Introduction

l-lysine is one of the essential amino acids in mammals, which have to intake l-lysine through their diet as they lack l-lysine biosynthesis. l-lysine is produced biotechnologically to meet the annual world market of about 2 million tons (Hermann 2003; Wendisch 2007; Yukawa and Inui 2013). Glucose-based fermentations with C. glutamicum are in place for efficient l-lysine production (Wendisch et al. 2006a, b; Hirasawa et al. 2012; Blombach et al. 2009b, Heider and Wendisch 2015). This microorganism was initially isolated as a natural l-glutamate producer (Abe et al. 1967). C. glutamicum is a non-pathogenic, facultative anaerobic Gram-positive bacterium found in soil, animal feces, fruits, and vegetables with an irregular rod shape (Eggeling and Bott 2005). The physiology, biochemistry, and genetics of l-lysine biosynthesis are well understood and have been employed in rational development of l-lysine producing strains. For example, the l-lysine production strain GRLys1 (Unthan et al. 2015) possesses two gene copies of the key enzyme of l-lysine biosynthesis, aspartokinase (Schrumpf et al. 1992; Cremer et al. 1991; Kalinowski et al. 1991), devoid of feedback inhibition by l-lysine (encoded by lysC T311I), two copies of the l-lysine biosynthesis genes asd, dapA, dapB, ddh, and lysA, and two copies of l-lysine export gene lysE (Vrljic et al. 1996). Side product formation is reduced due to a restricted homoserine dehydrogenase (encoded by hom V59A) (Eikmanns et al. 1991), anaplerosis is improved due to deletion of the PEP carboxykinase gene (Riedel et al. 2001), and a variant of the pyruvate carboxylase gene (Peters-Wendisch et al. 2001) and the strain was cured of prophages CGP1, CGP2, and CGP3 (Baumgart et al. 2013). A wealth of knowledge on l-lysine biosynthesis and production has accumulated (Wendisch et al. 2006a, b; Hirasawa et al. 2012; Blombach et al. 2009a; Heider and Wendisch 2015; Becker and Wittmann 2012; Wendisch 2014; Mitsuhashi 2014; Park et al. 2010).

C. glutamicum takes up and phosphorylates glucose primarily via the phosphoenolpyruvate (PEP)-dependent carbohydrate phosphotransferase system (PTS). In addition, glucose can be imported via the non-phosphorylating inositol permeases IolT1 and IolT2 (Lindner et al. 2011; Ikeda et al. 2011; Bäumchen et al. 2009). Intracellular glucose may be phosphorylated by ATP-dependent or polyphosphate-dependent glucokinases PpgK and Glk (Park et al. 2000; Lindner et al. 2010). While the PTS constitutes a high-affinity uptake system (Km of 14 μM), the permeases IolT1 and IolT2 have about hundred fold lower affinities for glucose (Km values of 2.8 mM and 1.9 mM, respectively). As the ATP- or polyphosphate-dependent glucokinases Glk and PpgK do not consume PEP, the availability of PEP for anabolism e.g., for biosynthesis of aromatic amino acids may increase (Tang et al. 2013; Zhang et al. 2015). Since high initial glucose concentrations are typical for biotechnological processes, use of the glucose permease and glucokinase pathway offers an interesting opportunity to enhance glucose utilization while maintaining high PEP availability. Commensurate with this notion, PpgK was shown to be important only at high glucose concentrations (Lindner et al. 2010). Glucose utilization by C. glutamicum is subject to transcriptional control (Toyoda et al. 2008, 2009a; Teramoto et al. 2011; Park et al. 2010; Gerstmeir et al. 2004). For example, SugR is a general repressor of the PTS, glycolysis and fermentative l-lactate dehydrogenase genes in C. glutamicum (Engels and Wendisch 2007; Engels et al. 2008; Teramoto et al. 2011; Gaigalat et al. 2007). Deletion of sugR accelerated glucose utilization (Blombach et al. 2009b; Bartek et al. 2010; Teramoto et al. 2011), while overexpression of sugR slowed glucose utilization (Engels et al. 2008). Due to co-utilization of protocatechuic acid, single cells of C. glutamicum grow faster with glucose in the highly diluted environments of a Lab Chip than cells in suspension (Unthan et al. 2014). Taken together, glucose uptake and utilization by C. glutamicum wild type as well as by l-lysine producing strains may be relevant targets for strain development.

l-lysine can be converted to other value-added compounds such as the diamine cadaverine (Mimitsuka et al. 2007; Tateno et al. 2009; Kind et al. 2014; Oh et al. 2015; Shimizu 2013; Leßmeier et al. 2015), 5-aminovalerate (5AVA) (Park et al. 2013) and l-PA (Fujii et al. 2002; Gatto et al. 2006; Neshich et al. 2013). C. glutamicum has been engineered for cadaverine production but not for production of l-PA. The non-proteinogenic amino acid l-PA is a precursor of immunosuppressants (Maddess et al. 2008), peptide antibiotics (Takayama et al. 1984), or piperidine alkaloids (Clevenstine et al. 1979). In plants, l-PA is a critical regulator of inducible plant immunity, mediating defense amplification and priming (Návarová et al. 2012). Lysine degradation yields l-PA as intermediate. dl-lysine and l-lysine are oxidatively deaminated by stereospecific l-lysine α-oxidases from Pseudomonas putida (Miller and Rodwell 1971) and Trichoderma viride (Kusakabe et al. 1980), respectively, to α-keto-ε-aminocaproate, which spontaneously dehydrates to Δ1-piperideine-2-carboxylate (P2C). P2C reductase yields l-PA from P2C (Payton and Chang 1982). As shown in Fig. 1, l-PA may also be derived from Δ1-piperideine-6-carboxylate (P6C) in an NADPH-dependent reduction reaction catalyzed by pyrroline 5-carboxylate reductase ProC (Fujii et al. 2002), an enzyme of l-proline biosynthesis in many bacteria including C. glutamicum (Ankri et al. 1996). P6C arises from α-aminoadipic semialdehyde (AASA) by spontaneous cyclization (Fig. 1). Four reactions give rise to AASA from l-lysine: (1) l-lysine cyclodeaminase (LCD) from streptomycetes such as S. hygroscopicus that produce polyketide immunosuppressants such as rapamycin or meridiamycin (Molnár et al. 1996); (2) l-lysine 6-aminotransferase (LAT) from Flavobacterium lutencens (Soda et al. 1968); and (3) l-lysine 6-dehydrogenase e.g., from Agrobacterium tumefaciens (Misono et al. 1989) or from Silicibacter pomeroyi (Neshich et al. 2013) (Fig. 1). Furthermore, l-lysine can be converted to AASA via saccharopine by l-lysine α-ketoglutarate reductase and saccharopine dehydrogenase from Rizoctonia leguminicola (Wickwire et al. 1990). l-PA accumulated intracellularly when Escherichia coli overproducing l-lysine 6-dehydrogenase from S. pomeroyi was grown in l-lysine-containing media (Neshich et al. 2013). l-PA was produced from l-lysine in biotransformations with E. coli overproducing either l-lysine α-oxidase from Scomber japonicus (Tani et al. 2015) or LAT (Fujii et al. 2002).

Fig. 1
figure 1

Metabolic pathway for conversion of l-lysine to l-PA in recombinant C. glutamicum. l-lysine 6-dehydrogenase (deaminating) (EC 1.4.1.18) from S. pomeroyi encoded by lysDH and pyrroline 5-carboxylate reductase (EC 1.2.1.41) from C. glutamicum encoded by proC were used. The combined reactions of l-lysine 6-dehydrogenase (EC 1.4.1.18) and the spontaneous cyclization of α-aminoadipic semialdehyde are sometimes referred to as l-lysine:NAD+ oxidoreductase (deaminating, cyclizing) (EC 1.4.1.15)

Full size image

In this study, l-lysine production from glucose was accelerated by overexpression of endogenous iolT2 encoding inositol permease and by heterologous expression of the glucokinase gene from B. subtilis. In addition, deletion of sugR led to derepression of glycolytic and PTS genes and deletion of ldhA precluded formation of l-lactate as by-product. This strain was used as basis for strain development for fermentative production of l-PA from glucose.

Materials and methods

Bacterial strains, plasmids, and growth conditions

The bacterial strains and plasmids are listed in Table 1. E. coli DH5α (Hanahan 1983) was used for vector construction. The precultivation of C. glutamicum and cultivation of E. coli were carried out in Luria-Bertani (LB) medium (Sambrook et al. 1989) inoculated from a fresh LB agar plate. For growth of C. glutamicum strains, the cells from the preculture were washed in CGXII minimal medium (Eggeling and Bott 2005) without glucose, before 50 ml of main CGXII culture with glucose was inoculated to an optical density of 0.5 (OD600). C. glutamicum precultures and cultures were incubated at 30 °C and 120 rpm in 500 ml baffled flasks. E. coli cultures were incubated at 37 °C and 200 rpm in 500 ml baffled flasks. The final glucose concentration varied between 0.5 % (w/v) and 4 % (w/v). For selection of pVWEx1 and pSH1, 25 μg ml−1 of kanamycin was added to cultures (Peters-Wendisch et al. 2001). For selection of pEKEx3, 100 μg ml−1 of spectinomycin was added to cultures (Stansen et al. 2005).

Table 1 Strains and plasmids used in this work
Full size table

Construction of expression vectors

Genes were amplified via PCR from genomic DNA of C. glutamicum ATCC 13032 (Kalinowski et al. 2003), Mycobacterium smegmatis mc2 155 (Pimentel-Schmitt et al. 2009), S. coelicolor A3(2) (van Wezel et al. 2005; Imriskova et al. 2005), E. coli K-12 MG 1655 (Meyer et al. 1997), B. subtilis subsp. subtilis str. 168 (Mesak et al. 2004), or S. pomeroyi DSS-3 (DSMZ-15171) (González et al. 2003). Genomic DNA was prepared following the same protocol described previously (Eikmanns et al. 1995). The primers for the vector constructions are listed in Table 2.

Table 2 List of primers used in this work
Full size table

Balanced expression of glucose permease and kinase genes

For the balanced expression of glucose permease and kinase genes, a combinatorial approach was performed by combining four ribosomal binding site (AAAGGAGG, AAAGGTGG, AAAGGCGG, and AAAGGGGG) and two transcriptional start site (ATG and GTG) options for the permease genes cloned together with the glucokinase gene, which was not varied. Primers pgk13, pgk14, pgk15, pgk16, pk35, and pk36 (s. Table 2) were used for PCR, and genes were joined via Gibson assembly to yield pEKEx3-glcK Bs -glcP Sco and pEKEx3-glcK Bs -iolT2 Cg vectors (Gibson 2011). These vectors were used to transform C. glutamicum Δhpr, and the best combinations were chosen according to fastest growth in minimal medium with 2 % (w/v) of glucose and different IPTG concentrations. IPTG concentrations of 100 μM were optimal for (pEKEx3-glcK Bs -glcP Sco) and 50 μM for (pEKEx3-glcK Bs-iolT2 Cg).

Deletion of sugR and ldhA

The two-step homologous recombination procedure was used to generate in-frame deletions of ΔsugR and ΔldhA in the strain GRLys1 (Rittmann et al. 2003). For that purpose, the vectors pK19mobsacB-ΔsugR (Engels et al. 2008) and pK19mobsacB-ΔldhA (Blombach et al. 2011) were used as described previously (Engels and Wendisch 2007). The deletions were verified by PCR using the primer pairs ΔsugR-Ver-fw/ΔsugR-Ver-rv and ΔldhA-Ver-fw/ΔldhAR-Ver-rv (Table 2).

Analytical procedures

Extracellular amino acids and carbohydrates were quantified using a high-pressure liquid chromatography system (1200 series, Agilent Technologies Deutschland GmbH, Böblingen, Germany). Cell cultures were centrifuged (13,000×g, 10 min), and the supernatant was used for analysis. For the detection of l-lysine, l-glutamate, l-valine and l-alanine, samples were derivatized with ortho-phthaldialdehyde and for detection of l-PA with fluorenylmethyl chloroformate (FMOC) as described (Schneider and Wendisch 2010). All amino acids were separated on a system consisting of a pre-column (LiChrospher 100 RP18 EC-5μ (40 × 4 mm), CS Chromatographie Service GmbH, Langerwehe, Germany) and a main column (LiChrospher 100 RP18 EC-5 μ (125 × 4 mm), CS Chromatographie Service GmbH), and detected with a fluorescence detector (FLD G1321A, 1200 series, Agilent Technologies). l-Asparagine was used as internal standard. For quantification of carbohydrates, a column for organic acids (300 × 8 mm, 10 μm particle size, 25 Å pore diameter, CS Chromatographie Service GmbH) and a refractive index detector (RID G1362A, 1200 series, Agilent Technologies) was used (Peters-Wendisch et al. 2014).

Transcriptome analysis using DNA microarrays

To identify gene expression changes due to the addition of l-PA to the growth medium, C. glutamicum WT was cultivated in CGXII minimal medium with 4 % (w/v) glucose supplemented with either 200 mM l-PA or 200 mM sodium chloride. Exponentially growing cells were harvested by centrifugation (4000 × g, 10 min, 4 °C) and kept at −80 °C. RNA isolation was performed as described (Wendisch 2003). DNA microarray analysis, synthesis of fluorescently labeled cDNA from total RNA, DNA microarray hybridization, and gene expression analysis were performed as described previously (Ishige et al. 2003; Lange et al. 2003). The data are available as Gene Expression Omnibus GSE79690 data set at http://www.ncbi.nlm.nih.gov/geo/.

Results

Potential of alternative glucose permeases to substitute for the glucose PTS

C. glutamicum not only takes up and phosphorylates glucose mainly via the high-affinity (Km of 14 μM) phosphotransferase system (PTS) but also can use inositol permeases IolT1 and IolT2, which have about hundred fold lower affinities for glucose and require subsequent phosphorylation by glucokinases (Lindner et al. 2011; Ikeda et al. 2015). The latter systems may be important at the high glucose concentrations employed in industrial fermentations. In order to test if uptake and phosphorylation via glucose permeases and glucokinase can fully substitute for the glucose PTS at medium to high glucose concentrations (1 % (w/v) to 4 % (w/v)), glucose permease and glucokinase genes from various organisms were expressed in the PTS-deficient C. glutamicum strain Δhpr (Lindner et al. 2011). The heat-stable protein HPr is an essential component of the PTS system, which is phosphorylated by PTS component EI and, subsequently, phospho-HPr transfers the phosphoryl group to the PTS permease component. The genes for the glucose permease from S. coelicolor A3(2) (GlcPSco) with a Km of 41 μM (van Wezel et al. 2005), the glucose permease from M. smegmatis mc2 155 (GlcPSmeg) with a Km of 19.4 μM (Pimentel-Schmitt et al. 2009), the galactose permease from E. coli MG1655 (GalPEco) with a Km of 10.2 μM (McDonald et al. 1997), and the inositol transporter 2 from C. glutamicum ATCC 13032 (IolT2Cg) with a Km of 1.9 mM (Lindner et al. 2011) were cloned into the IPTG-inducible expression vector pVWEx1 (Peters-Wendisch et al. 2001) and used to transform C. glutamicum Δhpr. In minimal medium with 4 % (w/v) of glucose, C. glutamicum WT (pVWEx1) grew with a growth rate of 0.42 ± 0.01 h−1 (Fig. 2), while C. glutamicum Δhpr(pVWEx1) hardly grew (0.03 ± 0.01 h−1), which is due to low expression of iolT1 and iolT2 in the absence of inositol (Krings et al. 2006). Of the glucose permease genes tested, overexpression of iolT2 Cg supported fastest growth (μ = 0.27 ± 0.01 h−1); however, even after variation of the IPTG concentration used for induction, the wild-type growth rate could not be restored. Thus, overexpression of glucose permease alone was not sufficient to substitute for the PTS in C. glutamicum Δhpr.

Fig. 2
figure 2

Influence of sugar permease gene overexpression on growth of C. glutamicum Δhpr in glucose minimal medium. Growth rates in minimal medium with 4 % (w/v) glucose at the indicated IPTG concentrations are given for strains WT(pVWEx1), Δhpr(pVWEx1), Δhpr(pVWEx1-glcP Sco), Δhpr(pVWEx1-glcP Smeg), Δhpr(pVWEx1-galP Eco), and Δhpr(pVWEx1-iolT Cg) as means and standard deviations of three replicates

Full size image

Combined overexpression of glucokinase and glucose permease genes

C. glutamicum possesses ATP-dependent glucokinase Glk (Park et al. 2000) and polyphosphate/ATP-dependent glucokinase PpgK with a Km for glucose of 1 mM (Lindner et al. 2010). Combined overexpression of ppgK with either iolT1 or iolT2 in a PTS-negative strain led to slightly slower glucose utilization rates than in the parent strain (Lindner et al. 2011). Thus, combinations of alternative glucokinases with glucose permeases were tested for their potential to fully compensate for the lack of the PTS. Genes encoding glucokinase from E. coli K-12 MG1655 (GlkEco) with Km of 0.78 mM (Meyer et al. 1997), the glucokinase from S. coelicolor A3(2) (GlkSco) with Km of 1.4 mM (Imriskova et al. 2005), the glucokinase from B. subtilis subsp. Subtilis str. 168 (GlcKBs) with Km of 0.24 mM (Mesak et al. 2004), the glucokinase (GlkCg) and the polyphosphate (PolyP)/ATP-dependent glucokinase from C. glutamicum (PpgKCg) with Km of 1 mM (Lindner et al. 2010) were cloned into the IPTG-inducible expression vector pEKEx3 (Stansen et al. 2005). The resulting vectors were used to transform C. glutamicum Δhpr(pVWEx1- iolT2 Cg), Δhpr(pVWEx1-galP Eco), and Δhpr(pVWEx1-glcP Sco), respectively. All strains were grown in CGXII minimal medium with 4 % (w/v) glucose (Fig. 3). The fastest growing strains were also cultivated at lower glucose concentrations (Fig. 3). At 1 % (w/v) and 0.5 % (w/v) glucose, none of the combinations allowed growth as fast as the wild-type control. With 4 % (w/v) glucose, the combination of IolT2Cg and GlcKBs allowed faster growth (μ = 0.35 h−1) than the wild-type control (μ = 0.32 h−1). With 2 % (w/v) glucose, the combination of GlcPSco/GlcKBs grew faster (μ = 0.37 h−1) than the wild-type control (μ = 0.32 h−1).

Fig. 3
figure 3

Influence of sugar permease and glucokinase gene overexpression on growth of C. glutamicum Δhpr in glucose minimal medium. Growth rates in minimal medium with the indicated glucose concentration (4 %, 2 %, 1 % or 0.5 % (w/v)) are given as means of three replicates. The strain C. glutamicum WT(pVWEx1)(pEKEx3) was used as positive control (labeled WT in the figure). The strain Δhpr(pVWEx1)(pEKEx3) was used as negative control (labeled Δhpr in the figure). Combinations of the indicated glucokinases and permeases were tested in Δhpr with permease genes expressed from plasmid pVWEx1 and glucokinase genes expressed from plasmid pEKEx3. A growth rate of zero is depicted if growth was not tested

Full size image

Varying translation initiation rates for balanced glucokinase and glucose permease gene expression

Since overexpression of genes for permeases and other membrane proteins often perturbs growth, various translation initiation sequences for expression of the permease genes glcP Sco and iolT2 Cg were tested for combined expression with glcK Bs in one vector. Expression of glcK Bs was kept constant, and translation initiation rates of the permease gene (glcP Sco or iolT2 Cg) were varied by modifying ribosomal binding sites (RBS) and the translational start codons (TLS). The pool of vectors carrying different RBS and TLS sequences was used to transform Δhpr, and 24 clones for each combination were assayed for growth in minimal medium containing 2 % (w/v) glucose. The fastest-growing strains were selected and the inserts of the vectors were sequenced. Vector pEKEx3-ScoBest carried the RBS sequence of AAAGGTGG and the TLS sequence of GTG (Fig. 4a) for expression of glcK Bs and glcP Sco, and vector pEKEx3-IolTBest carried the RBS sequence of AAAGGGGG and the TLS sequence of GTG (Fig. 4b) for expression of glcK Bs and iolT2 Cg.

Fig. 4
figure 4

Influence of varied ribosome binding site and translational start codon sequences of the permease gene glcP Sco (a) or iolT Cg (b) for expression in combination with glucokinase gene glcK Bs on growth in minimal medium with 2 % (w/v) glucose. Growth rates of 24 clones with varied glcP Sco (a) or iolT Cg (b) ribosome binding site and translational start codon sequences expressed in strain Δhpr in combination with glcK Bs are indicated. The translational start codon (underlined) and ribosome binding site (in italics) sequences for the fastest growing clones (highlighted in dark gray) are depicted, and differences to the consensus ribosome binding site are given in bold. The respective plasmids were named pEKEx3-ScoBest and pEKEx3-IolTBest, respectively

Full size image

To assay the effect of vectors pEKEx3-ScoBest and pEKEx3-IolTBest on glucose utilization by PTS-positive C. glutamicum, both vectors were used to transform C. glutamicum WT and growth, glucose consumption, and by-products formation in minimal medium with 2 % (w/v) glucose were compared to the empty vector carrying control strain (Table 3). Under these conditions, both WT(pEKEx3-ScoBest) and WT(pEKEx3-IolTBest) grew faster than the control strain with growth rates of 0.35 and 0.37 h−1, respectively, which were 9 and 15 % higher than that of WT(pEKEx3) (μ = 0.32 h−1). l-Lactate formation increased by about 30 % for WT(pEKEx3-ScoBest) and by about 50 % for WT(pEKEx3-IolTBest) as compared to WT(pEKEx3) (about 6 mM; Table 3 (A)). Since vectors pEKEx3-ScoBest and pEKEx3-IolTBest accelerated growth in glucose minimal medium by PTS-positive (WT) and PTS-negative (Δhpr) strains, both were used in subsequent experiments with the l-lysine-producing C. glutamicum strain GRLys1.

Table 3 Growth parameters of C. glutamicum strains (A) WT(pEKEx3), WT(pEKEx3-ScoBest), and WT(pEKEx3-IolTbest) and (B) WT(pVWEx1), WT(pVWEx1-ScoBest) and WT(pVWEx1-IolTbest) in minimal medium with 2 % (w/v) of glucose
Full size table

However, C. glutamicum WT transformed with the empty vector pEKEx3 (replication origin of pBL1, SpecR) grew slower than WT carrying the empty vector pVWEx1 (replication origin of pHM1519, KanR) reflecting different burdens by these plasmids. To exclude vector specific effects, vectors pVWEx1-IolTBest and pVWEx1-ScoBest were constructed and used to transform C. glutamicum WT. In line with the results obtained with pEKEx3-ScoBest and pEKEx3-IolTBest (Table 3 (A)), strains WT(pVWEx1-IolTBest) and WT(pVWEx1-ScoBest) grew faster in glucose minimal medium and showed improved glucose consumption rates as compared to WT(pVWEx1) (Table 3 (B) and Fig. S1).

Engineering C. glutamicum strain GRLys1 for increased l-lysine productivity

The l-lysine-producing strain GRLys1 was transformed with vectors pEKEx3, pEKEx3-ScoBest, and pEKEx3-IolTBest, and growth, Yx/s, lactate, and l-lysine production were monitored in 2 % (w/v) glucose minimal medium. GRLys1(pEKEx3) showed a growth rate of 0.30 ± 0.01 h−1, while GRLys1 (pEKEx3-ScoBest) and GRLys1 (pEKEx3-IolTBest) showed growth rates of 0.34 ± 0.01 and 0.35 ± 0.01 h−1, respectively (Fig. 5). However, the l-lysine titer decreased in comparison to GRLys1(pEKEx3) (21 ± 1 mM) as 20 ± 1 mM l-lysine was produced by GRLys1(pEKEx3-ScoBest) and 18 ± 1 mM by GRLys1(pEKEx3-IolTBest) (Fig. 5). Thus, faster growth reduced l-lysine formation. A similar trend is often observed in l-lysine strain development, i.e., an increase of the l-lysine titer is often accompanied by a reduction of the growth rate.

Fig. 5
figure 5

Growth and l-lysine production of various strains derived from C. glutamicum GRLys1. Growth rates, biomass yields (Yx/s), l-lactate titers, l-lysine titers, l-lysine volumetric productivities, and l-lysine-specific productivities are given as means and standard deviations of three replicates. The parameters of the control strain GRLys1(pEKEx3) carrying empty vector pEKEx3 are highlighted in black. The parameters of the control strain GRLys1(pVWEx1) carrying empty vector pVWEx1 are highlighted in dark gray. Parameters obtained for the strains that carry pEKEx3-ScoBest or pEKEx3-IolTBest are shown in white, while data for strains that carry pVWEx1-ScoBest or pEKEx3-IolTBest are shown in light gray. nd not detected

Full size image

To test if relieving glycolytic genes from transcriptional repression improved growth rates and l-lysine titers, the repressor gene sugR was deleted because deletion of sugR in C. glutamicum has been shown previously to increase expression of glycolysis genes and glucose utilization (Engels et al. 2008; Toyoda et al. 2008). The resulting strain GRLys1ΔsugR was transformed with the vectors pEKEx3-ScoBest, pEKEx3-IolTBest, as well as the empty vector pEKEx3. The absence of sugR led to higher titer and volumetric productivities of 30 ± 1 mM and 0.19 ± 0.01 g L−1 h−1, respectively, as compared to GRLys1(pEKEx3) (Fig. 5). However, deletion of sugR reduced the growth rate to 0.20 ± 0.01 h−1 (Fig. 5).

As consequence of overexpression of glcP Sco and glcK Bs in GRLys1ΔsugR, the resulting strain GRLys1ΔsugR(pEKEx3-ScoBest) grew faster (μ = 0.27 ± 0.01 h−1), but produced less l-lysine (25 ± 1 mM), which resulted in a volumetric productivity (0.19 ± 0.01 g L−1 h−1) comparable to the empty vector control GRLys1ΔsugR(pEKEx3) (Fig. 5). On the other hand, overexpression of iolT2 Cg and glcK Bs in GRLys1ΔsugR improved l-lysine titer by 57 % to 33 ± 1 mM and volumetric productivity by 22 % to 0.22 ± 0.01 g L−1 h−1 as compared to GRLys1(pEKEx3) (Fig. 5). Taken together, C. glutamicum GRLys1ΔsugR(pEKEx3-IolTBest) showed the fastest l-lysine production of the strains constructed here.

C. glutamicum GRLys1ΔsugR(pEKEx3-IolTBest) grew to lower biomass concentrations (3.2 ± 0.1 gCDW L−1) than GRLys1(pEKEx3) (5.3 ± 0.1 gCDW L−1) and produced more lactate (27 ± 2 mM) as by-product than GRLys1(pEKEx3) (4 ± 1 mM). Increased lactate formation is due to deletion of sugR and was also observed with C. glutamicum GRLys1ΔsugR(pEKEx3) that accumulated 22 ± 1 mM lactate. Since l-lactate is produced from pyruvate by NAD-dependent l-lactate dehydrogenase LdhA (Toyoda et al. 2009b), ldhA was deleted in strain GRLys1ΔsugR(pEKEx3-IolTBest). The resulting strain GRLys1ΔsugRΔldhA(pEKEx3-IolTBest) did not accumulate l-lactate but produced l-lysine with a lower titer (31 ± 1 mM) and a lower volumetric productivity (0.21 ± 0.01 mM g L−1 h−1) than the parental LdhA-positive strain (Fig. 5). Since GRLys1ΔsugRΔldhA(pEKEx3-IolTBest) and GRLys1ΔsugR(pEKEx3-IolTBest) grew to biomass concentrations of 3.2 ± 0.1 and 3.7 ± 0.1 gCDW L−1, respectively, deletion of ldhA improved biomass formation rather than l-lysine production. To exclude vector-specific effects, pVWEx1-based vectors were also tested (Fig. S1). C. glutamicum GRLys1(pVWEx1) showed comparable growth rate, biomass yield, l-lysine titer, volumetric and specific productivity as GRLys1(pEKEx3) (Fig. 5). Moreover, the production parameters obtained with strains GRLys1ΔsugR(pVWEx1-IolTBest) and GRLys1ΔsugRΔldhA(pVWEx1-IolTBest), respectively, were comparable to those of GRLys1ΔsugR(pEKEx3-IolTBest) and GRLys1ΔsugRΔldhA(pEKEx3-IolTBest) (Fig. 5).

Engineering C. glutamicum for production of l-PA

In order to test how accelerated l-lysine production can be translated into production of a l-lysine-derived value-added chemical, strains overproducing the non-proteinogenic amino acid l-PA, a precursor of immunosuppressants, peptide antibiotics, or piperidine alkaloids, were constructed. l-PA is not synthesized by C. glutamicum which is unable to catabolize l-lysine (Vrljic et al. 1996). However, l-PA was rapidly accumulated from the growth medium by corynebacteria but did not accumulate under hyperosmolar conditions (Frings et al. 1993). To determine the response of C. glutamicum to l-PA, the effect of extracellularly added l-PA on growth and global gene expression was analyzed. Growth of C. glutamicum WT in 4 % (w/v) glucose minimal medium was hardly affected by addition of l-PA, and an inhibition constant (K i) of about 3.5 M was extrapolated (Fig. 6a). The gene expression response to 200 mM l-PA added to the medium (Table 4) revealed increased expression of genes of carbon metabolism such as rpe (ribulose-5-phosphate-3-epimerase), pck (phosphoenolpyruvate carboxykinase), vanR (repressor of the vanillate utilization operon), and cdaS (cyclomaltodextrinase). On the other hand, malP coding for maltodextrin phosphorylase and lldD encoding quinone-dependent l-lactate dehydrogenases for l-lactate utilization (Stansen et al. 2005) showed decreased expression. The gene cg0569 encoding a cation transporting P-type ATPase, one of the 50 most abundant membrane proteins in C. glutamicum (Burkovski 2008), was highly expressed in the presence of l-PA. Neither the gene expression response nor the growth response to l-PA indicated that product inhibition may limit l-PA production by C. glutamicum.

Fig. 6
figure 6

Influence of l-PA on growth of C. glutamicum WT (a) and l-lysine and l-PA titers produced by various strains derived from C. glutamicum GRLys1 (b). a l-PA was added to the medium before inoculation. b Means and standard deviations of three replicates are given

Full size image
Table 4 Genes differentially expressed in C. glutamicum WT grown in glucose minimal medium in the presence of 200 mM l-PA as compared to sodium chloride
Full size table

A synthetic pathway for conversion of l-lysine to l-PA was employed (Fig. 1). l-Lysine can be oxidatively deaminated by l-lysine 6-dehydrogenase to yield α-aminoadipic semialdehyde, which after spontaneous cyclization to 1-piperidine 6-carboxylic acid can be reduced by pyrroline 5-carboxylate reductase to l-PA. While C. glutamicum possesses pyrroline 5-carboxylate reductase ProC (Ankri et al. 1996), corynebacterial genomes lack genes encoding l-lysine 6-dehydrogenase (Neshich et al. 2013). Accordingly, a vector for overexpression of endogenous proC and for expression of heterologous l-lysine 6-dehydrogenase gene lysDH from S. pomeroyi under the control of the constitutive C. glutamicum promoter pTuf was constructed and used to transform GRLys1(pEKEx3), GRLys1ΔsugR(pEKEx3-IolTBest), and GRLys1ΔsugRΔldhA(pEKEx3-IolTBest). While the empty vector carrying control strain GRLys1(pEKEx3)(pSH1) produced l-lysine but not l-PA in 2 % (w/v) glucose minimal medium, the strains carrying pSH1-lysDH-proC accumulated l-PA besides l-lysine after 48 h (Fig. 6b).

Notably, the four strains accumulated l-lysine to comparable concentrations (about 20 mM), and l-PA was produced in addition to l-lysine by strains GRLys1(pEKEx3)(pSH1-lysDH-proC), GRLys1ΔsugR(pEKEx3-IolTBest)(pSH1-lysDH-proC) and GRLys1ΔsugRΔldhA(pEKEx3-IolTBest)(pSH1-lysDH-proC) (Fig. 6b). This result may indicate that conversion of l-lysine to l-PA lowers the intracellular l-lysine concentration below a threshold relevant for genetic or allosteric regulation by l-lysine, as a similar metabolic pull effect was observed for the production of cadaverine employing l-lysine decarboxylase in Bacillus methanolicus (Naerdal et al. 2015). Indeed, the presence of pSH1-lysDH-proC in C. glutamicum WT(pEKEx3) reduced the growth rate from 0.32 ± 0.01 h−1 observed for WT(pEKEx3)(pSH1) to 0.17 ± 0.01 h−1 observed for WT(pSH1-lysDH-proC)(pEKEx3) (data not shown). While under these conditions l-PA was not produced, the availability of l-lysine for growth may have been perturbed.

The l-PA product titers achieved with GRLys1(pEKEx3)(pSH1-lysDH-proC), GRLys1ΔsugR(pEKEx3-IolTBest)(pSH1-lysDH-proC), and GRLys1ΔsugRΔldhA(pEKEx3-IolTBest)(pSH1-lysDH-proC), respectively, were 8 ± 1, 11 ± 1, and 14 ± 1 mM, respectively. Thus, while l-lysine could not be converted to l-PA completely, it is possible to produce l-PA in the grams per liter range using recombinant C. glutamicum.

Discussion

l-Lysine productivity of C. glutamicum on glucose media was improved here (a) by overexpressing genes for uptake and phosphorylation of glucose in addition to the endogenous PTS, (b) by relieving glycolysis genes from SugR-mediated repression, and (c) avoiding formation of l-lactate as by-product through disruption of ldhA. Subsequently, l-PA production was established.

In nature, C. glutamicum rarely encounters the high glucose concentrations prevalent in the early phases of biotechnological batch and fed-batch fermentations. While a high-affinity system for glucose utilization such as the PTS is favorable under most natural conditions, its capacity may limit glucose utilization under conditions of industrial l-lysine production. In line with this notion, replacing the endogenous PTS under low glucose conditions was achieved with the high-affinity glucose permease of S. coelicolor A3(2) (GlcPSco), which constitutes the major glucose uptake system of this organism (van Wezel et al. 2005). Intracellular glucose was phosphorylated by GlcK from B. subtilis, a high-affinity glucokinase with a Km for glucose of 0.24 mM (Mesak et al. 2004). Coupling of iolT2 overexpression with overexpression of endogenous glk or ppgK was less efficient (Fig. 3), which may be explained by lower affinities of the endogenous glucokinases and commensurates with earlier findings that e.g., deletion of ppgK only entails a growth disadvantage at high glucose concentrations (Lindner et al. 2010).

At 4 % (w/v) glucose, however, overexpression of the low affinity permease IolT2 in combination with overexpression of glcK from B. subtilis proved beneficial as a 10 % higher growth rate than the control strain C. glutamicum WT(pEKEx3) was reached (Fig. 3). This was accompanied by a higher specific glucose utilization rate (Qs), lower final biomass concentrations, and higher titers of l-lactate as by-product (Table 3). The problem of l-lactate formation was enhanced when sugR was deleted. This effect may be called overflow metabolism and is often observed when glycolysis capacity exceeds other parts of metabolism, in E. coli e.g., the TCA cycle (Veit et al. 2007) or the respiratory chain (Vemuri et al. 2006). This limitation may be due to transcriptional control by global regulators such as ArcA (Perrenoud and Sauer 2005). Overexpression of genes encoding glycolytic enzymes as applied in E. coli (Xie et al. 2014; Seol et al. 2015), B. subtilis (Liu et al. 2014), or S. coelicolor (Borodina et al. 2008) was also used for C. glutamicum, e.g., to improve the production of high value compounds like amino acids (Yamamoto et al. 2012; Reddy and Wendisch 2014), alcohols (Jojima et al. 2015; Yamamoto et al. 2013), or the organic acid d-lactate (Tsuge et al. 2015). Under these conditions or when a repressor gene of glycolysis genes was deleted, i.e., that of the DeoR-type transcriptional regulator SugR, high titers of l-lactate were formed as side effect of faster glucose utilization (Engels and Wendisch 2007; Engels et al. 2008; Teramoto et al. 2011). Likewise, sugR deletion accelerated l-valine production, but at the expense of increased l-lactate formation (Blombach et al. 2009a). Similarly, deletion of sugR in the l-lysine-producing strain GRLys1 increased l-lactate formation (22 ± 1 mM as compared to 4 ± 1 mM; Fig. 5). Consequently, the fermentative l-lactate dehydrogenase gene ldhA was deleted in GRLys1ΔsugR(pEKEx3-IolTBest) and l-lactate formation was abolished. This entailed more biomass formation rather than more l-lysine production since the final cell dry weight increased by 15 % while the l-lysine titer was slightly reduced. However, the volumetric l-lysine productivity remained almost unchanged (Fig. 5). Taken together, GRLys1ΔsugRΔldhA(pEKEx3-IolTBest) and GRLys1ΔsugRΔldhA(pVWEx1-IolTBest) are suitable strains for fast production of l-lysine and l-lysine derivatives from glucose without formation of l-lactate as by-product.

l-PA production from glucose using recombinant C. glutamicum strains was established here (Fig. 6). Chemical methods to obtain optically pure l-PA (and derivatives thereof) have been described and include diazotation of l-lysine or electrochemical oxidation of protected l-lysine, but also conversion of protected l-aspartic acid or l-glutamic acid derivatives to l-PA molecules substituted at the 4- or 5-position (Couty 1999). However, these reaction sequences are often long, e.g., 11 steps from l-glutamic acid to 4-oxopiperidine, and show low yields even for the one-step conversion of l-lysine to l-PA using disodium nitrosyl-pentacyanoferrate(II) as oxidant (Couty 1999). l-PA can be generated from racemic pipecolic acid via resolution by fractional crystallization of diastereomeric salts or via dynamic kinetic resolution by stereoselective amidases (Eichhorn et al. 1997), but it is difficult to achieve 100 % enantiomeric excess. In nature, l-PA occurs in l-lysine-catabolizing bacteria, e.g., in S. pomeroyi, and its biosynthesis is known to be induced in a l-lysine-dependent manner under high salt conditions (Neshich et al. 2013). l-PA accumulated intracellularly (to about 0.6 μg mg−1 protein) when E. coli BL21 expressing the l-lysine 6-dehydrogenase gene from S. pomeroyi was grown in l-lysine-containing medium (Neshich et al. 2013). Biotransformation of l-lysine to l-PA has been developed using E. coli overproducing LAT (about 3.9 g L−1 of optically pure l-PA (Fujii et al. 2002). Additional overexpression of the E. coli gene for lysine uptake, lysP, accelerated l-PA formation from l-lysine about five fold (Fujii et al. 2002). Biotransformation of racemic dl-lysine with E. coli overproducing l-lysine α-oxidase from S. japonicus yielded high titers (about 45 g L−1) of optically pure l-PA (Tani et al. 2015). The functionality of l-lysine cyclodeaminase from Actinoplanes friuliensis has been shown in C. glutamicum (Wagner et al. 2010), and the enzyme from Streptomyces. pristinaespiralis has been used in biotransformations of l-lysine to l-PA with recombinant E. coli (Tsotsou and Barbirato 2007). However, since l-lysine cyclodeaminases show low turnover numbers (about 0.6 s−1), only 4.5 mM l-PA was produced per hour using 1 g L−1 pure enzyme (Tsotsou and Barbirato 2007). All biotransformations with isolated enzymes or whole cells use l- or dl-lysine as substrate for conversion to l-PA, while to the best of our knowledge, a fermentative route for production of l-PA from glucose is presented here for the first time.

l-PA production from glucose proceeded with a yield of 0.09 ± 0.01 g g−1, and a volumetric productivity of 0.04 ± 0.01 g L−1 h−1 by the recombinant strains developed here (Fig. 6). Metabolic engineering often targets transport of products and/or by-products, which pertains to export of l-PA, re-uptake of l-PA, and export of l-lysine here. l-PA has been described to rapidly accumulate from the growth medium by corynebacteria; however, l-PA did not accumulate under hyperosmolar conditions (Frings et al. 1993). Avoiding l-PA re-uptake may not be of priority for l-PA production by C. glutamicum since the l-PA concentration did neither diminish once it was produced nor when added extracellularly (data not shown). On the other hand, improving l-PA export may be relevant to increase volumetric productivity since, for instance, lysine producing strains such as DM1933 and GRLys1 contain multiple copies of the lysine export gene lysE (Unthan et al. 2015). It is tempting to speculate that l-PA is exported by l-proline export systems since both heterocyclic secondary amino acids only differ in ring size. Genes and proteins for l-proline import are known (the l-proline-specific uptake system PutP, the proline/ectoine uptake system ProP, and the ectoine/proline/glycine betaine carrier EctP (Peter et al. 1998), but l-proline export has only been characterized biochemically. Osmoregulated channel(s) are responsible for efflux of various solutes upon sudden decrease in osmotic pressure and were shown to preferentially mediate efflux of the compatible solutes glycine betaine and l-proline (Ruffert et al. 1997). l-Proline and glycine betaine export occurred even in the absence of the genes mscL and yggB, which encode mechanosensitive channels (Nottebrock et al. 2003). YggB functions as major l-glutamate export system, its C-terminal domain is important for gating hysteresis in a voltage-dependent manner (Nakayama et al. 2016), and certain amino acid exchanges in YggB proved beneficial for l-glutamate export (Nakamura et al. 2007). By contrast, a genetic strategy to improve l-proline export is not evident as long as proteins/genes important for l-proline export remain unknown.

Full conversion of l-lysine to l-PA was not observed. Since conversion of l-lysine to l-PA occurs intracellularly, l-PA production suffers from l-lysine export by GRLys1 that overexpresses lysE from two chromosomal copies. Export of l-lysine is well studied and is due to a single protein, LysE (Vrljic et al. 1996). Thus, deletion of lysE in the l-PA-producing strain may be of advantage for l-PA production. The sum of l-lysine production and l-PA production was higher than production of l-lysine alone, which may be explained by a metabolic pull effect. Export of l-lysine and conversion of l-lysine to l-PA may lower the intracellular l-lysine concentration below a threshold relevant for negative genetic or allosteric regulation by l-lysine. A similar metabolic pull effect was observed for the production of cadaverine from l-lysine employing l-lysine decarboxylase in methylotrophic B. methanolicus (Naerdal et al. 2015). On the other hand, strains with increasing l-lysine productivity, i.e., GRLys1(pEKEx3), GRLys1ΔsugR(pEKEx3-IolTBest), and GRLys1ΔsugRΔldhA(pEKEx3-IolTBest), showed increasing l-PA production after transformation with pSH1-lysDH-proC (8 ± 1 mM, 11 ± 1 mM, and 14 ± 1 mM, Fig. 6), thus, supply of l-lysine as precursor positively influenced l-PA production. It can be expected that transformation of industrial l-lysine-producing strains with pSH1-lysDH-proC is a viable strategy to reach higher l-PA titers.