Glutamate synthase of Corynebacterium glutamicum is not essential for glutamate synthesis and is regulated by the nitrogen status

Most enteric bacteria and many other organisms have two primary pathways to facilitate the incorporation of ammonium into the key nitrogen donors for biosynthetic reactions, glutamate and glutamine (for review, see Merrick & Edwards, 1995 ; Magasanik, 1996 ; Reitzer, 1996 ). Glutamate dehydrogenase (GDH) catalyses the reductive amination of 2-oxoglutarate to form glutamate. In most micro-organisms, GDH is used under high ammonium supply, while the second pathway, the glutamine synthetase/glutamate synthase (GS/GOGAT) system allows the assimilation of ammonium when present in the medium at low concentrations (1 mM). In this pathway, glutamate is amidated, with consumption of ATP, to form glutamine by GS and the amide group is then transferred reductively to 2-oxoglutarate by GOGAT, resulting in the net conversion of oxoglutarate to glutamate. Since assimilation via GDH is bioenergetically more favourable, a tight regulation of the different enzyme activities ensures in many organisms the optimal nitrogen supply of the cell under the lowest consumption of energy.

Corynebacterium glutamicum is a Gram-positive soil bacterium, which belongs phylogenetically to the actinomycetes group and is related to important pathogens like Corynebacterium diphtheriae, Mycobacterium leprae and Mycobacterium tuberculosis, and to the antibiotic-producing streptomycetes. Due to its remarkable ability to excrete high amounts of glutamic acid under conditions of biotin limitation (Kinoshita et al., 1957 ; Gutmann et al., 1992 ), this bacterium is applied in fermentation processes on an industrial scale. By use of different C. glutamicum mutant strains, not only large amounts of L-glutamate (1000000 t per year) but also L-lysine (450000 t per year) are produced, in addition to smaller amounts of the industrially less important amino acids L-alanine, L-isoleucine and L-proline (Leuchtenberger, 1996 ).

Due to their putatively pivotal role in glutamate production, basic biochemical work on the ammonium-assimilating enzymes GDH, GS and GOGAT was carried out earlier (Shiio & Ozaki, 1970 ; Tachiki et al., 1981 ; Tochikura et al., 1984 ). However, with the isolation of the corresponding genes, a more detailed characterization of the different enzymes and their regulation was possible. The gdh gene was isolated by Börmann et al. (1992) . Deletion and overexpression of this gene revealed that GDH is not essential for glutamate formation with respect to growth and glutamate production in C. glutamicum (Börmann-El Kholy et al., 1993 ). GDH activity is not regulated in dependence on nitrogen availability, i.e. no alterations in specific enzyme activity were found when raising the ammonium concentration in the growth medium from 1 to 90 mM (Tesch et al., 1998 ). The glnA gene, encoding GS I, was isolated by Jakoby et al. (1997) . Site-directed mutagenesis experiments revealed that GS is regulated via adenylylation/deadenylylation. Whilst nitrogen starvation enhances GS activity (Jakoby et al., 1999 ; Tesch et al., 1999 ), the enzyme is down-regulated by adenylylation upon addition of ammonium (Jakoby et al., 1997 ). However, in vivo flux analysis by ¹⁵N nuclear magnetic resonance showed that also in the presence of 100 mM ammonium a surprising large fraction, 28%, of ammonium is assimilated via GS. It was assumed that the observed high GS activity is the result of a strong demand for glutamine as a nitrogen donor in amidotransferase reactions for cell-wall synthesis (Tesch et al., 1999 ). In contrast to GS, GOGAT is inactive when cells were grown with high ammonium supply (Tesch et al., 1999 ). The nucleotide sequence of the genes encoding the large and small GOGAT subunits α and ß, gltB and gltD, was published in GenBank by Kanno and co-workers in 1999 (accession no. AB024708) without further experimental evidence.

In this study we present data regarding the genetic organization of the gltB and gltD genes, their transcriptional regulation, and the effect of inactivation of these genes on GOGAT activity, growth with different nitrogen sources, methylammonium/ammonium uptake and glutamate production.

Bacterial strains, plasmids and growth conditions.
Strains and plasmids used in this study are shown in Table 1. Bacteria were routinely grown in LuriaBertani (LB) medium (Sambrook et al., 1989 ) at 30 °C (C. glutamicum) or 37 °C (Escherichia coli). If appropriate, antibiotics were added in standard concentrations (Sambrook et al., 1989 ). LB medium for C. glutamicum strains was supplemented with 2% glucose (final concentration). To study the effects of nitrogen starvation under highly comparable conditions, a standard inoculation scheme was applied. A fresh C. glutamicum culture in LB medium was used to inoculate minimal medium (Keilhauer et al., 1993 ) for overnight growth. This culture, with an overnight OD₆₀₀ of approximately 2530, was used to inoculate fresh minimal medium to an OD₆₀₀ of approximately 0·5, and cells were grown until the exponential growth phase was reached (OD₆₀₀ approximately 45). To induce nitrogen starvation, cells were harvested by centrifugation, and the pellet was resuspended in and transferred to minimal medium without nitrogen source. The nitrogen-deprived cells were incubated at 30 °C under aeration. To induce glutamate production, cells were grown in minimal medium (5 g ammonium sulfate, 5 g urea, 2 g KH₂PO₄ and 2 g K₂HPO₄ per l distilled water, adjusted with NaOH to pH 7·0; glucose, Ca²⁺, Mg²⁺ and trace element concentrations as indicated by Keilhauer et al., 1993 ) with 0·5 µg biotin l^-1 overnight; this preculture was used to inoculate minimal medium without biotin.

Table 1. Strains and plasmids used in this study

General molecular biology techniques.
For plasmid isolation, transformation and cloning standard techniques were used (Ausubel et al., 1987 ; Sambrook et al., 1989 ). DNA sequence determination was carried out using the ABI Prism BigDye Terminator Cycle Sequencing Ready Reaction Kit and an ABI 310 automated sequencer (PE Applied Biosystems). Transcriptional start sites were determined using the 5'-RACE system as recommended by the supplier (Roche Diagnostics).

RNA preparation and Northern hybridization analysis.
Total RNA was prepared after disruption of the C. glutamicum cells by glass beads using the RNeasy Mini Kit as recommended by the supplier (Qiagen). The RNA was either size-fractionated using agarose gels containing formaldehyde and blotted onto positively charged nylon membranes (Sambrook et al., 1989 ) or spotted directly onto nylon membranes using a Schleicher & Schuell Minifold I Dot Blotter. Hybridization of digoxigenin-labelled RNA probes was detected with a Fuji luminescent image analyser LAS1000 or Kodak X-OMAT X-ray films using alkaline-phosphatase-conjugated anti-digoxigenin Fab fragments and CSPD as light-emitting substrate as recommended by the supplier (Roche Diagnostics).

Digoxigenin labelling of DNA and gel retardation experiments.
Digoxigenin-labelled DNA fragments were prepared using digoxigenin-labelled primers for PCR (5'-DIG-CAAGTCGGGCTGCGATGG- 3' / 5' -DIG-AATCAGGCGCAGTACCTGC-3'). The amtR gene was heterologously expressed in E. coli DH5αmcr pUC11-1.8 and cell extracts were prepared by ultrasonic treatment. For the retardation experiments, digoxigenin-labelled PCR product was incubated in TEK buffer (10 mM Tris/HCl, pH 8·0, 10 mM KCl, 1 mM EDTA, 7 µM 2-mercaptoethanol) containing 200 µg BSA ml^-1 together with salmon sperm DNA (150 µg ml^-1) and various amounts of cell extract for 15 min on ice. The samples were loaded onto a 4% polyacrylamide gel, and electrophoresis and blotting were subsequently carried out as described by Sambrook et al. (1989) . Bands were detected with a Fuji luminescent image analyser LAS1000 or Kodak X-OMAT X-ray films using alkaline-phosphatase-conjugated anti-digoxigenin Fab fragments to probe the digoxigenin-labelled DNA and CSPD as light-emitting alkaline phosphatase substrate, as recommended by the supplier (Roche Diagnostics).

Construction of glt mutant strains.
For the disruption of glt genes in the C. glutamicum chromosome, internal gene fragments of gltB and gltD were amplified by PCR using chromosomal DNA of the wild-type as template and the primer combinations 5'-CCGTGGCGCGCAGCTTGC-3'/5'-CCCATGGACCACCGGCATG-3' for gltB and 5'-AGGCGCCTGCGTGCTCGG- 3 ' / 5 ' -GCGGTTTTGGGCGGTGAG-3' for gltD. The resulting 1·0 kb gltB and 0·5 kb gltD fragments were ligated to plasmid pK18mob. After transfer of these constructs in competent C. glutamicum cells via electroporation (van der Rest et al., 1999 ), kanamycin-resistant clones carried the plasmids integrated in the chromosome via homologous recombination since pK18mob derivatives cannot be replicated by C. glutamicum.

Construction of a lacZ reporter gene strain.
As the first step to exchange the chromosomal gltB gene for a gltBlacZ fusion, a DNA fragment comprising 564 bp upstream of the gltB gene and its first 593 bp was amplified via PCR using the following primers: 5'-CCGGAATCGTCTTTCAGGATC-3'/5'-GCATGCAGCAATGGCCGACTCCAGGC-3'. After phosphorylation, the PCR product was ligated to SmaI-restricted and dephosphorylated pUC18 DNA. In the resulting plasmid, pUCgltB, a XbaI and a XhoI restriction site were introduced directly downstream of position +6 via site-directed mutagenesis, leading to plasmid pUCgltBXbaI/XhoI. In the second step, the lacZ gene was amplified by PCR from plasmid pLacZi. With the primers 5'-TCTAGAATGACCGGATCCGGAGCTTG- 3 ' / 5 ' -CATGAGTTACGCGAAATACGGGCAGAC-3', a XbaI site was introduced at the 5' end and a XhoI site at the 3' end of the DNA fragment (shown in bold). The XbaI/XhoI-restricted PCR product was ligated to XbaI/XhoI-restricted and dephosphorylated pUCgltBXbaI/XhoI plasmid DNA, leading to pUCgltBlacZ. In the last step, the gltBlacZ fusion was isolated by EcoRI/SphI restriction and cloned into vector pK18mobsacB. Integration of the resulting plasmid pK18mobsacBgltBlacZ in the C. glutamicum chromosome and allelic exchange was carried out as described by Schäfer et al. (1994) . The correct chromosomal integration of the lacZ fusion in strain GltBlacZ was checked by PCR (data not shown).

Enzyme-activity measurements, transport assays and miscellaneous methods.
Cell extracts were prepared by ultrasonic treatment. The protein content of samples was determined using a modified Bradford assay (Roti-Quant, Roth). GOGAT activity was measured as described by Meers et al. (1970) , ß-galactosidase activity was assayed as described by Miller (1972) using cells permeabilized with 0·2% (final concentration) N-cetyl-N,N,N-trimethylammonium bromide (CTAB). Glutamate in the culture supernatant was determined according to Gutmann et al. (1992) . Uptake of [¹⁴C]methylammonium was measured using a rapid filtration approach (Siewe et al., 1996 ).

Computer-assisted DNA analyses.
Database searches were performed using the BLAST program (Altschul et al., 1990 ). For domain analyses the PROSITE software package (http//:www.expasy.ch) and the Pfam program (http//:www.sanger.ac.uk) were used.

Genetic organization of gltB and gltD
The nucleotide sequence of the genes encoding GOGAT was published recently (GenBank accession no. AB024708) without further details. First analyses showed that the gltBD cluster has a G+C content of 56·7 mol%, which is in accord with the overall 56·1 mol% G+C content of C. glutamicum genes published in databases (Nakamura et al., 1997 ). More interestingly, the start of gltD overlaps the stop codon of gltB by 1 bp, indicating a transcriptional coupling of the two genes.

From the deduced amino acid sequence, a number of putative functional important amino acid motifs could be identified (for overview, see Fig. 1). As, for example, in the Azospirillum brasilense GOGAT (Pelanda et al., 1993 ), six regions with the sequence GXG₂(G/A/P), which interact with the adenylate portion of FAD or NADPH, were observed in the α subunit (gltB gene product) and two in the ß subunit (gltD gene product). Bacterial GOGATs contain three distinct ironsulfur centres. A search for cysteine clusters potentially involved in FeS centre formation revealed neither an exact CX₃CPX₄CX₃C nor CX₂DX₂CX₃C motif; however, one cysteine cluster was found in the α and two in the ß subunit at positions 11201133, 4959 and 97105. When computer-based domain searches were carried out, a GOGAT domain, comprising a putative FMN-binding site and a FeS cluster, was identified at position 8061179, and a glycine-rich region from position 12851422 in the α subunit and two NAD-binding domains at positions 146174 and 279308 in the ß subunit.

(4K): Fig. 1. Chromosomal organization of the C. glutamicum gltBD operon. The position of DNA fragments used for the construction of gene disruption strains Glt1 and Glt2 is indicated by black bars; locations of primers used for RT-PCR are indicated by arrows. Additionally, the positions of glycine-rich FAD/NADPH-interacting regions (Gly), cysteine clusters (Cys), an FMN-binding domain (FMN) and two NAD-binding motifs (NAD) are shown.

Sequence homology searches revealed the highest identities of the amino acid sequence deduced from the C. glutamicum gltB gene with a hypothetical GOGAT from Mycobacterium tuberculosis (63% identity, SWISS-PROT P96218) and with the GOGAT α subunit from Streptomyces coelicolor (61% identity, SWISS-PROT Q9S2Y9). Identities obtained for the ß subunit were slightly lower, with 58% identity to a hypothetical GOGAT protein from M. tuberculosis (P96219) and 52% identity to the ß subunit of S. coelicolor (Q9S270).

Analysis of gltBD transcription
To study expression of gltB and gltD and its (putative) regulation, total RNA was isolated from C. glutamicum cells before and after nitrogen starvation and hybridized with a digoxigenin-labelled gltB anti-sense RNA probe. While no gltB transcript was detected in cells grown under nitrogen excess, hybridization signals were observed within 15 min after cells were transferred to a nitrogen-free medium (Fig. 2a). These results show that GOGAT is up-regulated on the level of transcription upon nitrogen starvation.

(52K): Fig. 2. (a) Northern hybridization experiments. Dot blots of total RNA isolated from wild-type cells before and 2, 15 and 30 min after nitrogen starvation (lanes 14) probed for gltB transcription (upper panel) and for 16S rRNA (lower panel) as a control. (b) RT-PCR using gltBD-specific primers (lane 2). As template, DNase-treated total RNA from nitrogen-starved cells was used; lane 1, control without reverse transcriptase; lane 3, DNA standard (100 bp ladder, from top to bottom, 1·4, 1·3, 1·2, 1·0, 0·9, 0·8, 0·7, 0·6, 0·5, 0·4. 0·3, 0·2 kb). The 0·25 kb RT-PCR product is indicated by an arrow.

To verify that gltB and gltD form an operon, RT-PCR was carried out with one primer annealing to the gltB gene (5'-AGCCTTCCGCTCAGGCTC-3') and one to the gltD nucleotide sequence (5'-CAGAATCATCGTTGATGCCG-3'; see also Fig. 1). While the control reaction without reverse transcriptase gave no product, a common 0·25 kb product was detected after RT-PCR, demonstrating the presence of a gltBD operon in C. glutamicum (Fig. 2b).

To determine the transcriptional start site of the gltBD operon, total RNA isolated from nitrogen-starved cells was used as a template in a RACE assay. The transcriptional start site was identified 41 bp upstream of gltB and a -10 and -35 consensus sequence typical for C. glutamicum promoter regions (Patek et al., 1996 ) were identified (Fig. 3). In addition, an insertion sequence which prefers conserved target sequences located adjacent to genes involved in aspartate and glutamate metabolism, ISCg2, is located 756 bp upstream of gltB (Quast et al., 1999 ). The role of this insertion sequence in C. glutamicum is unclear. A putative terminator structure with a free energy of 66·8 kJ mol^-1 was identified 16 bp downstream of the gltD stop codon spanning nucleotides 66336663.

(7K): Fig. 3. Physical map of the gltB upstream region. The AmtR-binding motif and the ATG start codon of the gltB gene are shown. TS indicates the transcriptional start site determined. The AmtR-binding motif is boxed. Sequences typical for the -10 and -35 promoter consensus sequences defined by Pátek et al. (1996) for C. glutamicum are shown on top of the corresponding gltB upstream sequence.

Regulation of gltBD expression by the AmtR repressor protein
Recently, a global repressor involved in nitrogen control in C. glutamicum, AmtR, was identified and an AmtR-binding motif, ATCTATAGN₁₄ATAG, was characterized. By database searches no obvious recognition motif could be identified upstream of gdh, encoding GDH and gltBD, encoding GOGAT (Jakoby et al., 2000 ). However, the obvious dependence of gltBD expression on nitrogen starvation prompted us to reinvestigate the promoter region of this operon. In fact, a rudimentary AmtR recognition sequence located on the opposite DNA strand at position -102 to -109 upstream of the gltB start codon was found, exhibiting the sequence ATCTATAG (Fig. 3).

To verify the assumption that AmtR might regulate gltBD expression, Northern hybridization experiments were carried out using RNA isolated from the wild-type and from amtR deletion strain MJ6-18 before and after induction of nitrogen starvation. Deregulated gltB expression was observed in strain MJ6-18, demonstrating the regulatory role of AmtR in the wild-type (Fig. 4a).

(87K): Fig. 4. Influence of AmtR on gltBD transcription. (a) Northern hybridization of total RNA isolated from amtR deletion strain MJ6-18 before and after 2, 15 and 30 min of nitrogen starvation (lanes 14) probed for gltB transcription (upper panel) and probed for 16S rRNA as a control (lower panel). (b) Gel retardation experiment using digoxigenin-labelled DNA (2 ng per lane). As a positive control (compare Jakoby et al., 2000 ), an amt fragment was used without (lane 1) and with 3 µg AmtR-containing cell extract (lane 2); lane 3, gltB upstream DNA without cell extract; lanes 47, gltB upstream DNA plus increasing amounts (3, 6, 12 and 15 µg) of DH5αmcr pUC11-1·8 extracts. (c) Titration control. For the experiment, 2 ng digoxigenin-labelled gltB upstream DNA (lane 1) was incubated with 3 µg AmtR-containing cell extract without the addition of unlabelled gltB fragment (lane 2) and with increasing amounts (2, 4, 10, 20 and 40 ng) of specific competitor DNA (lanes 37).

Moreover, in an independent approach, binding of AmtR to the gltB upstream region was shown by DNA retardation experiments. For this purpose, increasing amounts of AmtR-containing E. coli extracts were added to a digoxigenin-labelled PCR fragment spanning the region from -258 to -51 with respect to the gltB start codon. Addition of AmtR-containing E. coli extracts resulted in a retardation of the DNA fragment upstream of gltB, whilst a control extract without AmtR protein showed no shift of the corresponding DNA (Fig. 4b). As a control, a titration experiment was carried out. The shift of the digoxigenin-labelled PCR fragment was quenched by the addition of unlabelled DNA fragment, as expected for a specific interaction (Fig. 4c).

GOGAT activity
GOGAT activities were determined for C. glutamicum wild-type cells grown in different media. No activity was detectable when ammonium was present in high concentrations (100 mM). When cells were incubated for 3 h without nitrogen source, GOGAT activities of approximately 3 mU (mg protein)^-1 were determined, which are in accord with data reported previously (Börmann-El Kholy et al., 1993 ).

Analysis of a reporter gene strain
Since the results obtained for GOGAT showed only a low, hardly detectable enzymic activity, we tested, in addition to the GOGAT measurements, the activity of a reporter gene construct to obtain more reliable results. For this purpose, a gltBlacZ translational fusion was constructed and integrated in the C. glutamicum gltB locus. The resulting strain GltBlacZ was tested under different physiological conditions. When this strain was grown under nitrogen surplus (100 mM ) no ß-galactosidase activity was detected in permeabilized cells, while complete lack of nitrogen sources or limiting amounts of ammonium (4 mM), glutamine (4 mM) and urea (2 mM) induced ß-galactosidase. After 3 h incubation, activities of 6·1±0·8 mU (mg dry weight)^-1 were determined upon nitrogen starvation, 24·5±1·5 mU (mg dry weight)^-1 when cells were grown in the presence of ammonia, 14·8±1·2 mU (mg dry weight)^-1 with glutamine as nitrogen source, and 15·0±1·2 mU (mg dry weight)^-1 using limiting amounts of urea. These results indicate that synthesis of GOGAT is induced when cells grow in the presence of limiting nitrogen supply.

Analysis of glt mutant strains
To study the physiological role of GOGAT, both a gltB and a gltD insertion mutant were constructed (for overview, see Fig. 1). When transcription of the gltBD operon was tested in these mutants for control, only minor binding of a gltB probe in gltD disruption strain Glt2 was observed, indicating an enhanced degradation of the truncated messenger RNA, whilst no binding was detected when gltB mutant strain Glt1 was analysed (Fig. 5). As a consequence, GOGAT activity was not detectable in strains Glt1 and Glt2, irrespective of the nitrogen status of the cells (data not shown).

(88K): Fig. 5. Northern hybridization analysis of glt mutant strains Glt1 and Glt2: dot blots of total RNA isolated before and 2, 15 and 30 min after nitrogen starvation (lanes 14). Upper panel: wild-type, gltD mutant Glt2, and gltB mutant Glt1 probed for gltB transcription; lower panel: wild-type, Glt2, and Glt1 probed for 16S rRNA as control.

To study the effect of GOGAT deficiency, growth of strains Glt1 and Glt2 was tested in the presence of different nitrogen sources. No effect of the glt disruptions was observed when glutamine was used as sole nitrogen source, most likely due to the activity of glutamine-dependent transaminases (Tesch et al., 1999 ). When growth was tested in the presence of limiting concentrations of ammonium (2 mM), strains Glt1 and Glt2 exhibited the same final OD₆₀₀ as the wild-type. However, the doubling time of the glt mutant strains was increased from 2 h 45 min in the wild-type to 3 h 20 min in strains Glt1 and Glt2 (experiments were carried out in duplicate with identical results; data not shown). In E. coli, high-affinity methylammonium/ammonium transport was shown to be severely impaired in a gltB mutant strain (Servin-Gonzalez & Bastarrachea, 1984 ). Growth of strains Glt1 and Glt2 on limiting amounts of urea was also tested in order to exclude an indirect effect of the glt mutations on ammonium uptake resulting in slow growth. Urea is transported into the cell by a specific uptake system and degraded by urease to ammonium and carbon dioxide (Siewe et al., 1998 ). While the final OD₆₀₀ was identical for the different strains in minimal medium with 25 mM urea, the doubling time of the glt mutant strains was increased from 2 h 45 min in the wild-type to 3 h 15 min in strains Glt1 and Glt2 (experiments were carried out in duplicate with identical results; data not shown).

In other Gram-positive organisms such as Bacillus subtilis, glt gene transcription is under the control of carbon metabolism regulators as well (Faires et al., 1999 ). Therefore, in addition to growth tests with limiting nitrogen sources, growth tests were also performed with various carbon conditions. No growth defects were observed for glt mutant strains in comparison to the wild-type when cells were grown in the presence of acetate, glucose, glutamate, maltose, or sucrose (2% final concentrations; data not shown).

Methylammonium uptake in strains Glt1 and Glt2
In addition to the growth experiments using urea as nitrogen source, which indicated no connection of GOGAT activity and methylammonium/ammonium transport (see above), [¹⁴C]methylammonium transport measurements with wild-type cells and glt mutant strains were carried out. In contrast to E. coli, methylammonium uptake was not negatively affected by a gltB or gltD mutation. Both mutant strains showed no methylammonium uptake when grown with high amounts of nitrogen sources (500 mM ammonium), while transport was induced upon nitrogen limitation. After 3 h nitrogen starvation, the wild-type revealed a methylammonium uptake rate of 14·7 nmol min^-1 (mg dry weight)^-1, strain Glt1 of 38·8 nmol min^-1 (mg dry weight)^-1, and strain Glt2 of 25·5 nmol min^-1 (mg dry weight)^-1. Obviously, methylammonium/ammonium uptake is significantly enhanced in the GOGAT-deficient strains in order to circumvent problems arising from the less effective ammonium assimilation.

Effect of GOGAT deficiency on glutamate production
The C. glutamicum wild-type is able to excrete high amounts of glutamic acid under conditions of biotin limitation (Kinoshita et al., 1957 ; Gutmann et al., 1992 ). A prerequisite of this excretion is an efficient glutamate synthesis pathway. To investigate whether there is an influence of GOGAT activity on glutamate production, the wild-type ATCC 13032 and glt mutant strains Glt1 and Glt2 were grown under biotin limitation. When glutamate excreted in the medium was measured, no detrimental influence of the gltB and gltD mutation on glutamate production was observed. In all strains comparable amounts of glutamate were determined, e.g. 0·4 µmol (mg dry weight)^-1 produced overnight by the wild-type versus 0·4 µmol (mg dry weight)^-1 produced by Glt1 and 0·5 µmol (mg dry weight)^-1 excreted by strain Glt2. Obviously, GDH activity alone is sufficient for glutamate production in C. glutamicum.

In this study we demonstrated that the gltB and gltD genes form an operon in C. glutamicum. Transcription of this operon is under control of AmtR, a repressor of nitrogen-regulated genes in C. glutamicum, as shown by mutant strain analyses and gel retardation experiments. A corresponding AmtR-binding motif comprising the sequence ATCTATAG was identified upstream of the gltB start codon. The data on induction of gltBD transcription obtained in this study by Northern hybridization and reporter gene assays are in accord with previous observations that GOGAT is active under ammonium concentrations lower than 10 mM (Börmann-El Kholy et al., 1994; Tesch et al., 1998 ). As expected, glt mutant strains showed a prolonged doubling time in comparison to the wild-type when grown on limiting amounts of ammonium or urea. However, this effect was not very pronounced, e.g. when compared to E. coli, although no paralogues of gltB and gltD are present in the C. glutamicum genome sequence (B. Bathe, unpublished observation). A reason for this might be the fact that GDH is highly active in C. glutamicum (Börmann et al., 1992 ) and that transcription of the gdh gene is additionally enhanced upon nitrogen starvation (L. Nolden, unpublished observation) in combination with the enhanced uptake of ammonium via the Amt system observed for glt mutants in this study. By contrast, GOGAT can function as a backup system in a gdh mutant strain as shown previously (Börmann-El Kholy et al., 1993 ).

The same enzymic pathways, GDH and GS/GOGAT, are used for ammonium assimilation in the Gram-positive soil bacterium C. glutamicum and Gram-negative enterobacteria such as E. coli. However, transcriptional regulation of the corresponding genes differs significantly. For example, gltBD expression in E. coli is influenced by a number of regulators including LRP and CRP (for summary, see Reitzer, 1996 ) and gdhA transcription is repressed moderately by nitrogen limitation (Riba et al., 1988 ), while in this study we showed that gltBD transcription is regulated in C. glutamicum via AmtR and previous studies revealed no down-regulation of GDH activity in response to ammonium limitation (Börmann et al., 1992 ; Tesch et al., 1998 ). Also in comparison to Bacillus subtilis, the model organism for Gram-positive bacteria with low G+C content, assimilation of ammonium is organized differently in the high-G+C-content C. glutamicum. In B. subtilis, ammonium is assimilated exclusively via the GS/GOGAT pathway since GDH is absent in this organism (for review, see Schreier, 1993 ). Transcription of the gltAB operon encoding GOGAT requires a specific positive regulator, GltC, and an intact carbon control protein CcpA (Faires et al., 1999 ), and is repressed by TnrA (Belitsky et al., 2000 ). In summary, the results obtained for C. glutamicum support the statement of Fisher (1999) that in addition to the paradigm of nitrogen assimilation and regulation in enterobacteria, a number of different mechanisms might be realized in other organisms.

The authors wish to thank R. Krämer for his support and interest, and for critical reading of the manuscript, and B. Bathe (Degussa) for providing C. glutamicum genome information prior to publication. The excellent techical assistance of J. Strösser and Gregor Wersch is gratefully acknowledged. This work was supported by the Deutsche Forschungsgemeinschaft (grant BU 894/1-1).