Improvement of the Redox Balance Increases l-Valine Production by Corynebacterium glutamicum under Oxygen Deprivation Conditions

Abstract

Production of l-valine under oxygen deprivation conditions by Corynebacterium glutamicum lacking the lactate dehydrogenase gene ldhA and overexpressing the l-valine biosynthesis genes ilvBNCDE was repressed. This was attributed to imbalanced cofactor production and consumption in the overall l-valine synthesis pathway: two moles of NADH was generated and two moles of NADPH was consumed per mole of l-valine produced from one mole of glucose. In order to solve this cofactor imbalance, the coenzyme requirement for l-valine synthesis was converted from NADPH to NADH via modification of acetohydroxy acid isomeroreductase encoded by ilvC and introduction of Lysinibacillus sphaericus leucine dehydrogenase in place of endogenous transaminase B, encoded by ilvE. The intracellular NADH/NAD⁺ ratio significantly decreased, and glucose consumption and l-valine production drastically improved. Moreover, l-valine yield increased and succinate formation decreased concomitantly with the decreased intracellular redox state. These observations suggest that the intracellular NADH/NAD⁺ ratio, i.e., reoxidation of NADH, is the primary rate-limiting factor for l-valine production under oxygen deprivation conditions. The l-valine productivity and yield were even better and by-products derived from pyruvate further decreased as a result of a feedback resistance-inducing mutation in the acetohydroxy acid synthase encoded by ilvBN. The resultant strain produced 1,470 mM l-valine after 24 h with a yield of 0.63 mol mol of glucose⁻¹, and the l-valine productivity reached 1,940 mM after 48 h.

INTRODUCTION

Amino acids are essential nutrients for all living organisms and have been commercially utilized chiefly as flavor enhancers and food additives. Until the discovery of Corynebacterium glutamicum as a superior amino acid-producing microbe, amino acids were produced entirely by extraction methods or chemical synthesis. Current production, estimated to be in excess of two million tons per year, is predominantly from fermentation of sugars (14, 24). Conventional amino acid fermentation is an aerobic process requiring cell growth and, consequently, sufficient aeration or appropriate control of dissolved oxygen or redox potential of the culture medium in order to achieve high productivity (2). Moreover, under aerobic conditions, the carbon source, besides being the substrate from which the amino acids are directly produced, is also consumed during cell growth. In effect, the amount of substrate available for fermentation is diminished, implying lowered yields of desired products.

Under oxygen deprivation conditions, C. glutamicum cells do not grow but are able to metabolize sugars to organic acids. Their glucose consumption rates under these conditions are higher than those attained during aerobic cultivation (i.e., Pasteur effect) (19). These growth-arrested but metabolically active cells permit high product yields while suppressing by-product formation and can be used at higher cell densities than is possible under aerobic conditions, leading to high volumetric productivities (17, 18, 20, 32). In addition to high resistance to lignocellulose hydrolysate-derived fermentation inhibitors during oxygen deprivation (37), C. glutamicum strains with the added capability of utilizing pentose and hexose sugars simultaneously have been constructed (39, 40). Based on these observations, an innovative and versatile bioprocess that is simple, economical, and environment friendly can be developed to produce commodity chemicals.

In this study, we demonstrated production of l-valine, one of the branched-chain amino acids (BCAAs). The biosynthesis pathway for l-valine from pyruvate in C. glutamicum consists of four successive enzyme reactions (Fig. 1): acetohydroxy acid synthase (AHAS), encoded by ilvBN; acetohydroxy acid isomeroreductase (AHAIR), encoded by ilvC; dihydroxy acid dehydratase (DHAD), encoded by ilvD; and transaminase B (TA), encoded by ilvE (25, 35). The pathway is also respon-sible for the biosynthesis of other BCAAs (l-leucine and l-isoleucine) and d-pantothenate (35). The key enzyme among the four is AHAS, because it is subject to feedback inhibition (10, 25, 47), and the gene encoding it is controlled by an attenuation mechanism by the BCAAs (8). l-Valine-overproducing C. glutamicum strains have thus been constructed by overexpressing ilvBNCDE (4, 10, 16, 47) and derepressing feedback inhibition of AHAS (10, 47). Deletion or downregulation of competitive pathways (ilvA, leuA, panB, aceE, pqo, and pyc) also results in improved l-valine productivity because of the channeling of pyruvate away from these competitive side reactions as well as away from biomass formation (4, 10, 16).

Fig 1.

Metabolic pathway of C. glutamicum under oxygen deprivation conditions (bold arrows, predominant flow) and the biosynthetic pathway of l-valine. GAPDH, glyceraldehyde-3-phosphate dehydrogenase; PEPC, phosphoenolpyruvate carboxylase; PEPCK, phosphoenolpyruvate carboxykinase; PC, pyruvate carboxylase; OD, oxaloacetate decarboxylase; MDH, malate dehydrogenase; SDH, succinate dehydrogenase; MQ, menaquinone; NDH, NADH dehydrogenase; LDH, lactate dehydrogenase; AHAS, acetohydroxy acid synthase; AHAIR, acetohydroxy acid isomeroreductase (AHAIR*, NAD-preferring mutant); DHAD, dihydroxy acid dehydratase; TA, transaminase B; LeuDH, leucine dehydrogenase (L. sphaericus).

As shown in Fig. 1, there exists an imbalance in cofactor production versus consumption in the l-valine synthesis pathway: the synthesis of one mole of l-valine from one mole of glucose generates two moles of NADH via glycolysis but consumes a total of two moles of NADPH, one at the AHAIR reaction and the other at the regeneration of l-glutamate (5) as an amino-group donor for the TA reaction (26, 27) (Fig. 1). Conversion of NADH to NADPH is presumed to be difficult in C. glutamicum because this organism possesses no chromosomally encoded nicotinamide nucleotide transhydrogenase to catalyze the reversible interconversion between NADH and NADPH (21), and NADPH generation from NADH via malic enzyme would play only a minor role (3, 12). Thus, this cofactor imbalance is a suspected rate-limiting factor in l-valine production by C. glutamicum. The importance of NADPH supply for l-valine production is well documented (3, 4), and surplus NADH generation would be unfavorable for metabolism as well as xylose fermentation by yeast, which ceased under anaerobic conditions due to cofactor imbalance (7). As reported previously (18, 20, 33), the intracellular NADH/NAD⁺ ratio of C. glutamicum strains lacking pathways to reoxidize NADH anaerobically is increased during oxygen deprivation, which in turn inhibits glyceraldehyde-3-phosphate dehydrogenase (GAPDH) activity and, consequently, lowers glucose consumption.

In the present study, we demonstrate improved l-valine production in C. glutamicum strains in which the cofactor requirement for l-valine synthesis is converted from NADPH into NADH via modification of the coenzyme specificity of AHAIR and introduction of NAD-specific leucine dehydrogenase (LeuDH) derived from Lysinibacillus sphaericus (synonym: Bacillus sphaericus) in place of TA. Both the mutant AHAIR and LeuDH consumed NADH in their reactions, unifying cofactor production and consumption during l-valine synthesis (Fig. 1). As a consequence, the intracellular NADH/NAD⁺ ratio was so significantly decreased that sugar consumption and both l-valine productivity and yield were drastically increased under oxygen deprivation conditions. Furthermore, a feedback-resistant mutant AHAS, which was cloned from a mutant strain resistant to a valine analogue, showed even better l-valine productivity and yield.

MATERIALS AND METHODS

Bacterial strains, media, and cultivation conditions.

All bacterial strains used in this study are listed in Table 1. Escherichia coli strains were grown at 37°C in Luria-Bertani medium (38). C. glutamicum R (51) and its recombinants were grown at 33°C in complex medium (A medium) or minimal medium (BT medium) with 4% glucose (18). The final concentrations of antibiotics used were as follows: for E. coli, kanamycin, 50 μg ml⁻¹, and chloramphenicol, 50 μg ml⁻¹, and for C. glutamicum, kanamycin, 50 μg ml⁻¹, and chloramphenicol, 5 μg ml⁻¹. L. sphaericus was grown at 30°C in NBRC 802 medium (polypeptone [10 g liter⁻¹], yeast extract [2 g liter⁻¹], MgSO₄ · 7H₂O [1 g liter⁻¹]).

Table 1.

Bacterial strains and plasmids used in this study

Strain or plasmid	Relevant characteristicsa	Source or reference
Strains
E. coli
JM109	recA1 endA1 gyrA96 thi hsdR17 (rK− mK+) e14− (mcrA) supE44 relA1 Δ(lac-pro0AB)/F− (traD36 proAB+lacIqlacZΔM15)	Takara Bio Inc.
JM110	dam dcm supE44 hsdR17 thi leu rpsL1 lacY galK galT ara tonA thr tsx Δ(lac-proAB)/F′ (traD36 proAB+ lacIqlacZΔM15)	38
C. glutamicum R		51
ΔilvC strain	ilvC-null mutant by transposon-mediated mutagenesis	This work
BNC™/ΔilvC	ΔilvC harboring pCRB-BNC™	This work
ΔLDH	Markerless ldhA null mutant	17
BNC/ΔLDH	ΔLDH harboring pCRB-BNC	This work
BNCDE/ΔLDH	ΔLDH harboring pCRB-BNC and pCRB-DE	This work
BNC™DE/ΔLDH	ΔLDH harboring pCRB-BNC™ and pCRB-DE	This work
BNCDLD/ΔLDH	ΔLDH harboring pCRB-BNC and pCRB-DLD	This work
BNC™DLD/ΔLDH	ΔLDH harboring pCRB-BNC™ and pCRB-DLD	This work
BNGEC™DLD/ΔLDH	ΔLDH harboring pCRB-BNGEC™ and pCRB-DLD	This work
L. sphaericus NBRC3525	Synonym: Bacillus sphaericus	NBRCb
Plasmids
pHSG298	Kmr; α-lac multicloning site. E. coli cloning vector	Takara Bio Inc.
pHSG398	Cmr; α-lac multicloning site. E. coli cloning vector	Takara Bio Inc.
pKK223-3	Apr; expression vector under the control of tac promoter	Pharmacia
pCASE1	Plasmid identified in Corynebacterium casei JCM 12072	45
pCG1	Plasmid identified in C. glutamicum	GenBank accession no. AB027714
pCRB12	Kmr; α-lac multicloning site. E. coli-Corynebacterium sp. shuttle vector derived from pHSG298 and pCG1	This work
pCRB21	Cmr; α-lac multicloning site. E. coli-Corynebacterium sp. shuttle vector derived from pHSG398 and pCASE1	This work
pCRB22	Kmr; α-lac multicloning site; E. coli-Corynebacterium sp. shuttle vector derived from pHSG298 and pCASE1	This work
pCRB207	Kmr; expression vector under the control of gapA promoter.	This work
	E. coli-Corynebacterium sp. shuttle vector derived from pCRB22
pCRB-BNC	PgapA-ilvBNC (C. glutamicum) in pCRB21	This work
pCRB-BNGEC	PgapA-ilvBNGEC (C. glutamicum) in pCRB21	This work
pCRB-BNC™	PgapA-ilvBNC™ (C. glutamicum) in pCRB21	This work
pCRB-BNGEC™	PgapA-ilvBNGEC™ (C. glutamicum) in pCRB21	This work
pCRB-DE	Ptac-ilvD (C. glutamicum) and Ptac-ilvE (C. glutamicum) in pCRB12	This work
pCRB-DLD	Ptac-ilvD (C. glutamicum) and Ptac-LeuDH (L. sphaericus) in pCRB12	This work

^ailvC™, NAD-preferring mutant ilvC; ilvN^GE, feedback-resistant mutant ilvN; α-lac multicloning site, N-terminal fragment of β-galactosidase gene lacZ of E. coli (α fragment) containing the multicloning site region.

^bNITE Biological Resource Center.

Cloning of a feedback inhibition-resistant mutant of AHAS.

C. glutamicum was exposed to 300 μg ml⁻¹ N-methyl-N′-nitro-N-nitrosoguanidine at 30°C for 1 h following cultivation in A medium (ca. 10⁹ cells ml⁻¹). After being washed with BT medium, the treated cells were plated on BT plates containing 4% glucose and 4% dl-α-aminobutyrate, an analogue of valine, and incubated at 30°C for 5 days. Mutant strains that grew on the valine analogue plate were screened by l-valine productivity after cultivation in A medium at 33°C for 48 h. The nucleotide sequences of chromosomal ilvN genes encoding the regulatory subunit of AHAS from mutant strains with high l-valine productivity were analyzed, and mutation points were identified. AHAS activity of the resulting mutant strains and its resistance to l-valine inhibition were evaluated.

Construction of plasmids and strains.

General DNA manipulations were performed according to reference (38). All plasmids and primers used in this study are listed in Table 1 and in Table S1 in the supplemental material. pCRB207 was constructed from E. coli plasmid pHSG298 and coryneform plasmid pCASE1 (45). The entire pHSG298 and the replication origin of pCASE1 were amplified by PCR with primers P1 and P2 and primers P3 and P4, respectively. Each amplicon was digested with BglII and ligated, resulting in pCRB22. The promoter region of C. glutamicum gapA (amplified by PCR from chromosomal DNA with primers P5 and P6 and digested with SalI) was inserted into the SalI site of pCRB22, and subsequently the rrnB terminator region (amplified from plasmid pKK223-3 with primers P7 and P8 and digested with NcoI and BspHI) was inserted into the NcoI site of pCRB22 carrying the gapA promoter.

pCRB12 was derived from pHSG298 and coryneform plasmid pCG1 (GenBank accession no. AB027714). The entire pHSG298 and the replication origin of pCG1 were amplified by PCR with primers P1 and P2 and primers P9 and P10, respectively. Each resulting amplified product was digested with BglII and ligated.

pCRB21 was constructed from pHSG398 and pCASE1. The entire pHSG398 and the replication origin of pCASE1 were amplified by PCR with primers P11 and P12 and primers P3 and P4, respectively. Each resulting amplicon was digested with BglII and ligated.

pCRB-BNC was constructed as follows. The DNA fragments of ilvBN and ilvC were amplified by PCR from C. glutamicum chromosomal DNA using primers 1 and 2 and primers 3 and 4, respectively. Each amplified DNA fragment was digested with MluI and BspHI or NcoI and ligated to pCRB207 digested with NcoI. The resulting plasmid (harboring the 5′ region of ilvB and the 3′ region of ilvC) was digested with MluI and ligated with an MluI-digested PCR fragment, which was generated using primers 1 and 4, containing the middle region of the ilvBNC operon. Finally, full-length ilvBNC with the gapA promoter and the rrnB terminator was excised with BamHI from pCRB207 carrying ilvBNC and ligated into the BamHI site of pCRB21. pCRB-BN^GEC was also constructed, using chromosomal DNA of the valine analogue-resistant mutant as a template for PCR.

Site-directed mutagenesis of ilvC was introduced by PCR using pCRB-BNC and pCRB-BN^GEC as the templates. Primers 5 and 6 were used for the introduction of S34G mutation into ilvC, and subsequent L48E and R49F mutations on ilvC were introduced through primers 7 and 8, resulting in two plasmids harboring ilvC™ (triple S34G L48E R49F mutation), called pCRB-BNC™ and pCRB-BN^GEC™, respectively.

The ilvD, ilvE (C. glutamicum), and LeuDH (L. sphaericus) genes were amplified by PCR from each chromosomal DNA with primers 9 and 10, primers 13 and 14, and primers 17 and 18, respectively. Each amplified product was digested with SmaI and ligated into pKK223-3 digested with SmaI. The resulting plasmids were used as templates to amplify tac promoter-linked ilvD, ilvE, and the LeuDH gene with primers 11 and 12, 15 and 16, and 19 and 20, respectively. The amplified DNA fragments containing tac promoter and each gene were digested with Sse8387I (ilvD) or NheI (ilvE and the LeuDH gene). The digested tac promoter-linked ilvD was ligated into Sse8387I-digested pCRB12, and subsequently tac promoter-linked ilvE or the LeuDH gene was inserted into the NheI site of pCRB12 harboring ilvD. The resulting plasmids were called pCRB-DE (containing ilvD and ilvE) and pCRB-DLD (containing ilvD and the LeuDH gene as well), respectively.

C. glutamicum ΔilvC, the ilvC-null mutant, was constructed by transposon mutagenesis as previously described (42). Plasmids were transformed into C. glutamicum by electroporation (46).

Enzyme assays.

Crude extracts of the cells were prepared according to the method of Leyval et al. (25). Cultivated cells were harvested by centrifugation (5,000 × g; 4°C; 10 min). Cell pellets were washed twice with 2% KCl and resuspended in disruption buffer (100 mM potassium phosphate [pH 7.3], 0.5 mM dithiothreitol, and 20 vol/vol [percent] glycerol). The resulting cell suspensions were sonicated using an ultrasonic homogenizer (Bioruptor UCD-200T; Cosmo Bio, Tokyo, Japan) in an ice water bath for 15 min. Cell debris was removed by centrifugation (20,000 × g; 4°C; 10 min).

AHAS activity was determined at 30°C by monitoring pyruvate decrease at 333 nm (ε = 17.5 M⁻¹ cm⁻¹) (15) by using DU800 spectrophotometer (Beckman Coulter, Inc., Brea, CA). The reaction mixture contained 100 mM potassium phosphate (pH 7.5), 50 mM sodium pyruvate, 10 mM MgCl₂, 0.1 mM thiamine pyrophosphate, 0.1 mM flavin adenine dinucleotide, and crude extract. One unit of AHAS activity was defined as the activity needed to form 1 μmol of 2-acetolactate (i.e., decrease 2 μmol of pyruvate) per min.

AHAIR activity was measured at 30°C as the decrease of NADPH absorbance at 340 nm (25). The assay system contained 100 mM potassium phosphate (pH 7.5), 10 mM 2-acetolactate, 5 mM MgCl₂, 0.2 mM NADPH, and crude extract. 2-Acetolactate was synthesized by hydrolysis of ethyl-2-acetoxy-2-methylacetoacetate (Aldrich Chemical Company, Inc., Milwaukee, WI) (49). One unit of AHAIR activity was defined as the activity necessary to oxidize 1 μmol of NADPH per min.

DHAD activity was assayed at 30°C by forming of 2-ketoisovalerate at 240 nm (ε = 190 M⁻¹ cm⁻¹) (11). The assay mixture contained 50 mM Tris-HCl (pH 8.0), 6 mM 2,3-dihydroxyisovalerate, 10 mM MgCl₂, and crude extract. 2,3-Dihydroxyisovalerate was purchased as a custom-made reagent from Wako Pure Chemical Industries, Ltd. (Osaka, Japan). One unit of DHAD activity was defined as the amount of activity necessary to form 1 μmol of 2-ketoisovalerate per min.

TA activity was determined by quantification of l-valine produced by amination of 2-ketoisovalerate using l-glutamate as an amino-group donor (25). The reaction mixture contained 100 mM Tris-HCl (pH 9.0), 0.25 mM pyridoxal-5′-phosphate, 5 mM sodium 2-ketoisovalerate, 10 mM sodium l-glutamate, and crude extract. The assay was performed at 30°C; 200-μl samples were taken at intervals, and the reaction was terminated by mixing with 60 μl of 21% perchloric acid. l-Valine formation was quantified by high-pressure liquid chromatography (HPLC) after neutralization with 5 M KOH and centrifugation (20,000 × g, 4°C; 10 min). One unit of TA activity was defined as the amount of activity necessary to form 1 μmol of l-valine per min.

LeuDH activity was measured as reductive amination at 30°C by the change of NADH absorbance at 340 nm. The assay system contained 100 mM glycine-NaOH (pH 9.5), 200 mM NH₄Cl, 10 mM sodium 2-ketoisovalerate, 0.2 mM NADH, and crude extract. One unit of LeuDH activity was defined as the activity required to oxidize 1 μmol of NADH per min.

Purification and kinetic analysis of AHAIR.

All purification steps were performed at 4°C. Wild-type and mutant AHAIRs were purified from BNC/ΔLDH and BNC™/ΔilvC, respectively. The standard buffer used for purification consisted of 50 mM Tris-HCl (pH 7.5), 5 mM MgCl₂, 1 mM dithiothreitol, and 1 mM EDTA. Cultivated cells suspended with the standard buffer were disrupted by sonication, and cell debris was removed by centrifugation (18,000 × g, 4°C; 1 h). Each AHAIR was purified from the resultant supernatant by ammonium sulfate fractionation (over 40% saturation), hydrophobic chromatography (HiLoad 16/10 phenyl Sepharose high-performance; GE Healthcare, Bucks, United Kingdom), and anion exchange chromatography (HiPrep 16/10 DEAE FF; GE Healthcare).

For kinetic analysis, the enzyme reaction was performed using purified AHAIRs, and NAD(P)H was monitored at 340 nm with a DU800 spectrophotometer or a SpectraMax M2^e (Molecular Devices, Inc., CA). The ranges of the substrate concentrations were 0 to 20 mM 2-acetolactate and 0 to 1 mM NAD(P)H. For the determination of K_m and k_cat, kinetic data were fitted to a Hanes-Woolf plot (25). Inhibition of NADP⁺ for the AHAIR reaction using NADH as a cofactor was evaluated with reaction mixtures containing 0.2 mM NADH and variable amounts of NADP⁺. Each experiment was performed in triplicate.

Conditions for l-valine production.

l-Valine production during oxygen deprivation was carried out as previously described (18). C. glutamicum strains were aerobically cultivated at 33°C for 14 h in A medium with 4% glucose. Harvested cells were washed with BT medium without urea (BT-U medium) and resuspended in 50 ml BT-U medium containing 200 to 400 mM glucose. This cell suspension was incubated at 33°C without aeration and with gentle agitation (100 rpm). The pH of the reaction solution was maintained at 7.5 by supplementing with NH₃. When necessary, glucose was added to the reaction solution before its depletion. Due to the addition of NH₃ solution and glucose, the reaction volume significantly increased throughout the reaction for the l-valine production. Glucose, l-valine, and by-product concentrations were thus corrected for the increase in volume and subsequent dilution by the original reaction volume. Each experiment was performed in triplicate.

Analytical procedures.

Protein concentration was determined by the procedure of Bradford with a protein assay kit (Bio-Rad Laboratories, Inc., Hercules, CA) using bovine gamma globulin as the standard. Glucose concentration was measured by an enzyme electrode glucose sensor (BF-4; Oji Scientific Instruments, Hyogo, Japan). Organic acids and 2-ketoisovalerate concentrations were determined by HPLC (8020; Tosoh Corporation, Tokyo, Japan) with TSKgel OApak-A columns (Tosoh Corporation) and a UV detector at 210 nm. Amino acid concentrations were determined by HPLC using a Prominence 20A chromatograph (Shimazu Corporation, Kyoto, Japan) equipped with a Shim-pack Amino-Na column (Shimazu Corporation) and a spectrofluorometer after derivatization with ο-phthalaldehyde according to the manufacture's protocol. When l-valine titers exceeded the solubility limit and the l-valine was partially precipitated, samples were taken as suspensions, and l-valine concentrations were measured after dilution and dissolution.

Intracellular metabolites were extracted from C. glutamicum cells as follows. Samples (25 μl) were taken 2 h after the reaction started and immediately quenched by mixing with 1.0 ml cold methanol (−80°C). The resultant cell suspension (0.5 ml) was mixed vigorously with 0.5 ml chloroform and 0.5 ml H₂O (−20°C) to disrupt cells, and after being incubated for 60 min at −20°C, the sample solution was centrifuged (20,000 × g, 4°C; 5 min). An aliquot of the upper layer (50 μl) was mixed with 50 μl H₂O or authentic standard mixture solution (5.0 μM each) and centrifuged (20,000 × g, 4°C; 5 min). The resultant supernatant was analyzed by liquid chromatography-tandem mass spectrometry (LC-MS/MS) using HPLC on a Prominence 20A chromatograph coupled with a linear ion trap mass spectrometer (4000 Q Trap; Applied Biosystems/MDS SCIEX) as previously described (9). Data were obtained from five independent culture samples. A factor of 1.8 ml/g (dry weight) of cells was assumed as the cell volume for the calculation of intracellular concentrations.

RESULTS

Mutant AHAS (IlvN^GE) is resistant to l-valine inhibition.

Seven different mutants of IlvN, a regulatory subunit of AHAS, were isolated from 330 mutants resistant to the valine analogue dl-α-aminobutyrate. The most feedback-resistant mutant retained almost full AHAS activity even in the presence of 10 mM l-valine, while the rest exhibited more than 10% reduced activity in the presence of 1 mM l-valine. In contrast, wild-type AHAS activity was inhibited by ca. 50% by 1 mM l-valine. The most feedback-resistant mutant IlvN had a single substitution at position 156 of glycine (GGA) to glutamate (GAA). This G156E mutant was termed IlvN^GE and used for l-valine production.

Coenzyme preference of triple mutant AHAIR (IlvC™) is reversed.

The structure of C. glutamicum AHAIR with NADPH has not been reported. Accordingly based on previous reports (1, 36, 41), the proposed nicotinamide nucleotide binding site of this enzyme was identified from the amino acid sequence alignment with homologs and modified. As shown in Fig. 2, several NAD(P)-requiring enzymes commonly possess the β-α-β motif, known as the Rossmann fold, in coenzyme-binding domains. NAD-specific enzymes exhibit a highly conserved GXGXXGXXXG sequence (where X is any amino acid), whereas NADP-specific enzymes have a GXGXXAXXXA sequence on the motifs. The first two glycine residues of either sequence contribute to the tight folding of the β-α-β motif for interaction with coenzymes. The latter two glycine or alanine residues could confer local conformational changes associated with the different coenzyme specificities of each sequence. In addition, some amino acid residues on the β-α-β motif play an important role in discriminating between NAD and NADP. Positively charged arginine or lysine residues associate with the negatively charged 2′-phosphate group of NADP by electrostatic interaction. Conversely, glutamate or aspartate residues form hydrogen bonds to the 2′- and/or 3′-hydroxyl groups of NAD, and their negative charge also repels the 2′-phosphate group of NADP via charge-charge repulsion. Based on these observations, a total of three amino acid residues were substituted by site-directed mutagenesis in this study: S34G, L48E, and R49F (Fig. 2). The mutation S34G corresponds to the fourth glycine residue on the NAD-specific conserved GXGXXGXXXG sequence, while L48E and R49F eliminate NADP-specific positively charged arginine and introduce NAD-specific negatively charged glutamate. The resultant triple mutant of AHAIR was designated IlvC™.

Fig 2.

Amino acid sequence alignment of nicotinamide coenzyme-binding enzymes around the β-α-β motif in the Rossmann fold domain and the design of NAD-preferring mutant AHAIR of C. glutamicum. Conserved amino acid residues specific for NAD or NADP binding sites are indicated as follows: ●, the fingerprint regions of the β-α-β-fold (GXGXXGXXXG, NAD specific; GXGXXAXXXA, NADP specific); −, negatively charged amino acid residues (NAD specific); +, positively charged amino acid residues (NADP specific). Substituted amino acid residues (S34G, L48E, and R49F) of the mutant AHAIR of C. glutamicum are shaded. Swiss-Prot accession codes for dihydrolipoamide dehydrogenase: Saccharomyces cerevisiae, P09624; E. coli, P0A9P0; Pseudomonas fluorescens, P14218; C. glutamicum, A4QB06. Swiss-Prot accession codes for glutathione reductase: S. cerevisiae, P41921; E. coli, P06715; Pseudomonas aeruginosa, P23189. Swiss-Prot accession codes for acetohydroxy acid isomeroreductase: E. coli, P05793; Buchnera aphidicola, Q9RQ51; P. aeruginosa, Q9HVA2; C. glutamicum, A4QDN4. ¹⁾, reference 1; cylinders, α-helixes; arrows, β-strands.

Wild-type and mutant AHAIRs were purified for kinetic analysis from the crude extracts of strains overexpressing each AHAIR. Relative to activities in crude extracts, the specific activities of wild-type and mutant AHAIR increased 7.3- and 8.6-fold, respectively, upon purification, and each resultant purified AHAIR was shown to be homogenous by sodium dodecyl sulfate-polyacrylamide gel electrophoresis (SDS-PAGE). Tables 2 and 3 show kinetic constants determined for each substrate and coenzyme of the wild-type and mutant AHAIRs. Wild-type AHAIR was NADP specific mainly because of an ∼100-fold-lower K_m (higher affinity) for NADPH than NADH (Table 3). In contrast, k_cat/K_m of mutant AHAIR with NADPH drastically dropped to less than 1%, whereas that obtained with NADH was ca. 60% that of the wild-type enzyme (Table 2). These properties of the mutant enzyme were due to 20-fold-increased K_m and 40-fold-decreased k_cat for NADPH, in contrast to the kinetic parameters for NADH, which did not significantly change (Table 3). Consequently, k_cat/K_m of the mutant AHAIR using NADH was 10-fold higher than that using NADPH; i.e., the coenzyme specificity of the mutant AHAIR was reversed even though its NADH-specific activity was not improved. This “NAD-preferring” mutant was used for l-valine production.

Table 2.

Kinetic parameters of wild-type and mutant AHAIRs for 2-acetolactate^a

AHAIR	NADH			NADPH
	Km (mM)	kcat (s−1)	kcat/Km (s−1 M−1)	Km (mM)	kcat (s−1)	kcat/Km (s−1 M−1)
Wild-type	2.60 ± 0.22	0.460 ± 0.028	1.77 × 102	0.95 ± 0.09	1.38 ± 0.07	1.45 × 103
Mutant	3.14 ± 0.11	0.330 ± 0.009	1.05 × 102	1.48 ± 0.10	0.0153 ± 0.0003	1.03 × 10

^aSubstrate concentrations were 0.2 mM NAD(P)H and variable amounts of 2-acetolactate. The data are averages from triplicate experiments.

Table 3.

Kinetic parameters of wild-type and mutant AHAIRs for NADH or NADPH^a

AHAIR	NADH			NADPH
	Km (μM)	kcat (s−1)	kcat/Km (s−1 M−1)	Km (μM)	kcat (s−1)	kcat/Km (s−1 M−1)
Wild-type	415 ± 31	1.26 ± 0.06	3.04 × 103	6.53 ± 2.56	1.21 ± 0.02	2.12 × 105
Mutant	500 ± 13	0.938 ± 0.012	1.88 × 103	121 ± 14	0.0291 ± 0.0012	2.42 × 102

^aSubstrate concentrations were 10 mM 2-acetolactate and variable amounts of NAD(P)H. The data are averages from triplicate experiments.

The raised K_m of the mutant for NADPH had another effect on the enzyme reaction when NADH was used as a coenzyme. As shown in Fig. 3, wild-type AHAIR activity using NADH was markedly decreased by NADP⁺. In contrast, the mutant suffered only slightly from this inhibition (K_i calculated of wild-type AHAIR for NADP⁺ was 3.7 μM, whereas that of the mutant was more than 1,000 μM). This was due to competition between NADP⁺ and NADH for binding to the enzymes because of the much higher affinity of the wild-type enzyme for NADP than for NAD and the significantly decreased affinity of the mutant for NADP. Besides reversal of coenzyme preference, this decreased inhibition of the mutant by NADP⁺ with NADH as a cofactor can be advantageous for in vivo reaction where both NAD(H) and NADP(H) exist.

Fig 3.

Relative activity of AHAIR using NADH as a cofactor in the presence of NADP⁺. ●, wild type; □, mutant. Substrate concentrations were 10 mM 2-acetolactate, 0.2 mM NADH, and variable amounts of NADP⁺. Specific activities of the wild-type and mutant AHAIRs in this reaction (without NADP⁺) were 0.71 U/mg and 0.45 U/mg, respectively. The data are averages from triplicate experiments.

l-Valine-producing strains show elevated activities of l-valine synthesis enzymes.

Since wild-type C. glutamicum produces lactate as a major product under oxygen deprivation conditions, the productivity of which reaches ca. 90% of consumed glucose (17, 18, 32), a mutant deficient in the ldhA gene, encoding lactate dehydrogenase (LDH), must be used as platform strain for l-valine production to minimize lactate production (Fig. 1). In order to enhance the l-valine synthesis pathway, ilvBN, ilvC, ilvD, and ilvE of C. glutamicum were overexpressed in the platform host using plasmids. In addition, three other mutant or exogenous genes were introduced in place of native genes: feedback-resistant mutant ilvBN^GE, NAD-preferring mutant ilvC™, and the L. sphaericus LeuDH gene (Fig. 1).

Table 4 shows the enzyme activities of the resulting l-valine-producing strains. The AHAS activities of strains overexpressing wild-type ilvBN (BNCDE/ΔLDH, BNC™DE/ΔLDH, BNCDLD/ΔLDH, and BNC™DLD/ΔLDH) were increased ca. 50- to 64-fold compared to that of the platform host strain (ΔLDH). The AHAS activity of BN^GEC™DLD/ΔLDH harboring mutant ilvBN^GE (42-fold higher than that of the host) was slightly lower than those of strains overexpressing wild-type ilvBN. However, in the presence of 10 mM l-valine, the activity of the mutant AHAS (0.75 U/mg) was comparable to those of the wild-type (ca. 0.58 to 0.80 U/mg) because of its feedback resistance. The AHAIR activity increased 41- and 45-fold upon overexpressing wild-type ilvC in BNCDE/ΔLDH and BNCDLD/ΔLDH, respectively. It was difficult to evaluate whether the mutant enzyme (IlvC™) was overexpressed, as its enzyme activity using NADPH drastically decreased (Tables 2 and 3). Instead, the overexpression of AHAIRs was verified by SDS-PAGE: levels of expression of wild-type (BNCDE/ΔLDH and BNCDLD/ΔLDH) and mutant (BNC™DE/ΔLDH, BNC™DLD/ΔLDH, and BN^GEC™DLD/ΔLDH) AHAIRs were identical and markedly higher than that of the host (see Fig. S1 in the supplemental material). The DHAD activities of every l-valine-producing strain overexpressing ilvD increased ca. 1.5- to 2.0-fold. TA activity was also enhanced 3.6- and 4.8-fold in BNCDE/ΔLDH and BNC™DE/ΔLDH compared with ΔLDH. A small amount of LeuDH activity (oxidation of NADH) was detected even in the host, which should not have this enzyme, but the activity significantly increased (ca. 38- to 41-fold) in BNCDLD/ΔLDH, BNC™DLD/ΔLDH, and BN^GEC™DLD/ΔLDH by introduction of the LeuDH gene. These results confirmed that every l-valine biosynthesis gene introduced was successfully overexpressed in the l-valine-producing strains.

Table 4.

Enzyme activities for l-valine biosynthesis by C. glutamicum strains

Strain	Enzyme activitya
	AHAS (U/mg)	AHAIR (U/mg)	DHAD (mU/mg)	TA (mU/mg)	LeuDH (U/mg)
ΔLDH	0.020 ± 0.006 (0.010 ± 0.004)	0.009 ± 0.001	29.1 ± 9.9	8.9 ± 1.8	0.22 ± 0.06
BNCDE/ΔLDH	1.24 ± 0.16 (0.80 ± 0.09)	0.369 ± 0.050	57.0 ± 18.1	32.3 ± 5.0	0.12 ± 0.08
BNC™DE/ΔLDH	1.29 ± 0.21 (0.77 ± 0.10)	0.009 ± 0.002	46.2 ± 3.8	43.0 ± 12.7	0.17 ± 0.04
BNCDLD/ΔLDH	1.22 ± 0.17 (0.78 ± 0.11)	0.406 ± 0.035	43.1 ± 17.5	8.2 ± 0.5	9.03 ± 0.95
BNC™DLD/ΔLDH	1.01 ± 0.06 (0.58 ± 0.08)	0.008 ± 0.003	47.6 ± 12.4	7.8 ± 0.1	8.94 ± 0.93
BNGEC™DLD/ΔLDH	0.85 ± 0.20 (0.75 ± 0.19)	0.026 ± 0.004	51.2 ± 6.5	8.4 ± 0.7	8.46 ± 0.14

^aThe data are averages from triplicate experiments. Values in parentheses are residual activities in the presence of 10 mM l-valine.

The NAD-preferring mutant AHAIR (IlvC™) improves redox balance, increasing l-valine production.

Table 5 shows l-valine production, by-product formation, and glucose consumption of each strain (stoichiometric formulae for synthesis of each product under oxygen deprivation are given in Fig. S2 in the supplemental material). A BNCDE/ΔLDH strain overexpressing four wild-type genes for l-valine synthesis consumed 166 mM glucose and produced 54 mM l-valine after 24 h. In contrast, introduction of the mutant AHAIR (IlvC™) in order to solve cofactor imbalance in the l-valine synthesis significantly increased glucose consumption and l-valine production: BNC™DE/ΔLDH consumed 912 mM glucose and produced 239 mM l-valine within 24 h, values which were 5.5- and 4.4-fold higher than those obtained with BNCDE/ΔLDH, respectively. Similarly, when LeuDH was added (see below), the glucose consumption and l-valine production of BNC™DLD/ΔLDH increased 4.9- and 6.1-fold, respectively, compared with those of BNCDLD/ΔLDH. Thus, the NAD-preferring mutant AHAIR is very effective for l-valine production under oxygen deprivation.

Table 5.

l-Valine production, by-product formation, and glucose consumption of C. glutamicum strains after 24 h of oxygen deprivation

Strain	Concn (mM) (yield [%])a
	l-Valine	Alanine	Ketoisovalerate	Lactate	Succinate	Acetate	Glucose consumption
ΔLDH	5.4 ± 1.0 (4.8 ± 0.4)	14.9 ± 1.5 (6.7 ± 0.2)	ND	5.2 ± 0.1 (2.3 ± 0.3)	111 ± 9 (49.3 ± 2.0)	48.1 ± 3.4 (21.5 ± 1.5)	112 ± 13
BNCDE/ΔLDH	53.9 ± 1.3 (32.5 ± 1.4)	2.3 ± 0.5 (0.7 ± 0.2)	0.7 ± 0.0 (0.4 ± 0.0)	3.8 ± 1.2 (1.1 ± 0.3)	142 ± 5 (42.9 ± 1.4)	20.3 ± 0.6 (6.1 ± 0.3)	166 ± 8
BNC™DE/ΔLDH	239 ± 25 (26.2 ± 2.2)	3.0 ± 0.5 (0.2 ± 0.0)	164 ± 21 (18.0 ± 2.1)	2.9 ± 0.5 (0.2 ± 0.0)	632 ± 53 (34.6 ± 1.9)	62.3 ± 11.9 (3.4 ± 0.6)	912 ± 39
BNCDLD/ΔLDH	190 ± 12 (43.0 ± 1.7)	53.9 ± 10.9 (6.1 ± 1.3)	ND	7.0 ± 1.3 (0.8 ± 0.2)	278 ± 16 (31.5 ± 0.8)	41.5 ± 2.3 (4.7 ± 0.4)	442 ± 18
BNC™DLD/ΔLDH	1,170 ± 50 (54.1 ± 2.7)	140 ± 9 (3.2 ± 0.2)	ND	ND	1,010 ± 110 (23.4 ± 2.5)	233 ± 27 (5.4 ± 0.6)	2,160 ± 10
BNGEC™DLD/ΔLDH	1,470 ± 50 (63.0 ± 1.5)	37.4 ± 2.5 (0.8 ± 0.1)	ND	ND	928 ± 26 (19.9 ± 0.3)	108 ± 4 (2.3 ± 0.1)	2,330 ± 30

^al-Valine production, by-product formation, and glucose consumption values were corrected for dilution caused by the addition of NH₃ solution and glucose throughout the reaction. Yields in parentheses are expressed as a percentage of the theoretical yield (100% means 1 mol of l-valine and ketoisovalerate or 2 mol of alanine and organic acids produced per mol of glucose consumed). Cell concentration was 40 g (dry weight) cell liter⁻¹. The data are averages from three independent experiments. ND, not detected.

Table 6 shows intracellular intermediates in the l-valine-producing strains. The results show a tendency for the mutant AHAIR to cause a decrease in the metabolites upstream of GAPDH (dihydroxyacetone phosphate and glyceraldehyde-3-phosphate) and an increase in those downstream of GAPDH (1,3-bisphosphoglycerate and pyruvate). Moreover, as shown in Table 7, the intracellular NADH/NAD⁺ ratio was significantly decreased by the mutant AHAIR from 2.19 to 1.35 with TA and from 2.12 to 0.52 with LeuDH. These observations suggested that GAPDH should be the rate-limiting step in the overall l-valine synthesis pathway under oxygen deprivation, and this step can be accelerated by a decrease in the intracellular NADH/NAD⁺ ratio. It follows, therefore, that the drastic increases in glucose consumption and concomitant l-valine production by the NAD-preferring mutant AHAIR under oxygen deprivation were predominantly caused by the reduced intracellular NADH/NAD⁺ ratio.

Table 6.

Intracellular concentrations of metabolic intermediates from glycolysis and l-valine synthesis pathway at 2 h during l-valine production under oxygen deprivation conditions

Strain	Metabolite concn (mM)a
	DHAP	GAP	BPG	Pyruvate	AL	DHIV	KIV
BNCDE/ΔLDH	6.10 ± 0.47	0.40 ± 0.04	0.06 ± 0.01	1.93 ± 0.38	0.20 ± 0.02	0.09 ± 0.01	8.62 ± 0.40
BNC™DE/ΔLDH	5.15 ± 0.27	0.44 ± 0.02	0.14 ± 0.02	8.79 ± 0.80	6.00 ± 0.52	0.46 ± 0.02	43.10 ± 5.30
BNCDLD/ΔLDH	7.51 ± 0.72	0.43 ± 0.05	0.07 ± 0.01	2.41 ± 0.27	0.26 ± 0.05	0.31 ± 0.05	ND
BNC™DLD/ΔLDH	2.98 ± 0.11	0.26 ± 0.02	0.17 ± 0.02	21.20 ± 5.80	46.40 ± 7.80	5.19 ± 0.32	0.90 ± 0.07
BNGEC™DLD/ΔLDH	3.02 ± 0.15	0.31 ± 0.04	0.16 ± 0.02	0.67 ± 0.08	63.70 ± 23.10	11.00 ± 0.70	0.53 ± 0.08

^aDHAP, dihydroxyacetone phosphate; GAP, glyceraldehyde-3-phosphate; BPG, 1,3-bisphosphoglycerate; AL, 2-acetolactate; DHIV, 2,3-dihydroxyisovalerate; KIV, 2-ketoisovalerate. The data are from five analytical replicates from one experiment. ND, not detected.

Table 7.

Intracellular NADH/NAD⁺ ratio at 2 h during l-valine production under oxygen deprivation conditions

^aThe data are from five analytical replicates from one experiment.

LeuDH increases l-valine production by improving redox balance and accelerating amination.

LeuDH effectively catalyzes the reversible deamination of l-leucine and other BCAAs (l-valine and l-isoleucine) using NAD⁺ as a cofactor (Fig. 1) (30). Hence, introduction of LeuDH in place of endogenous TA should improve the intracellular redox balance by reoxidizing NADH, increasing l-valine production. In fact, glucose consumption and l-valine productivity were improved by LeuDH (Table 5). The LeuDH-harboring strain BNCDLD/ΔLDH consumed 442 mM glucose and produced 190 mM l-valine, values that are 2.7- and 3.5-fold higher than those for BNCDE/ΔLDH, respectively. LeuDH also worked efficiently in combination with NAD-preferring mutant AHAIR. Glucose consumption and l-valine productivity of BNC™DLD/ΔLDH were 2,160 mM and 1,170 mM, respectively, which represented 2.4- and 4.9-fold increases, respectively, over values for BNC™DE/ΔLDH. Moreover, l-valine yield from glucose consumption was improved by LeuDH from 33 to 43% with the wild-type AHAIR and from 26 to 54% with the mutant AHAIR. These results demonstrated that LeuDH was strongly effective for l-valine production under oxygen deprivation.

As shown in Table 7, despite introduction of LeuDH, the intracellular NADH/NAD⁺ ratio of BNCDLD/ΔLDH was almost identical to that of BNCDE/ΔLDH (2.19 and 2.12, respectively). However, with the NAD-preferring mutant AHAIR, the redox ratio was significantly reduced by LeuDH from 1.35 (BNC™DE/ΔLDH) to 0.52 (BNC™DLD/ΔLDH). LeuDH therefore decreased the intracellular NADH/NAD⁺ ratio due to reoxidization of NADH and contributed to an increase in glucose consumption and l-valine production.

Furthermore, it is noteworthy that 2-ketoisovalerate, a precursor of l-valine before amination (Fig. 1), was detected extracellularly only in BNCDE/ΔLDH and BNC™DE/ΔLDH, both of which lacked LeuDH (Table 5). As shown in Table 6, intracellular concentrations of 2-ketoisovalerate in these strains (8.6 and 43.1 mM, respectively) were also much higher than those in other strains overexpressing LeuDH (less than 1.0 mM). These results suggested that LeuDH catalyzed amination of 2-ketoisovalerate much more efficiently than TA under oxygen deprivation, and the resultant reduction in intracellular accumulation of the keto acid accelerated the overall reaction rate for l-valine synthesis. The superiority of LeuDH to TA is prominent especially in the strains overexpressing mutant AHAIR, where glucose consumption was markedly increased. In the case of BNC™DE/ΔLDH, conversion of 2-ketoisovarelate to l-valine by TA was absolutely insufficient, and large amounts of the keto acid were excreted extracellularly. This by-product formation was equivalent to 18% from glucose consumption and 69% of l-valine productivity (Table 5). In contrast, intracellular 2-ketoisovarelate accumulation was significantly decreased, and extracellular excretion of the keto acid was not detected in BNC™DLD/ΔLDH despite its higher glucose consumption (Tables 5 and 6). Collectively, LeuDH improved l-valine production by not only improving the intracellular redox balance but also accelerating the amination reaction of 2-ketoisovalerate.

Feedback-resistant mutant AHAS improves l-valine production and limits formation of by-products.

In order to derepress the feedback inhibition, the feedback-resistant mutant AHAS (IlvBN^GE) was introduced into the l-valine-producing strain in addition to the mutant AHAIR and LeuDH. As shown in Table 5 and Fig. 4, BN^GEC™DLD/ΔLDH produced 1,470 mM l-valine within 24 h, which was ca. 25% higher than the value for BNC™DLD/ΔLDH, with a volumetric productivity of 61.2 mmol liter⁻¹ h⁻¹ (the rate during the first 2 h was 118 mmol liter⁻¹ h⁻¹). After 48 h, the l-valine productivity reached 1,940 mM. Moreover, the mutant AHAS led to an increase in l-valine yield from glucose consumption from 54 to 63% as well as a decrease in coproduction of alanine, succinate, and acetate. The analysis of intracellular intermediates shows that pyruvate accumulation was significantly decreased in the strain overexpressing the mutant AHAS (Table 6). These results suggested that the feedback-resistant mutant AHAS efficiently catalyzed the conversion from pyruvate to 2-acetolactate in vivo during l-valine production, which led to reduced accumulation of intracellular pyruvate. Consequently, the feedback-resistant mutant AHAS was effective for improvement of l-valine productivity and yield as well as minimization of by-product formation derived from pyruvate.

Fig 4.

l-Valine production by BN^GEC™DLD/ΔLDH under oxygen deprivation conditions. ■, l-valine; ♦, succinate; ▲, acetate; ●, alanine; □, glucose. l-Valine production, by-product formation, and glucose concentration were corrected for dilution caused by the addition of NH₃ solution and glucose throughout the reaction. The data are averages from three independent experiments.

DISCUSSION

Cofactor production and consumption are imbalanced in the overall pathway for l-valine synthesis (Fig. 1). The resultant surplus NADH generation is suspected to be a key issue affecting l-valine production, particularly under oxygen deprivation conditions, where NADH cannot be oxidized by respiration, in addition to the supply of NADPH. In order to circumvent this cofactor imbalance, the coenzyme specificity of AHAIR was reversed, and LeuDH was introduced in place of TA.

The mutant AHAIR (IlvC™) constructed in this study preferred NAD over NADP (Tables 2 and 3), while k_cat/K_m of the mutant was inferior to that of the wile-type even when NADH was used as a cofactor (Table 2). Thus, it seems that the wild-type reoxidizes NADH and contributes to l-valine production more effectively than the mutant under oxygen deprivation conditions. However, NADH-dependent activity of the wild-type AHAIR was strongly inhibited by NADP⁺, whereas the mutant was only slightly inhibited (Fig. 3). The intracellular NADP⁺ concentration determined during l-valine production with oxygen deprivation was 170 to 300 μM: at this concentration, the wild-type AHAIR activity using NADH is decreased to less than 10%, resulting in the NAD-dependent activity of the mutant becoming higher than that of the wild type. Therefore, the mutant AHAIR was able to work efficiently for l-valine production using NADH in vivo under oxygen deprivation conditions.

LeuDH is an amino acid dehydrogenase which catalyzes reversible addition and elimination of an amino group on amino acids using NAD(P) as a cofactor (i.e., the oxidative deamination or the reductive amination). Although amino acid dehydrogenases have been used for in vitro synthesis of various optical active or artificial amino acids (31), their usability in biological amino acid fermentation is not very well recognized, except that of alanine dehydrogenase (AlaDH) (13, 20, 23). AlaDH is NAD specific and thus enables more efficient alanine production with smaller amounts of oxygen (23) and contributes to maintenance of the glycolytic flow by reoxidizing NADH (13). Therefore, the NAD-specific LeuDH is expected to be effective for l-valine synthesis under oxygen deprivation. Indeed, LeuDH increased l-valine production (Table 5) by improvement of the intracellular redox balance (Table 7) as well as through a much more efficient amination than TA (Tables 5 and 6). This is presumably due to differences in amino-group donors: LeuDH uses NH₃, whereas the donor for TA is primary l-glutamate (26, 27), and NADPH is required for regeneration of l-glutamate (5). In other words, by using LeuDH instead of TA, an amino-group donor and a reduced cofactor needed for l-valine synthesis were changed from l-glutamate and NADPH to NH₃ and NADH, respectively (Fig. 1). This suggests a drawback of TA, which may be limited by a supply of l-glutamate. On the other hand, sufficient amounts of NH₃ for LeuDH can be easily supplied by adding NH₃ to maintain the pH of the reaction solution. Intracellularly accumulated NADH, another cofactor for LeuDH, under oxygen deprivation conditions would also serve as a driving force to shift a reaction equilibrium toward l-valine formation.

As expected, the NAD-preferring mutant AHAIR (IlvC™) and LeuDH significantly reduced the intracellular NADH/NAD⁺ ratio (Table 7), resulting in drastic increases in glucose consumption and l-valine productivity (Table 5). Both glucose consumption and l-valine productivity showed very good correlation with the NADH/NAD⁺ ratio of each strain, except in BNC™DE/ΔLDH, where l-valine production was limited due to poor amination by TA. These observations suggest that the intracellular redox state, which affects GAPDH activity, is the primary rate-limiting factor for glucose metabolism as well as l-valine synthesis during oxygen deprivation, which agrees with previous results (18, 20, 33). Moreover, l-valine yield from glucose consumption also improved proportionally with a decrease in the intracellular redox state except in BNC™DE/ΔLDH, whereas succinate formation was reduced concomitant with improvement of the redox state (Tables 5 and 7). Note that succinate synthesis through oxaloacetate and the reductive arm of the tricarboxylic acid cycle consume an additional one mole of NADH (6) (Fig. 1). Hence, succinate should be coproduced in order to reoxidize surplus NADH generated by l-valine synthesis, in which cofactor production and consumption are imbalanced, under oxygen deprivation conditions. Although BN^(GE)C™DLD/ΔLDH, in which redox reactions for l-valine synthesis are theoretically balanced, produced some succinate, this may reflect the disturbed intracellular redox balance resulting from small amounts of unexpected carbon flux. Taken together, in l-valine production under oxygen deprivation conditions, improvement of the intracellular redox balance is essential to not only increase l-valine productivity but also decrease succinate formation and increase l-valine yield.

In addition to improvement of the redox balance, the feedback-resistant mutant IlvN^GE was used for l-valine production. This mutant enzyme obtained in this study had a single substitution, G156E. Almost all feedback-resistant IlvN mutants reported so far are classified into two groups: in one group, an amino acid alteration occurs near the N-terminal region, and in the other, the C-terminal half of IlvN is truncated (10, 22, 29, 47). The capacity to bind l-valine is lost in the C-terminal truncations, due to instability of IlvN dimer interaction, since l-valine should come into contact with each IlvN between dimers as well as E. coli IlvH, which is a homolog of IlvN (29). It is similarly lost around the N-terminal mutations; the N terminus was proposed as the l-valine binding site (22). This functional change is the presumed reason for the reduced sensitivity to l-valine feedback inhibition. G156E is different from both the N-terminal mutations and the C-terminal truncations: this mutation is near the C terminus, where peptide chains of E. coli IlvH, homologous to IlvN, intersect each other upon dimer formation (22). This mutation, in which the small side chain of glycine is substituted with the bulky glutamate, should consequently result in instability of dimer interaction as well.

The feedback-resistant mutant AHAS (IlvBN^GE) resulted in a further increase in l-valine productivity and yield. BN^GEC™DLD/ΔLDH accordingly produced 1,470 mM after 24 h, with a yield of 63% from glucose, and 1,940 mM after 48 h (Table 5 and Fig. 4). These productivities are much higher than those obtained in previous studies, e.g., 412 mM with a yield of 75% after 74 h, 240 mM with a yield of 86% after 46 h (4), and 250 mM with a yield of 67% (44) with C. glutamicum. In the case of E. coli, l-valine productivity was up to 518 mM with a yield of 33.8% after 29.5 h (34). Several successful cases have been reported so far where mutation of cofactor requirement in order to improve the redox balance caused an increase in production. For example, ethanol fermentation by Saccharomyces cerevisiae was improved by 14% by switching cofactor requirement of xylose reductase from NADPH to NADH (48) and by 20% by switching xylitol dehydrogenase from NAD⁺ to NADP⁺ dependence (28). Also, a C. glutamicum strain whose GAPDH was substituted by one directly generating NADPH required for l-lysine synthesis showed a 2.2-fold increase in its l-lysine production relative to its NADH-generating parent (43). Separately, introducing transhydrogenase improved the l-lysine production of C. glutamicum up to 4-fold (21). However, drastic improvement such as the 20-fold (from 54 mM to 1,170 mM) increase in l-valine production realized in this study has not previously been attributed to improved cofactor imbalance. Several by-products were coproduced, but these could be reduced by genetic disruption based on previous analysis of each biosynthetic pathway (e.g., for alanine, alaT and avtA [26, 27]; for succinate, ppc [18]; for acetate, pta-ack and ctfA [50]), resulting in further increased l-valine yield. In conclusion, our bioprocess of C. glutamicum under oxygen deprivation was able to produce considerable amounts of l-valine and can be applied to the efficient production of many kinds of commodity chemicals, including other amino acids.