Вы находитесь на странице: 1из 54

CHAPTER 1

Biotechnological Manufacture of Lysine


Walter Pfefferle 1 Bettina Mckel Brigitte Bathe Achim Marx 2
Degussa AG, Feed Additives Division, R&D Feed Additives/Biotechnology, Kantstrasse 2, 33790 Halle-Kuensebeck, Germany 1 E-mail: walter.pfefferle@degussa.com 2 E-mail: achim.marx@degussa.com

l-Lysine has been manufactured using Corynebacterium glutamicum for more than 40 years. Nowadays production exceeds 600,000 tons per year. Based on conventionally bred strains, further improvement of lysine productivity has been achieved by genetic engineering. Pyruvate carboxylase, aspartate kinase, dihydrodipicolinate synthase, homoserine dehydrogenase and the specific lysine exporter were shown to be key enzymes for lysine production and were characterized in detail. Their combined engineering led to a striking increase in lysine formation. Pathway modeling with data emerging from 13C-isotope experiments revealed a coordinated flux through pentose phosphate cycle and tricarboxylic acid cycle and intensive futile cycling between C3 compounds of glycolysis and C4 compounds of tricarboxylic acid cycle. Process economics have been optimized by developing repeated fed-batch techniques and technical continuous fermentations. In addition, on-line metabolic pathway analysis or flow cytometry may help to improve the fermentation performance. Finally, the availability of the Corynebacterium glutamicum genome sequence has a major impact on the improvement of the biotechnological manufacture of lysine. In this context, all genes of the carbon flow from sugar uptake to lysine secretion have been identified and are accessible to manipulation. The whole sequence information gives access to post genome technologies such as transcriptome analysis, investigation of the proteome and the active metabolic network. These multi-parallel working technologies will accelerate the generation of knowledge. For the first time there is a chance of understanding the overall picture of the physiological state of lysine overproduction in a technical environment.
Keywords. l-Lysine, Fermentation, Strain development, Corynebacterium glutamicum, Scale down, Metabolic engineering

1 2 2.1 2.2 2.3 2.4 2.5

Introduction

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . 60 . . . . . . . . . . . . . . . . . . . . . . . . . 62 . . 62 . . 63 . . 68 . . 70 . . 73

Strain Development

Historical Routes . . . . . . . . . . . . . . . . . . . . . . . . . Genetic Engineering of the Key Enzymes Aspartate Kinase and Dihydrodipicolinate Synthase . . . . . . . . . . . . . . . Availability of Oxaloacetate as a Precursor for the Aspartic Acid Family . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Availability of NADPH for Biosynthetic Reductive Power . . . Enhancement of Secretion: Influence on Activity of Amino Acid Exporters by the Lipid Environment . . . . . . . . . . . . . . .

Advances in Biochemical Engineering/ Biotechnology, Vol. 79 Managing Editor: T. Scheper Springer-Verlag Berlin Heidelberg 2003

60 2.6 2.7 3

W. Pfefferle et al.

Combination of Enhancement of Enzymes for the Lysine Biosynthetic Pathway . . . . . . . . . . . . . . . . . 77 Enhancement of Stress Resistance . . . . . . . . . . . . . . . . . 79 Generation of Basic Knowledge for the Understanding of Lysine Production with Corynebacterium glutamicum for Further Strain and Process Development . . . . . . . . . . . . . . . . . . . . . . 81 Corynebacterium glutamicum Genome Analysis Identification of All Genes Leading from Sugar Uptake to Lysine Secretion . . . . Expression Profiling as a Tool for Strain and Process Optimization Pathway Modeling with Data Emerging from 13C-Isotope Experiments and Applications for Biochemical and Metabolic Engineering . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Genotype and Metabolic Phenotype . . . . . . . . . . . . . . . . Short Review on Isotope Analysis . . . . . . . . . . . . . . . . . . Homologous and Heterologous Glutamate Dehydrogenase Mutant Amphibolic Function of the Glyoxylate Cycle Anaplerosis and Link to C4-Compound Decarboxylation . . . . . . . . . . . . . . Comparison of in-Vivo and in-Vitro Characterization of Enzyme Activities . . . . . . . . . . . . . . . . . . . . . . . . . . Application of Fluxomics Exploitation of Metabolic Flux Information . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 81 86 87 87 88 89 92 94 95

3.1 3.2 3.3 3.3.1 3.3.2 3.3.3 3.3.4 3.3.5 3.3.6 4 4.1 4.2 4.3 4.3.1 4.3.2 5 6 7

Optimization of the Lysine Fermentation Process . . . . . . . . . 95 Improvement of Process Economics by Process Intensification Scale up and Scale down Techniques . . . . . . . . . . . . . . Process Control . . . . . . . . . . . . . . . . . . . . . . . . . . Physiological State Estimation and Optimization of Nutrient Supply . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . Single Cell Analysis by Flow Cytometry . . . . . . . . . . . . . . . 95 . . 97 . . 99 . . 99 . . 103

Large Scale Lysine Manufacture . . . . . . . . . . . . . . . . . . . 103 Outlook . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 106 References . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 107

1 Introduction
l-Lysine is an essential amino acid that has to be available in sufficient amounts in feed-stuffs to meet the nutritional requirements of the animals. Especially feed which is based on corn, wheat or barley is poor in lysine. Therefore, supplementation of a lysine-rich source is necessary to increase the efficacy of the feed (Fig. 1).

Biotechnological Manufacture of Lysine

61

Fig. 1. Demand of lysine, threonine and methionine in piglet feed and natural content of these amino acids in plant biomass (digestible amino acids)

This supplementation can be realized by adding feed ingredients which exhibit significant higher lysine content like soybean meal or by directly adding lysine. The advantage of supplementing lysine is that intake of other essential amino acids is not increased concomitantly. For example, addition of 0.5% lysine increases the protein quality of the feed as effectively as 20% soy meal. Since the nitrogen of any superfluously added non-limiting amino acids of a high protein diet is degraded to ammonia which is excreted by the animals, the supplementation with l-lysine makes it possible to reduce the amount of protein supplement thereby decreasing environmental pollution by manure. For the last ten years a tremendous growth in the market has taken place (Fig. 2). In 2000, the world-wide production of l-lysine used as a feed additive was approx. 550,000 tons and the market still shows a growth potential of 710% per year. Since only l-lysine is effective as a feed additive, all todays manufacturing processes use the fermentation route [1]. The sole production organism is Corynebacterium glutamicum (including subspecies, Table 1), whose capability of secreting amino acids was discovered by Kinoshita et al. [2, 3] and Udaka [4]. Corynebacterium glutamicum is a gram-positive, irregularly shaped, nonmotile, non-spore forming aerobic rod-like bacterium having a peptido-glycan crosslinked via meso-diaminopimelic acid, a cell wall with an arabino-galactan and mycolic acids showing 26 to 36 carbons [6]. Corynebacterium glutamicum shows a GC content of 54.1%. Large scale production with Corynebacterium glutamicum started as early as 1958 at Kyowa Hakkos plant in Japan. Other companies joined the business and subsequently, during 4 decades of production with Corynebacterium glutamicum,

62
600

W. Pfefferle et al.

Annual Lysine Production [103 tons]

500

400

300

200

100

0 1970 1975 1980 1985 1990 1995 2000 2005

Year

Fig. 2. Fermentative lysine production in tons per year during the last 3 decades

Table 1. Subspecies belonging to Corynebacterium glutamicum ATCC13032 according to [5]

Corynebacterium glutamicum ATCC13032 Brevibacterium flavum ATCC14067 Brevibacterium lactofermentum ATCC13869 Corynebacterium lilium ATCC15990 Brevibacterium divaricatum ATCC14020

the biotechnological manufacture of l-lysine has been improved constantly by strain improvement and process engineering. These optimizations made possible the availability of the large amounts of lysine required nowadays.

2 Strain Development
2.1 Historical Routes

Soon after discovery of the ability of Corynebacterium glutamicum to secrete amino acids, large scale production of lysine was started using mutants auxotrophic for amino acids. Even these early mutants displayed a remarkable lysine productivity. Within 70 hours, one of the first generation strains such as the ho-

Biotechnological Manufacture of Lysine

63

moserine auxotrophic strain ATCC13287, patented in 1961 as US 2,979,439 [7], yielded 44 g L1 with a conversion yield of approx. 26% g lysine HCl (g sugar)1 [8]. Further strain development was carried out by introducing additional amino acid auxotrophies, vitamin auxotrophies and resistances to antimetabolites. These strains produce lysine up to a conversion yield of more than 50% based on sugar. A summary is given by Leuchtenberger [1]. As early as1970 [8], there were efforts to overcome the disadvantage of the auxotrophic strains that fermentation broths need substantial and defined growth factor supplementation. Looking for strains with lower requirements for the demanded compounds led to so-called leaky strains; a leaky biosynthetic pathway is still functional but can only provide very limited amounts of the required metabolite. So this metabolite will become growth limiting without showing the typical metabolic pattern of starvation. In consequence, the intracellular level will be very low thus avoiding feedback inhibition and repression of a key enzyme of the desired production pathway, respectively. Examples are strains with reduced activity of citrate synthase [9] or reduced homoserine dehydrogenase [10, 11]. Sugimoto et al. [12] revealed that leaky mutations may be caused by single base changes leading to relevant amino acid replacements in the catalytic domain of the enzyme.
2.2 Genetic Engineering of the Key Enzymes Aspartate Kinase and Dihydrodipicolinate Synthase

The first reaction initiating l-lysine synthesis is catalyzed by the aspartate kinase which has been shown to be a key enzyme in l-lysine biosynthesis (Fig. 3). In Escherichia coli, three isoenzymes with activity and expression controlled by different regulatory mechanisms are present. In contrast to the situation of E. coli, only one aspartate kinase has been detected in C. glutamicum. In the early days of lysine production with C. glutamicum, it became apparent that the aspartate kinase of C. glutamicum is inhibited by the two end products of the emerging pathways, l-lysine and l-threonine and that the main control of the carbon flow into the pathway is mediated by this concerted feedback inhibition. In the screening process to obtain feedback resistant strains, mutants resistant to the mixtures of the l-lysine analogue AEC (amino-ethyl-cysteine) and l-threonine were isolated [13]. The reduction of the susceptibility of aspartate kinase to feedback inhibition is the most important step in the development of l-lysine producing strains. In 1990, the corresponding gene of an AEC resistant mutant, the feedback resistant aspartate kinase of strain DM 581 with a LysC activity completely insensitive to feedback inhibition was isolated [14]. The mutation leading to AEC resistance of the lysC gene was elucidated [15]. These studies showed that the lysC locus encoding aspartate kinase is composed of two overlapping genes, lysCa and lysCb with lysCb being responsible for the feedback inhibition of the enzyme [15, 16]. Also the enzyme has an interesting structure as shown in Fig. 4.

64

W. Pfefferle et al.

Fig. 3. The split pathway of l-lysine biosynthesis in C. glutamicum

Biotechnological Manufacture of Lysine

65
allosteric center lysCb subunit

native aspartate kinase

lysC a a subunit catalytic domain


Fig. 4. Model of the native aspartate kinase with the four identical polypeptide sequences of two lysCb subunits and the C-terminal part of lysCa arranged as the allosteric center

Two a-subunits and two b-subunits of 421 and 171 amino acid residues respectively together form the active enzyme with the sequence of the b-subunits being identical to the carboxy terminal part of the a-subunits. Sequencing of a number of lysC genes from strains carrying a feedback released aspartate kinase provided further evidence that lysCb encodes the regulatory feature of the enzyme [1519]. With such an insensitive LysC protein, C. glutamicum secretes some l-lysine. Cloning a lysC(fbr) gene on a high copy plasmid in the Corynebacterium glutamicum wild type strain ATCC 13032 results in the accumulation of about 38 mM l-lysine compared to no lysine secretion of the parent strain [17]. In a recent study using site-directed mutagenesis, a set of mutant lysC genes was constructed. In these mutant genes, nucleotide substitutions of codon 279 (threonine) of the a-subunit (corresponding to amino acid residue 30 of the bsubunit) revealed nine mutants with different amino acid mutations at the mutation point. For the mutant with a proline residue introduced at position 279, the largest reduction of susceptibility to l-lysine inhibition was observed [19]. This mutation is very likely to have the strongest effect on the structure of the enzyme of all amino acid residues chosen in this study. Since, in proline residues, the side chain is fixed covalently to the main chain of the protein, there is less conformational freedom in the backbone of the protein chain [20]. Jetten et al. [21] analyzed the effect of the copy number of a feedback released aspartate kinase gene on l-lysine production and growth in Brevibacterium lactofermentum ATCC 21799. The strain B. lactofermentum ATCC21799 was chosen because of its high resistance to AEC and high lysine productivity [22] resulting from a feedback released aspartate kinase [21]. In this study, the parental strain was compared to a strain carrying two copies of the mutant lysC gene, showing that the growth rates of the two strains were indistinguishable, but the strain with the second lysC copy produced significantly more l-lysine. With the

66

W. Pfefferle et al.

lysC (fbr) gene on a high copy number plasmid, the resulting strain did not grow as well as the parent strain, but produced larger amounts of lysine. However, the lysine production rate was lower. These results made it clear that, in some cases, a feedback released aspartate kinase is not sufficient to overcome the first limiting step in l-lysine production, but the optimal enzyme activity has to be adjusted to maximize the production rate. The third step in the l-lysine biosynthetic pathway starting from aspartate is the condensation of l-aspartate semialdehyde with pyruvate to l-2,3-dihydrodipicolinate. This reaction is catalyzed by the dihydrodipicolinate synthase. The synthase is located at the branch point of metabolic distribution to either lysine or threonine and competes with homoserine dehydrogenase for their common substrate l-aspartate semialdehyde (Fig. 3). Cloning and overexpression of the dapA gene encoding dihydrodipicolinate synthase revealed that DapA can provide a powerful target in the development of l-lysine producers [17, 2325]. Overexpression of only dapA in the wild type strain ATCC13032 resulted in the production of already about 11 mM l-lysine [17] which is a similar effect like the enhancement of the aspartate kinase. To analyze the function of DapA in metabolic flux toward l-lysine, several strains were constructed with different levels of dapA overexpression [26]. It was shown that the degree of dapA expression is linked to a decreased metabolic flux towards the homoserine dehydrogenase. Strains with a second copy of the dapA gene inserted into the genome, with the dapA gene cloned onto a low copy number plasmid and on a multi-copy plasmid were elucidated. An increased copy number of the dapA gene leads to increased DapA activity and results in a stepwise increase of l-lysine formation. However, interestingly, the growth rate of the resulting strains is reduced with increased DapA activity. This is similar to the effect of lysC overexpression in C. lactofermentum observed by Jetten et al. [18]. A global response of the metabolism to the DapA activity became apparent (Table 2). Besides increasing the copy number of a gene there are several more ways to modulate gene expression in prokaryotes. One of these is to modify transcription initiation signals. However, there is still limited knowledge available about promoter structure in C. glutamicum and the way how to optimize the transcription signal to enhance gene expression. Studying promoter structure and function in
Table 2. The effect of various dapA copy numbers on growth rate, internal alanine concentrations and l-lysine excretion rates. Dihydrodipicolinate synthase activity (mmol min1 mg1 protein). Excretion rates (nmol min1 (mg dry weight)1)

Strain 13032 13032::dapA 13032/pKW3::dapA 13032/pJC23

dapA Copies 1 2 6 20

Synthase activity 0.051 0.072 0.250 0.630

Growth rate (h1) 0.43 0.37 0.36 0.22

Intracellular alanine (mM) 3 6 8 9

Excretion rates 0.0 0.25 2.70 3.80

Biotechnological Manufacture of Lysine

67

Fig. 5. DNA sequence of the dapA promoter region as described by Vasicova et al. (1999). The

arrows mark an inverted repeat in the 35 region. Replacement bases are indicated in bold face and underlined. Sequences of wild type P-dapA, mutant MC20 and MA16 are described. TS marks the transcription start (underlined A)

C. glutamicum showed that the activity of the dapA promoter is moderately high compared to other promoters of the bacterium [27]. In comparison to E. coli or B. subtilis, there is a low conservation of the 10 in C. glutamicum promoters and in some cases the 35 region is missing [28]. A mutational analysis of the dapA promoter was performed to identify regions and the particular nucleotides important for its function. It was found that a stretch of six Ts at position 55 to 50 and an extended 10 region were the most important elements in the promoter function. For the dapA promoter the extended 10 motive was defined (AGGTAACCTTG) and no 35 region. This situation is similar to E. coli sigma 70 depending promoters with an extended 10 region. Vasicova et al., [28] made it clear, that the dapA promoter activity can be increased 3.9 times by introducing single nucleotide exchanges in the extended 10 region (Fig. 5). Subsequent studies with C. glutamicum MH2022B mutants carrying a second copy of the dapA gene with the wild type promoter or the MA16 or MC20 mutation were performed (Fig. 6). As expected, the DapA activity of the strains was significantly enhanced compared to the parent strain MH2022B resulting in enhanced l-lysine production [29].

Fig. 6. The effect of a second dapA copy with different dapA promoter mutations on dihydrodipicolinate synthase activity and l-lysine production. Dihydrodipicolinate synthase activity (mmol min1 mg1). l-lysine production (g L1). Black columns represent the dihydrodipicolinate synthase activity. Grey columns indicate the amount of l-lysine HCl produced

68
2.3 Availability of Oxaloacetate as a Precursor for the Aspartic Acid Family

W. Pfefferle et al.

Genetic and biochemical investigations on the complex network of anaplerotic carboxylation (Fig. 7) and opposed decarboxylation have been reviewed recently [21, 30]. Oxaloacetate is a direct precursor for aspartic acid, therefore its availability is of central importance for the production of lysine. As early as1969, Ozaki and Shiio [32] investigated the regulation of phosphoenolpyruvate carboxylase and pyruvate kinase in Brevibacterium lactofermentum to understand this branching point of glucose metabolism. Enhancement of phosphoenolpyruvate carboxylase was shown to be positive for the formation of aspartic acid derived amino acids in Brevibacterium lactofermentum [33, 34]. Later on, it was shown that in Corynebacterium glutamicum, phosphoenolpyruvate carboxylase is dispensable for growth and lysine production [35]. Therefore, it was obvious that another anaplerotic reaction has to be responsible for replenishing the oxaloacetate pool. The most prominent candidate was pyruvate carboxylase, an enzyme which catalyzes the conversion of pyruvate and carbon dioxide to oxaloacetate under hydrolysis of ATP.

Fig. 7. Complex network of anaplerotic carboxylation of C3-compounds and decarboxylation

of C4-compounds. In C. glutamicum the anaplerotic carboxylation and opposed decarboxylation is potentially catalyzed by five enzymes. In addition, pyruvate kinase and malate dehydrogenase might catalyze the conversion of phosphoenolpyruvate (PEP) into pyruvate and oxaloacetate into malate, respectively. This way the operation of futile cycles might become possible. (From [31]. Kindly provided by Dr. L. Eggeling, Research Center Juelich, Germany)

Biotechnological Manufacture of Lysine

69

Initially, pyruvate carboxylase activity could only be demonstrated by 13C-isotope investigations [36, 37] but recently, significant in-vitro activities were measured and it was shown that growth is decreased in pyruvate carboxylase deletion mutants [38]. Overexpression of this enzyme confirmed that this anaplerotic reaction plays a key role in lysine biosynthesis. Enhancement of pyc in Corynebacterium glutamicum MH2022B increased lysine production by 50% [39]. The computer-aided 13C-isotope analysis indicated for the first time for C. glutamicum that not only C3-compounds of glycolysis are carboxylated to C4-compounds of the tricarboxylic acid cycle but at the same time a decarboxylation of C4- to C3-compounds takes place [40, 41] (Fig. 8). A detailed sensitivity analysis revealed a high sensitivity between the 13C-enrichment in position oxaloacetate C-2 and the bidirectional flux between C4- and C3-compounds [31]. This bidirectional interconversion represents a futile cycle which can result in the consumption of energy without contribution to biosynthesis [43, 44]. This substrate cycle or futile cycle contributes to the high flexibility of the central metabolism of C. glutamicum and was identified as a target for metabolic engineering. To improve metabolite production the net flux has to be

Fig. 8. Net supply of C4- and C3-compounds (black bar) for different physiological states which is the result of simultaneously operating (white bar) carboxylation of C3-compounds and decarboxylation of C4-compounds which has been described elsewhere [31, 40, 41].All values are expressed as percentage of the molar flux of glucose uptake. The particular physiological states for isogenic mutant strains were characterized in chemostat culture [31, 40, 41]. The isocitrate dehydrogenase mutant was cultivated in a shake flask batch culture [42]. A, lysine production with 0.29 g lysineHCl (g glucose)1; B, lysine production with 0.18 g lysineHCl (g glucose)1; C, growth; D, glutamate production; E, growth for an isocitrate dehydrogenase deletion mutant

70

W. Pfefferle et al.

increased. The strategy to improve lysine overproduction by inactivating energy dissipating futile cycles has already been validated experimentally for C. glutamicum by inactivating the phosphoenolpyruvate carboxykinase [45]. Recently, detailed 13C-isotope analysis revealed which particular enzyme catalyzes which portion of the total in-vivo flux of carboxylation and decarboxylation [46]. The net flux of decarboxylation of C4-compounds turned out to be maximum for an isocitrate dehydrogenase mutant of C. glutamicum (Fig. 8). For this physiological state the decarboxylation of C4-compounds plays an essential role as a link between the flux through pyruvate dehydrogenase and glyoxylate cycle to maintain the oxidative metabolism of glucose when the tricarboxylic acid cycle is blocked. In contrast to the situation of the isocitrate dehydrogenase mutant, the net flux of decarboxylation of C4-compounds was decreased to maximize the net supply of C4-compounds in a homologous glutamate dehydrogenase mutant with a high requirement for C4-compounds due to the high lysine yield (Fig. 17; Fig. 8; cf. Sect. 3.3). In conclusion, a significant decarboxylation of C4-compounds was observed although this was not expected for cells metabolizing glucose. A decrease in the decarboxylation of C4-compounds in combination with increasing lysine yield indicates that the futile cycle between carboxylation of C3-compounds and decarboxylation of C4-compounds is decreased in relation to a higher demand for C4-compounds. The decarboxylation of C4-compounds can potentially contribute to the NADPH synthesis when the flux is catalyzed by the malic enzyme. In this case, it becomes obvious that especially for the overproduction of lysine, a compromise between the optimization of C4-compound formation and NADPH supply must be achieved since the intracellular provision of both intermediates might represent a bottleneck for lysine formation. For Aspergillus nidulans [47] and C. glutamicum [48] it has been proposed that the malic enzyme is of crucial importance for NADPH synthesis. In contrast, 13C-isotope investigations for the lysine model strain MH2022B revealed that, at least for a high lysine yield achieved with the homologous glutamate dehydrogenase mutant, the decarboxylation of C4-compounds was low compared to the overall NADPH synthesis by the pentose phosphate pathway and tricarboxylic acid cycle. This indicates only a minor contribution of the malic enzyme to NADPH synthesis in C. glutamicum for this particular physiological state of lysine overproduction. A minor contribution of the malic enzyme to intracellular NADPH supply has also been observed in relation to the overproduction of riboflavin by B. subtilis [44].
2.4 Availability of NADPH for Biosynthetic Reductive Power NADPH Metabolism in C. glutamicum

One use of 13C-isotope experiments based metabolic flux analysis which is described in detail in Sect. 3.3 is that, without making assumptions about bacterial energetics, the carbon flux estimations can be used to establish a NADPH balance. From the 13C-isotope analysis based flux distributions for different physiological states [41] the NADPH supply by glucose 6-phosphate dehydrogenase, 6-phosphogluconate dehydrogenase and isocitrate dehydrogenase was calcu-

Biotechnological Manufacture of Lysine

71

lated.A comparison of this supply to the requirement of the synthesis of biomass building blocks and products revealed an unassigned NADPH consumption which varied according to the physiological state under investigation (Fig. 9). Since the unassigned NADPH consumption (flux of 16 to 67% relative to the glucose uptake rate) was minimal for the observed maximum lysine yield NADPH limitation of lysine production is likely. Recently, the increase of nicotinamide nucleotide transhydrogenase activity in microbial cells has been identified as a means for removing the bottle-neck of limited NADPH supply and thereby facilitating the formation of amino acids and microbial products in general [49]. A similarly unusually high capacity for the reoxidation of NADPH has been revealed by 13C-isotope investigations for C. glutamicum in batch culture [50] and for riboflavin-producing Bacillus subtilis in chemostat culture [44]. This indicates that in both gram-positive bacteria, in-vivo NADPH consumption processes must be active which are still unassigned to a metabolic function and are not related to the synthesis of biomass building blocks and products. Since the NADPH requirement for the synthesis of biomass building blocks and products for C. glutamicum was partially derived from data for the gramnegative bacterium Escherichia coli [40] and it has been confirmed that grampositive bacteria such as B. subtilis contain more peptidoglycan [51] the requirement of NADPH might be higher than indicated in Fig. 9. A correction of the data would result in an increase in the NADPH requirement equal to a flux of only 4.6% relative to the glucose uptake rate which is less than the observed unassigned NADPH consumption of up to a flux of 67% relative to the glucose uptake rate. It is likely that the unassigned NADPH consumption is required for cellular processes which depend on the activity of NADPH oxidizing enzymes like glutathione reductase [52]. Processes such as the detoxification of reactive oxygen intermediates have been described for Saccharomyces cerevisiae [53] and E. coli [54]. Furthermore, for C. glutamicum, an NADPH oxidase activity has been described recently [55] and in particular an in-vitro activity of 16.5 mU (mg protein)1 has been observed upon incubation of crude extract with NADPH [56] which corresponds to a NADPH consumption flux of about 30% relative to the glucose uptake rate referring to the above mentioned metabolic flux data. NADPH is consumed by the lysine biosynthetic pathway in four steps either directly in the dihydrodipicolinate dehydrogenase and diaminopimelate dehydrogenase reaction or indirectly by the reactions aspartate aminotransferase and succinylaminoketopimelate aminotransferase. It turned out that the overexpression of a homologous glutamate dehydrogenase as described in Sect. 3.3 led to an improvement in lysine formation most likely because a better channeling of NADPH into the amination reactions of lysine synthesis.A scenario could be that the cytoplasmatic glutamate concentration was increased and thereby the cytoplasmatic substrate availability for lysine synthesis specific aspartate aminotransferase and succinylaminoketopimelate aminotransferase has been optimized. Recently an attempt to make glutamate formation dependent on NADH by introduction of a heterologous glutamate dehydrogenase from P. assaccharolyticus has been described [31]. Initially it was expected that lysine formation would be increased for the heterologous glutamate dehydrogenase mutant since

72

W. Pfefferle et al.

Fig. 9. Some NADPH fluxes in Corynebacterium glutamicum indicated with the unit mole of

NADPH per 100 mol of glucose taken up by the bacterial cells. An unexpected and still unassigned NADPH consumption flux (bar with lines) was calculated from the supply of NADPH by pentose phosphate pathway (black bar) plus tricarboxylic acid cycle (bar with rectangles) minus the requirement of NADPH which was derived from product synthesis (bar with diamonds) and C. glutamicum biomass formation (white bar) as described elsewhere [40].All values are expressed as percentage of the molar flux of glucose uptake. The particular physiological states for isogenic mutant strains were characterized in chemostat culture [31, 40, 41]: A, lysine production with 0.29 g lysine HCl (g glucose)1; B, lysine production with 0.18 g lysine HCl (g glucose)1; C, growth; D, glutamate production

Biotechnological Manufacture of Lysine

73

NADH was required for glutamate synthesis by the heterologous NADH-dependent glutamate dehydrogenase to circumvent a NADPH limitation of lysine production. However, complex adjustments of some cytoplasmatic pools in response to the introduction of the heterologous glutamate dehydrogenase [57] might have led to the observed decrease in pentose phosphate cycle flux. In this way, the overall NADPH supply for enzyme reactions, such as for diaminopimelate dehydrogenase, which are directly dependent on NADPH might be reduced and might have limited lysine formation. Further changes in cytoplasmatic pools of important intermediates might have influenced the regulatory network as a response to the introduction of a heterologous enzyme. The in-vivo monitoring of the cytoplasmatic concentration of intermediates as was published for lysine production with C. glutamicum [58] might contribute to an understanding of how changes in cytoplasmatic pool sizes influence metabolic flux patterns and lysine production. This kind of investigation, called metabolome analysis, has recently been introduced for E. coli [59], and might be a valuable tool for understanding and engineering the central metabolism of industrial microorganisms, especially for lysine producing C. glutamicum.
2.5 Enhancement of Secretion: Influence on Activity of Amino Acid Exporters by the Lipid Environment

Detailed knowledge of the mechanism by which l-lysine crosses the permeability barrier of the plasma membrane is also significant for considerations concerning process efficiency and yield. Studying the l-lysine export of C. glutamicum, Brer et al. [60, 61] showed that l-lysine is secreted by a carrier mediated secondary transport process. However, the molecular basis for lysine export was unknown until 1996. One important step towards the cloning of the gene was the isolation of a l-lysine excretion-defective mutant [62]. Cloning the gene encoding the lysine carrier protein LysE from C. glutamicum provided the breakthrough [63]. The comparatively small protein of 236 amino acids in size or 25.4 kDa performs the translocation of llysine. In a lysE disruption mutant, l-lysine excretion is blocked, which proves that no second function can substitute the LysE-mediated lysine export [64].With overexpression of lysE, the excretion rate for l-lysine is enhanced five times to a rate of 3.76 nmol min1 (mg dry weight)1 compared to the wild type strain. These findings indeed show that lysE is a potential target in strain development, avoiding the bottleneck of product excretion. The LysE protein represents a new family of transport proteins consisting of 5 transmembrane spanning helices.A large hydrophilic loop in the middle of the polypeptide chain is probably involved in the export of the positively charged llysine (Fig. 10). The translocation mechanism of l-lysine has been studied in detail and it has been discovered that the driving force for the entire translocation process is the membrane potential DY with the export activity being modulated by the pH and the l-lysine gradient between the cytosol and the external medium.

74

W. Pfefferle et al.

Fig. 10. Model of the topology of the lysE product encoding the l-lysine exporter of C. glu-

tamicum. Conserved aminoacyl residues proposed to be of major importance for translocation are given in the single-letter code

A model describing the l-lysine translocation differentiated four distinct steps: (i) The negatively charged LysE carrier protein is loaded with l-lysine and two hydroxyl ions. (ii) l-Lysine is translocated across the membrane. (iii) The substrate l-lysine and the associated hydroxyl ions are released on the outside of the membrane. (iv) In the last step, the carrier is reoriented to be ready to be loaded up again on the inside of the membrane. In this model the driving force DY is required for the reorientation of the carrier [61] (Fig. 11). However, is lysE the only key to modulate the export rate of lysine producers? It has been shown that different lysine-producing strains show differences in export capabilities. The idea was that mutations in the lysE gene should be responsible for the variations in lysine export. Cloning and sequencing of the genes of several C. glutamicum strains made it clear that the lysE gene itself is not the only source to modulate export capabilities of production strains [65]. But what is the story behind this evidence? It is well known that the lipid environment determines the fluidity of the cell wall and the activity of transport proteins. Detailed studies indicate that lipids play a supporting role in the stabilization of transport proteins in a native and active conformation [66]. Integral membrane proteins usually require specific lipids for their optimal activity and are inhibited by other lipid species. The activity of transport proteins

Biotechnological Manufacture of Lysine

75

Fig. 11. Putative kinetic mechanism of l-lysine efflux in C. glutamicum. Abbreviations: lys+,

lysine base; C+, carrier; Dy, transmembrane potential

is modulated by bulk physical characteristics of the lipid bilayer, while specific requirements rarely appear. The influence of the lipid content of the membrane on amino acid efflux in Corynebacteria has been intensively studied for the glutamate excretion (cf. chapter Metabolic Engineering of Glutamate Production) and interestingly there is also an effect on the activity of the lysine secretion system. Recently two genes (dtsR1 and dtsR2) were cloned whose expression influence the lipid transporters of the cell. Another protein encoded by the accBC gene is a two-domain protein composed of a N-terminal biotin carboxylase and a C-terminal biotin-carboxyl-carrier protein. The high homology to an acyl-coenzyme A carboxylase from Mycobacterium leprae led to the conclusion that the identified gene is a subunit of the AccBCDA complex [67]. Thus, it has been assumed that DtsR1 forms a complex with the biotin-containing protein AccBC, as does DtsR2 [30]. Evaluation of sequence identities lead to the conclusion that AccBC is the alpha-chain of acetyl-coenzyme A-carboxylase and DtsR1 and DtsR2 function as the b-subunit of the enzyme. There is experimental evidence that DtsR1/AccBC and DtsR2/AccBC complexes are formed in vivo (Fig. 12).

76

W. Pfefferle et al.

Fig. 12. Probable structure of accCBDA complex in C. glutamicum. In E. coli four sub-

units are encoded by separated genes. In C. glutamicum two subunits were detected. DtsR1, DtsR2 and AccDA form homologous proteins and may function as the carboxyltranferase- acetyl-/propionyl binding domain of the acetyl-CoA-carboxylase or propionyl-CoAcarboxylase

Besides the dtsR1 and dtsR2 genes, a third orthologous gene accDA has been cloned and identified [68]. Now there is experimental evidence that the genes encoding the acetyl-coenzyme A-carboxylase have been identified. Enzyme measurements made it clear that a complex of AccBC/AccDA shows acetyl-coenzyme A-carboxylase activity (Eggeling, personal communication). Furthermore, in addition to acetyl-coenzyme A, propionyl-coenzyme A can be used as a substrate of the enzyme complex, indicating that beside malonyl-coenzyme A, methylmanoyl-coenzyme A is formed (Eggeling, personal communication).Acyl-coenzyme A carboxylase is known to provide building blocks for fatty-acids and mycolic acids. Interestingly, amplifying the dtsR1 gene in an l-lysine producing strain leads to enhanced l-lysine production. The idea is that the changed lipid environment of the carrier protein is responsible for enhanced l-lysine efflux. Similar results were obtained by amplifying accDA and the concerted amplification of both accBC and accDA in the l-lysine producing strain MH2022B (DSM5715) [68, 69] (Table 3). These findings show that modulating the lipid environment of transport proteins can be used to enhance l-lysine secretion. Further studies will reveal the molecular mechanism of the interaction between LysE and the lipids of the cell membrane in C. glutamicum.

Biotechnological Manufacture of Lysine Table 3. l-Lysine secretion of recombinants of Coryne-

77

bacterium glutamicum MH2022B carrying different plasmids Strain DSM5175 DSM5715/pZ1accBC DSM5715/pZ1accDA OD660 31.4 27.6 43.1 l-Lysine (g L1) 7.2 9.6 8.0

The expression of the lysE gene encoding the lysine carrier protein is regulated by the lysG gene located immediately adjacent to the lysE gene on the opposite DNA strand [64]. LysG is homologue to a group of DNA regulators which belong to the LysR type transcriptional regulators family [70] and acts as a positive control element, activating the transcription of lysE and displaying autoregulatory activities. One characteristic of this type of activator is that they act in response to small molecules. The molecule controlling lysG activity is probably l-lysine. Beside lysE and the lipid environment of lysE, lysG is indeed a powerful target for modulating the excretion capacity of l-lysine producing strains.
2.6 Combination of Enhancement of Enzymes for the Lysine Biosynthetic Pathway

Several successful attempts have been made to enhance l-lysine productivity of strains by means of amplification of l-lysine biosynthesis genes or introduction of mutant genes in coryneform bacteria. It has been shown that aspartate kinase and dihydrodipicolinate synthase are the most important enzymes in regulating the carbon flux throughout the l-lysine biosynthetic pathway. The dapA gene amplification by genetic engineering methods has two effects on the cell, as described above, not only l-lysine production is enhanced, but also growth speed is decreased resulting in increased intracellular pyruvate availability favorable for l-lysine synthesis. The amplification of both the feedback released aspartate kinase together with the dihydrodipicolinate synthase is used to further enhance l-lysine formation [17] but leaving the problem of growth decrease still existing. The amplification of single genes is a well known tool and proved to be successful not only for the amplification of l-lysine specific genes. Several genes of the central metabolic pathways from C. glutamicum like phosphoenolpyruvate carboxylase or pyruvate carboxylase have been used to enhance l-lysine productivity of several strains [17, 38, 71]. The combination of several genes from the l-lysine biosynthetic pathway is also applied to achieve an improved metabolic flux in the pathway to l-lysine. Amplification of dapB alone does not have any effect on l-lysine formation [17], but in combination with dapA and lysC(fbr) overexpression the positive effect of dapB on lysine production became apparent [72].

78

W. Pfefferle et al.

Table 4. Enhancement of dapB restores growth and accelerates lysine formation in strains with amplified lysC(fbr) and dapA (according to [72]).Abbreviation: OD600, optical density at 600 nm

Bacterial strain/ amplified gene

Amount of produced l-Lysine (g L1) 24 h 72 h 29.8 34.5 35.0 36.5 45.0

Growth OD600

AJ11082 +lysC fbr +lysC fbr dapB +lysC fbr dapA +lysC fbr dapA dapB

22.0 16.8 23.3 19.7 23.0

0.450 0.398 0.440 0.360 0.425

In this study the authors showed that the concerted overexpression of 5 genes of the l-lysine biosynthetic pathway, lysC, dapA, dapB, ddh and lysA enhanced llysine accumulation from 29.8 g L1 to 47.0 g L1 after 72 h incubation with only slightly reduced growth from 0.450 to 0.409 (OD). They report that dapB is important for restoring growth rate. Enhancing lysC(fbr) in a moderate lysine producer lead to reduced growth and increased yield as expected; lysine formation after 24 h was slowed down by 24 h due to poor growth rate. Additional overexpression of dapB restored growth and overcompensated productivity losses (Table 4). Nevertheless, the effect of high dihydrodipicolinate reductase activity on the metabolic flux is not obvious and still unknown. There is no explanation why dapB amplification is able to overcome the growth reduction resulting from overexpression of aspartate kinase. Beside the already mentioned steps in l-lysine biosynthesis there are some more genes whose expression is capable to influence l-lysine productivity dramatically.A new aspect in enhancing l-lysine productivity is the attempt to modulate the expression of several genes from the central metabolism, the l-lysine biosynthetic pathway and furthermore the l-lysine export carrier gene lysE. Further work showed that overexpression of the recently cloned pyc gene encoding pyruvate carboxylase in addition to the enhanced activity of dapA and dapB in combination with lysC(fbr) increases l-lysine production. Overexpression of pyruvate carboxylase (pyc), dihydrodipicolinate synthase (dapA) and reductase (dapB) together with the gene encoding the l-lysine export carrier lysE results in substantial increased accumulation of l-lysine (Fig. 13). The parent strain MH2022B with amplified pyc gene, accumulates 11.3 g L1 l-lysine HCl after 72 h whereas the strain with amplified pyc, dapA, dapB and lysE gene accumulates up to 17.6 g L1 [73]. This strong increase in l-lysine formation is due to the combination of redirecting the carbon flow in the most important rate limiting steps of C4-compound supply, specific lysine biosynthesis and product secretion. Ohnishi et al. demonstrated a high-speed fermentation yielding 80 g L1 l-lysine HCl in 27 h combining a deregulated aspartate kinase with a mutated pyruvate carboxylase and homoserine dehydrogenase both bearing a single amino acid exchange [74]. These findings make clear that there are not only one or two

Biotechnological Manufacture of Lysine

79

Fig. 13. l-Lysine secretion of recombinants of Corynebacterium glutamicum MH2022B. Cells were cultivated on production medium containing 50 g glucose l1. l-Lysine HCl concentration was determined in supernatants after 72 h. Strain MH2022B is a leucine auxotrophic and AEC-resistant mutant derived from C. glutamicum ATCC 13032. Pyc (pyruvate carboxylase), dapA (dihydrodipicolinate synthase), dapB (dihydrodipicolinate reductase), lysE (lysine carrier), lysC(Fbr) (feedback resistant aspartate kinase)

rate limiting steps in l-lysine production of the cell; in consequence amplification of single genes which were supposed to be limiting will lead to new bottlenecks to be enlarged.
2.7 Enhancement of Stress Resistance

Technical manufacturing of lysine is performed in fermentation media with high osmotic pressure up to 2 osmol. Therefore an essential requirement for production strains is their osmostress resistance. As a soil bacterium Corynebacterium glutamicum is well equipped to face hyperosmotic environments. As early as 1990, Kawahara reported that supplementation of glycine betaine stimulated growth of Corynebacterium glutamicum in media of inhibitory osmotic strength and that this osmoprotectant had a positive influence on lysine fermentation [75, 76]. Transport studies showed that Corynebacterium glutamicum can use the compatible solutes glycine betaine, proline and ectoine for osmoprotection [77]. In the meantime, a complex osmoprotective network has become known in this organism (Fig. 14). For lysine production, the constitutively expressed but strongly regulated glycine betaine transport system BetP is the most important: on the one hand it

80

W. Pfefferle et al.

Fig. 14. Present knowledge of the osmoprotective network in Corynebacterium glutamicum.

Kindly provided by Dr. R. Krmer, University Cologne, Germany

shows a very high maximum activity and on the other hand glycine betaine is cheap and readily available in molasses which is a natural substrate or co-substrate for lysine fermentation. A more general aspect of stress tolerance is covered by the approach of Kimura et al. [78]. The idea is that enhanced expression of heat shock proteins or of the alternative sigma-32 factor which activates the synthesis of heat shock proteins (for a summary of the E. coli heat shock response see [79]) will elevate the stress tolerance of the production strains towards product concentration, temperature and osmotic stress. They report that heterologue amplification of the E. coli rpoH gene in Brevibacterium lactofermentum leads to enhanced lysine formation.

Biotechnological Manufacture of Lysine

81

3 Generation of Basic Knowledge for Understanding of Lysine Production with Corynebacterium glutamicum for Further Strain and Process Development
3.1 Corynebacterium glutamicum Genome Analysis Identification of All Genes Leading from Sugar Uptake to Lysine Secretion

Over the past 10 years, the roots of genetic engineering enabled the molecular biologists to modulate the pathways critical for l-lysine production. As described in Sect. 2.2 the aim is to tune the metabolic flux in C. glutamicum production strains towards a highly efficient conversion of the substrate glucose into the end product l-lysine. Methods such as transposon mutagenesis, DNA transfer by electroporation [80] conjugational transfer of plasmids [81, 82] or disruption mutagenesis and gene replacement [83, 84] have been successfully established for corynebacteria to make strain optimization by genetic engineering the method of choice. A variety of genes of C. glutamicum have been cloned by complementation of heterologous E. coli mutants or using the PCR technique [85]. Furthermore, the impact of most of the genes of the lysine biosynthetic pathway on amino acid production has been elucidated [17, 21]. However, despite the success of metabolic engineering, there are still some secrets remaining in the central metabolism of the coryneform bacteria which are not only of importance for the scientific world, but are of major interest for industrial biotechnology. There is only a little knowledge about the regulatory networks of C. glutamicum limiting sugar consumption and influencing growth performance. The most powerful tool for filling this gap in information is the analysis of the whole genome sequence. Nowadays, sequencing of the entire genome of an organism is possible within a limited time and with a minimum of effort. The entire genomes of 27 microorganisms have been sequenced and data are accessible in the public databases, the annotation of 17 genomes is in progress and more than 70 genome sequencing projects will be completed within the first year of the new millennium (http://www.tigr.org, http://www.ncbi.nlm.nih.gov/PMGifs/Genomes/bact.html, http://geta.life.uiuc.edu/~nikos/prokaryagenomes.html).With the development of high throughput sequencing capabilities, genome sequencing has become affordable making information available about all genes defining an organism such as C. glutamicum. In 1999, the amino acid producing companies BASF, Degussa, and KyowaHakko completed the sequencing of the Corynebacterium glutamicum genome [8689]. In principle, there are two different ways of starting a sequencing project, the whole genome shotgun sequencing and sequencing based on ordered cosmid/bac libraries [90].

82 Shotgun approach:

W. Pfefferle et al.

Here the whole genome is randomly sheared and DNA fragments of about 1.5 kb are subcloned into plasmids which are sequenced. These small inserts often contain only a part of a gene thus being compatible with the metabolism of the cell. Using this technology, a significantly higher amount of DNA has to be sequenced to provide sufficient redundancy (about 10 fold) so that computational assembly is also possible with repetitive sequences. In recent times, this disadvantage is compensated for by lower sequencing costs and more powerful bioinformatics. Ordered cosmid/BAC libraries: The other approach is based on cosmid or BAC libraries which are screened for minimal overlaps by hybridization techniques. In this way the necessary redundancy is minimized.
3.1.1 Sequencing the Genome of the Amino-Acid Producing C. glutamicum

For deciphering the C. glutamicum genome, Degussa AG applied the cosmid strategy using a physical map of the genome as a starting point [91]. It is known that many genes from C. glutamicum are effectively expressed in E. coli leading to toxic effects making the subcloning of many regions of the genome on high copy number vectors very difficult or nearly impossible. Similar experiences were made sequencing the Bacillus subtilis genome [92]. By sequencing the Mycobacterium tuberculosis genome, several repetitive DNA elements (e.g., insertion elements) were identified. In addition, about 10% of the coding capacity of the genome is represented by two different polymorphic repetitive sequences [93]. Because Mycobacterium tuberculosis is evolutionarily related to Corynebacterium glutamicum, similar facts were expected for this sequencing project. Therefore, the physical map provided an important framework throughout the whole sequencing project. The strategy was divided into 6 steps [94]: (i) The C. glutamicum genome was cut using three rare cutting restriction enzymes SwaI, PacI and PmeI and with the PFGE technique, a physical map of the genome was established. About 40 genes known from the public databases were localized on the map by hybridization techniques (Fig. 15). (ii) A cosmid library of about 2300 cosmids was established. (iii) The cosmid library was screened and minimal overlapping clones were selected from the library by hybridization techniques using the physical map as a guideline throughout the project. (iv) Sequencing of the 3,282 Kbp genome of C. glutamicum ATCC 13032 by a cosmid-by-cosmid sequencing strategy.A set of 98 cosmids covering about 90% of the entire genome were subcloned and sequenced. (v) Assembling of the data, identification of the remaining gaps. (vi) Closing of the remaining gaps in the 3.282 Kbp sequence was mainly performed by establishing a BAC library and subsequently sequencing the remaining gaps on BAC templates.

Biotechnological Manufacture of Lysine

83

Fig. 15. Overview of the physical and genetic map of the chromosome of Corynebacterium glu-

tamicum ATCC 13032. Localization of the restriction fragments derived from SwaI-cut DNA are shown as segments of the circle. Selected genes are arranged around the circle due to their genomic localization

Computer analysis of the genome of the lysine producer Corynebacterium glutamicum indicates that the present knowledge concerning the genes of the most important pathways has not been completed. Publicly available microbial genomes have shown that alternative reactions often appear to be required or certain steps are missing. The alignment of biochemical pathways from different species is an important step towards the understanding of bacterial physiology [9597]. It is of great value for pathway engineering to get detailed knowledge about the genes defining the central pathways of the metabolism. It has been shown that the adaptability of organisms is underestimated if only textbook knowledge regarding the

84

W. Pfefferle et al.

fundamental pathways is considered [98100]. These pathways rarely occur as previously defined [101102]. Since the data on genome sequences started to accumulate, a number of isoenzymes have been detected and identified [103] and it has been shown that some genes are lacking in various organisms. The comparison of the genes encoding the enzymes of the glycolytic pathway from 17 sequenced genomes revealed an interesting insight into the variability of glycolysis [98, 99, 104]. The genome sequence of C. glutamicum has been analyzed to identify all genes involved in glycolysis and compared to the data with the genome annotation of Mycobacterium tuberculosis one of the most closely related bacteria with its genome sequence already published [93]. The glycolytic pathway includes some reactions that, in different species, are catalyzed by non-orthologous enzymes, namely phosphofructokinase, fructosebisphosphate aldolase and phosphoglycerate mutases [98, 102] and glyceraldehyde-3-phosphate dehydrogenase (Fig. 16). As in Mycobacterium tuberculosis, one gene encoding phosphofructokinase A and one encoding phosphofructokinase B have both been found in C. glutamicum [105, 106]. In the Bacillus subtilis genome, only the 6-phosphofructokinase family B is present. The phosphofructokinase isoenzymes are regulated differently by allosteric activators. Family A enzymes (main 6-phosphofructokinase) are activated by any nucleotide diphosphate and inhibited by PEP. Whereas family B (which supplies only 10% of the activity of phosphofructokinase family A) is inhibited by ATP. 6-Phosphofructokinase activity has been measured in Brevibacterium flavum strains, indicating that there is a different mode of regulation of this enzyme activity. It has been found, that in B. flavum, the phosphofructokinase activity is inhibited by ADP only [107]. These findings have to be clarified, which can now easily be performed with the sequence of the two genes at hand. Similar to Mycobacterium tuberculosis, a second glyceraldehyde-3-phosphate dehydrogenase gene (gap) has also been detected in C. glutamicum, indicating that the metabolisms of C. glutamicum and M. tuberculosis are closely related. The finding that two open reading frames annotated as gpmA and gpmB encoding phosphoglycerate mutase are present in C. glutamicum was unexpected . Both homologous genes have been identified in the analyzed genome of the soil bacterium C. glutamcium [108]. An open reading frame with significant homology to the poxB gene from E. coli indicated the existence of this enzymatic activity in C. glutamicum. The poxB gene encodes the pyruvate oxidase which catalyzes the oxidation of pyruvate to acetate. It became apparent that the disruption of the poxB gene in a l-lysine producing strain C. glutamicum MH2022B results in an increase in l-lysine formation from 9.5 g L1 to 12.9 g L1 [109]. These findings show that genome analysis provides further insights into metabolic pathways. For the pentose phosphate cycle and tricarboxylic acid cycle, all the missing genes of the pathway have been identified. Also, the missing genes in the succinylase route of the lysine specific pathway, dapC and dapF are now available. Thus both branches of the lysine biosynthesis, the 4-step tetradihydrodipicolinate succinylase pathway and the one-step meso-diaminopimelate dehydrogenase pathway are accessible for thorough investigation. Shaw-Reid et al. [110] inacti-

Biotechnological Manufacture of Lysine

85

Fig. 16. Glycolytic enzymes found in organisms with completely sequenced genomes. In C. glu-

tamicum a second gap gene was found and three open reading frames with homology to the gpm genes. The enzymes are listed under E. coli gene names. Black arrows (thick lines) indicate practically irreversible reactions, thin lines indicate reversible reactions

86

W. Pfefferle et al.

vated the desuccinylase gene dapE and reported that in shake flasks there was no influence on lysine biosynthesis indicating that the succinylase pathway is dispensable for lysine overproduction. In addition, they state that under these conditions wild type meso-diaminopimelate dehydrogenase activity is not rate limiting for lysine biosynthesis. Since the dehydrogenase pathway uses NADPH as reductive power and the succinylase pathway glutamate (synthesized from a-ketoglutarate and NADPH), it will be interesting to see how high yielding lysine producers respond to enhancing either of them under fermentation conditions. To sum up the whole pathway leading from substrate uptake by the PTS system to the end product, l-lysine is now ready to be studied on the molecular level. Metabolic pathways of C. glutamicum have been studied over the years by geneticists and biologists. With the knowledge of the genome sequence, the presence of until now undetected genes becomes apparent. It must be considered that the presence of a gene sequence does not necessarily ensure that the corresponding enzyme is functional in the organism. To answer this question and to find out which role the enzyme might play in the metabolism, the successful approach using genetic tools, biochemical characterization, and metabolic flux analysis will be further applied. New software tools for the analysis of metabolic pathways will give more hints coming from the computational analysis of genome data [111]. Furthermore, it will be an interesting task to compare the data from the C. glutamicum genome with the Corynebacterium diphtheriae genome sequenced at the Sanger Center [112] to elucidate the gene structure in both organisms and to find differences in the metabolic pathways of these two closely related bacteria.
3.2 Expression Profiling as a Tool for Strain and Process Optimization

Advances in microarray technology enable massive parallel data mining from hybridization based expression profiling on the genomic scale. The availability of genome data makes possible the production of whole genome chips for transcriptome analysis of C. glutamicum. DNA arrays have already provided some important insights into gene function in a variety of microorganisms like Saccharomyces cerevisiae, Streptococcus pneumoniae, Haemophilus influenzae, Mycobacterium tuberculosis or Escherichia coli [113115]. A twofold change in expression is detectable and a sensitivity of one to five transcripts per cell is achievable with this technology [114]. A variety of different systems is available ranging from DNA arrays on nylon membranes to high density arrays on silica chips. Printing technologies are developing fast and different formats including spotted PCR fragments and oligonucleotide arrays are available [117]. There are obvious advantages of DNA chips, since chips allow miniaturization of the assay, parallelism of the analysis and automation of the assay procedure, making expression profiling an appropriate tool also in fermentation monitoring. Analyzing fermentation with expression profiling will lead to a better understanding of the physiological status lysine production. Fermentation phases of

Biotechnological Manufacture of Lysine

87

optimal growth can be compared to phases of optimal lysine production. Studying different conditions in general will reveal clusters of coordinated expressed genes; the pattern of clusters should be typical for a physiological state. This way genes and regulatory networks which are relevant to, e.g., exponential growth, maximal lysine overproduction, osmotic stress, toxification, or oxygen depletion, can be identified and investigated in detail. In consequence, they become targets for site specific manipulation to achieve further improvements. The combined use of different postgenomic techniques such as DNA chips, 2D gel-electrophoresis and old-fashioned physiology will provide the means to understand the cells regulators networks [118]. A second application field of expression profiling is the monitoring of routine production. Applying DNA chips in production will help to assess the culture quality of inocula and seed fermenters as well as the lysine formation capability in a prolonged fermentation mode. In addition, the fermentation may be run at the edge of the mass transfer capacity of the fermentation tank due to a precise monitoring of the physiological state of the cell. There are some basic requirements DNA chips have to match with regard to this technology as a tool in routine production monitoring, e.g., for the monitoring of process related target genes in fermentation [119]: Costs of such a diagnostic DNA-chip system have to be low A quantity of DNA chips of uniform quality must be available Improved and reliable methods for RNA isolation and labeling Easy handling of the array during hybridization (automation required) Short hybridization time Sensitive detection system Data handling and data evaluation tools for detection of the crucial steps in fermentative production and to detect changes of the expression profile which might influence the quality of the production process.

3.3 Pathway Modeling with Data Emerging from 13C-Isotope Experiments and Applications for Biochemical and Metabolic Engineering 3.3.1 Genotype and Metabolic Phenotype

Very recently, bioinformatics programs for the improvement of lysine production by C. glutamicum have been initiated ([111] cf. Sect. 3.1). Once the C. glutamicum genome has been completely sequenced and annotated an in-silico metabolic phenotype may be defined. With the use of this approach recently for Haemophilus influenzae Rd metabolic capabilities were assessed and different optimal phenotypes arising from the same metabolic genotype were derived, each with different constraint features [120]. Especially, high flexibility in the metabolic system to generate redox equivalents has been demonstrated computationally. This kind of metabolic flexibility has recently been revealed experimentally for C. glutamicum [31]. In parallel to phenotype characterization by

88

W. Pfefferle et al.

transcriptome and proteome analysis, the physiological state of bacterial cells is characterized by an array of metabolic fluxes through all reactions that occur in an organism which is called fluxome [121] as a particular feature of the whole metabolic phenotype. Transcriptome analysis will in future reveal to which extent genes are switched on and off for a particular physiological state. The actual use of the single reaction steps or whole metabolic pathways will be quantified by characterization of the fluxome by metabolic flux analysis. The combination of both techniques will have great impact on advanced process control. In the past, metabolic flux analysis research into lysine production by C. glutamicum was mainly focused on the supply of carbon precursors and NADPH for product formation. Without the availability of transcriptome information, this approach was severely dependent on assumptions of the metabolic network architecture [56]. Detailed reviews on the biochemistry of C. glutamicum central metabolism are available as a basis for the establishment of metabolic models [30, 56, 122124]. Stoichiometric balances were used to estimate theoretical maximum lysine yield on glucose of 0.45 to 0.75 g lysine HCL (g glucose)1 depending on the model assumptions as, e.g., for minimized biomass formation or restrictions by the metabolic network structure [56, 125]. Assuming a flux limitation due to succinyl-coenzyme A supply, the theoretical maximum yield was estimated to be 0.56 g lysine HCl (g glucose)1 [126]. Some uncertainty is introduced into the estimation of metabolic fluxes with the use of stoichiometric models, since bacterial energetics are still subject to reassessment [127] and general assumptions on the substrate demand to cover bacterial energetics may not be valid for particular physiological states. Nevertheless, with the help of published data on the efficiency of the electron transport chain for C. glutamicum [55, 128, 129] and with progress in the field of bioinformatics as has been described for the analysis of the electron transport chain in B. subtilis [130] it might become possible to establish reliable metabolic flux models for particular physiological states changing during the course of lysine fermentation. The question of which of the alternative cytochrome oxidases are switched on or off in a particular physiological state will be answered by transcriptome analysis. When metabolic flux analysis is combined with fast DNA micro-array-techniques, which have recently been applied to investigate E. coli physiology [131], changes in the structure of the metabolic network might be monitored during the course of fermentations. Models for process monitoring and control might be adjusted properly, especially the clustering of genes might be monitored and together with the identification of metabolic patterns by fluxome analysis might be used as an important input for physiological state estimation.
3.3.2 Short Review on Isotope Analysis

Recently, isotope analysis has been used to determine metabolic fluxes for C. glutamicum in general [132134] and for lysine production in particular [31, 36, 40, 41, 50, 135, 136]. Isotope investigations were used to reduce the number of assumptions on the metabolic network which must be made for the flux estimation by exclusive stoichiometric balancing. For the growth on fructose it was shown

Biotechnological Manufacture of Lysine

89

that the pentose phosphate cycle flux was lower than for growth on glucose [134]. As a compensation for the lack of NADPH supply by the pentose phosphate cycle, the NADPH generating flux through the malic enzyme seemed to be increased and the ratio of cytoplasmatic NADH versus NAD was increased. For lysine production and glutamate production the split ratio of the pentose phosphate cycle versus glycolysis was 70 versus 30 and 20 versus 80, respectively [135]. The split ratio of tricarboxylic acid cycle versus anaplerotic carboxylation which leads to C4-compound synthesis was shown to be 52 versus 48 for lysine production [137]. Recently, the above mentioned split ratios for lysine production were roughly confirmed by extensive 13C-isotope investigations for various fermentation runs [31, 40, 41] which indicates that at least for a subset of the metabolic network some split ratios can be predetermined by 13C-isotope analysis. In combination with transcriptome analysis, it should be possible in future to resolve singularity problems for the establishment of stoichiometric models for process control. Bidirectional metabolic fluxes are exclusively accessible by isotope analysis. The investigation of bidirectional fluxes is useful for the identification of targets for metabolic engineering [31, 41] and mandatory during the phase of the establishment of metabolic models. For routine metabolic flux analysis, bidirectional fluxes cannot be monitored continuously and the use of simplified models turned out to be valuable for the control of lysine fermentation [138]. Very recently, very comprehensive approaches which make use of the 13C-enrichment in biomass constituents like amino acids were introduced to access not only some but all major fluxes of central metabolism for lysine production with C. glutamicum [40, 139] and riboflavin production with B. subtilis [44]. This comprehensive treatment of all the metabolic branch-points, rather than a portion of the central metabolism, allows us to reveal metabolic patterns which might be used for the design of metabolic and biochemical engineering strategies. An extensive validation of a stoichiometric metabolic model by 13C-isotope analysis for the lysine producer strain MH2022B has been published [31, 40, 41, 50, 140]. In summary, the comparison of six metabolic patterns which were revealed for C. glutamicum by 13C-isotope analysis showed that in C. glutamicum some metabolic fluxes depend severely on the physiological state of the cells: (1) coordinated flux through pentose phosphate cycle and tricarboxylic acid cycle, (2) still unassigned NADPH consumption or in other words high capacity for the reoxidation of NADPH, and (3) futile cycling between C3-compounds of glycolysis and C4compounds of tricarboxylic acid cycle.
3.3.3 Homologous and Heterologous Glutamate Dehydrogenase Mutant

An extensive validation of a stoichiometric metabolic model by 13C-isotope analysis for the lysine producer strain MH2022B has been published [31, 40, 41, 50, 140]. The computer-aided overall analysis of the 13C-isotope data for two isogenic glutamate dehydrogenase mutants of the lysine producer strain C. glutamicum MH2022B [31] revealed that the pentose phosphate cycle flux was only high for a high demand of NADPH and a low tricarboxylic acid cycle flux. The

90

W. Pfefferle et al.

a
Fig. 17. In-vivo flux distribution in the central metabolism for C. glutamicum MH2022B (a) using its NADP-dependent glutamate dehydrogenase (homologous C. glutamicum mutant MH2022DBgdh pEK1.9gdh-1) and (b) using the NAD-dependent glutamate dehydrogenase of P. asaccharolyticus (heterologous mutant C. glutamicum MH2022BDgdh pEKEXpgdh). Numbers in oval symbols near thick lines give the estimated net fluxes; those near the thin arrows give the measured fluxes required for biomass synthesis. Numbers in hexagons give the estimated additional exchange fluxes in reversible reactions. All fluxes are given as molar

Biotechnological Manufacture of Lysine

91

b
metabolite flux expressed as a percentage of the glucose uptake rate, which was (a) 26.9 nmol (mg bio-dry mass)1 min1 and (b) 23.0 nmol (mg bio-dry mass)1 min1. Abbreviations: g6p, glucose 6-phosphate; f16 bp, fructose 1,6-bisphosphate; f6p, fructose 6-phosphate; gap, glyceraldehyde 3-phosphate; pep, phosphoenolpyruvate; pyr, pyruvate; oaa, oxaloacetate; fum, fumarate; icit, isocitrate, akg, a-ketoglutarate; p5p, pentose 5-phosphate; s7p, sedoheptulose 7-phosphate; e4p, erythrose 4-phosphate. (From [31] with permission of Academic Press)

92

W. Pfefferle et al.

observed metabolic patterns indicated that the pentose phosphate cycle flux was adjusted to the tricarboxylic acid cycle flux. In the heterologous glutamate dehydrogenase mutant NADH was required for glutamate synthesis by a heterologous NADH-dependent glutamate dehydrogenase form Peptostreptococcus asaccharolyticus. When a high tricarboxylic acid cycle flux occurred (Fig. 17) resulting from a high demand of NADH for glutamate synthesis much NADPH was supplied by the isocitrate dehydrogenase as a reaction step in the tricarboxylic acid cycle pathway. In response to this the NADPH formation by the pentose phosphate cycle was low. This indicates that the reoxidation of NADPH might limit the pentose phosphate cycle flux in C. glutamicum as has been described for liver tissue [141]. For an increased reoxidation of NADPH in the homologous glutamate dehydrogenase mutant during lysine production in which much NADPH was needed for glutamate synthesis by the NADPH-dependent glutamate dehydrogenase from C. glutamicum and in which a higher lysine yield was observed than in the heterologous mutant resulting in a high NADPH demand the pentose phosphate cycle flux was increased and at the same time the tricarboxylic acid cycle flux was decreased (Fig. 17).
3.3.4 Amphibolic Function of the Glyoxylate Cycle Anaplerosis and Link to C4-Compound Decarboxylation

An illustrative example for validation of metabolic models is that for an isocitrate dehydrogenase deletion mutant of C. glutamicum ATCC 13032. It was shown that a blocked tricarboxylic acid cycle was bypassed by high glyoxylate cycle activity and conversion of dicarboxylic acids into pyruvate or phosphoenolpyruvate as intermediates of glycolysis which afterwards entered the glyoxylate cycle sequence for a second time [42]. In this way, pyruvate was oxidized to carbon dioxide and redox equivalents were provided for energy generation.At the same time, in response to the blocked tricarboxylic acid cycle, the pentose phosphate cycle flux was high to provide sufficient amounts of NADPH. This is in agreement with the generally observed inverse coordination of tricarboxylic acid cycle and pentose phosphate cycle in C. glutamicum. In general, a problem for the establishment of a stoichiometric model for lysine fermentation arises when two alternative pathways are available which can operate in parallel as is the case for the tricarboxylic acid cycle and the glyoxylate cycle. In this case, singularity problems prevent mathematical modeling of the metabolic network.An analysis of the sensitivity between the 13C-enrichment in position oxaloacetate C-4 and the two alternative anaplerotic pathways of glyoxylate cycle and anaplerotic carboxylation for a particular data set [40] showed that a low 13C enrichment in oxaloacetate C-4 indicates a high glyoxylate cycle flux (Fig. 18). Such a low 13C enrichment in oxaloacetate C-4 was observed experimentally for the above mentioned isocitrate dehydrogenase mutant. The computer-aided overall analysis of the 13C-enrichment data for this mutant resulted in a metabolic flux distribution with a high glyoxylate cycle flux. The glyoxylate cycle fulfils an amphibolic function in this mutant because on the one hand it provides ox-

Biotechnological Manufacture of Lysine

93

Fig. 18. Influence of the net flux through the two alternative metabolic pathways anaplerotic carboxylation of C3-compounds and the glyoxylate cycle on the 13C-enrichment in position oxaloacetate C-4 (OAA C-4). The sensitivity was investigated for a particular data set [40]. Flux values are expressed as percentage of the molar flux of glucose uptake rate. (From [42])

aloacetate which contributes to the supply of biomass building blocks so that it can be called anaplerotic [142]. On the other hand, the glyoxylate cycle catalyzes the interconversion of isocitrate to oxaloacetate and malate providing two C4-compounds which are subsequently decarboxylated to acetyl-coenzyme A and can enter the glyoxylate cycle again. By this metabolic cycle, glucose can be oxidized totally to CO2 in eight steps which provide as much NADH and FADH2 as an intact tricarboxylic acid cycle would do [143]. All other 13C-isotope investigations so far undertaken for the strain MH2022B during cultivation on glucose have not revealed significant glyoxylate cycle flux at all which means that the glyoxylate cycle pathway can be excluded from a stoichiometric model for process control as long as no additional metabolic engineering is carried out. As described for the isocitrate dehydrogenase mutant where the glyoxylate cycle is active even during growth on glucose metabolic engineering resulted in an overall response of the metabolic network so that the stoichiometric model for the particular strain has to be reevaluated by isotope investigations.

94
3.3.5 Comparison of in-Vivo and in-Vitro Characterization of Enzyme Activities

W. Pfefferle et al.

Metabolic flux analysis was used for the determination of the in-vivo activity of enzyme reactions of the split lysine synthesis pathway (cf. Sect. 2.2).Although the in-vitro activity of the succinylase pathway enzymes and of the diaminopimelate dehydrogenase are in the range of up to 100 mU (mg protein)1 and 1 U (mg protein)1 respectively, at the end of a batch fermentation when ammonia concentration is very low, the in-vivo metabolic flux of about 40 mU (mg protein)1 is exclusively catalyzed by the succinylase pathway enzymes for which lower invitro activity values had been determined [140]. Here it becomes obvious that cytoplasmatic substrate availability determines flux split ratios since the Michaelis-Menten constant of diaminopimelate dehydrogenase for ammonia was determined to be high with about 36 mM [144]. In a chemostat culture with an ammonia concentration of 150 mM and 600 mM, the split ratio of succinylase to dehydrogenase flux was about 2.8 to 1 and 1.1 to 1 respectively [31, 40], which again stresses the importance of substrate availability as a key factor which influences metabolic flux split ratios during lysine production. As is valid for the diaminopimelate dehydrogenase, the isocitrate dehydrogenase, in-vitro and invivo activities of up to 1 U (mg protein)1 [145] and up to 50 mU (mg protein)1 [31, 40, 41] have been published for comparable physiological states, respectively, which again stresses that the in-vitro activity does not give any information on the actual in-vivo metabolic activity of the reaction step as a part of the whole metabolic network. Fluxome research has been applied to analyze the in-vivo split ratio between low- and high-affinity metabolic pathways which operate in parallel. Isotope investigations with the use of 13C and 15N lead to the elucidation of the metabolic fluxes through the split lysine biosynthetic pathway [140] and of intracellular glutamate formation [146], respectively. It was revealed that, with progressing fermentation time and decreasing extracellular ammonia concentration, the invivo dehydrogenase pathway activity decreased indicating that the high-affinity route succinylase pathway became predominant. Intracellular glutamate formation might be crucial for an optimal operation of the succinylase pathway of lysine biosynthesis. It was found that glutamate formation was catalyzed to one third by the high-affinity route glutamine synthetase. This reaction was especially active to provide glutamine N-5 atoms for the synthesis of biomass components. The other two thirds of intracellular glutamate synthesis was catalyzed by the glutamate dehydrogenase reaction. The significant in-vivo activity of high-affinity routes indicates that reduced extracellular ammonia concentrations could be sufficient for ammonia input. Furthermore, by 15N-isotope data analysis, a bidirectional metabolic flux was quantified in addition to the net synthesis of glutamate from ammonia and a-ketoglutarate by the glutamate dehydrogenase reaction which was more than half the net flux. Metabolic engineering could target a reduction of bidirectional fluxes in order to increase the net flux in the desired direction. The same applies to the formation of C4-compounds, as described in Sect. 2.3. Summarizing, these few examples of fluxome research by metabolic flux analy-

Biotechnological Manufacture of Lysine

95

sis in C. glutamicum during growth and amino acid overproduction illustrate how important it is to establish metabolic models for particular strains and physiological states stressing the differences between in-vitro and in-vivo enzyme activities.
3.3.6 Application of Fluxomics Exploitation of Metabolic Flux Information

One use of stoichiometric metabolic flux models is that, knowing all requirements for the synthesis of biomass and product, it should be possible to design proper substrate feed profiles with which neither overflow metabolism nor precursor limitation of biomass and lysine synthesis occur. Especially the demand for NADPH, oxaloacetate, pyruvate, and glutamate should not become limiting in lysine production process. For the strain MH2022B it was revealed that, with increasing lysine formation, the net synthesis of oxalocatetate increased because the reaction rate of the oxaloacetate consuming decarboxylations decreased (cf. Sect. 2.3; [31, 41]). In parallel, the total NADPH supply increased with enhanced lysine formation mainly by an increase in pentose phosphate cycle flux (cf. Sect. 2.4; [31, 41]). It should be stressed that conclusions about metabolic patterns can only be extrapolated partially and that especially during scale-up it must be considered that cell physiology in general and especially metabolic fluxes change rapidly when the micro-environment of bacterial cells varies within the different spatial zones of a large bioreactor [147, 148].

4 Optimization of the Lysine Fermentation Process


4.1 Improvement of Process Economics by Process Intensification

Lysine is mainly produced in fed-batch mode and from data on the specific rates for sugar consumption and lysine formation which were published in a study on the use of osmoprotectants for optimization of lysine production experimental values on the apparent yield of up to 0.62 g lysine HCl (g glucose)1 were determined for resting cells [75, 76]. This result indicates that quite high experimental product yields can be achieved which are even higher than theoretically determined values (cf. Sect. 3.3) based on particular assumptions which might not apply to the above mentioned physiological state. A target for further improvement of process economics is to repeat batch or fed-batch fermentations without the need of repeated preparation of the bioreactor and new inocula. In order to improve the overall productivity of lysine fermentation, a portion of about 60 to 95% of the final culture volume might be removed and replaced by fresh medium. Using this mode of repeated batch or fed-batch fermentation successive fermentation runs can be carried out without introducing fresh inoculum and down time for the preparation of the sterile bioreactor. Up to 2000 only repeated batch for the production of lysine has been described

96

W. Pfefferle et al.

[149151]. Only few examples for repeated fed-batch as for phenylalanine production have been reported [152] where strain instability was revealed as a major problem. Strain instability becomes a major problem when lysine production strains are used which are auxotrophic for amino acids [153]. With a stable model strain for which acetylcysteine-resistance and feed-back resistance of aspartate kinase has been determined, in two successive fed-batch runs a lysine HCl concentration of 53.4 and 55.6 g L1, respectively was measured [154]. The lysine HCl yield on sucrose was 0.209 and 0.228 g g1 and the volumetric productivity was 1.71 and 2.01 g L1 h1, respectively (Fig. 19).With about 30 h of pure fermentation time per single fed-batch run, the preparation time of 10 h could be saved for the second fed-batch run. Thus with one repetition, a 15% increase in overall productivity was achieved for this repeated fed-batch fermentation where the performances of the two successive fed-batch runs were comparable. The next step for the improvement of process economics is the establishment of continuous culture fermentation but up to now has not yet been published for large-scale lysine production. Investigations on glutamate production revealed that the volumetric productivity with 5 to 8 g L1 h1 for continuous cultivations was at least twice the value for comparable fed-batch fermentations [155]. For a two-step continuous culture system for lysine production where for up to 400 hours in the growth bioreactor and production bioreactor pH is controlled at 7.2 and 6.2, respectively, the productivity through the whole process has been

Fig. 19. Repeated fed-batch process for lysine production with a stable model strain for which acetylcysteine-resistance and feed-back resistance of aspartate kinase was determined. Time course of Optical density at 535 nm (1) and concentration (g L1) of ammonia (2), dextrose (3), and lysine HCl (4)

Biotechnological Manufacture of Lysine

97

shown to be about twice that of a batch type cultivation [156]. In small-scale continuous culture fermentations for the production of lysine, maximum apparent yields for lysineHCl on glucose in basic research experiments with model strains were 0.25 g g1 [157], 0.29 g g1 [31], and 0.30 g g1 [158] which indicates a maximum intrinsic product yield of 0.41 g g1.A maximum of 1.0 g lysine HCl L1 h1 was achieved for a strain requiring homoserine and leucine and having a resistance to an isoleucine analogue [157]. Continuous culture experiments by Hirao et al. with a strain derived from ATCC13032 by selecting mutants resistant to S-(2-aminoethyl)-l-cysteine, rifampicin, streptomycin, and 6-azauracil resulted after optimization of the culture redox potential kinetics in a maximum apparent yield for lysine HCl on glucose of about 0.40 g g1 and a maximum of 5.6 g lysine HCl L1 h1 [159]. This clearly indicates that there is potential for the development of more economic processes once limitations such as insufficient oxygen supply have been overcome. In a more detailed analysis of the culture redox potential, a strategy for the addition of dithiothreitole has been used to optimize lysine production in batch fermentations [160, 161]. By the introduction of simultaneous carbon and phosphate limitation, continuous fermentation for lysine production has been optimized [162]. Compared to simple carbon limitation, with the double limitation, the volumetric productivity increased from 3.18 to 3.75 g lysine HCl L1 h1. The biomass specific product formation rate increased from 0.084 to 0.240 g lysine HCl (g bio-dry mass)1 h1 so that for the production of the same amount of lysine, 65% less biomass is produced in the double limitation process. Furthermore, it was observed that with the use of the double limitation culture, instability as typically observed for the continuous cultivation of amino acid production strains could be avoided for up to 30 generations. In conclusion, process intensification for lysine production has been achieved in the past by introduction of fed-batch and continuous culture fermentation as well as by optimization of oxygen availability. Modern techniques as fluxomics, transcriptomics, and single cell analysis by cytometry might help to extend the cultivation time of this culture modes e.g. by detection of the inoculum viability and metabolic activity before successive repeats are started.
4.2 Scale up and Scale down Techniques

Lysine fermentations are performed in very large tanks with a working volume of several hundred cubic meters. Fermentations performed in such large tanks may be severely limited by mass transport phenomena such as mixing and oxygen transfer. Einsele [163] reports that mixing time in large fermenters can easily exceed 100 s. Due to different kinetics of mass transfer, mixing and consumption gradients of dissolved oxygen, pH and substrates are formed to which the microorganisms are exposed during their circulation in large vessels. Even in comparatively small vessels of 20 m3, strong oxygen profiles have been reported [164]. These gradients affect the overall performance negatively so that scaling up becomes a very important factor in bioengineering.

98

W. Pfefferle et al.

Adjusting a bioprocess to a large tank at the production site is very costly. Therefore, there is a series of reports which deal with the investigation of insufficient mixing in lab fermenters simulating fluctuating conditions that occur on the production scale. Especially the effects of changes in oxygen and substrate availability were investigated in detail in a one fermenter system [165169]. Another quite common experimental design is to connect 2 lab tanks in series in which defined but different conditions are set [170172]. The aim of scale down investigations is to mimic the large-scale microenvironment of the bacterial cell in a small-scale bioreactor. A special reactor design for investigation mixing time effects in one reactor is reported by Schilling et al. [147]. This scale down lab tank is equipped with cylindrical disks and Rushton turbines leading to reduced axial mixing (Fig. 20). Due to the reactor design, the apparent mixing time q90 was prolonged to130 s compared to 10 s of the reference reactor. Comparison studies with a leucine-auxotrophic lysine strain were performed in fed-batch fermentations with feeding of l-leucine. Table 5 shows that the lysine fermentation in the scale-down reactor was significantly slowed down indicating that varying availability of the process limiting substrate directly affects the fermentation performance. Since the authors report that these findings are reflected in a lower specific activity of enzymes such as citrate synthase, aspartate kinase and phosphoenol pyruvate carboxylase, the cellular sensing and responding to substrate gradients is of general interest.

Fig. 20. Scale down lab tank. A prolonged mixing time (q90 =130 s) is achieved by cylindrical

disks which form compartments and thus reduce axial mixing. Potential pH and pO2 gradients are monitored using probes on two levels

Biotechnological Manufacture of Lysine

99

Table 5. Experimental results after 53 h cultivation of parallel exponential fed-batch cultiva-

tions with the leucine auxotrophic Corynebacterium glutamicum DSM5715 in a standard and a scale down reactor with prolonged mixing time Reference Reactor q90 =10 s Sugar Consumption Lysine Production Total CO2-evolution Dry cell weight 4.48 kg 0.82 kg 3.22 kg 0.33 kg Model Reactor q90 =130 s 3.85 kg 0.72 kg 2.07 kg 0.31 kg

A scale down system which allows us to investigate these assumed rapid responses of the microorganism to changing environments is a reactor design where a lab fermenter is connected to a plugged flow reactor bypass without sparging leading to oxygen depletion [148]. Sampling along the plugged flow reactor with subsequent expression profiling and proteome analysis can follow the quick reactions which take place inside the microorganisms. Besides working on the elucidation of the influence of gradients on microbial physiology improved stirring systems with strong axial mixing have been developed. This way spatial gradients are considerably reduced improving the fermentation performance in large tanks.
4.3 Process Control 4.3.1 Physiological State Estimation and Optimization of Nutrient Supply

For the metabolic state of overproduction Neijssel and Tempest [173, 174; for summary 175] developed a concept of overflow metabolism. They report that carbon sufficient chemostat cultures often metabolize the substrate inefficiently leading to excretion of energy rich compounds like acetate, pyruvate, 2-oxoglutarate or gluconic acid. Characteristics of the investigated conditions are quite similar to lysine production conditions: Carbon source in excess Growth limitation of the culture No strict regulation of substrate uptake Although lysine overproduction itself can be understood as a metabolic state of favorable overflow metabolism, the above mentioned characteristics suggest that lysine fermentations may also suffer from concomitant by-product formation. This was confirmed by Hadj Sassi et al. [176] and Hua et al. [177], who found that under fully aerobic conditions in cultures with growth limitations and carbon excess, significant formation of organic acids like acetate and pyruvate occurs.

100

W. Pfefferle et al.

The concept of monitoring and control of the physiological state of cell cultures has been reviewed recently [178]. For phenylalanine production with E. coli especially exact monitoring of specific growth rate enabled one to adjust the glucose feed properly and changes in metabolic rate ratios as oxygen to glucose uptake and in feed ratios as ammonia to glucose feed revealed the onset of growth decline and by-product formation, respectively [179]. Figure 21 shows that the typical E. coli by-product acetate is secreted when substrate consumption rate exceeds a so-called critical glucose consumption rate. Below this rate, acetate already present in the medium is readily co-metabolized.

b
Fig. 21. Acetic acid formation in phenylalanine fed-batch fermentation with E. coli AT2471 de-

pendent on specific glucose uptake rate (GUR): a discrimination of the time-field of fermentation based on acetic acid excretion, b confirmation of the time-fields in a fed-batch fermentation GUR (1); acetic acid (2). After [179]

Biotechnological Manufacture of Lysine

101

In general, action was taken to counteract the onset of overflow metabolism by glucose feed reduction. For C. glutamicum chemostat fermentations were used to determine the critical growth rate for the onset of by-product formation [180]. The strategy of adjusting the sugar feeding in such a way that maximum sugar uptake is established during the whole microbial cultivation has recently been applied to fermentations with yeast [181]. For technical lysine fed-batch fermentations with Corynebacterium glutamicum, Miwa et al. [182] describe a method to control the substrate feed efficiently in order to avoid overflow metabolism. To obtain low concentrations of the carbon source in the fermentation broth, the increase in the pH or in the dissolved oxygen content is controlled by intermittently adding the feed solution at a computer-calculated flow rate. In this way, the limitation signals trigger the next substrate dosage (Fig. 22). A similar method is described by Pfefferle et al. [183] who ensure that during carbon feeding, the culture of a technical lysine fermentation is sugar limited. With this method, they are able to achieve high coefficients for lysine (total dry matter)1 indicating elevated lysine yields and low by-product formation. An extension of physiological state estimation is the online metabolic pathway analysis which has recently been applied to the optimum control of lysine production with C. glutamicum [138]. The online metabolic flux analysis for fermentation monitoring and control has recently been reviewed [184]. With advances in fuzzy and neuronal network supervisory control systems for lysine production, trends from experimental data might be extracted, production rules might be defined, and proper feed profiles might be chosen as has been described for glutamate production with C. glutamicum [185]. For leucine auxotrophic lysine production strains, a central feature of the improvement of lysine production is the optimization of the initial concentration of leucine [186, 187]. The optimization of the subsequent feed-profile for leucine addition resulted in an ideal course of growth rate and led to improved process performance [187]. Fluctuations in the composition of complex media components, e.g., corn steep liquor, which are used for the large-scale production of amino acids, might prevent a stable and efficient industrial process performance because of the input of growth limiting components as leucine fluctuates. Recently, this problem was observed for threonine production with E. coli where the iron input from corn steep liquor varied with different corn steep liquor lots [188]. The growth level was made independent of this iron input fluctuation by addition of iron to the medium.At the same time, the growth of the methionine-auxotrophic threonine production strain had to be restricted in order to ensure maximum threonine production. A reproducible restriction of growth was achieved by reduction of the methionine concentration in the medium. As a consequence, an efficient industrial process with high productivity and stability was established which was no longer dependent on iron limitation or fluctuations of corn steep liquor quality. This kind of process adaptation in order to compensate fluctuations in raw material quality might be useful for the stabilization of lysine production processes as well. Optimization of medium composition using a genetic algorithm and control of glucose concentration at about 9 g L1 by online high performance liquid chro-

102

W. Pfefferle et al.

Substrate depletion pO2

Time

pH

Time Feed pulse calculated on demand

1 st addition

2 nd addition

3 rd addition

Substrate concentration in the fermenter

Time
Fig. 22. In fed-batch fermentation the dosage of the feed solution is controlled by substrate de-

pletion signals (according to [182]). The increase of pH and pO2 over a threshold value trigger another feed pulse, the resulting sugar concentration in the fermentation broth remains below 2 g L1

Biotechnological Manufacture of Lysine

103

matography resulted in a volumetric productivity of up to 8.3 mmol L1 h1 and a yield of lysine on glucose of 0.32 mol mol1 [189]. Further improvement of the product yield to 0.38 mol mol1 as a consequence of increased growth linked lysine formation was achieved by introduction of glucose limitation in the preculture which might lead to the activation of glucose-starvation-stimulons as was observed for E. coli [190]. Here it becomes obvious that also the control of precultures might be a target for the improvement of lysine overproduction. Especially, when the analysis of transcriptome and fluxome is established for process monitoring.
4.3.2 Single Cell Analysis by Flow Cytometry

Recently, flow cytometry has been used to quantify the response of C. glutamicum to antibiotics [191] and to measure intracellular pH as well as transmembrane potential [192]. This kind of monitoring of single cell properties might, in future, improve process monitoring and control especially because changes in the composition of a bacterial population might even be detected at an early stage of fermentation so that early decisions to counteract them can be made. Since single cells are analyzed it might be possible to observe physiological state changes much earlier than for conventional measurements. The reason is that the uptake of oxygen or sugar can only be measured on conventional devices once a sufficient cell density has been achieved to be in the measurement range of the conventional measurement devices. Applications of flow cytometry to E. coli fermentations for the investigation of sheer stress phenomena have been described recently [193]. First results for lysine producing C. glutamicum strain MH2022B are available and these clearly indicate distinct subpopulations of single cells and cell aggregates of different sizes (Fig. 23). On the single-cell level, intact cells, depolarized cells, and permeabilized cells were differentiated by staining with propidium iodide and bis-oxonole as described in detail for E. coli [193]. For realtime process monitoring, it might be interesting to measure the size of the subpopulation of aggregates since these are formed when adverse effects appear in lysine fermentation. In conclusion, advances in the monitoring of the physiological state of lysine producing C. glutamicum cultures by online metabolic flux analysis and optimization of nutrient supply in precultures and main cultures contributed to the improvement of lysine production. Extended physiological state estimation by fluxomics, transcriptomics, and by cytometry analysis might help to achieve further process intensification.

5 Large Scale Lysine Manufacture


Corynebacterium glutamicum is able to utilize as substrate sucrose and dextrose. Therefore, the major C-sources for lysine fermentation are pure sucrose, beet molasses, cane molasses, high test molasses (inverted cane molasses), and dextrose prepared from starch hydrolysate. In addition, technologies for the utilization of acetic acid and ethanol as co-substrates are available. The utilization

104

W. Pfefferle et al.

c
Fig. 23. Single cell analysis of C. glutamicum strain MH2022B. In graph a and b on the x-axis sideward scatter and on the y-axis forward scatter is plotted. In graph a, the total population is split into a comet tail (A1) and a bimodal subpopulation which is typical for rod-shaped bacteria where cells are detected which crossed the laser beam with a maximum (A2) or minimum (A3) cross-section. In graph b, the comet tail subpopulation is reduced which indicates that a smaller part of the population is made up by aggregates. In graph c, fluorescence signals from propidium iodide (PI; x-axis) and bis-oxonole (BOX; y-axis) stained cells are indicated (cf. text). C. glutamicum cells were stained with 0.1 mg BOX mL1 and 0.05 mg PI mL1. The analysis was restricted to cells of the bimodal subpopulation as indicated by the box in graph b. Notice that most of the C. glutamicum cells are depolarized (C2) and only a few cells are classified as live (C1) or permeabilized (C3)

Biotechnological Manufacture of Lysine

105

of molasses is gradually declining because its waste burden leads to increased costs for waste disposal in the lysine plant. Because of the relatively low market price for lysine of around 1.32 $ (kg lysine HCl)1 there is great pressure to reduce costs. The most important part of the manufacturing costs of lysine production is the C-source; therefore it is not astonishing that sugar suppliers and lysine producers are subject to close alliances: either the plants are situated close together so that the sugar is provided across the fence so to speak, or they form joint ventures, or C-source suppliers install their own lysine technology. Beside the price, the economical utilization of the C-source is very important. In technical processes conversion yields of 0.450.50 g lysine HCl (g sugar)1 have been reported (e.g., FR2645172 [194]). A typical fermentation graph of a lysine fed-batch fermentation is shown in Fig. 24. Looking at lysine manufacturers world-wide, there is a clear tendency toward bigger plants with very high capacities and thus using their economy of scale. Nowadays, lysine plants use large scale fermenters with up to 500-m3 volume or even more. Finally, work up has a major impact on manufacturing costs in lysine production. Lysine fermentation usually yields lysine sulfate at the end of fermentation. The classical way of work up follows a series of steps leading to a crystalline product of lysine HCl. For lysine producers, reduced investment and running costs of downstreaming are decisive, low costs for waste disposal are also of great importance. Since lysine is predominantly used as a feed additive, various more economical downstream processes have been developed yielding new lysine preparations which are well established on the market (Table 6) indicating that the advantage

200 Lys*HCl [g/l]; Optical Density

25

160

20

80

10

40

0 0 10
Lysine

0 20
OD

30
Sugar

40

50

Fermentation time [h]

Fig. 24. Technical lysine fed-batch fermentation; sugar feeding is lower than the maximal sugar consumption rate resulting in a carbon limitation during the feed period

Sugar [g/l]

120 [relU]

15

106

W. Pfefferle et al.

Table 6. Commercially available lysine preparations for feed use based on different downstream technologies

Product preparation

Lysine-base Downstreaming content [%] steps Biomass separation, Ion exchange, Drying

Wastes

Advantages

Lysine HCl [195] 78.8

Liquid lysine [196] Liquid lysine sulfate [197]

Approx. 50 2030

Biomass separation, Evaporation Evaporation

Biomass, No Organic acids, crystallization Salts, Ammonia, Water Biomass, Salts No drying step Almost none No drying, only a concentration step No wastes, No separation step

Granulated lysine 4050 sulfate [198]

Evaporation, Spray granulation

Almost none

of the new product preparations exists on the suppliers side but also on the customers side.

6 Outlook
Recently in the field of microbiology, an enormous amount of data based on genome sequencing became available for strain development and bioengineering. Therefore, it is not surprising that the genome of the sole economically important lysine producer Corynebacterium glutamicum has also been sequenced. The availability of these data gives access to post genome technologies such as expression profiling with DNA chips and investigation of protein patterns with 2Dgelelectrophoresis. Metabolic flux analysis in combination with metabolome analysis will help to characterize the actual in-vivo conversion rates through the active metabolic network. In this way, process control will be adjusted in a correct way and new strategies for metabolic engineering will be developed. These multi-parallel working technologies will accelerate the generation of knowledge and elucidate new global regulatory networks. For the first time, there is a chance of us being able to understand the overall picture of the physiological state of lysine overproduction in a technical environment. Additionally, technologies like flow cytometry will give deep insights into the heterogeneity of microbial population under lysine producing conditions based on information at the single-cell level. Especially the possibility for online controlling of technical fermentations will be very attractive for further process engineering. Combination of the knowledge emerging from these new technologies will surely lead to further improvements in lysine production. Therefore, despite the fact that efforts have already been made for four decades for optimizing lysine technology, additional major improvements can be expected.

Biotechnological Manufacture of Lysine

107

Acknowledgement. The authors thank Dr. C. J. Hewitt for experimental advice and discussions on C. glutamicum flow cytometry analysis and Dr. L. Eggeling for the contribution of unpublished results. Mrs. C. Aug is acknowledged for excellent technical assistance. The authors thank Dr. M. Farwick and Dr. R. Krmer for their helpful comments. Dr. R. Krmer contributed Fig. 14.

7 References
1. Leuchtenberger W (1996) Amino acids technical production and use. In: Rehm HJ, Reed G (eds) Biotechnology, VCH, Weinheim Vol 6, p 465 2. Kinoshita S, Udaka S, ShimonoM (1957) J Gen Appl Microbiol 3:193 3. Kinoshita S, Nakayama K Kitada S (1958) J Gen Appl Microbiol 4:128 4. Udaka S (1960) J Bacteriol 79:754 5. Liebl W, Ehrmann M, Ludwig W, Schleifer KH (1991) Int J Syst Bacteriol 41:255 6. Goodfellow M, Collins MD, Minniken DE (1976) J Gen Microbiol 96:351 7. Kitada S; Nakayama K; Kinoshita S (1961) US Patent 2979439 8. Kyowa-Hakko (1970) GB Patent 1186988 9. Shio I, Ozaki H, Ujigawa-Takeda K (1982) Agric Biol Chem 46:101 10. Shio I and Sano KJ (1969) Gen Appl Microbiol 15:267 11. Hilliger M and Hertel T (1997) J Basic Microbiol 37:29 12. Sugimoto M, Ogawa Y, Suzuki T, Tanaka A, Matsui H (1997) US Patent 5688671 13. Sano K, Shiio I (1970) J Gen Appl Microbiol 16:373 14. Thierbach G, Kalinowski J, Bachmann B, Phler A (1990) Appl Microbiol Biotechnol 32:443 15. Kalinowski J, Cremer J, Bachmann B, Eggeling L, Sahm H, Phler A (1991) Mol Microbiol 5:1197 16. Follettie MT, Peoples OP, Agoropoulou C, Sinskey AJ (1993) J Bacteriol 175:4096 17. Cremer J, Eggeling L, Sahm H (1991) Appl Environ Microbiol 57:1746 18. Jetten MSM, Follettie MT, Sinskey AJ (1995) Appl Microbiol Biotechnol 43:76 19. Sugimoto M, Tanaka A, Suzuki T, Matsui H, Nakamori H and Takagi H (1997) Biosci Biotechnol Biochem 61:1760 20. Branden C, Tooze J (1991) Introduction to protein structure. Garland Publishing Inc, New York London 21. Jetten MSM, Sinskey AJ (1995) Crit Rev Biotechnol 15:73 22. Tosaka O, Takinami K (1978) Agric Biol Chem 42:95 23. Katsumata, R Mizukami T, Oka T (1986) EP Patent 197335 24. Bonassie S, Oreglia J, Sucard AM (1990) NAR 18:6421 25. Pisabarro A, Malumbres M, Mateos LM, Oguiza JA, Martin JF (1993) J Bacteriol 175: 2743 26. Eggeling L, Oberle S, Sahm H (1998) Appl Microbiol Biotechnol 49:24 27. Patek M, Eikmanns B, Patek J, Sahm H (1996) Microbiol 142:1297 28. Vasicova P, Patek M, Nesvera J, Sahm H, Eikmanns B (1999) J Bacteriol 181:6188 29. Mckel B, Pfefferle W, Kreutzer C, Hans S, Rieping M, Eggeling L, Sahm H (1999) EU Patent Application EP1067192 30. Eggeling L, Sahm H (1999) Appl Microbiol Biotechnol 52:146 31. Marx A, Eikmanns BJ, Sahm H, de Graaf AA, Eggeling L (1999) Metabol Eng 1:35 32. Ozaki H, Shiio I (1969) J Biochem 66:297 33. Sano K, Ito K, Miwa K, Nakamori S (1985) EU Patent Application EP143195 34. Sano K, Ito K, Miwa K, Nakamori S (1987) Agric Biol Chem 51, 597 35. Peters-Wendisch PG, Eikmanns B, Thierbach G, Bachmann B, Sahm H (1993) FEMS Microbiol Lett 112:269 36. Park SM, Shaw-Reid C, Sinskey AJ, Stephanopoulos G (1997) Appl Microbiol Biotechnol 47:430

108

W. Pfefferle et al.

37. Peters-Wendisch PG, Wendisch VF, de Graaf AA, Eikmanns BJ, Sahm H (1996) Arch Microbiol 165:387 38. Peters-Wendisch PG, Kreutzer C, Kalinowski J, Patek M, Sahm H, Eikmanns BJ (1998) Microbiology 134:1095 39. Peters-Wendisch PG, Eikmanns B, Sahm H (1999) DE Patent Application 19831609 40. Marx A, de Graaf AA, Wiechert W, Eggeling L, Sahm H (1996) Biotechnol Bioeng 49:111 41. Marx A, Striegel K, de Graaf AA, Sahm H, Eggeling L (1997) Biotechnol Bioeng 56:168 42. Marx A, de Graaf AA, Wiechert W, Eggeling L, Sahm H (1998) Preprints of the 7th International Conference on Computer Applications in Biotechnology Osaka, Japan, p 387 43. Katz J, Wals P, Lee W-NP (1993) J Biol Chem 268:25509 44. Sauer U, Hatzimanikatis V, Bailey JE, Hochuli M, Szyperski T, Wthrich K (1997) Nature Biotechnol 15:448 45. Riedel C, Eikmanns B, Mckel B, Sahm H (1999) EU Patent Application EP1094111 46. Petersen S, de Graaf AA, Eggeling L, Mllney M, Wiechert W, Sahm H (2000) J Biol Chem 275:35932 47. Wynn JP, Ratledge C (1997) Microbiol 143:253 48. Cocaign-Bousquet M, Lindley ND (1995) Enzyme Microbiol Technol 17:260 49. Kojima H, Totsuka K (1995) Patent Application WO9511985 50. Sonntag K, Schwinde J, de Graaf AA, Marx A, Eikmanns BJ, Wiechert W, Sahm H (1995) Appl Microbiol Biotechnol 44:489 51. Sauer U, Hatzimanikatis V, Hohmann H-P, Manneberg M, van Loon APGM, Bailey JE (1996) Appl Environ Microbiol 62:3687 52. Meister A (1995) Glutathione metabolism In: Packer L (ed) Methods in enzymology 251: Monothiols and dithiols, protein thiols, and thiyl radicals. Academic, Orlando, p 3 53. Muller EGD (1996) Mol Biol Cell 7:1805 54. Holland, D, Faltin Z, Perl A, Ben-Hayyim G, Eshdat Y (1994) FEBS Lett 337:52 55. Matsushita K, Yamamoto T, Toyama H, Adachi O (1998) Biocsi Biotechnol Biochem 62: 1968 56. Vallino JJ (1991) PhD thesis, Massachusetts Institute of Technology 57. Moritz B, Striegel K, de Graaf AA, Sahm H (2000) Eur J Biochem 267:3442 58. Hartbrich A, Schmitz G, Weuster-Botz D, de Graaf AA, Wandrey C (1996) Biotechnol Bioeng 51:624 59. Tweeddale H, Notley-McRobb L, Ferenci T (1998) J Bacteriol 180:5109 60. Brer S and Krmer R (1991) Eur J Biochem 202:137 61. Brer S, Eggeling L, Krmer R (1993) Appl Environ Microbiol 59:316 62. Vrljic M, Kronemeyer W, Sahm H, Eggeling L (1995) J Bacteriol 177:4021 63. Vrljic M, Eggeling L, Sahm H (1995) Patent Application DE 19548222 64. Vrljic M, Sahm H, Eggeling L (1996) Molec Microbiol 22:815 65. Vrljic M (1997), PhD thesis, Forschungszentrum Juelich 66. Int Veld G, Driessen AJM, Konings WN (1993) FEMS Micobiol 12:293 67. Jger W, Peters-Wendisch PG, Kalinowski J, Phler A (1996) Arch Microbiol 166:76 68. Tilg Y, Eikmanns B, Eggeling L, Sahm H, Mckel B (1999) EU Patent Application EP1055725 69. Tilg Y, Eikmanns B, Eggeling L, Sahm H, Mckel B (1999) EU Patent Application EP1055730 70. Schell M (1993) Annu Rev Microbiol 47:597 71. Eikmanns BJ, Follettie MT, Griot MU, Sinskey AJ (1989) Mol Gen Genet 218:330 72. Otsuna S, Sugimoto M, Izui M, Hayakawa A, Nakano E, Kobayashi M, Yoshihara Y, Nakamatsu T (1996) Patent Application WO96/40934 73. Kreutzer C, Mckel B, Pfefferle W, Eggeling L, Sahm H, Patek M (1999) EU Patent Application EP1067193 74. Ohnishi J, Mitsuhashi S, Hayashi M, Ando S, Yokoi H, Ochiai K, Ikeda M (2002) Appl Microbiol Biotechnol 58:217 75. Kawahara Y; Nakamura T; Yoshihara Y; Ikeda S; Yoshii H (1990) Appl Microbiol Biotechnol 34:340

Biotechnological Manufacture of Lysine

109

76. Kawahara Y; Yoshihara Y; Ikeda S; Yoshii H; Hirose Y (1990) Appl Microbiol Biotechnol 34:87 77. Farwick M, Siewe RM, Krmer R (1995) J Bacteriol 177, 4690 78. Kimura E, Yakoshi C, Osumi T, Nakamatsu T (1998) Patent Application JP1998234371 79. Bukau B (1993) Mol Microbiol 9:671 80. Haynes JA, Britz ML (1990) J Gen Microbiol 136:255 81. Schfer A, Kalinowski J, Simon R, Seep-Feldhaus AH, Phler A (1990) J Bacteriol 172:1663 82. Schfer A, Kalinowski J, Phler A (1994) Appl Environ Microbiol 60:756 83. Jger W, Schfer A, Puehler A, Labes G, Wohlleben W (1992) J Bacteriol 174:5462 84. Schwarzer A, Phler A (1991) Biotechnology 9:84 85. Nampoothiri M, Pandey A (1998) Proc Biochem 33:147 86. Mckel B, Marx A, Hermann T, Farwick M, Pfefferle W (1999) Transkript 1011:41 87. Kyowa Hakko (1999) Jpn Chem Week 40:8 88. Pompejus M, Krger B, Schrder H, Zelder O, Haberhauer G (2001) Patent Application WO0100843 89. Nakagawa S, Mizoguchi H, Ando S, Hayashi M, Ochiai K, Yokoi H, Tateishi N, Senoh A, Ikeda M, Ozaki A (2001) EU Patent Application EP1108790 90. Frangeul L, Nelson KE, Buchrieser C, Danchin A, Glaser P, Kunst F (1999) Microbiology. 145:2625 91. Bathe B, Kalinowski J, Phler A (1996) Mol Gen Genet 252:255 92. Kunst F, Ogasawara N, Moszer I,Albertini AM,Alloni G,Azevedo V, Bertero MG, Bessieres P, Bolotin A, Borchert S, Borriss R, Boursier L, Brans A, Braun M, Brignell SC, Bron S, Brouillet S, Bruschi CV, Caldwell B, Capuano V, Carter NM, Choi SK, Codani JJ, Connerton IF, Cummings NJ, Daniel RA, Denizot F, Devine KM, Dsterhft A, Ehrlich SD, Emmerson PT, Entian KD, Errington J, Fabret C, Ferrari E, Foulger D, Fritz C, Fujita M, Fujita Y, Fuma S, Galizzi A, Galleron N, Ghim SY, Glaser P, Goffeau A, Golightly EJ, Grandi G, Guiseppi G, Guy BJ, Haga K, Haiech J, Harwood CR, Hnaut A, Hilbert H, Holsappel S, Hosono S, Hullo MF, Itaya M, Jones L, Joris B, Karamata D, Kasahara Y, Klaerr-Blanchard M, Klein C, Kobayash Y, Koetter P, Koningstein G, Krogh S, Kumano M, Kurita K, Lapidus A, Lardinois S, Lauber J, Lazarevic V, Lee SM, Levine A, Liu H, Masuda S, Maul C, Mdigue C, Medina N, Mellado RP, M Mizuno, Moestl D, Nakai S, Noback M, Noone D, OReilly M, Ogawa K, Ogiwara A, Oudega B, Park SH, Parro V, Pohl TM, Portetelle D, Porwollik S, Prescott AM, Presecan E, Pujic P, Purnelle B, Rapoport G, Rey M, Reynolds S, Rieger M, Rivolta C, Rocha E, Roche B, Rose M, Sadaie Y, Sato T, Scanlan E, Schleich S, Schroeter R, Scoffone F, Sekiguchi J, Sekowska A, Seror SJ, Serror P, Shin BS, Soldo B, Sorokin A, Tacconi E, Takagi T, Takahashi H, Takemaru K, Takeuchi M, Tamakoshi A, Tanaka T, Terpstra P, Tognoni A, Tosato V, Uchiyama S, Vandenbol M, Vannier F, Vassarotti A, Viari A, Wambutt R, Wedler E, Wedler H, Weitzenegger T, Winters P, Wipat A, Yamamoto H, Yamane K, Yasumoto K, Yata K, Yoshida K Yoshikawa HF, Zumstein E, Yoshikawa, Danchin A (1997) Nature 390:249 93. Cole ST, Brosch R, Parkhill J, Garnier T, Churcher C, Harris D, Gordon SV, Eiglmeier K, Gas S, Barry CE 3rd, Tekaia F, Badcock K, Basham D, Brown D, Chillingworth T, Connor R, Davies R, Devlin K, Feltwell T, Gentles S, Hamlin N, Holroyd S, Hornsby T, Jagels K, Krogh A, McLean J, Moule S, Murphy L, Oliver K, Osborne J, Quail M AM, Rajandream MA, Rogers J, Rutter S, Seeger K, Skelton J, Squares R, Squares S, Sulston JE, Taylor K, Whitehead S, Barrell BG (1998) Nature 393:537 94. Tauch A, Homann I, Mormann S, Rberg S, Billault A, Bathe B, Brand S, Brockmann-Gretza O, Ruckert C, Schischka N, Wrenger C, Hoheisel J, Mockel B, Huthmacher K, Pfefferle W, Puhler A, Kalinowski J (2002) J Biotechnol 95:25 95. Forst CV, Schulten K (1999) J Comput Biol 6:343 96. Karp PD, Krummenacker M, Paley S, Wagg J (1999) Trends Biotechnol 17:275 97. Snel B, Bork P, Huynen MA (1999) Nat Genet 21:108 98. Dandekar T, Schuster S, Snel B, Huynen M, Bork, P (1999) Biochem J 343:115 99. Cordwell SJ (1999) Arch Microbiol 172:269 100. Downs DM, Escalante-Semerena JC (2000) 20:47

110
101. 102. 103. 104. 105. 106. 107. 108. 109. 110. 111. 112. 113. 114. 115. 116. 117. 118. 119. 120. 121. 122. 123. 124. 125. 126. 127. 128. 129. 130. 131. 132. 133. 134. 135. 136. 137. 138. 139. 140. 141. 142. 143. 144.

W. Pfefferle et al. Koonin EV, Mushegian AR, Rudd KE (1996) Curr Biol 6:404 Koonin EV, Tatusov RL, Galperin MY (1998) Curr Opin Struc Biol 8:355 Galperin MY, Koonin EV (1999) Curr Opin Biotechnol 10:571 http://www.bork.embl-heidelberg.de/Genome/glycolytic_enzymes/ Mckel B; Pfefferle W (1999) Patent Application EP1103613 Mckel B; Pfefferle W (1999) Patent Application EP1106622 Sugimoto S and Shiio I (1989) Agric Biol Chem 53:1261 Mckel B; Pfefferle W (1999) EU Patent Application EP1104812 Mckel B, Weissenborn A, Pfefferle W, Kalinowski J, Bathe B, Phler A (1999) Microb Comp Genomics 4:111 Shaw-Reid CA, McCormick MM, Sinskey AJ, Stephanopoulos G (1999) Appl Microbiol Biotechnol 51:325 Ogata H, Goto S, Sato K, Fujibuchi W, Bono H, Kanehisa M (1999) NAR 27:29 http://www.sanger.ac.uk/Projects/C_diphteriae Wodicka L, Dong H, Mittmann M, Ho MH, Lockhart DJ (1997) Nature Biotech 15:1359 De Saizieu A, Certa U, Warrington J, Gray C, Keck W, Mous J (1998) Nature Biotechnol 16:45 Gingeras TR, Ghandour G,Wang E, Berno A, Small PM, Drobniewski F,Alland D, Desmond E, Holodniy M, Drenkow J (1998) Genome Res 8:435 Richmond CS, Glasner JD, Mau R, Jin H, Blattner FR (1999) Nucleic Acid Research 27:3821 Shena M, Heller RA, Theriault TP, Konrad K, Lachenmeier E, Davis RW (1998) TIBTECH 16:301 vanBogelen RA, Greis KD, Blumenthal RM, Tani TH, Matthews RG (1999) Trends Microbiol 7:320 Farwick M, Hermann T, Bathe B, Mckel B, Ermantraut E, Ellinger T, Huthmacher K, Pfefferle W (2000) Biotechnology 2000, Berlin, Germany, p 225 Edwards JS, Palsson BO (1999) J Biol Chem 274:17410 Sauer U, Lasko DR, Fiaux J, Hochuli M, Glaser R, Szyperski T,Wthrich K, Bailey JE (1999) J Bacteriol 181:6679 Krmer R (1994) FEMS Microbiol Rev 13:75 Krmer R (1996) J Biotechnol 45:1 Sahm H, Eggeling L, Eikmanns B, Krmer R (1995) FEMS Microbiol Rev 16:243 Vallino JJ, Stephanopoulos G (1993) Biotechnol Bioeng 41:633 van Gulik WM, Heijnen JJ (1995) Biotechnol Bioeng 48:681 Neijssel OM, Teixeira de Mattos MJ (1994) Mol Microbiol 13:179 Kawahara Y, Tanaka T, Ikeda S, Sone N (1988) Agric Biol Chem 52:1979 Shvinka YE, Viestur UE, Toma MK (1979) Microbiol 48:4 Sauer U, Bailey JE (1999) Biotechnol Bioeng 64:750 Tao H, Bausch C, Richmond C, Blattner FR, Conway T (1999) J Bacteriol 181:6425 Walker TE, Han CH, Kollman VH, London RE, Matwiyoff NA (1982) J Biol Chem 257:1189 Rollin C, Morgant V, Guyonvarch A, Guerquin-Kern J-L (1995) Eur J Biochem 227:488 Dominguez H, Rollin C, Guyonvarch A, Guerquin-Kern J-L, Cocaign-Bousquet M, Lindley ND (1998) Eur J Biochem 254:96 Ishino S, Shimomura-Nishimuta J, Yamaguchi K, Shirahata K, Araki K (1991) J Gen Appl Microbiol 37:157 Park SM, Sinskey AJ, Stephanopoulos G (1997) Biotechnol Bioeng 55:864 Inbar L, Lapidot A (1987) Eur J Biochem 162:621 Takiguchi N, Shimizu H, Shioya S (1997) Biotechnol Bioeng 55:170 Wiechert W, Siefke C, Marx A, de Graaf AA (1997) Biotechnol Bioeng 55:118 Sonntag K, Eggeling L, de Graaf AA, Sahm H (1993) Eur J Biochem 213:1325 Novello F, McLean P (1968) Biochem J 107:775 Kornberg HL (1966) Anaplerotic sequences and their role im metabolism. In: Campbell PN, Greville GD (eds) Essays in biochemistry, Academic, New York, vol 2, p 1 Stryer L (1988) Biochemistry, 3rd edn. Freeman, New York Misono H, Togawa H, Yamamoto T, Soda K (1979) J Bacteriol 137:22

Biotechnological Manufacture of Lysine

111

145. Eikmanns BJ, Rittmann D, Sahm H (1995) J Bacteriol 177:774 146. Tesch M, de Graaf AA, Sahm H (1999) Appl Environ Microbiol 65:1099 147. Schilling BM, Pfefferle W, Bachmann B, Leuchtenberger W, Deckwer WD (1999) Biotechnol Bioeng 64:599 148. Schweder T, Krger E, Xu B, Jrgen B, Blomsten G, Enfors SO, Hecker M (1999) Biotech Bioeng 65:151 149. Pilat P, Culik K, Paleckova F (1982) CS Patent 208076 150. Pham CB (1994) PH Patent 27995 151. Pham CB, Matsumura M, Kataoka H (1995) Microb Util Renewable Resour 9:539 152. Kim D-Il (1995) Nonmunjip Sanop Kwahak Kisul Yonguso (Inha Taehakkyo) 23:523 153. Kiss RD, Stephanopoulos G (1992) Biotechnol Bioeng 40:75 154. (2000) Res Discl 431:427 155. Koyoma Y, Ishii T, Kawahara Y, Koyama Y, Shimizu E, Yoshioka T (1998) EU Patent Application EP0844308 A2 980527 156. Lim H, Son J (1992) KR Patent 9209511 157. Oh N-S, Sernetz M (1993) Appl Microbiol Biotechnol 39:691 158. Kiss RD, Stephanopoulos G (1992) Biotechnol Bioeng 39:565 159. Hirao T, Nakano T, Azuma T, Sugimoto M, Nakanishi T (1989) Appl Microbiol Biotechnol 32:269 160. Kwong SCW, Rao G (1991) Biotechnol Bioeng 38:1034 161. Kwong SCW, Rao G (1991) Biotechnol Bioeng 40:851 162. de Hollander JA, Eswilder FR, Noordover JAC (1997) EU Patent Application EP 0796916 163. Einsele A (1978) Proc Biochem 7:1 164. Oosterhuis NMG, Kossen NWF (1984) Biotechnol Bioeng 26:546 165. Cleland N, Enfors SO (1987) Bioproc Eng 2:115 166. Heinzle E, Moes J, Dunn IJ (1985) Biotechnol Lett 7:235 167. Sweere APJ, Mesters JR, Janse L, Kossen NWF (1988) Biotech Bioeng 31:567 168. Namdev PK, Yegneswaran, PK, Thompson BG, Gray MR (1991) Can J Chem Eng 69:513 169. Vadar F, Lilly MD (1982) Eur J Appl Microbiol Biotechnol 14:203 170. Oosterhuis NMG (1983) Biotech Lett 3:141 171. Amanullah A, Baba A, McFarlane CM, Emery AN, Nienow AW (1993) In: Nienow AW (ed) Bioreactor and bioprocess fluid dynamics. Mechanical Engineering Publications, London, p 381 172. Sweere APJ, Janse L, Luyben KCAM, Kossen NWF (1988) Biotech Bioeng 31:567 173. Neissel OM and Tempest DW (1979) Symp Soc Gen Microbiol 29:53 174. Tempest DW, Neissel OM (1992) FEMS Microbiol Lett 100:169 175. Neijssel OM, Teixeira de Mattos MJ, Tempest DW (1996). In: Neidhardt FC (ed) Escherichia coli and Salmonella typhimurium. American Society for Microbiology, Washington 176. Hadj Sassi A, Fauvart L, Deschamps AM, Lebeault JM (1998) Biochem Eng J 1:85 177. Hua Q, Fu PC, Yang C, Shimizu K (1998) Biochem Eng J 2:89 178. Konstantinov KB (1996) Biotechnol Bioeng 52:271 179. Konstantinov KB, Nishio N, Yoshida T (1990) J Ferm Bioeng 70:253260 180. Cocaign-Bousquet M, Guyonvarch A, Lindley, ND (1996) Appl Environ Microbiol 62:429 181. Oh G, Moo-Young M, Chisti Y (1998) Biochem Eng J 1:211 182. Miwa H, Tamura K, Koyama Y, Tsuruta M, Tosaka O, Shimazaki K, Nakamura T, Nakayama T (1992) FR Patent 2669935 183. Pfefferle W, Lotter H, Friedrich H, Degener W (1993) EU Patent EP532 867 184. Shioya S, Shimizu H, Takiguchi N (1999) In: Lee SY, Papoutsakis ET (eds) Metabolic engineering. Marcel Dekker, New York, p 227 185. Kitsuta Y, Kishimoto M (1994) Biotechnol Bioeng 44:87 186. Patek M, Krumbach K, Eggeling L, Sahm H (1994) Appl Environ Microbiol 60:133 187. Shimizu H, Takiguchi N, Tanaka H, Shioya S (1999) Metabol Eng 1:299 188. Okamoto K, Ikeda MJ (2000) Biosci Bioeng 89:87 189. Weuster-Botz D, Kelle R, Frantzen M, Wandrey C (1997) Biotechnol Prog 13:387 190. Nystrm T (1994) Mol Microbiol 12:833

112
191. 192. 193. 194. 195. 196. 197. 198.

W. Pfefferle et al.: Biotechnological Manufacture of Lysine Kijima N, Goyal D, Takada A, Wachi M (1998) Appl Microbiol Biotechnol 50:227 Leyval D, Debay F, Engasser J-M, Goergen J-L (1997) J Microbiol Methods 121 Hewitt CJ, Boon LA, McFarlane CM, Nienow AW (1998) Biotechnol Bioeng 59:612 Kim S, Mo J, Cho YJ, Kim SJ, Lee JH, Oh JW, Park, NH (1990) FR Patent 2645172 Fechter W, Dienst JH, Le Patourel JF (1995) ZA Patent 9409059 Lucq P, Domont C (1993) EP Patent 534865 Uffmann K, Binder M (2002) US Patent 6340486 Hfler A, Alt HC, Klasen CJ, Friedrich H, Hertz U, Mrl L, Schtte R (1997) Patent Application EP809940

Received: April 2002

Вам также может понравиться