Rational Design of a Corynebacterium glutamicum Pantothenate ...

APPLIED AND ENVIRONMENTAL MICROBIOLOGY, June 2005, p. 3255–3268 0099-2240/05/$08.00⫹0 doi:10.1128/AEM.71.6.3255–3268.2005 Copyright © 2005, American Society for Microbiology. All Rights Reserved.

Vol. 71, No. 6

Rational Design of a Corynebacterium glutamicum Pantothenate Production Strain and Its Characterization by Metabolic Flux Analysis and Genome-Wide Transcriptional Profiling Andrea T. Hu ¨ser,1 Christophe Chassagnole,2,3† Nic D. Lindley,2 Muriel Merkamm,4 Armel Guyonvarch,4 Veronika Elisˇákova´,5 Miroslav Pa´tek,5 Jo ¨rn Kalinowski,6,7 1 1 Iris Brune, Alfred Pu ¨hler, and Andreas Tauch7* Lehrstuhl fu ¨r Genetik, Universita ¨t Bielefeld, Universita ¨tsstraße 25, D-33615 Bielefeld, Germany1; Laboratoire Biotechnologie-Bioproce´de´s, UMR INSA/CNRS N°5504 & UMR INSA/INRA 792, Centre de Bioingeńierie Gilbert Durand, Institut National de Sciences Appliqueés, 135 Avenue de Rangueil, F-31077 Toulouse Cedex 4, France2; CRITT Bio-Industries, 135 Avenue de Rangueil, F-31077 Toulouse Cedex 4, France3; Institut de Geńe´tique et Microbiologie, Universite´ Paris-Sud, Centre d’Orsay, F-91405 Orsay Cedex, France4; Institute of Microbiology, Academy of Sciences of the Czech Republic, Vı´den ˇska ´ 1083, CZ-14220 Praha 4, Czech Republic5; Institut fu ¨r Innovationstransfer an der Universita ¨t Bielefeld GmbH, Gescha ¨ftsbereich BioTech, Universita ¨t Bielefeld, Universita ¨tsstraße 25, D-33615 Bielefeld, Germany6; and Institut fu ¨r Genomforschung, Universita ¨t Bielefeld, Universita ¨tsstraße 25, D-33615 Bielefeld, Germany7 Received 19 October 2004/Accepted 18 December 2004

A “second-generation” production strain was derived from a Corynebacterium glutamicum pantothenate producer by rational design to assess its potential to synthesize and accumulate the vitamin pantothenate by batch cultivation. The new pantothenate production strain carries a deletion of the ilvA gene to abolish isoleucine synthesis, the promoter down-mutation P-ilvEM3 to attenuate ilvE gene expression and thereby increase ketoisovalerate availability, and two compatible plasmids to overexpress the ilvBNCD genes and duplicated copies of the panBC operon. Production assays in shake flasks revealed that the P-ilvEM3 mutation and the duplication of the panBC operon had cumulative effects on pantothenate production. During pH-regulated batch cultivation, accumulation of 8 mM pantothenate was achieved, which is the highest value reported for C. glutamicum. Metabolic flux analysis during the fermentation demonstrated that the P-ilvEM3 mutation successfully reoriented the carbon flux towards pantothenate biosynthesis. Despite this repartition of the carbon flux, ketoisovalerate not converted to pantothenate was excreted by the cell and dissipated as by-products (ketoisocaproate, DL-2,3,-dihydroxy-isovalerate, ketopantoate, pantoate), which are indicative of saturation of the pantothenate biosynthetic pathway. Genome-wide expression analysis of the production strain during batch cultivation was performed by whole-genome DNA microarray hybridization and agglomerative hierarchical clustering, which detected the enhanced expression of genes involved in leucine biosynthesis, in serine and glycine formation, in regeneration of methylenetetrahydrofolate, in de novo synthesis of nicotinic acid mononucleotide, and in a complete pathway of acyl coenzyme A conversion. Our strategy not only successfully improved pantothenate production by genetically modified C. glutamicum strains but also revealed new constraints in attaining high productivity. Pantothenate, also known as vitamin B5, is a necessary precursor for the biosynthesis of coenzyme A and phosphopantetheine, the prosthetic group of acyl carrier proteins, both of which are vital for a multitude of metabolic processes in all living cells (13). Pantothenate is synthesized by bacteria, fungi, and plants, but it is a nutritional requirement in mammals, including livestock and humans. The present technology for industrial pantothenate production relies largely on chemical synthesis from bulk chemicals, but manufacturing of pantothenate was also demonstrated by chemical conversion utilizing an enzymatic resolution process for the synthesis of D-pantolactone, the key intermediate in calcium pantothenate production

(18), and by biotechnological approaches using recombinant Escherichia coli strains (8). Pantothenate biosynthesis and production have also been investigated in the gram-positive soil bacterium Corynebacterium glutamicum, which is widely used for large-scale fermentative production of amino acids, such as L-glutamate and L-lysine (10). In C. glutamicum ATCC 13032, four enzymes are involved in the biosynthesis of pantothenate from the precursors ketoisovalerate and aspartate (15). The pathway for pantothenate biosynthesis in C. glutamicum and its integration into the synthesis of branched-chain amino acids is depicted in Fig. 1. The first reaction in the biosynthesis of pantothenate is catalyzed by the panB-encoded ketopantoate hydroxymethyltransferase (36) that converts ketoisovalerate (a precursor of valine and leucine biosynthesis [Fig. 1]) into ketopantoate using 5,10methylenetetrahydrofolate as a cofactor. Subsequently, ketopantoate is reduced to pantoate by a ketopantoate reductase activity. In C. glutamicum, the ketopantoate reductase activity

* Corresponding author. Mailing address: Institut fu ¨r Genomforschung, Universita¨t Bielefeld, Universita¨tsstraße 25, D-33615 Bielefeld, Germany. Phone: 49 521 106 5605. Fax: 49 521 106 5626. E-mail: [email protected]. † Present address: Physiomics plc, Magdalen Centre, Oxford Science Park, Oxford OX4 4GA, United Kingdom. 3255

3256

¨ SER ET AL HU

APPL. ENVIRON. MICROBIOL.

FIG. 1. Pantothenate biosynthetic pathway in C. glutamicum ATCC 13032 and its integration into the synthesis of branched-chain amino acids. The genes and enzymes involved in the biosynthetic steps are indicated. Abbreviations: TDH, threonine dehydratase; AHAS, acetohydroxy acid synthase; AHAIR, acetohydroxy acid isomeroreductase; DHAD, dihydroxy acid dehydratase; BCAT, branched-chain amino acid transaminase; KPHMT, ketopantoate hydroxymethyltransferase; PS, pantothenate synthetase; PK, pantothenate kinase; ADC, aspartate-␣-decarboxylase; IPMS, isopropylmalate synthase.

is encoded solely by the ilvC gene (25), which is also involved in the common pathway for the synthesis of the branched-chain amino acids isoleucine and valine and encodes acetohydroxy acid isomeroreductase (5, 19). Aspartate is converted into ␤-alanine by the product of the panD gene encoding aspartate␣-decarboxylase (8). The biosynthetic pathway is completed by the ATP-consuming condensation of ␤-alanine with pantoate that is catalyzed by pantothenate synthetase, encoded by the panC gene (36). C. glutamicum is a promising microorganism for examination of pantothenate overproduction since not only the molecular physiology of amino acid biosynthesis in general but also the accumulation of valine has been analyzed in detail (32, 37). The biosynthesis of valine involves many enzymes required for the production of pantothenate (Fig. 1), and the enzymatic activities and their regulation have been studied extensively in a valine-producing C. glutamicum strain (21). Furthermore, a “first-generation” pantothenate producer was developed directly from the wild-type strain C. glutamicum ATCC 13032 (36). Two important genetic features for obtaining substantial pantothenate accumulation by C. glutamicum were a chromosomal deletion of the ilvA gene, encoding threonine dehydratase (Fig. 1), and combined overexpression of the ilvBNCD and panBC genes on two compatible plasmids. Using this type of production strain, up to 1 g per liter of pantothenate accumulated in the culture supernatant (36). Analysis of different production strains suggested that increased availability of ketoisovalerate was mandatory for enhanced pantothenate synthesis by C. glutamicum (36). The successful use of ilvBNCD overexpression to obtain pantothe-

nate production was obviously due to increased ketoisovalerate availability (Fig. 1), since only then did panBC overexpression result in substantial accumulation of pantothenate. Carbon flux analysis of a “first-generation” production strain of C. glutamicum during batch cultivation with ␤-alanine supplementation revealed that the flux towards valine was 10-fold higher than that directed to pantothenate, indicating that significant improvements of strain design could be obtained only if the carbon flux at the ketoisovalerate branch point of the pathway were modulated efficiently (3). In the present work, we developed a “second-generation” C. glutamicum pantothenate producer by introducing two further genetic modifications into the “first-generation” strain (36). These modifications increased the availability and utilization of ketoisovalerate, particularly by means of promoter down-mutation of the ilvE gene in the chromosome and by duplication of the panBC operon on an expression vector. The resulting C. glutamicum strain was examined in detail by carbon flux analysis during batch cultivation and additionally by genome-wide transcriptional profiling using a whole-genome DNA microarray (11) that was based on the recently deciphered genome sequence of C. glutamicum ATCC 13032 (15). This detailed investigation of the newly designed pantothenate producer at the metabolic flux level and at the transcriptional level revealed novel constraints regarding improved production of pantothenate. MATERIALS AND METHODS Bacterial strains and growth conditions. Bacterial strains and plasmids used in this study are listed in Table 1. E. coli strains were routinely grown in LB medium

C. GLUTAMICUM PANTOTHENATE PRODUCTION

VOL. 71, 2005

3257

TABLE 1. Bacterial strains and plasmids used in this study Strain or plasmid

Relevant characteristicsa

E. coli DH5␣ C. glutamicum strains ATCC 13032⌬ilvA ATCC 13032⌬ilvA P-ilvEM3

⌬(lacZYA-argF) recA1 endA1 hsdR supE44 gyrA1 relA1 deoR thi-1 (⌽80dlacZ⌬M15)

BRL

ilvA deletion mutant of ATCC 13032 ATCC 13032⌬ilvA carrying the P-ilvEM3 mutation in the ilvE promoter

36 This study

Plasmids pK18mobsacB pK18mobsacB-ilvEM3 pJC1 pJC1ilvBNCD pCR2.1 pKK388-1 pMM36 pMM37 pMM55

E. coli cloning vector with sacB gene; Kmr pK18mobsacB containing ilvE with P-ilvEM3 mutation; Kmr E. coli-C. glutamicum shuttle vector; Kmr pJC1 containing ilvBNC and ilvD; Kmr E. coli cloning vector for PCR products; Apr Kmr E. coli expression vector with P-trc promoter; Apr E. coli-C. glutamicum shuttle vector containing trpA terminator; Cmr panBC under P-trc control in pMM36; Cmr Two copies of panBC under P-trc control in pMM36; Cmr

39 This study 6 36 Invitrogen 2 25 This study This study

a

Source or reference

Abbreviations: Kmr, kanamycin resistance; Apr, ampicillin resistance; Cmr, chloramphenicol resistance.

(38) at 37°C. C. glutamicum strains were grown at 34°C, either in brain heart infusion rich medium (Difco) or in BMCG chemically defined medium (9) containing 4% (wt/vol) glucose and supplemented with 50 mg/liter isoleucine and 20 mM ␤-alanine. The antibiotics used for selection were kanamycin (25 ␮g/ml for C. glutamicum and E. coli), chloramphenicol (6 ␮g/ml for C. glutamicum and 30 ␮g/ml for E. coli), and ampicillin (100 ␮g/ml for E. coli). General molecular biology techniques. Standard molecular biology procedures described by Sambrook et al. (38) were used. Enzymes were purchased from Promega and were used as recommended by the manufacturer. Plasmid DNA was prepared with a Wizard DNA extraction kit (Promega) by using a 1-h pretreatment at 37°C for C. glutamicum in cell resuspension solution (50 mM Tris-HCl [pH 7.5], 10 mM EDTA) containing 20 mg/ml lysozyme. DNA fragments were isolated from agarose gels with a JETSORB gel extraction kit (Genomed). Preparation of C. glutamicum genomic DNA was performed as described by Reyes et al. (33). In vitro amplification of DNA was achieved by PCR using thermostable Taq DNA polymerase (Promega) under the conditions recommended by the manufacturer. Standard transformation of C. glutamicum cells with replicative plasmids was carried out by using a protocol of Bonamy et al. (1), and integration of nonreplicative plasmids was carried out as described by van der Rest et al. (48). Construction of the P-ilvEM3 mutant. Mutation of the promoter of the ilvE gene (P-ilvE) was accomplished by site-directed mutagenesis using essentially the method of Ito et al. (12) and the following four oligonucleotide primers: Pilve (5⬘-TCGAATTCTTTGCGCAGACCCG-3⬘), PilveR (5⬘-CTGAATTCCACCG GTGAAGTAAG-3⬘), PilveM1 (5⬘-GTAGCAGGTGYNCCTTAAAATCCATG ACG-3⬘), and PilveM1R (5⬘-CGTCATGGATTTTAAGGNRCACCTGCTAC3⬘) (Y ⫽ C or T; R ⫽ A or G; N ⫽ A, C, G, or T; EcoRI sites are underlined). Two PCR fragments located upstream and downstream of the central promoter region of ilvE were amplified. These PCR fragments, delimited by primers Pilve-PilveM1R (670 bp) and PilveM1-PilveR (560 bp), overlapped in the degenerate mutational complementary primers PilveM1 and PilveM1R. The purified PCR fragments were mixed, and the final large DNA fragment (1,210 bp) was synthesized by using the outer primers. The selected mutational primers introduced various base substitutions into the ⫺10 promoter region of ilvE. The mixture of the fragments was cloned after ligation into the EcoRI site of the vector pK18mobsacB (Table 1). Mutations in individual clones were defined subsequently by DNA sequencing. The specific clone pK18mobsacB-ilvEM3 containing the mutation TACCTT 3 CTCCTT of the ⫺10 sequence was chosen for further work. This mutation nearly completely abolished the activity of the ilvE promoter and thus the expression of the gene. The construct was transferred to C. glutamicum ATCC 13032⌬ilvA by electrotransformation, and the clones in which a double recombination event occurred were identified using positive selection based on the conditional lethal effect of the sacB gene in C. glutamicum (39). The resulting C. glutamicum clone, free of the vector sequences and carrying the P-ilvEM3 mutation in the chromosome, was confirmed by PCR and DNA sequencing. Construction of the panBC expression vectors pMM37 and pMM55. The following two PCR primers were chosen based on the nucleotide sequence of the panBC genes from C. glutamicum (GenBank accession no. X96580): PanBC1 (5⬘-TCATGACCATGTCAGGCATTGATGC-3⬘) containing an RcaI site (un-

derlined) overlapping the translation initiation codon of panB and PanBC2 (5⬘-CTCGAGCTAGAGCTCGATATTGTC-3⬘), which created an XhoI site (underlined) downstream of the stop codon of panC. The 1,663-bp PCR product was cloned in E. coli into the vector pCR2.1 using a TA cloning kit (Invitrogen) and was subsequently verified by DNA sequencing. The panBC genes were reisolated from the resulting plasmid as a 1,657-bp RcaI-XhoI DNA fragment and cloned downstream of the P-trc hybrid promoter region into the E. coli-C. glutamicum shuttle vector pMM36 (Table 1) to obtain plasmid pMM37. This construction was performed by three-part ligation between the 360-bp BamHINcoI P-trc fragment from pKK388-1 (Table 1), the 1,657-bp RcaI-XhoI panBC fragment, and the pMM36 vector digested by BamHI and XhoI. The 2,037-bp BamHI-BglII fragment, containing panBC under the control of the P-trc promoter and the trpA terminator region, was reisolated from pMM37 and reinserted into the BamHI site of pMM37. Thus, the resulting plasmid, pMM55, contained two copies of panBC from C. glutamicum under the control of P-trc promoters and trpA terminator regions. Pantothenate production assay. Pantothenate production assays were performed essentially as described by Tauch et al. (46), except that the transformed C. glutamicum strains were grown in BMCG medium containing 50 mg/liter isoleucine, 20 mM ␤-alanine, 25 ␮g/ml kanamycin, and 6 ␮g/ml chloramphenicol. Pantothenate production was measured in the C. glutamicum culture supernatants after 48 h of cultivation by a microbiological assay with the pantothenate auxotrophic strain Lactobacillus plantarum ATCC 8014 (8). Metabolic flux analysis. A previously established stoichiometric method was used for metabolic flux analysis (4). This method was modified as follows. Three fluxes were added to take into account the accumulation of novel metabolites shown to accumulate in the medium. These fluxes were (i) a lactate output from pyruvate via lactate dehydrogenase, (ii) 2,3-dihydroxy-isovalerate (the direct precursor of ketoisovalerate) accumulation corresponding to the consumption of one NADPH, and (iii) an output of ketoisocaproate (the precursor of leucine) directly from ketoisovalerate but without consumption of additional NADPH equivalents as in leucine synthesis. Cultivation, determination of biomass, and extracellular metabolite measurements. The production strain C. glutamicum ATCC 13032⌬ilvA P-ilvEM3 (pJC1ilvBNCD)(pMM55) and the reference strain C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1)(pMM36) with the same genetic background but empty plasmids (Table 1) were used in pH-regulated batch cultivations. The cultivation medium and the cultivation conditions were those described previously (4), except that 1 mM leucine and valine were added to the medium to ensure good growth despite attenuation of the expression of the branched-chain amino acid transaminase IlvE. The analytical methods used to quantify biomass concentrations and extracellular metabolites have been described previously (3, 4). Ketopantoate and DL-pantoate were kindly provided by Degussa AG. DL-2,3-Dihydroxy-isovalerate sodium salt and ␣-acetolactate were synthesized as described by Leyval et al. (21). RNA isolation and labeling of hybridization probes. At different characteristic phases of the batch cultivation (7 h, 10 h, 14 h, 18 h, 22 h, and 30 h) samples of the production strain and the reference strain were frozen directly in liquid nitrogen, thereby rapidly quenching cell activity to ensure acceptable RNA extraction without degradation artifacts. Subsequently, the C. glutamicum cells were disrupted with a Ribolyser instrument (Hybaid), and total RNA was iso-

3258

¨ SER ET AL HU

APPL. ENVIRON. MICROBIOL. TABLE 2. Pantothenate production of plasmid-carrying C. glutamicum strains

C. glutamicum straina

ATCC ATCC ATCC ATCC ATCC

Pantothenate production (mM)b

Characteristics of vectors

13032⌬ilvA(pJC1)(pMM36) 13032⌬ilvA(pJC1ilvBNCD)(pMM37) 13032⌬ilvA(pJC1ilvBNCD)(pMM55) 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM37) 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55)

Controls ilvBNCD ilvBNCD ilvBNCD ilvBNCD

⫹ ⫹ ⫹ ⫹

⬍0.04 1.4 ⫾ 0.2 3.8 ⫾ 0.3 2.9 ⫾ 0.2 4.6 ⫾ 0.2

P-trc-panBC (P-trc-panBC) ⫻ 2 P-trc-panBC (P-trc-panBC) ⫻ 2

a

Strains and plasmids are described in Table 1. The growth conditions and pantothenate quantification method are described in Materials and Methods. The data are means ⫾ standard errors of four independent determinations. b

lated using an RNeasy mini kit (QIAGEN). Fluorescently labeled cDNA copies of total RNA from the production strain and the reference strains, as well as a “reference pool” of all samples, were prepared by an indirect labeling technique (11). For this, amino-modified hexamer primers and aminoallyl-modified nucleotides (dUTP) were incorporated during a first-strand reverse transcription. The cDNA was labeled with Cy3 and Cy5 monofunctional NHS esters (Amersham Biosciences). DNA microarray hybridization. The PCR product-based genomic DNA microarray of C. glutamicum ATCC 13032 containing four replicates per gene was used in this study (11). For each DNA microarray hybridization differently labeled cDNA probes, one from a specific sample and one from the “reference pool,” were combined, vacuum dried, and dissolved in 200 ␮l DIG EasyHyb hybridization solution (Roche Diagnostics). The DNA microarrays were prehybridized with DIG EasyHyb for 45 min at 45°C, washed in MilliQ H2O for 1 min, immersed in ethanol for 10 s, and finally centrifuged for 3 min at 185 ⫻ g. The dried DNA microarrays were placed into preheated (55°C) hybridization chambers of an automated slide processor (Amersham Biosciences). The combined hybridization probes were denatured by incubation at 65°C for 5 min, and 200 ␮l-samples were injected into the automated slide processor hybridization chambers with microliter syringes (Hamilton Deutschland). During the 16 h of hybridization at 45°C, the probes were mixed 60 times per h by pump movement of 50 ␮l to ensure equal delivery of the hybridization probe on the slide surface (24). Following the hybridization procedure, the DNA microarrays were washed and dried as described previously (11). Signal detection and data analysis. Signal acquisition was performed with a ScanArray 4000 microarray scanner (Perkin-Elmer). The two fluorophores were excited and detected separately, and the resulting data were stored with a pixel size of 10 ␮m. Spot finding, signal background segmentation, and intensity quantification were performed with the ImaGene 5.0 software package (BioDiscovery). Normalization and determination of t test statistics were performed with the EMMA microarray data analysis software (7). Normalization was performed with the LOWESS function, which computes the logarithmic intensity ratio and the logarithmic mean signal intensity for each spot. Data for cluster analysis were selected according to the following criteria: (i) at least six of eight replicates per gene were detected by spot finding after DNA microarray hybridization in duplicate; (ii) the P value derived from t test statistics and indicating the significance of differential gene expression compared to the “reference pool” was less than 0.001; and (iii) further extraction of genes (CDS) showing significant changes in expression regarding the complete batch cultivation by filtering with the following function, where the sd.ratio was set to 1.2:

sd. filtered(CDS) ⫽

冦

true :

sd(CDS) ⬎ sd.ratio ៮ sd

false :

sd(CDS) ⬍ sd.ratio ៮ sd

冧

The agglomerative clusters were grouped by using the Ward’s minimum-variance method (49) with the Pearson correlation coefficient as distances. Real-time RT-PCR assays. Real-time reverse transcription (RT)-PCRs were performed by using a QuantiTect SYBR Green RT-PCR kit (QIAGEN) and a LightCycler instrument (Roche Diagnostics) to validate the results of the DNA microarray hybridization. Product verification was performed by melting curve analysis. Differences in gene expression were determined by comparing the crossing points of samples of the production strain and the reference strain measured in duplicate. Crossing points were calculated with the LightCycler software, version 3 (Roche Diagnostics).

RESULTS Rational design of a C. glutamicum pantothenate production strain. A “second-generation” C. glutamicum pantothenate producer was constructed to increase the availability and utilization of ketoisovalerate for pantothenate formation. To do this, a promoter down-mutation of the ilvE gene was established by a PCR-based site-directed mutagenesis procedure, as outlined in Materials and Methods. Mutational PCR primers introduced various base substitutions into the ⫺10 promoter region of the ilvE gene, as verified by DNA sequencing. The specific clone P-ilvEM3 carried the mutation TACCTT 3 CTCCTT within the ⫺10 promoter region of ilvE. This defined mutation almost completely abolished the activity of the ilvE promoter and thus the expression of the ilvE gene in C. glutamicum (data not shown). Consequently, the corresponding plasmid clone, pK18mobsacB-ilvEM3 (Table 1), was chosen for subsequent strain construction, in which the wild-type ilvE gene of C. glutamicum ATCC 13032⌬ilvA (Table 1) was replaced in the chromosome by the P-ilvEM3 mutation, using the sacB-mediated positive selection procedure (39); this finally resulted in C. glutamicum ATCC 13032⌬ilvA P-ilvEM3. Furthermore, a duplication of the panBC operon of C. glutamicum ATCC 13032 (36) was cloned on the expression vector pMM36 in such a way that the resulting plasmid, pMM55, carried two copies of the panBC operon, each under separate control of the P-trc promoter (Table 1). Following chromosomal manipulation of C. glutamicum and plasmid construction, both C. glutamicum ATCC 13032⌬ilvA and its P-ilvEM3 derivative were cotransformed by two compatible plasmids, carrying the ilvBNC-ilvD genes (pJC1ilvBNCD) and the panBC genes (one copy of panBC under P-trc control in pMM37 and two copies of panBC under P-trc control in pMM55). This set of C. glutamicum strains allowed us to determine separately the effects on pantothenate production of attenuated ilvE expression by the P-ilvEM3 mutation and of enhanced panBC expression by pMM55. Thus, C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55) carrying both the P-ilvEM3 mutation and pMM55 was of special interest regarding its potential to produce pantothenate. Pantothenate production by plasmid-carrying C. glutamicum strains. Pantothenate production assays were performed after 48 h of shake flask cultivation of a defined set of plasmidcarrying C. glutamicum strains (Table 2). C. glutamicum ATCC 13032⌬ilvA(pJC1)(pMM36) carrying the empty cloning vectors served as a control. At least a twofold increase in pantothenate production by C. glutamicum ATCC 13032⌬ilvA

VOL. 71, 2005

(pJC1ilvBNCD) was observed when a second P-trc-panBC copy was present in the expression vector pMM55 (3.8 mM with the pMM55 plasmid versus 1.4 mM with pMM37). We also observed a twofold increase in pantothenate formation when the P-ilvEM3 mutation was introduced into the chromosome of C. glutamicum ATCC 13032⌬ilvA cotransformed by pJC1ilvBNCD and pMM37 (2.9 mM versus 1.4 mM) (Table 2). In the presence of pMM55, which carried two copies of Ptrc-panBC, the production of pantothenate by C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55) was even greater (4.6 mM versus 3.8 mM for the ⌬ilvA strain carrying the pMM55 plasmid). The increased panBC expression by pMM55 and the attenuated ilvE expression by PilvEM3 obviously had a cumulative effect on pantothenate formation by C. glutamicum, indicating that not only the precursor availability but also the enzymatic activities encoded by panB and/or panC are limiting steps for pantothenate production. The newly designed strain C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55) was further investigated by batch cultivation along with metabolic flux analysis and by genome-wide transcriptional profiling to determine other strain-specific characteristics relevant for improved pantothenate accumulation. Fermentation analysis of C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55). pH-regulated batch cultivation of C. glutamicum ATCC 13032⌬ilvA P-ilvEM3 (pJC1ilvBNCD)(pMM55) was undertaken, in which the extent of biomass development was limited by the availability of the branched-chain amino acids isoleucine, leucine, and valine and in which pantothenate production was supported by ␤-alanine supplementation. Both isoleucine and leucine were completely depleted from the medium, and valine was partially consumed (Fig. 2D). The level of biomass accumulation was comparable to that observed in C. glutamicum strains containing the wildtype ilvE gene (4), indicating that isoleucine was the dominant constraint regarding growth. Substrate consumption profiles (glucose, isoleucine, leucine, and ␤-alanine) and extracellular product accumulation profiles throughout the cultivation (pantothenate, ketoisovalerate, DL-2,3-dihydroxy-isovalerate, ketoisocaproate, pantoate, ketopantoate, pyruvate, lactate, valine, glycine, alanine, and leucine) are summarized in Fig. 2. The final pantothenate concentration attained was 8 mM (Fig. 2B), which is slightly higher than the value previously obtained (6.9 mM) for a “first-generation” C. glutamicum production strain with overexpression of both ilvBNCD and panBC genes (4). Furthermore, the specific rates of metabolite production were significantly increased, leading to a decreased duration of the fermentation (53 h instead of 72 h). The accumulation profiles for the other cometabolites show some similarities to those observed with the “first-generation” strain, including, notably, those for ␤-alanine (Fig. 2B) and glycine (Fig. 2D) that are directly linked to pantothenate production since they are involved in the supply of essential precursors. Leucine and alanine (Fig. 2D), which are alternate metabolites derived from pyruvate, also show similar accumulation characteristics, but the final concentration of valine (5 mM), the principal product in the “first-generation” strain (with a 10fold-higher flux to valine than that transformed to pantothenate), was significantly diminished (Fig. 2D), indicating that


3259

the P-ilvEM3 mutation successfully reoriented the metabolic flux. This switch in flux distribution at the level of ketoisovalerate did, however, only slightly increase the final accumulation of pantothenate. Instead, pantothenate pathway intermediates (ketopantoate, pantoate) (Fig. 2C) were observed to accumulate, and there was also increased accumulation of novel products (e.g., DL-2,3-dihydroxy-isovalerate, an intermediate of the valine pathway) (Fig. 2B). It should also be noted that ketoisocaproate, an intermediate of the leucine pathway, accumulated during the exponential growth phase (Fig. 2C) concomitant with leucine uptake from the medium and that lactate (Fig. 2C) accumulated late in the culture. Neither of these coproducts was observed in previous fermentations of C. glutamicum pantothenate producers containing plasmids leading to overexpression of the ilvBNCD and panBC genes (3, 4). Metabolic flux analysis of the pyruvate node. The flux distribution at the pyruvate node is shown in Fig. 3 as both specific rates (Fig. 3A) and percent distribution (Fig. 3B). Approximately 80% of the glucose flux was found to reach pyruvate, and the value increased further to 95% in the latter period of the cultivation as metabolite depletion for biomass synthesis no longer occurred. The rate at which carbon flux towards pyruvate took place was, however, somewhat higher than that in other C. glutamicum strains examined to date, notably during the stationary phase (4). Despite this, the flux oriented towards acetolactate was not significantly higher than that seen previously since an apparently increased flux of pyruvate towards the tricarboxylic acid cycle was observed from flux calculations based on carbon balance equations taking into account all metabolites that accumulated in the medium. In such models the CO2 not accounted for in the stoichiometric matrix is absorbed by the tricarboxylic acid cycle flux. Thus, to some extent these variations can be considered to be dependent upon NADPH requirements and the resulting apparent flux through the pentose phosphate pathway (PPP). The metabolites produced here required less NADPH, leading to a calculated PPP flux somewhat lower than that necessary in the “first-generation” strain. Thus, part of the carbon loss associated with CO2 production in the PPP was avoided, leading to an increased flux into the tricarboxylic acid cycle of the “second-generation” strain. In conclusion, the carbon flux control at the pyruvate node, at least for the input to the isoleucine-valine and pantothenate pathways, was not significantly modified despite the increased rate of the pyruvate synthesis, implying that the pathway flux is controlled by the internal characteristics of the biosynthetic pathway itself rather than by control over pyruvate metabolism. Metabolic flux analysis of the ketoisovalerate node. The flux distribution at the ketoisovalerate node regarding the specific rates and the percentages of flux is shown in Fig. 4. If one examines this flux distribution for the entire cultivation, the main feature observed for C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55) is the important decrease in the flux directed towards valine synthesis (Fig. 4A). In previous studies (4) this flux represented close to 90% of the overall flux of pyruvate via ketoisovalerate, and it was shown here to be lower than that associated with pantothenate accumulation. This increased availability of ketoisovalerate did not, however, significantly modify the pantothenate flux (there was

3260

¨ SER ET AL HU


FIG. 2. Evolution of different substrates and products throughout cultivation of C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD) (pMM55). Biomass accumulation and glucose consumption are shown in panel A. Most of the results are expressed in mM; the only exception is biomass, which is expressed in grams per liter.

a twofold increase in the specific rate of pantothenate accumulation but no gain in the flux percentage); rather, it led to increased accumulation of coproducts, notably ketoisovalerate (Fig. 2B). Closer examination of these data indicated that the flux repartition could be divided into three major periods of the batch cultivation: the early exponential growth phase (0 h to 15 h), the late exponential phase (15 h to 30 h), and the stationary phase (after 30 h). In the first period, the major prod-

uct is ketoisocaproate, a direct precursor of leucine, and this period is concomitant with the leucine uptake period (Fig. 2C). Effectively, after 15 h no more flux towards ketoisocaproate was observed (Fig. 4A), although growth continued, presumably by alternative aminotransferase reactions. During the late exponential phase, ketoisovalerate became the major metabolite that accumulated (Fig. 2B). During this period an increase in the pantothenate, ketopantoate, valine, and leucine fluxes was observed (Fig. 4A and B). During the final period of batch

VOL. 71, 2005


3261

FIG. 3. Metabolic flux distribution at the pyruvate node throughout cultivation of C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD) (pMM55). (A) Absolute flux values expressed as mmol/(g [dry weight] ⫻ h). (B) Relative values expressed as molar percentages of the glucose consumption rate. Fluxes from glycolysis and the phosphotransferase system arriving at pyruvate }, from pyruvate to acetolactate via acetohydroxy acid synthase E, from pyruvate to acetyl-CoA via pyruvate dehydrogenase ■, and from pyruvate to biomass Œ were determined. The flux from pyruvate to alanine is not shown since the value was less than 2% throughout cultivation. The flux to biomass includes the direct flux from pyruvate and the anaplerotic flux leading to oxaloacetate (OAA).

cultivation, a shift occurred, in which the rate of ketoisovalerate synthesis diminished despite a constant flux through acetolactate (Fig. 4A) due to the excretion of DL-2,3-dihydroxyisovalerate (Fig. 2B), a direct precursor of ketoisovalerate. During the same period the output fluxes for the pantothenate pathway intermediates ketopantoate and pantoate were also found to increase. Genome-wide transcriptional analysis of C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55). Complementary to the metabolic flux analysis, the pantothenate producer C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD) (pMM55) was also characterized at the transcriptional level by using the recently developed whole-genome DNA microarray of C. glutamicum (11). The control strain C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1)(pMM36), carrying the same chromosomal modifications but empty cloning vectors, was cultivated under identical conditions in a batch fermentation and served as a reference for the detection of relative changes in gene expression in the production strain. At six times during the batch cultivations (7 h, 10 h, 14 h, 18 h, 22 h, and 30 h), C. glutamicum cells were sampled and frozen directly in liquid nitrogen to ensure accurate extraction of total RNA without degradation artifacts. Aliquots of the 12 RNA samples were finally combined to obtain the so-called “reference pool,” and each specific RNA sample from both the pantothenate producer and the control strain was separately cohybridized with the “reference pool” on the DNA microarray in order to optimize data normalization (51). Following signal detection and data acquisition, genes showing differential expression compared with the “reference pool” were selected according to the

criteria defined in Materials and Methods. The resulting data were filtered further to identify those genes that were differentially expressed during the whole batch cultivation process. This filtering resulted in the identification of 349 C. glutamicum genes whose expression data were used finally for visualization by agglomerative hierarchical clustering (Fig. 5). Hierarchical clustering enabled us to detect three obvious clusters (clusters A to C) among the genes selected from DNA microarray hybridizations. It was very striking that genes belonging to clusters A or C showed opposite expression profiles during the batch cultivation and that the switch in gene expression occurred between 14 h and 18 h of the fermentation (Fig. 5). This overall change in gene expression might coincide, therefore, with the flux repartition detected between the early exponential (0 h to 15 h) and late exponential growth phases (15 h to 30 h) of the batch cultivation and with the simultaneous depletion of the amino acid isoleucine, which represented the dominant constraint regarding growth of the ⌬ilvA mutants. A closer examination of the hybridization data revealed that cluster A consists of 117 genes whose relative expression was enhanced during the early exponential growth phase. Although this observation holds for both the pantothenate producer and the control strain, the relative gene expression in the production strain was increased compared with that in the “reference pool.” According to the genome annotation of C. glutamicum ATCC 13032 (15), a putative function was assigned to 51 genes in cluster A. The gene products probably participate in several metabolic pathways of C. glutamicum ATCC 13032, including leucine biosynthesis, serine, glycine, and methionine forma-

3262

¨ SER ET AL HU


FIG. 4. Metabolic flux distribution at the ketoisovalerate node throughout cultivation of C. glutamicum ATCC 13032⌬ilvA P-ilvEM3 (pJC1ilvBNCD)(pMM55). (A) V1, V2, V3, V4, V5, V6, V7, and V8 are absolute values expressed as mmol/(g [dry weight] ⫻ h). (B) V9, V10, V11, V12, and V13 are absolute values expressed as mmol/(g [dry weight] ⫻ h). (C) V2, V3, V4, V5, V7, V8, V10, V12, and V13 are relative values expressed as molar percentages of the glucose consumption rate. The following fluxes were determined: from pyruvate to DL-2,3-dihydroxyisovalerate (V1); DL-2,3-dihydroxy-isovalerate excretion (V2); from DL-2,3-dihydroxy-isovalerate to ketoisovalerate (V3); valine excretion (V4); ketoisovalerate excretion (V5); from ketoisovalerate to ketoisocaproate (V6); leucine excretion (V7); ketoisocaproate excretion (V8); from ketoisovalerate to ketopantoate (V9); ketopantoate excretion (V10); from ketopantoate to pantoate (V11); pantoate excretion (V12); and pantothenate excretion (V13). The following enzyme reactions corresponded to the fluxes: V1, acetohydroxy acid synthase and acetohydroxy acid isomeroreductase; V2, DL-2,3-dihydroxy-isovalerate export; V3, dihydroxy acid dehydratase; V4, branched-chain amino acid transaminase and valine export; V5, ketoisovalerate export; V6, isopropylmalate synthase, isopropylmalate isomerase, and 3-isopropylmalate dehydrogenase; V7, branched-chain amino acid transaminase and leucine export; V8, ketoisocaproate export; V9, ketopantoate hydroxymethyltransferase; V10, ketopantoate export; V11, acetohydroxy acid isomeroreductase; V12, pantoate export; and V13, pantothenate synthetase and pantothenate export.

tion, methylenetetrahydrofolate regeneration, de novo synthesis of nicotinic acid mononucleotide, and the conversion of acyl coenzyme A (acyl-CoA) (Fig. 5). Cluster C is roughly twofold larger than cluster A and comprises 211 genes whose expression is characterized by a relative increase in the late exponential growth phase of fermentation. This cluster might be divided into subclusters, especially subclusters comprising genes with a lower expression level in the production strain in the early exponential growth phase (e.g., cluster C1). Interestingly, the coaA gene encoding pantothenate kinase was grouped into cluster C1 (Fig. 5). Pantothenate kinase catalyzes the first step of coenzyme A synthesis using pantothenate as a precursor. At least in E. coli, this reaction plays a crucial role in controlling

the intracellular coenzyme A concentration (44). A more detailed classification of the genes belonging to cluster C in the metabolic pathways of C. glutamicum ATCC 13032 was hampered mainly by the lack of substantial functional predictions so that putative functions were assigned to only 69 members of the cluster. The expression profile of genes belonging to cluster B was markedly different than those of clusters A and C since the relative gene expression was significantly higher in the production strain during both growth phases of the batch cultivation (Fig. 5). Cluster B contains 21 genes, including the plasmidencoded ilvBNCD and panBC genes. Further functions were assigned to only three genes, including genes encoding two

VOL. 71, 2005


3263

FIG. 5. Agglomerative hierarchical clustering of the DNA microarray hybridization data. RNA samples from both the pantothenate producer (P) C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55) and the control strain (C) C. glutamicum ATCC 13032⌬ilvA P-ilvEM3 (pJC1)(pMM36) were cohybridized with a “reference pool.” Samples were taken from batch cultivations at six times (7 h, 10 h, 14 h, 18 h, 22 h, and 30 h). The identified gene clusters, clusters A to C, are indicated; a specific region of cluster C is marked C1. An excerpt of the most relevant genes belonging to clusters A and C is shown on the right. All members of cluster B are specified by a relative increase in gene expression in the pantothenate producer throughout the fermentation.

3264

¨ SER ET AL HU


FIG. 6. Relative expression of plasmid-encoded genes in C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55). Gene expression was measured with the LightCycler instrument using total RNA from the pantothenate production strain and the reference strain with empty cloning vectors. Relative gene expression was normalized using the values determined for the reference strain. The values are means of four measurements, and the error bars indicate standard deviations.

putative acetyltransferases and the alternative sigma factor C (Fig. 5). Characterization of expression of the plasmid-encoded ilvBNC-ilvD and panBC genes by real-time RT-PCR. Due to their separate clustering, expression of the plasmid-encoded ilvBNC-ilvD and panBC genes was also examined by real-time RT-PCR since the dynamic measurement range of this method provides more precise detection of relative expression levels in the pantothenate producer C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55) compared with the reference strain C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1) (pMM36). The relative expression levels of ilvBNC-ilvD and panBC in the production strain measured at the characteristic sampling points of batch cultivation are illustrated in Fig. 6. The resulting data from the RT-PCR assays could be divided into three groups with respect to the relative expression units detected. (i) The ilvBNC gene cluster, under control of its native promoter on pJC1ilvBNCD (36), showed a more or less constant level of enhanced gene expression (approximately 10-fold) in the production strain, with the exception of a slight increase in relative expression at 14 h. This time is close to the end of the early exponential growth phase (0 h to 15 h) of the batch cultivation. (ii) The relative increase in ilvD gene expression was slightly lower than the relative increase in expression of the ilvBNC genes, but the expression exhibited a similar course during the fermentation and also exhibited a maximum at 14 h. The ilvD gene formed a separate transcription unit on pJC1ilvBNCD since it was cloned on pJC1 with its native promoter region (36). (iii) The panBC operon constituted a third group of genes with a different expression profile, probably since both copies of panBC are under control of the constitutive P-trc promoter on pMM55. The relative expression of the

panBC operon was strongly enhanced (between 50-fold and 80-fold) in the production strain and revealed a further noticeable increase in transcription in the late exponential phase of the batch cultivation (15 h to 30 h). These RT-PCR assays confirmed clearly that there was substantial overexpression of both the ilvBNCD and panBC genes throughout cultivation of C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD) (pMM55). DISCUSSION Physiological characterization of the new pantothenate producer and perspectives for strain design deduced from metabolic flux analysis. The major limitation of pantothenate production by genetically modified C. glutamicum strains was linked previously to the 10-fold-higher flux of ketoisovalerate towards valine than towards pantothenate (3). In the “secondgeneration” production strain C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55), the important flux towards valine was diminished significantly by the P-ilvEM3 mutation, thereby increasing the availability of ketoisovalerate for the pantothenate biosynthetic pathway. However, the increased flux towards pantothenate was only twofold higher than that previously described (4), which was consistent with the twofold overexpression of the panBC operon. Carbon flux analysis also revealed that much of the remaining ketoisovalerate not converted to pantothenate was either excreted by the cell or dissipated as by-products within associated pathways. Effectively, ketoisocaproate, DL-2,3-dihydroxy-isovalerate, ketopantoate, and pantoate have been quantitatively recovered in the culture supernatant, indicating saturation of the pantothenate biosynthetic pathway. These data suggest that the level of expression

VOL. 71, 2005

of the panBC genes has a controlling effect on pathway flux and that enhanced expression of panBC would probably lead to higher rates of carbon conversion to pantothenate. It also remains to be seen whether further amplification of the panBC operon along with the ilvBNCD genes can significantly modify the pathway flux in such a way that the rate of acetolactate synthesis can be improved. Despite the twofold increase in the pantothenate pathway flux, the final concentration of pantothenate increased only slightly compared with that of the “first-generation” production strain (4). This might indicate rate limitation at the transport level of pantothenate, which could also have a controlling effect on pathway flux. To date, no information is available concerning the pantothenate export system of C. glutamicum ATCC 13032. However, the corresponding genetic determinant represents an attractive target to explore its potential for improving pantothenate accumulation. A similar strategy has been used successfully in amino acid production by C. glutamicum since overexpression of the threonine exporter gene thrE considerably improved threonine accumulation by reducing its intracellular degradation and increasing its export (42). Since the availability of sufficient ␤-alanine was found to be mandatory for substantial pantothenate production by C. glutamicum (36), the identification and subsequent genetic engineering of the ␤-alanine import system in a production strain also might be promising. A further physiological characteristic of the “second-generation” production strain during batch cultivation was a shift in which the rate of ketoisovalerate synthesis decreased despite a constant flux through acetolactate due to the excretion of DL2,3-dihydroxy-isovalerate. This shift cannot be explained by the inhibition of dihydroxy acid dehydratase (IlvD) activity (EC. 4.2.1.9) by valine or leucine since only low concentrations of these amino acids occurred if the inhibition constants for these amino acids (170 and 120 mM, respectively) were taken into consideration (21). A possible cause for the excretion of DL2,3-dihydroxy-isovalerate could be the accumulation of the downstream metabolites ketoisovalerate and ketopantoate. Indeed, the two downstream activities dihydroxy acid dehydratase (IlvD) and ketopantoate hydroxymethyltransferase (PanB) were probably in near equilibrium. No experimental data are available for these reactions, but there are data for similar reactions; e.g., in the case of dehydratase activities, the phosphopyruvate hydratase (EC 4.2.1.11) has a Keq of 4.6 at 38°C and pH 7.1 (22). For the methyltransferase family using similar cofactors, Schirch et al. (40) obtained a value of 0.083 for the glycine hydroxymethyltransferase (EC 2.1.2.1) at 35°C and pH 7.5, and a value of 0.333 has been reported by Wilson and Snell (50) for the D-alanine 2-hydroxymethyltransferase (EC 2.1.2.7) at 38°C and pH 7.5. On the other hand, the final step of the pantothenate biosynthetic pathway could probably be considered reversible under physiological conditions since Zheng and Blanchard (52) have shown that the pantothenate synthetase of Mycobacterium tuberculosis is capable of AMP-dependent hydrolysis of pantothenate. Even though the equilibrium constant of this reaction has not been determined so far, Schuegraf et al. (41) obtained Keq values ranging from 2.14 to 38.3 at 38°C for pH values of 6.91 and 7.70, respectively, for a similar carbon-nitrogen bond-forming reaction catalyzed by the arginosuccinate synthetase (EC 6.3.4.5). In case of substantial pan-


3265

tothenate accumulation by a C. glutamicum production strain, near equilibrium of this step would therefore result in accumulation of the upstream metabolite(s), corresponding to the observed accumulation of pantoate and ketopantoate. However, these results are more likely to be interpreted as a poor export capacity for pantothenate in C. glutamicum. Physiological characterization of the new pantothenate producer and perspectives for strain design deduced from transcriptomics. DNA microarray hybridization and subsequent clustering of the resulting data provided another way to characterize the physiology of the “second-generation” pantothenate producer. Cluster B contains the plasmid-encoded ilvBNCD and panBC genes that showed enhanced expression during the whole fermentation process compared with the reference strain. A distinctive feature of the panBC genes was a further noticeable increase in expression from 14 h to 30 h of cultivation. A likely explanation for this effect is a decrease in panBC gene expression in the reference strain due to the transition from early to late exponential growth, which led to a relative increase in panBC expression in the pantothenate producer that was controlled by the constitutive P-trc promoter. In principle, these data demonstrated clearly that overexpression of the plasmid-encoded ilvBNCD and panBC genes was substantial in C. glutamicum, at least at the transcriptional level, and that pJC1ilvBNCD and pMM55 are convenient tools for improving pantothenate production. A remarkable outcome of hierarchical clustering is that expression of a set of genes involved in amino acid biosynthesis was increased in the pantothenate production strain during the early exponential growth phase. This set of genes comprises the leuBCD genes of the leucine biosynthetic pathway, which participates in the conversion of isopropylmalate to ketoisocaproate (15, 29). This might be consistent with the observation that ketoisocaproate, the direct precursor of leucine, is the major accumulation product during the first growth phase during cultivation. However, this growth phase was also characterized by the concomitant uptake of leucine from the medium, which was depleted after 15 h of cultivation. It was thus surprising to detect enhanced expression of the leuBCD genes in the production strain since at least the activity of the leuB promoter was shown to be significantly reduced when leucine was present in the growth medium (29), which is indicative of negative regulation at the transcriptional level. Likewise, the enzyme activities of LeuB and LeuCD are generally reduced upon addition of leucine to the growth medium (28). It is likely that the enhanced expression of leuBCD along with the accumulation of ketoisocaproate might be provoked by the rather unusual intracellular conditions in the production strain that are specified by the increased availability of ketoisovalerate, the precursor of leucine synthesis, and the simultaneous attenuation of ilvE gene expression, which is also involved in the final enzymatic reaction of this pathway (23). Attenuation of gene expression in the complete leucine biosynthetic pathway of a pantothenate production strain seems therefore worthwhile to reduce the synthesis of by-products directly derived from ketoisovalerate. The fact that the leuA gene, which is involved in the first step of leucine synthesis, was not among the genes in cluster A might be due to its different regulatory pattern of expression since it is most likely controlled by a transcriptional attenuation mechanism (28).

3266

¨ SER ET AL HU

Another set of genes belonging to cluster A (serA, glyA, and metE) is involved in the biosynthesis of serine, glycine, and methionine in C. glutamicum ATCC 13032. The serA gene encodes the 3-phosphoglycerate dehydrogenase that is responsible for the initial step in the synthesis of serine (31), which is converted to glycine by the glyA gene product, the serine hydroxymethyltransferase (42). The enhanced expression of both serA and glyA might be consistent with the observed accumulation of glycine during batch cultivation. In principle, the enzymatic reaction catalyzed by GlyA also results in regeneration of the one-carbon unit 5,10-methylenetetrahydrofolate that is required as a cofactor for the conversion of ketoisovalerate to ketopantoate by the PanB enzyme. The accumulation of glycine might indicate, therefore, that the availability of 5,10-methylenetetrahydrofolate is a limiting (co)factor for substantial pantothenate production (3). Targeted genetic experiments indicated that the serine hydroxymethyltransferase seems to be the only enzyme responsible for the regeneration of 5,10-methylenetetrahydrofolate in C. glutamicum ATCC 13032 (42). To overcome the problems of glycine accumulation and cofactor limitation, it would be interesting to couple directly glycine degradation to 5,10-methylenetetrahydrofolate regeneration via the glycine cleavage system, a multienzyme complex consisting of the GcvTHP components in E. coli (45). Since this system is not encoded by the C. glutamicum ATCC 13032 genome (15), it will be necessary to express the E. coli genes in the heterologous host and to examine whether pantothenate production can be substantially improved in such a recombinant C. glutamicum strain. The fact that the glycine cleavage system of E. coli is an attractive target for the design of production strains was shown recently in a patented process for the fermentative production of pantothenate with recombinant E. coli strains overexpressing the gcvTHP genes (34). The changes in gene expression of metE coding for methionine synthase, which catalyzes the final step of the methionine pathway in C. glutamicum (35), might be linked to glycine accumulation via the enzyme’s cofactor requirement. MetE catalyzes the methyl transfer from 5-methyltetrahydrofolate to homocysteine, the direct precursor of methionine. The cofactor of this reaction, 5-methyltetrahydrofolate, results from the enzymatic reduction of 5,10-methylenetetrahydrofolate by the metF gene product (35). The enhanced expression of metE might be indicative of an increased demand for methionine due to inefficient conversion of homocysteine caused by putative cofactor limitation. Cluster A also comprises the cg1214, cg1215, and cg1216 coding regions. Sequence annotation of the deduced proteins encoded by cg1216 and cg1215 revealed similarity to NadA, a quinolinate synthase, and NadC, a nicotinate-nucleotide pyrophosphorylase (15). These proteins catalyze consecutive steps in the biosynthesis of nicotinic acid mononucleotide using the precursors L-aspartate and dihydroxyacetone phosphate (30). The third gene in this gene cluster, cg1214, was predicted to encode a putative cysteine desulfurase (15). Cysteine desulfurases are responsible for the trafficking of sulfur in such a way that they transfer the sulfur from L-cysteine to iron-sulfur clusters of several proteins. In E. coli it was shown that quinolinate synthase A contains such an iron-sulfur cluster (26). It is currently not clear why de novo synthesis of nicotinic acid mononucleotide would be necessary in the “second-generation”


production strain of C. glutamicum. One might speculate that some NAD degradation occurs and that overexpression of nadAC is a way to maintain an adequate NAD concentration in cells of the pantothenate production strain. Additionally, a set of genes involved in the conversion of acyl-CoA showed enhanced expression in the “second-generation” production strain. The most prominent genes are dtsR1 and dtsR2, which encode carboxyltransferase domains of acylCoA carboxylases (20). The DtsR proteins were suggested to form biotin-containing enzyme complexes with the AccBC subunit (14, 20), encoded by the accBC gene that was also identified among the members of cluster A. In principle, these proteins are able to catalyze the conversion of propionyl-CoA to D-methylmalonyl-CoA. It is worth mentioning that the mcmAB genes exhibited a relative increase in expression. These genes most likely encode both subunits of the methylmalonyl-CoA mutase that generally coverts L-methylmalonylCoA to succinyl-CoA (15). The enzyme (methylmalonyl-CoA epimerase) mediating the important interconversion between the D and L configurations of methylmalonyl-CoA is probably encoded by the cg1373 gene, whose expression also increased throughout the fermentation. Furthermore, the expression of genes following the entry of succinyl-CoA into the tricarboxylic acid cycle (sucCD and sdhABC) was enhanced. Consequently, these data suggest that there is a complete pathway from propionyl-CoA to succinyl-CoA (16) that appeared to be differentially expressed in the pantothenate production strain and that could be involved in channeling of acyl-CoA into the tricarboxylic acid cycle. This methylmalonyl pathway can also be involved in the degradation of ketoisovalerate (16), which is a major accumulation product during fermentation. Further genes essential for this degradative pathway are not present in the C. glutamicum genome, actually excluding the possibility that ketoisovalerate degradation is the cause for the observed expression pattern. On the other hand, both ketoisovalerate and ketoisocaproate, which also accumulated during fermentation, might be converted to their corresponding short-chain fatty acid CoA esters, isobutyryl-CoA and isovaleryl-CoA, in the pantothenate production strain. Such conversions of keto acids have been observed in Bacillus subtilis in the presence of high intracellular concentrations of ketoisovalerate and ketoisocaproate caused by supplementation of the growth medium with branched-chain amino acids (17). In B. subtilis, the corresponding decarboxylation and activation reaction is catalyzed by the specific branched-chain keto acid decarboxylase, but also nonspecific decarboxylases, such as the pyruvate dehydrogenase, were assumed to mediate this conversion (27). It is worth mentioning that the expression of the aceE gene encoding the E1 subunit of the unusual pyruvate dehydrogenase complex of C. glutamicum (15) is enhanced in the pantothenate production strain. However, differential expression of genes involved in the metabolism of (acyl-)CoA in the pantothenate producer might be more complex since the expression of coaA, encoding pantothenate kinase involved in the initial step of coenzyme A biosynthesis, was decreased, whereas the expression of fas-IB, encoding fatty acid synthase IB, actA, encoding acyl-CoA:CoA transferase, and tesB2, encoding acyl-CoA thioesterase, was enhanced in the production strain. The C. glutamicum genome contains two genes coding for type I fatty acid synthases, Fas-IA and Fas-IB (15), but the actual physio-


VOL. 71, 2005

logical function of Fas-IB is currently unknown. During growth of C. glutamicum on acetate and propionate as carbon sources, the expression patterns of the fas genes characteristically differed, since only the expression of fas-IB along with the dtsR1 and dtsR2 genes was affected (11). The ActA enzyme catalyzes the reversible transfer of the coenzyme A moiety from acetylCoA to various acids, such as succinate, butyrate, and hydroxybutyrate (43). Thus, C. glutamicum might be able to synthesize tricarboxylic acid cycle intermediates from different acyl-CoA substrates. A probable physiological role of acyl-CoA thioesterase is to prevent the abnormal accumulation of acyl-CoA and to eliminate “irregular” acyl-CoA molecules from the acylCoA pool of the cell, thereby preventing their incorporation into lipids (47, 53). It would be interesting to identify the source of acyl-CoA and thus to reconcile the increased expression of the methylmalonyl pathway with the specific characteristics of carbon flux and metabolite accumulation in the pantothenate producer, to identify further targets for the rational design of production strains. In conclusion, a detailed analysis of the “second-generation” pantothenate producer C. glutamicum ATCC 13032⌬ilvA P-ilvEM3(pJC1ilvBNCD)(pMM55) suggested that a number of metabolic phenomena are involved in the improved production of pantothenate, which require further investigation. The use of both metabolic flux analysis and DNA microarray technology for the physiological characterization of a rationally designed production strain provided an impressive example for the detection of previously unsuspected complexities in the metabolic response. Further examination of the best production strain will hopefully allow us to determine and circumvent the next limiting step(s) for pantothenate production in C. glutamicum. ACKNOWLEDGMENTS We thank S. Delauney (Nancy) for supplying DL-2,3-dihydroxyisovalerate sodium salt, Degussa AG for providing ketopantoate and DL-pantoate, and M. Dondrup (Bielefeld) for help with clustering. This project received financial support from the European Union (research contract QLK3-CT-2000-00497). REFERENCES 1. Bonamy, C., A. Guyonvarch, O. Reyes, F. David, and G. Leblon. 1990. Interspecies electro-transformation in corynebacteria. FEMS Microbiol. Lett. 66:263–269. 2. Brosius, J. 1988. Expression vectors employing lambda-, trp-, lac- and lppderived promoters, p. 205–225. In R. L. Rodriguez and D. T. Denhardt (ed.), Vectors: a survey of molecular cloning vectors and their uses. Butterworth, Boston, MA. 3. Chassagnole, C., F. Le´tisse, A. Diano, and N. D. Lindley. 2002. Carbon flux analysis in a pantothenate overproducing Corynebacterium glutamicum strain. Mol. Biol. Rep. 29:129–134. 4. Chassagnole, C., A. Diano, F. Le´tisse, and N. D. Lindley. 2003. Metabolic network analysis during fed-batch cultivation of Corynebacterium glutamicum for pantothenic acid production: first quantitative data and analysis of byproduct formation. J. Biotechnol. 104:261–272. 5. Cordes, C., B. Mo ¨ckel, L. Eggeling, and H. Sahm. 1992. Cloning, organization and functional analysis of ilvA, ilvB and ilvC genes from Corynebacterium glutamicum. Gene 112:113–116. 6. Cremer, J., L. Eggeling, and H. Sahm. 1990. Cloning the dapA dapB cluster of the lysine-secreting bacterium Corynebacterium glutamicum. Mol. Gen. Genet. 220:478–480. 7. Dondrup, M., A. Goesmann, D. Bartels, J. Kalinowski, L. Krause, B. Linke, O. Rupp, A. Sczyrba, A. Pu ¨hler, and F. Meyer. 2003. EMMA: a platform for consistent storage and efficient analysis of microarray data. J. Biotechnol. 106:135–146. 8. Dusch, N., A. Pu ¨hler, and J. Kalinowski. 1999. Expression of the Corynebacterium glutamicum panD gene encoding L-aspartate-␣-decarboxylase leads to pantothenate overproduction in Escherichia coli. Appl. Environ. Microbiol. 65:1530–1539.

3267

9. Guillouet, S., and J. M. Engasser. 1995. Sodium and proline accumulation in Corynebacterium glutamicum as a response to an osmotic saline upshock. Appl. Microbiol. Biotechnol. 43:315–320. 10. Hermann, T. 2003. Industrial production of amino acids by coryneform bacteria. J. Biotechnol. 104:155–172. 11. Hu ¨ser, A. T., A. Becker, I. Brune, M. Dondrup, J. Kalinowski, J. Plassmeier, A. Pu ¨hler, I. Wiegra ¨be, and A. Tauch. 2003. Development of a Corynebacterium glutamicum DNA microarray and validation by genome-wide expression profiling during growth with propionate as carbon source. J. Biotechnol. 106:269–286. 12. Ito, W., H. Ishiguro, and Y. Kurosawa. 1991. A general method for introducing a series of mutations into cloned DNA using the polymerase chain reaction. Gene 102:67–70. 13. Jackowski, S. 1996. Biosynthesis of pantothenic acid and coenzyme A, p. 687–694. In F. C. Neidhardt, R. Curtiss III, J. L. Ingraham, E. C. C. Lin, K. B. Low, B. Magasanik, W. S. Reznikoff, M. Riley, M. Schaechter, and H. E. Umbarger (ed.), Escherichia coli and Salmonella: cellular and molecular biology. ASM Press, Washington, D.C. 14. Ja ¨ger, W., P. G. Peters-Wendisch, J. Kalinowski, and A. Pu ¨hler. 1996. A Corynebacterium glutamicum gene encoding a two-domain protein similar to biotin carboxylases and biotin-carboxyl-carrier proteins. Arch. Microbiol. 166:76–82. 15. Kalinowski, J., B. Bathe, D. Bartels, N. Bischoff, M. Bott, A. Burkovski, N. Dusch, L. Eggeling, B. J. Eikmanns, L. Gaigalat, A. Goesmann, M. Hartmann, K. Huthmacher, R. Kra ¨mer, B. Linke, A. C. Mc Hardy, F. Meyer, B. Mo ¨ckel, W. Pfefferle, A. Pu ¨hler, D. A. Rey, C. Ru ¨ckert, O. Rupp, H. Sahm, V. F. Wendisch, I. Wiegra ¨be, and A. Tauch. 2003. The complete Corynebacterium glutamicum ATCC 13032 genome sequence and its impact on the production of L-aspartate-derived amino acids and vitamins. J. Biotechnol. 104:5–25. 16. Kamoun, P. 1992. Valine is a precursor of propionyl-CoA. Trends Biochem. Sci. 17:175–176. 17. Kaneda, T. 1977. Fatty acids of the genus Bacillus: an example of branchedchain preference. Bacteriol. Rev. 41:391–418. 18. Kataoka, M., K. Shimizu, K. Sakamoto, H. Yamada, and S. Shimizu. 1995. Lactonohydrolase-catalyzed optical resolution of pantoyl lactone: selection of a potent enzyme producer and optimization of culture and reaction conditions for practical resolution. Appl. Microbiol. Biotechnol. 44:333–338. 19. Keilhauer, C., L. Eggeling, and H. Sahm. 1993. Isoleucine synthesis in Corynebacterium glutamicum: molecular analysis of the ilvB-ilvN-ilvC operon. J. Bacteriol. 175:5595–5603. 20. Kimura, E. 2002. Triggering mechanism of L-glutamate overproduction by DtsR1 in coryneform bacteria. J. Biosci. Bioeng. 94:545–551. 21. Leyval, D., D. Uy, S. Delaunay, J. L. Goergen, and J. M. Engasser. 2003. Characterisation of the enzyme activities involved in the valine biosynthetic pathway in a valine-producing strain of Corynebacterium glutamicum. J. Biotechnol. 104:241–252. 22. Lowry, O. H., and J. V. Passonneau. 1964. The relationships between substrates and enzymes of glycolysis in brain. J. Biol. Chem. 239:31–42. 23. McHardy, A. C., A. Tauch, C. Ru ¨ckert, A. Pu ¨hler, and J. Kalinowski. 2003. Genome-based analysis of biosynthetic aminotransferase genes of Corynebacterium glutamicum. J. Biotechnol. 104:229–240. 24. McQuain, M. K., K. Seale, J. Peek, T. S. Fisher, S. Levy, M. A. Stremler, and F. R. Haselton. 2003. Chaotic mixer improves microarray hybridization. Anal. Biochem. 325:215–226. 25. Merkamm, M., C. Chassagnole, N. D. Lindley, and A. Guyonvarch. 2003. Ketopantoate reductase activity is only encoded by ilvC in Corynebacterium glutamicum. J. Biotechnol. 104:253–260. 26. Mihara, H., and N. Esaki. 2002. Bacterial cysteine desulfurases: their function and mechanisms. Appl. Microbiol. Biotechnol. 60:12–23. 27. Oku, H., and T. Kaneda. 1988. Biosynthesis of branched-chain fatty acids in Bacillus subtilis. J. Biol. Chem. 263:18386–18396. 28. Pa ´tek, M., K. Krumbach, L. Eggeling, and H. Sahm. 1994. Leucine synthesis in Corynebacterium glutamicum: enzyme activities, structure of leuA, and effect of leuA inactivation on lysine synthesis. Appl. Environ. Microbiol. 60: 133–140. 29. Pa ´tek, M., J. Hochmannova ´, M. Jelıńkova ´, J. Nesˇvera, and L. Eggeling. 1998. Analysis of the leuB gene from Corynebacterium glutamicum. Appl. Microbiol. Biotechnol. 50:42–47. 30. Penfound, T., and J. W. Foster. 1996. Biosynthesis and recycling of NAD, p. 721–730. In F. C. Neidhardt, R. Curtiss III, J. L. Ingraham, E. C. C. Lin, K. B. Low, B. Magasanik, W. S. Reznikoff, M. Riley, M. Schaechter, and H. E. Umbarger (ed.), Escherichia coli and Salmonella: cellular and molecular biology. ASM Press, Washington, D.C. 31. Peters-Wendisch, P., R. Netzer, L. Eggeling, and H. Sahm. 2002. 3-Phosphoglycerate dehydrogenase from Corynebacterium glutamicum: the C-terminal domain is not essential for activity but is required for inhibition by L-serine. Appl. Microbiol. Biotechnol. 60:437–441. 32. Radmacher, E., A. Vaitsikova, U. Burger, K. Krumbach, H. Sahm, and L. Eggeling. 2002. Linking central metabolism with increased pathway flux: L-valine accumulation by Corynebacterium glutamicum. Appl. Environ. Microbiol. 68:2246–2250.

3268

¨ SER ET AL HU

33. Reyes, O., A. Guyonvarch, C. Bonamy, V. Salti, F. David, and G. Leblon. 1991. ‘Integron’-bearing vectors: a method suitable for stable chromosomal integration in highly restrictive corynebacteria. Gene 107:61–68. 34. Rieping, M., and T. Hermann. 11 January 2002. A process for the fermentative preparation of D-pantothenic acid and/or its salts. International patent application WO 02/072838 A2. 35. Ru ¨ckert, C., A. Pu ¨hler, and J. Kalinowski. 2003. Genome-wide analysis of the L-methionine biosynthetic pathway in Corynebacterium glutamicum by targeted gene deletion and homologous complementation. J. Biotechnol. 104:213–228. 36. Sahm, H., and L. Eggeling. 1999. D-Pantothenate synthesis in Corynebacterium glutamicum and use of panBC and genes encoding L-valine synthesis for D-pantothenate overproduction. Appl. Environ. Microbiol. 65:1973–1979. 37. Sahm, H., L. Eggeling, and A. A. de Graaf. 2000. Pathway analysis and metabolic engineering in Corynebacterium glutamicum. Biol. Chem. 381: 899–910. 38. Sambrook, J., E. F. Fritsch, and T. Maniatis. 1989. Molecular cloning: a laboratory manual, 2nd ed. Cold Spring Harbor Laboratory, Cold Spring Harbor, N.Y. 39. Scha ¨fer, A., A. Tauch, W. Ja ¨ger, J. Kalinowski, G. Thierbach, and A. Pu ¨hler. 1994. Small mobilizable multi-purpose cloning vectors derived from the Escherichia coli plasmids pK18 and pK19: selection of defined deletions in the chromosome of Corynebacterium glutamicum. Gene 145:69–73. 40. Schirch, L. V., C. M. Tatum, Jr., and S. J. Benkovic. 1977. Serine transhydroxymethylase: evidence for a sequential random mechanism. Biochemistry 16:410–419. 41. Schuegraf, A., S. Ratner, and R. C. Warner. 1960. Free energy changes of the argininosuccinate synthetase reaction and of the hydrolysis of the inner pyrophosphate bond of adenosine triphosphate. J. Biol. Chem. 235:3597– 3602. 42. Simic, P., J. Willuhn, H. Sahm, and L. Eggeling. 2002. Identification of glyA (encoding serine hydroxymethyltransferase) and its use together with the exporter ThrE to increase L-threonine accumulation by Corynebacterium glutamicum. Appl. Environ. Microbiol. 68:3321–3327. 43. So ¨hling, B., and G. Gottschalk. 1996. Molecular analysis of the anaerobic


44.

45.

46.

47. 48.

49. 50. 51.

52.

53.

succinate degradation pathway in Clostridium kluyveri. J. Bacteriol. 178: 871–880. Song, W. J., and S. Jackowski. 1992. Cloning, sequencing, and expression of the pantothenate kinase (coaA) gene of Escherichia coli. J. Bacteriol. 174: 6411–6417. Staufer, G. V. 1996. Biosynthesis of serine, glycine and one-carbon units, p. 506–513. In F. C. Neidhardt, R. Curtiss III, J. L. Ingraham, E. C. C. Lin, K. B. Low, B. Magasanik, W. S. Reznikoff, M. Riley, M. Schaechter, and H. E. Umbarger (ed.), Escherichia coli and Salmonella: cellular and molecular biology. ASM Press, Washington, D.C. Tauch, A., S. Go ¨tker, A. Pu ¨hler, J. Kalinowski, and G. Thierbach. 2002. The alanine racemase gene alr is an alternative to antibiotic resistance genes in cloning systems for industrial Corynebacterium glutamicum strains. J. Biotechnol. 99:79–91. Tilton, G., J. Shockey, and J. Browse. 2000. Two families of acyl-CoA thioesterases in Arabidopsis. Biochem. Soc. Trans. 28:946–947. van der Rest, M. E., C. Lange, and D. Molenaar. 1999. A heat shock following electroporation induces highly efficient transformation of Corynebacterium glutamicum with xenogeneic DNA. Appl. Microbiol. Biotechnol. 52:541–545. Ward, Jr., J. H. 1963. Hierarchical grouping to optimize an objective function. J. Am. Stat. Assoc. 58:236–244. Wilson, E. M., and E. E. Snell. 1962. Metabolism of ␣-methylserine. I. ␣-Methylserine hydroxymethyltransferase. J. Biol. Chem. 237:3171–3179. Yang, Y. H., S. Dudoit, P. Luu, D. M. Lin, V. Peng, J. Nagai, and T. P. Speed. 2002. Normalization for cDNA microarray data: a robust composite method addressing single and multiple slide systematic variations. Nucleic Acids Res. 30:e15. Zheng, R., and J. S. Blanchard. 2001. Steady-state and pre-steady-state kinetic analysis of Mycobacterium tuberculosis pantothenate synthetase. Biochemistry 40:12904–12912. Zheng, Z., Q. Gong, T. Liu, Y. Deng, J.-C. Chen, and G.-Q. Chen. 2004. Thioesterase II of Escherichia coli plays an important role in 3-hydroxydecanoic acid production. Appl. Environ. Microbiol. 70:3807–3813.