the last spectra no peaks were observed, thus ruling out any role for this amino acid. Whereas alanine was shown to play a major role in nitrogen metabolism in Agaricus bisporus (2), apparently it does not do so in C. glutamicum under the conditions studied. The strong label in the α-amino nitrogen of glutamate was in part expected from the high levels of this metabolite known to be present in C. glutamicum (24), but it also points at a significant activity of GDH in vivo since this large pool was labeled to half-maximum already after ca. 20 min. The observed incorporation of ¹⁵N label into glutamine N_δ was surprisingly fast, i.e., achieving half-maximum after ca. 15 min. The relatively strong intensity of this signal suggests that C. glutamicum wild type has, in addition to the large glutamate pool, a very significant glutamine pool as well. Given the large amount of ¹⁵NH₄⁺ added at time zero, the ammonium NMR signal is only weak. Furthermore, it demonstrated a curious behavior: it continuously diminished and changed its sign after ca. 1 h. While this may suggest some uptake phenomenon, it rather reflects a steadily decreasing NOE factor for ammonium during the experiment. Since this factor for ¹⁵N upon proton decoupling is negative, the point of zero-crossing corresponds to an average NOE factor of −1 for ammonium (3).

FIG. 3.

Series of in vivo ¹⁵N-NMR spectra obtained with C. glutamicum ATCC 13032 before (0 min) and in subsequent 15-min intervals after the addition of [¹⁵N]ammonium.

The series of in vivo ¹⁵N-NMR spectra obtained with the GDH mutant are shown in Fig. 4. From the good signal-to-noise ratio in the experiment with the wild type, it was deemed appropriate to employ a time resolution of only 7.5 min per spectrum for the GDH mutant. The in vivo experiment revealed surprising features of ammonium assimilation in this strain compared to the wild type. Although signals from the same nitrogen pools were observed, their relative intensities and the labeling time constants differed markedly from those of the wild type, indicating significantly altered pools as well as nitrogen fluxes. The α-amino nitrogen signal, which showed 50% labeling after 25 min, was strongly reduced relative to that of the amido nitrogen of glutamine, which showed 50% labeling already after 10 min. This seems indicative of a reduced glutamate pool and a strongly increased flux into the glutamine pool in the GDH mutant. Signals from other amino acids were not observed, possibly due to the low signal-to-noise ratio of the spectra. The ammonium signal now remained at an approximately constant negative level, indicating an NOE factor between 0 and −1.

FIG. 4.

Series of in vivo ¹⁵N-NMR spectra obtained with the GDH mutant of C. glutamicum before (0 min) and in subsequent 7.5-min intervals after the addition of [¹⁵N]ammonium.

Compounds observed in extract spectra and deconvolution of overlapping signals.

¹⁵N-NMR spectra of ethanolic cell extracts, due to their superior resolution and signal-to-noise ratio, allowed identification of signals that were not detectable in the in vivo spectra. Figure 5 presents an illustrative cell extract spectrum of a sample taken from the GDH mutant fermentation. Confirming the in vivo measurements (Fig. 4), the most prominent peaks stem from the amido nitrogen of glutamine at −264.3 ppm, from the amino nitrogens of both glutamate and glutamine at −335.5 ppm, and from ammonium at −355.5 ppm. Smaller signals, which were not yet visible in the in vivo spectrum taken at the corresponding time point due to the limited signal-to-noise ratio, are from the amino nitrogen of proline (−320.7 ppm), the ɛ-amino nitrogen of lysine (−343.5 ppm), and the α-amino nitrogens of alanine (−333.1 ppm) and valine (−339.6 ppm). The glutamine signal at −264.3 ppm partially overlaps with two unidentified signals at −264.1 and −264.0 ppm (one of them probably asparagine), and the glutamate-glutamine peak at −335.4 ppm overlaps with two small unidentified signals at −334.9 and −335.9 ppm. Another small signal (tentatively assigned to aspartate [18]) appears at −336.8 ppm. These latter three peaks were not resolved from the glutamate-glutamine peak at −335 ppm in the in vivo spectra. Therefore, their relative areas were determined from the extract spectra and subsequently used to correct the glutamate and glutamine peak areas determined from the in vivo spectra. The linearly increasing relative total overlapping signal for the GDH mutant after 2 h amounted up to ca. 20% at −355 ppm and 7% at −264 ppm, whereas for the wild type it reached 25% at −355 ppm and 15% at −264 ppm.

FIG. 5.

Illustrative ¹⁵N-NMR spectrum of a cell extract taken from the GDH mutant of C. glutamicum 120 min after the addition of the [¹⁵N]ammonium. Assignments are based on the literature (25).

In the extract ¹⁵N spectra from both the wild type and the GDH mutant, glutamine and glutamate were the only ammonium assimilatory products observed during the first 5 min after application of the ¹⁵N label (data not shown). This supports the conclusion from the in vivo experiments that these two compounds and no others are the primary ammonium assimilation products in C. glutamicum.

Relative ¹⁵N enrichments and pool sizes indicate a key role of glutamine.

The in vivo ¹⁵N data represent only relative peak areas. They must be scaled to absolute intracellular concentrations to enable the further analysis necessary to determine the nitrogen fluxes. This scaling cannot be done in a straightforward manner because the in vivo intensities are strongly mediated by the NOEs that are impossible to predict theoretically or to mimic experimentally by using standards. Therefore, glutamate and glutamine were purified from a number of ethanolic cell extracts taken at representative time points, and the ¹⁵N enrichment in the various positions were determined. For this purpose, heteronuclear ¹H spin-echo difference spectroscopy measurements were performed as described in the Materials and Methods section. The results are given in Table 3.

TABLE 3.

¹⁵N enrichments in glutamine and glutamate isolated from combined cell extracts taken at the indicated time points as determined by ¹H-NMR

Strain	Time point(s) (min) of combined extracts	% 15N enrichmenta
		Glutamate, Nα	Glutamine
			Nα	Nδ
Wild type	3 + 4 + 5	6.9 ± 0.7 (9.8)	ND (1.0)	6.1 ± 0.6 (12.1)
	25 + 30	37.5 ± 1.8 (41.2)	15.5 ± 1.5 (17.3)	48.0 ± 3.0 (49.7)
	165	73.6 ± 1.0 (76.2)	ND (73.9)	76.8 ± 3.0 (77.8)
GDH mu-tant	30	25.8 ± 3.0 (28.7)	14.1 ± 1.5 (17.1)	64.1 ± 3.0 (65.7)

^aNumbers in parentheses represent the corresponding labeling values as predicted by the final fit of the model (Fig. 1) to the experimental data. ND, not determined due to severe spectral overlap. 

The data for the wild type confirm the unexpected qualitative observation made with the in vivo experiment, i.e., that the amido nitrogen of glutamine received a very significant labeling, at a slightly faster rate than the combined α-amino nitrogen pool of glutamate and glutamine. This indicates significant activity of GS, since this reaction directly assimilates highly enriched ammonium into glutamine N_δ (cf. Fig. 1). The data also reveal that the N_α of glutamine was labeled at a much slower rate than the N_α of glutamate and the N_δ of glutamine. This observation, which could not be made from the in vivo data, is in accordance with N_α being labeled from a secondary source that needs time to get labeled itself, i.e., glutamate N_α. The relatively strong labeling of the latter is in accordance with high GDH activity.

The measurements for the GDH mutant also clearly confirm the in vivo observation that the labeling of the α-amino nitrogen pool of glutamate and glutamine was delayed compared to the wild type, whereas the labeling of the glutamine amido nitrogen was much faster. Apparently, glutamate N_α received little labeling, i.e., the ammonium assimilation primarily proceeded via glutamine in the GDH mutant strain.

In addition to the ¹⁵N isotopic enrichments, the pool sizes in vivo of glutamate and glutamine must be known to enable further analysis. These pool sizes were determined from the ethanolic cell extracts both by NMR and by HPLC. For the wild type, the glutamate pool size was 230 ± 20 μmol/g (dry weight) and the glutamine pool size was 54 ± 10 μmol/g (dry weight). The corresponding values for the GDH mutant were 96 ± 15 and 78 ± 15 μmol/g (dry weight), respectively. These findings lend even more support to the qualitative statements about the nitrogen flux distribution given above.

Fluxes in the ammonium assimilatory pathways.

The in vivo ¹⁵N-NMR data, calibrated with absolute pool sizes and ¹⁵N enrichments determined from samples taken at key time points as described above, could now directly be converted to amounts of nitrogen accumulated per gram of biomass and per minute. The differential equation flux model was then fitted to the experimental data of the wild type and the GDH mutant as described above to estimate the model parameters. The results of these fits, which are estimated values for the nitrogen fluxes, are given in Table 4.

TABLE 4.

Estimated nitrogen fluxes in wild-type and GDH mutant C. glutamicum resulting from the fit of the ammonium assimilation flux model (Fig. 1) to the experimental in vivo ¹⁵N-NMR data deconvoluted and calibrated as described in the text^a

Flux	Function	Strain
		Wild type	GDH mutant
Net fluxes (μmol/min · g [dry wt])
F0	Net NH4+ uptake	6.40	6.30
F1	GDH	4.61	0.00 (0.01)
F2	GOGAT	0.01 (0.98)	3.40 (0.82)
F3	GS	1.79 (0.17)	6.30
F4	Glu transaminases and incorporation	4.45	3.20
F5	Gln incorporation	0.17	0.20
F6	Gln amidotrans-ferases	1.61	2.69
Exchange fluxes (μmol/min · g [dry wt])
F1X	GDH	2.77 (0.54)	0.00
F2X	GOGAT	0.00	0.81 (0.49)
Pool sizes (μmol/g [dry wt])
Glutamate		250	82
Glutamine		49	93

^aThe standard deviation is given in parentheses where applicable. 

For both fermentations, the model fit to the data showed that the equilibration of ¹⁵N ammonium over the cell membrane was complete within 2 min, indicating that a very rapid diffusion of ammonia over the C. glutamicum cytoplasmic membrane took place in vivo. The experimentally determined NMR peak areas, as well as the corresponding values predicted from the fit, are shown in Fig. 6. The percents ¹⁵N enrichments in glutamine and glutamate predicted from the fit are given in Table 3 for comparison with the corresponding experimental values.

FIG. 6.

Experimental (squares, N_α of glutamate and glutamine; circles, N_δ of glutamine) and predicted NMR peak areas for the wild-type ATCC 13032 (A) and the GDH mutant (B) strains corresponding to the model fit results given in Table 4. Peak areas were corrected for overlap by using the ¹⁵N-NMR results obtained with ethanolic extracts. The downward trend in the peak area data reflects the dilution of biomass in the fermenter due to sample taking procedures.

From the in vitro enzyme data (Table 2), as well as the qualitatively interpreted in vivo data (Fig. 3) for the wild type, the principal pathways of ammonium assimilation are expected to be GDH and GS, though it seems at first unclear why glutamine should play a significant role. The flux estimation result does show that ammonium assimilation via the glutamine pool was significant in the wild type, 1.79 μmol/min · g (dry weight) flux through GS (F₃) compared to 4.61 μmol/min · g (dry weight) net flux through the GDH reaction (F₁). The fit result states that, most likely, 90% of this nitrogen is incorporated into the biomass (nucleotides, aromatic amino acids, and glucosamine) via glutamine amidotransferase reactions (F₆). This provides an explanation for the apparently significant role of glutamine. It must be noted here that the statistical correlation coefficient between F₂ and F₆ was −0.98 for this model in the case that F₆ was taken as independent flux instead of F₃ (data not shown). Therefore, the relatively high standard deviation for F₂ (Table 4) indicates that the possibility cannot be ruled out that a substantial fraction of the glutamine-targeted nitrogen flux might have been diverted to glutamate over the GOGAT reaction (F₂) instead of glutamine amidotransferases (F₆). Since the detection limit for GOGAT activity in the enzymatic measurements was about 3 mU/mg of protein, the fact that no GOGAT activity could be demonstrated for the wild type also does not exclude that a minor flux (up to ca. 1.5 μmol/min · g [dry weight]) over GOGAT could have been present in vivo. In each case there is no doubt that glutamine played a key role since the total flux F₃ (GS) through this pool was determined with only a 10% error (Table 4). Thus, also the flux over GDH was well determined. The relatively large GDH exchange flux of 2.77 μmol/min · g (dry weight) shows that the enzyme catalyzed a reversible reaction, with a forward flux of 7.35 μmol/min · g (dry weight) and a simultaneous reverse flux of 2.77 μmol/min · g (dry weight). This means that the enzyme had a significant overcapacity of glutamate synthesis, which can also be concluded when comparing the high GDH specific activity (1.8 U/mg of protein; Table 2) to the actual in vivo flux size. For the GDH mutant, the ammonium nitrogen was exclusively assimilated via the GS reaction (F₃, 6.30 μmol/min · g [dry weight]) as expected. The GOGAT reaction (F₂) transferred 54% of this nitrogen to glutamate while the glutamine amidotransferases (F₆), as in the wild-type situation, were strongly active (2.69 μmol/min · g [dry weight]). Again, the rather high standard deviation of F₂ (0.82 μmol/min · g [dry weight]) indicates that the N-flux distribution over F₆ and F₂ could only be resolved to a limited extent by the in vivo NMR data. The fact that a non-zero exchange flux, F_2X (0.81 μmol/min · g [dry weight]), was found indicates that the GOGAT reaction could be operating in a reversible manner in vivo. However, the rather large standard deviation of the estimate (0.49 μmol/min · g [dry weight]) still leaves a 5% probability that the enzyme had no reverse activity.

DISCUSSION

In previous in vivo ¹⁵N-NMR studies on concentrated cell suspensions taken from growing B. lactofermentum cultures, the prominent signals observed were from ¹⁵N-glutamate, ¹⁵N-glutamine, ¹⁵N-alanine, and ¹⁵N-lysine (13). These signals were also observed in the present in vivo ¹⁵N-NMR study on C. glutamicum cultivated at a high cell density under well-defined steady-state conditions in the NMR bioreactor system. However, the strong signal of N-acetylglutamine observed in packed cell spectra of B. lactofermentum (13) was not present in the spectra shown in the present study. Since in that study considerable time was needed to concentrate the cells and to measure the spectra (130- to 200-min data accumulation time per spectrum), it must be concluded that the cells were not in a defined state during those experiments. In experiments mimicking those of Haran et al. (13), signals of N-acetylglutamine were also observed with C. glutamicum. Therefore, the occurrence of this compound in glutamate-producing bacteria might be related to the presence of anaerobiosis or of cell lysis.

Comparison of the ¹⁵N-NMR spectrum of an ethanolic cell extract (Fig. 5) with the corresponding in vivo ¹⁵N-NMR spectra (Fig. 4) shows that the ratio of the glutamine N_δ to the glutamate-plus-glutamine N_α signal intensity differs strongly in the two cases. This can be explained by a difference in NOE factors in the in vivo versus the extract NMR measurement. The NOE results upon proton broadband decoupling as used in this study from ¹⁵N relaxation by dipolar interaction with protons. Since NOE factors for ¹⁵N are negative and may range from zero to a theoretical maximum of −4.93, the NOE may even cause nulling of signals under specific conditions (3), as was indeed observed in this study (Fig. 3). The NOE factor, when added to the original signal intensity (+1.0), results in a maximum NOE enhancement of −3.93. When the NOE factor is around −1.0, a strong sensitivity of the observed signal towards this factor results since, for example, a minor increase in NOE factor from −1.1 to −1.2 would produce a doubling of the observed ¹⁵N-NMR signal from −0.1 to −0.2 (relative intensity). Paramagnetic ions exert a significant influence on the ¹⁵N NOE by their contributing to the relaxation of the ¹⁵N nucleus. This can lead to a variable behavior of ¹⁵N signal intensities depending on environmental conditions such as medium mineral composition and cell density (1, 15) as observed in this study (cf. Fig. 3, 4, and 5). Most likely, the fact that paramagnetic ions bind much more easily to the α-amino (and the carboxylate) than to an amide group causes a reduction of the NOE factor for the former in vivo.

The specific activities of the NH₄⁺-assimilating enzymes in the wild type under conditions of NH₄⁺ abundancy as determined in this study are comparable to previously reported values (5, 41, 46, 49). The specific activities of GDH and GOGAT determined under nitrogen limitation conditions in this study were likewise comparable to values measured at low ammonium concentrations (11, 50). However, the specific GS activity found under N-limiting conditions (ca. 11 U/mg of protein) was 5- to 10-fold higher than that reported in comparable studies (47, 49). This could be a consequence of the fact that in none of those studies did C. glutamicum grow strictly in ammonium-limited conditions with carbon abundance, as was the case in the present work.

The observed regulatory behavior of GS and GOGAT in C. glutamicum ATCC 13032 in this study seems to be common to many microorganisms possessing both GDH and GS-GOGAT (33) and is similar to that observed by Kim et al. (21) and Sung et al. (47) for C. glutamicum ATCC 13058 and Brevibacterium flavum. The results of the present study therefore are consistent with GS being regulated directly by the nitrogen availability in the medium. The regulation mechanism possibly proceeds via adenylation of GS at high NH₄⁺ concentrations (35), a finding comparable to the known mechanism in Escherichia coli (37).

The observed weak dependency of GDH on the ammonium availability in the medium corroborates the finding of Sung et al. (47) for GDH in B. flavum but differs from that found with C. glutamicum ATCC 13058, where the GDH activity significantly decreased with increasing NH₄⁺ concentrations (21).

The in vivo NMR experiments presented in this study revealed that the glutamine pool was the second largest cytoplasmic amino acid pool in wild-type C. glutamicum ATCC 13032. Since it was observed that glutamine was easily hydrolyzed in an acidic environment (data not shown), the most likely reason that the important presence of glutamine in this bacterium has remained undetected thus far is because most studies of intracellular metabolites used perchloric acid extraction (10), thereby hydrolyzing the glutamine to glutamate. Differences in employed ammonium concentrations are unlikely to have caused elevated glutamine levels, since the steady-state ammonium concentration in our study (ca. 40 mM) was not very much different from that employed in most previous studies (typically 20 to 40 mM). Since, moreover, these concentrations are well in excess of the K_m values of GDH and GOGAT for NH₄⁺, it is also unlikely that even the 100 mM ammonium pulse applied in the present study has influenced the glutamate and glutamine pool sizes.

The flux analysis revealed the surprising fact that during chemostat growth under carbon limitation and NH₄⁺ abundance, nearly 28% of the NH₄⁺ was assimilated via the GS reaction in glutamine, while 72% was assimilated by the GDH reaction via glutamate. Thus, this study uncovered a previously unrecognized, important role of glutamine in NH₄⁺ metabolism of C. glutamicum. The value of 28% is approximately twice as high as the glutamine requirement for biomass synthesis in E. coli, which was reported to be 13% of the total assimilated nitrogen (37). This difference can be explained by the different cell wall compositions of gram-positive C. glutamicum and gram-negative E. coli. Glutamine functions as a nitrogen donor in the synthesis of carbamoyl phosphate, histidine, purines, and glucosamine-6-phosphate, a precursor of peptidoglycan (12). The fact that the amount of peptidoglycan in gram-positive bacteria is three- to sevenfold higher than that of gram-negative bacteria (39) may thus well be responsible for the strong use of glutamine for biomass synthesis of C. glutamicum observed in the present study. In fact, by using published data on C. glutamicum biomass composition (30) and taking into account the higher peptidoglycan content that can be calculated from the diaminopimelate content (146 μmol/g [dry weight]) of C. glutamicum, the amount of glutamine needed for the synthesis of a 1-g biomass can be calculated as ca. 2,000 μmol/g (dry weight). Glutamine amidotransferases make up ca. 1,800 μmol/g (dry weight) of this amount. The total requirement of nitrogen can likewise be calculated from the data of Marx et al. (30) modified for the higher peptidoglycan content; it amounts to ca. 9,500 μmol/g (dry weight). Thus, according to this calculation, 21% of the total nitrogen is assimilated via glutamine by direct incorporation or the above-mentioned glutamine amido-transferring enzymes (represented by F₅ and F₆, respectively, in our flux model [Fig. 1]). This compares rather well with the value of 28% found in the present in vivo study (Table 4).

The specific GOGAT activity in the GDH mutant C. glutamicum was relatively high and, in marked contrast to the wild type, showed only a weak dependency on the availability of NH₄⁺ (Table 2). This was also observed by Börmann-El Kholy (5). A comparably altered regulatory behavior of GOGAT has also been described for a GDH defect mutant of B. flavum (47). Since the GOGAT activity in the GDH mutant of C. glutamicum was essentially independent of ammonium availability in the medium, its regulation mechanism must be different from that of GS. The fact that the glutamate pool in the mutant was found to be much lower than in the wild type in the present study suggests that a reduced glutamate concentration may provide the signal for induction of GOGAT expression in C. glutamicum. The observation that glutamate or glutamate-generating nitrogen substrates (arginine, proline, or histidine) repress GOGAT expression in Salmonella typhimurium and Klebsiella aerogenes during N-limited growth (6, 7) is in agreement with this hypothesis. The possibility that the observed elevated GOGAT activity in the GDH mutant at high extracellular NH₄⁺ concentrations might be a consequence of a severe limitation in ammonium transmembrane transport in this strain can be definitively ruled out. Indeed, the flux analysis performed in this study indicated that a very fast equilibration of ammonium across the C. glutamicum membrane took place, i.e., the initial concentration gradient of 100 mM was compensated within 2 min. Literature data on this subject are scarce. While the ammonium uptake system in C. glutamicum has been characterized (23, 24), this system is not expressed at high ammonium concentrations as pertinent in the present work. Therefore, net ammonium uptake must be a consequence of ammonia diffusion across the cell membrane. A review (22) mentions ammonia permeability coefficient values for Nitella clavata (9 · 10⁻⁶ m/s) and Klebsiella pneumoniae (5 · 10⁻⁷ m/s). Using the average of the two values (4.75 · 10⁻⁶ m/s) for the permeability coefficient, assuming a volume-to-area ratio of 2 · 10⁻⁷ m for C. glutamicum (modelled as an ellipsoid with axes of 2 and 1 μm), taking into account that at neutral pH the equilibrium ratio of the NH₄⁺ and NH₃ concentrations is 180 and applying Fick’s First Law of Diffusion, we obtain d[NH₄⁺]_in/dt = (60 · 4.75 · 10⁻⁶) · ([NH₄⁺]_ex − [NH₄⁺]_in)/(2 · 10⁻⁷ · 180), from which a time constant for the equilibration of ammonium across the cell membrane of 0.13 min can be calculated. This agrees well with the value estimated by the flux modelling, given the uncertainty of the applied value for the permeability coefficient.

The results of the in vivo ¹⁵N-NMR nitrogen flux analysis of the GDH mutant of C. glutamicum were completely in accordance with the pattern expected from the enzyme data (Table 2), i.e., the absence of GDH flux, the high activity of GS, and the glutamate synthesis by operation of GOGAT. This demonstrates the reliability of the present approach. Again surprising was the even higher flux over glutamine amidotransferase fluxes (F₆), which amounted to almost 43% of total assimilated nitrogen (Table 4). The confidence region of this glutamine amido-transferring flux estimation, however, is rather large because the estimation of F₆ was highly correlated with that of the GOGAT flux F₂ which showed a moderately high standard deviation (Table 4). Thus, a fraction of 25 to 30% of the total assimilated nitrogen, which was also found for the wild type and which is not very much different from the calculated value of 21%, does not contradict the glutamine amidotransferase flux estimate for the GDH mutant. Our observation that the GOGAT reaction was partly reversible (Table 4) is at variance with the irreversibility demonstrated for the enzyme (referred to as glutamate synthase) from Azospirillum brasilense (52). However, in that study it was mentioned that the A. brasilense enzyme differed from the glutamate synthases from K. aerogenes and E. coli, which were weakly reversible, and the C. glutamicum enzyme may likewise be different in this respect. Moreover, the precision of our estimate does not allow us to fully exclude an irreversibility of the C. glutamicum GOGAT in vivo.

We observed that the GDH mutant grew much slower to high cell density than the wild type. This observation, together with the flux analysis results, identifies GOGAT as the growth-limiting bottleneck in the GDH mutant. This phenotype remained concealed in strain characterization studies of this mutant by Börmann-El Kholy (5), possibly because this author used a combination of ammonium and urea as nitrogen source instead of only NH₄⁺ as in the present study. This difference in nitrogen source composition may have influenced the regulation state of the NH₄⁺-assimilating enzymes in the two C. glutamicum strains.