Metabolic engineering of Corynebacterium glutamicum for enhanced production of 5-aminovaleric acid

Background 5-Aminovaleric acid (5AVA) is an important five-carbon platform chemical that can be used for the synthesis of polymers and other chemicals of industrial interest. Enzymatic conversion of l-lysine to 5AVA has been achieved by employing lysine 2-monooxygenase encoded by the davB gene and 5-aminovaleramidase encoded by the davA gene. Additionally, a recombinant Escherichia coli strain expressing the davB and davA genes has been developed for bioconversion of l-lysine to 5AVA. To use glucose and xylose derived from lignocellulosic biomass as substrates, rather than l-lysine as a substrate, we previously examined direct fermentative production of 5AVA from glucose by metabolically engineered E. coli strains. However, the yield and productivity of 5AVA achieved by recombinant E. coli strains remain very low. Thus, Corynebacterium glutamicum, a highly efficient l-lysine producing microorganism, should be useful in the development of direct fermentative production of 5AVA using l-lysine as a precursor for 5AVA. Here, we report the development of metabolically engineered C. glutamicum strains for enhanced fermentative production of 5AVA from glucose. Results Various expression vectors containing different promoters and origins of replication were examined for optimal expression of Pseudomonas putida davB and davA genes encoding lysine 2-monooxygenase and delta-aminovaleramidase, respectively. Among them, expression of the C. glutamicum codon-optimized davA gene fused with His6-Tag at its N-Terminal and the davB gene as an operon under a strong synthetic H36 promoter (plasmid p36davAB3) in C. glutamicum enabled the most efficient production of 5AVA. Flask culture and fed-batch culture of this strain produced 6.9 and 19.7 g/L (together with 11.9 g/L glutaric acid as major byproduct) of 5AVA, respectively. Homology modeling suggested that endogenous gamma-aminobutyrate aminotransferase encoded by the gabT gene might be responsible for the conversion of 5AVA to glutaric acid in recombinant C. glutamicum. Fed-batch culture of a C. glutamicum gabT mutant-harboring p36davAB3 produced 33.1 g/L 5AVA with much reduced (2.0 g/L) production of glutaric acid. Conclusions Corynebacterium glutamicum was successfully engineered to produce 5AVA from glucose by optimizing the expression of two key enzymes, lysine 2-monooxygenase and delta-aminovaleramidase. In addition, production of glutaric acid, a major byproduct, was significantly reduced by employing C. glutamicum gabT mutant as a host strain. The metabolically engineered C. glutamicum strains developed in this study should be useful for enhanced fermentative production of the novel C5 platform chemical 5AVA from renewable resources. Electronic supplementary material The online version of this article (doi:10.1186/s12934-016-0566-8) contains supplementary material, which is available to authorized users.

A non-proteinogenic ω-amino acid, 5-aminovaleric acid (5AVA), has attracted attention as a five carbon (C5) platform chemical because of its potential in polymer synthesis [18][19][20][21]. 5AVA can be used to produce δ-valerolactam (2-piperidone) via intramolecular dehydrative cyclization and can be further processed for synthesis of bio-based nylons, such as nylon-5 and nylon-6,5 [18,20]. Enzymatic conversion of l-lysine to 5AVA has been achieved by employing lysine 2-monooxygenase (E.C. 1.13.12.2, encoded by the davB gene) and 5-aminovaleramidase (E.C. 3.5.1.30, encoded by the davA gene) [19]. We recently reported the development of a wholecell bioconversion process for conversion of l-lysine to 5AVA by employing recombinant Escherichia coli strains expressing lysine 2-monooxygenase and 5-aminovaleramidase as whole cell biocatalysts [20,21]. However, it is obviously desirable to use glucose derived from non-food lignocellulosic biomass as a substrate rather than l-lysine [1]. There have been reports on the development of metabolically engineered microorganisms for the production of C3 and C4 ω-amino acids, such as β-alanine [22] and γ-aminobutyrate [23], from glucose. We and others also examined the possibility of producing the C5 ω-amino acid 5AVA by metabolic engineering of E. coli, but the yield and productivity of 5AVA remain very low [18,20]. Thus, it is necessary to develop a new strategy for more efficient production of 5AVA.
Corynebacterium glutamicum is an organism widely used for the production of amino acids, proteins, monomers for plastic materials, and compounds for cosmetics [24,25]. Additionally, C. glutamicum has been successfully engineered to produce a different C5-platform chemical, cadaverine (1,5-pentanediamine), and was shown to be a promising host for producing this chemical using different carbon sources, such as glucose and xylose [26][27][28][29][30]. Because C. glutamicum strains capable of producing l-lysine at very high levels have already been commercialized, we decided to exploit C. glutamicum as a host strain for the production of 5AVA.
In this study, we designed and introduced a synthetic pathway for the production of 5AVA into an l-lysineoverproducing C. glutamicum strain. The synthetic pathway consists of two key enzymes, lysine 2-monooxygenase encoded by the davB gene and 5-aminovaleramidase encoded by the davA gene, responsible for the conversion of l-lysine to 5AVA. Various expression systems including vectors and promoters were examined for the most efficient production of 5AVA in C. glutamicum. A reaction responsible for unexpected formation of glutaric acid as a major byproduct was identified and subsequently deleted. The final engineered C. glutamicum strain was used for enhanced production of 5AVA by fermentation with much reduced glutaric acid.

Results and discussion
Construction of the 5AVA synthesis pathway in C. glutamicum by expressing the davAB genes under the control of the tac promoter Although 5AVA exists as an intermediate in amino acid degradation pathways in Pseudomonads, metabolic engineering for enhanced production of 5AVA requires strong metabolic flux from the chosen intermediate metabolite toward 5AVA as indicated by many successful examples of systems metabolic engineering [16,17]. Recombinant E. coli strains employing the davAB genes from the l-lysine catabolic pathway of Pseudomonads were previously shown to produce 5AVA, although at low yield and productivity [18,21]. Because E. coli strains have a relatively low capacity to provide l-lysine as a precursor for 5AVA, high-level production of 5AVA could not be achieved in recombinant E. coli strains even though the l-lysine catabolic pathway via 5-aminovaleramide provides the shortest route from l-lysine to 5AVA [18,21]. Thus, C. glutamicum, the well-known, highly efficient l-lysine producing bacterium, was selected as a host strain for 5AVA production in this study to provide strong metabolic flux from glucose to l-lysine, the direct precursor of 5AVA [5,6]. This is strategically advantageous for producing compounds using l-lysine as a direct precursor.
To extend the metabolic flux from glucose toward 5AVA beyond l-lysine (Fig. 1a), P. putida ATCC 12633 strain. The metabolically engineered C. glutamicum strains developed in this study should be useful for enhanced fermentative production of the novel C5 platform chemical 5AVA from renewable resources.
Keywords: 5-Aminovaleric acid, Corynebacterium glutamicum, l-Lysine, Metabolic engineering, Glutaric acid davAB genes [20,21] were cloned in two different E. coli-C. glutamicum shuttle vectors (Additional file 5: Table S1; Fig. 1b) and expressed in C. glutamicum BE strain (KCTC 12390BP). In plasmids pKCA212davAB and pJS30 (Fig. 1b), the davA and davB genes were organized in an operon and expressed under the control of the tac promoter, with an additional lacUV5 ribosome-binding site (tttcacacaggaaaca) for the davB gene residing between the coding sequences of the two genes. Plasmid pKCA-212davAB was constructed based on an in-house shuttle vector, pKCA212-MCS, whereas pJS30 was derived from pEKEx1 [31]. Plasmid pEKEx1 contains a pBL1 origin of replication, having ~30 copy numbers per cell in C. glutamicum [32], and pKCA212-MCS contains a pCC1 origin of replication, also having ~30 to ~40 copies per cell [33]. It was found that C. glutamicum BE could produce 17.2 g/L l-lysine (yield of 325 mmol/mol glucose) in shake-flask cultivation in 44 h (Fig. 2). It was found through HPLC analysis of metabolites excreted into culture medium that expression of the davAB genes in C. glutamicum BE resulted in conversion of l-lysine to 5AVA, whereas the non-engineered strain did not convert any l-lysine into 5AVA (Fig. 2).
Slightly decreased l-lysine production was observed in recombinant C. glutamicum BE strains expressing the davAB genes, although l-lysine was still the major product in both engineered strains tested. Expression of the davAB genes using the shuttle vector pEKEx1 (pJS30) was more effective at producing 5AVA than using pKCA212-MCS (pKCA212davAB) (Fig. 2). The C. glutamicum BE strain harboring pKCA212davAB produced 13.4 g/L l-lysine (yield of 271.2 mmol/mol) and 58 mg/L 5AVA (yield of 1.5 mmol/mol), whereas the C. glutamicum BE strain harboring pJS30 produced 12.4 g/L l-lysine (yield of 258.2 mmol/mol) and 641 mg/L 5AVA (yield of 16.7 mmol/mol) from glucose. These results demonstrate that heterologous expression of the davAB genes from gram-negative P. putida correctly functioned to produce 5AVA from glucose using l-lysine as a 5AVA precursor in gram-positive C. glutamicum.
Although the C. glutamicum BE strain harboring pKCA212davAB or pJS30 successfully produced 5AVA Fig. 1 Metabolic engineering strategies for 5AVA production using C. glutamicum. Heterologous expression of the P. putida davB gene (encoding l-lysine 2-monooxygenase) and the davA gene (encoding delta-aminovaleramidase) results in conversion of l-lysine into 5AVA. 5AVA5, 5-aminovalerate; ASP5 l-aspartate; ASP-P aspartyl phosphate; ASP-SA aspartate semialdehyde; LYS l-lysine from glucose, most of l-lysine was not converted into 5AVA, which suggests that metabolic flux from l-lysine to 5AVA was still quite weak as a result of inefficient expression of the davAB genes. Thus, we investigated whether 5AVA production could be enhanced by employing C. glutamicum codon-optimized davAB genes. The C. glutamicum BE strain harboring pJS38, which expresses C. glutamicum codon-optimized davAB genes, produced 3.0 g/L 5AVA in flask cultivation, which represented an increase of almost 370 % over that produced by the C. glutamicum BE strain harboring pJS30 (Fig. 2). However, 7.9 g/L l-lysine still remained in the culture medium of C. glutamicum BE strain (pJS38). The yield for l-lysine obtained by C. glutamicum BE (pJS38) was 193.9 mmol/ mol glucose, whereas that for 5AVA was 92.3 mmol/mol glucose.
Notably, 0.14 mg/L glutaric acid was detected in the culture medium of C. glutamicum BE (pJS38), even though 5AVA aminotransferase and glutarate-semialdehyde dehydrogenase (encoded by davT and davD, respectively), which are the key enzymes for further conversion of 5AVA into glutaric acid using α-ketoglutarate as an amine acceptor [18,21], were not expressed in this strain. This result strongly suggests that endogenous enzymes homologous to 5AVA aminotransferase and glutarate-semialdehyde dehydrogenase might be involved in further conversion of 5AVA into glutaric acid in C. glutamicum.
Examination of 5AVA production by engineered C. glutamicum strain expressing the davAB genes under the control of the tac promoter in fed-batch fermentation We then investigated the capability of C. glutamicum BE (pJS38) to produce 5AVA by fed-batch fermentation in a 5-L fermentor (Fig. 3a, c) to examine its potential for large-scale production of 5AVA. During fed-batch cultivation of C. glutamicum BE (pJS38), l-lysine concentration reached 21.8 g/L after 94 h of cultivation, and then decreased beyond this point. C. glutamicum BE (pJS38) produced 17.8 g/L 5AVA in 191 h, with an overall yield and productivity of 0.07 g/g (107.3 mmol/mol) and 0.09 g/L/h, respectively (Fig. 3c). The total input of glucose was 790 g (4.4 mol), and the total amount of produced 5AVA was 55.3 g (471 mmol), with the final volume of 3.1 L. The maximum specific growth rate was 0.23 h −1 , and the observed maximum specific productivity was 9.2 mg/g/h. Initial increases in l-lysine levels followed by a decrease in titer along with constant production of 5AVA indicated that l-lysine-production flux might be strong initially before slowing down in the latter half of cultivation. The maximum l-lysine-specific productivity was 56.7 mg/g/h initially, but gradually decreased to Growth and production characteristics of the C. glutamicum BE strain harboring the pKCA212davAB, pJS30, or pJS38 vectors after 44 h of shake-flask cultivation (n = 3, error bars = SD). The C. glutamicum BE strain containing no plasmids was used as a non-engineered control. a The final OD 600 at the end of cultivation is shown for all strains tested, and an experimentally determined correlation factor (0.28) was used to determine the biomass yield (Y X/S ). b The production characteristics include the final titers for l-lysine (light-grey bars), 5AVA (dark-grey bars), and glutaric acid (black bars). c The molar yields from glucose for l-lysine (white bars), 5AVA (grey bars), and glutaric acid (black bars) are also shown. When appropriate, gene expression was induced by addition of IPTG at a final concentration of 0.5 mM when the growth reached an OD 600 of 0.5-0.6 zero. The amount of the major byproduct, glutaric acid, also increased steadily throughout the entire cultivation period until the concentration reached 5.3 g/L at the end of cultivation (Fig. 3c).
The maximum OD 600 reached was 95.4, corresponding to the measured dry cell weight concentration of 24.5 g/L, after 189 h of fermentation (Fig. 3a). These results demonstrated that C. glutamicum BE (pJS38) was able to successfully produce 5AVA from renewable resources in a laboratory-scale bioreactor. However, conversion of l-lysine toward 5AVA needs to be further enhanced through stronger expression of the davAB genes.

Construction of engineered C. glutamicum strain expressing the davAB genes under the control of constitutive promoters to improve 5AVA production
In addition to the tac promoter used in pEKEx1, promoters for SOD (NCgl2826; E.C. 1.15.1.1) and the transcription factor Tu (Tuf; Ncgl0480; E.C. 3.6.5.3) have also been widely used in metabolic engineering of C. glutamicum because of their capabilities to support strong gene expression at the chromosome level [34]. Although the sod promoter is known to exhibit weaker plasmidbased expression than the tac promoter [35], weak and medium-strength expression driven by constitutive promoters might be more effective than higher expression levels for producing chemicals of interest under different circumstances [15]. Moreover, use of the constitutive promoters circumvents the requirement for costly additives such as IPTG. Therefore, we replaced the lacI Q gene and tac promoter in pJS38 with P sod or P tuf (Additional file 5: Table S1). These constructs were then introduced into the C. glutamicum BE strain, and their functions were investigated by flask cultivation. However, expression of the davAB genes under control of the sod promoter (pJS59) and tuf promoter (pJS60) did not result in higher 5AVA production relative to that produced by the parent construct, pJS38 containing the tac promoter (Figs. 2, 4). C. glutamicum BE (pJS59) and C. glutamicum BE (pJS60) produced 556 and 587 mg/L 5AVA, Fig. 3 Production of 5AVA from glucose by fed-batch cultures of (a, c) C. glutamicum BE (pJS38) and (b, d) C. glutamicum BE (p36davAB3). Characteristics of the fed-batch cultivation profile, including growth (filled circles, OD 600 ), residual sugar (empty circles; g/L), l-lysine (filled diamonds), 5AVA (magenta diamonds), and glutaric acid (green triangles) production titers, are plotted against the cultivation time respectively, with no glutaric acid observed in the culture media after 44 h of flask cultivation.
We also investigated the newly designed synthetic promoter active in C. glutamicum, the H36 promoter [36], to see if it can possibly improve 5AVA production; it was successfully employed for the expression of glutamate decarboxylase (GAD) and lysine decarboxylase (LDC) in C. glutamicum strains, resulting in high-level production of gamma-aminobutyrate (GABA) and cadaverine, respectively [30,37].
We replaced the lacI Q and the tac promoter in pJS38 with the P H36 promoter to construct p36davAB2. However, cultivation of C. glutamicum BE harboring p36davAB2 also did not improve 5AVA production, which resulted in production of 661 mg/L 5AVA under the same culture condition (Fig. 4). These results indicate that expression of the davAB genes was still not strong enough to enable efficient conversion of l-lysine to 5AVA.
To continue exploring expression vector systems for possible improvement of 5AVA titer, we noticed that the strong H36 promoter originally developed with a different backbone vector, pCES208 [36,38], might not be optimal for pEKEx1. Engineered C. glutamicum strains harboring a pCES208-based plasmid for expression of target genes under strong synthetic promoters, such as H30 and H36, have been reported to efficiently produce GABA and cadaverine from renewable resources [30,37]. Therefore, we transferred codon-optimized versions of the davAB genes into the pCES208 vector system.
The new construct, p36davAB1, was further modified by inserting a His 6 -Tag into the N-terminal of davA gene, resulting in p36davAB3. This was done because there have been reports showing that His 6 -tagged constructs can sometimes be expressed more efficiently [39,40]. These constructs were transformed into the C. glutamicum BE strain and assessed by flask cultivation. Whereas C. glutamicum BE (p36davAB1) produced only 0.4 g/L 5AVA along with 11.7 g/L l-lysine, C. glutamicum BE (p36davAB3) produced 6.9 g/L 5AVA, with 5.5 g/L l-lysine remaining unconverted (Fig. 5). The 5AVA concentration obtained represents a 130 % increase over that (Fig. 2) obtained with C. glutamicum BE (pJS38). Interestingly, the construct containing the His 6 -tagged variant produced substantially more 5AVA compared to that produced using the construct lacking the His-tag, possibly because of the improved stability afforded by the 5′ modification, which resulted in higher expression of the davAB genes in the recombinant C. glutamicum BE strain (Additional file 1: Figure S1). Comparison of mRNA folding energies (ΔG) with the RNA secondary structure prediction program Mfold (http://unafold.rna. albany.edu/?q=mfold/download-mfold) suggested that the ΔG for the first 30 nucleotides starting from the +1 site of the H36 promoter in p36davAB2 is −6.00 kcal/ mol, which is much lower than ΔG of −0.06 kcal/mol obtained in p36davAB3. The higher ΔG in p36davAB3 indicates that less stable mRNA produced by p36davAB3 might allow the translation machinery to bind more easily than much stable mRNA produced by p36davAB2. Fig. 4 Growth and production characteristics of the C. glutamicum BE strain harboring the pJS59, pJS60, or p36davAB2 vector after 44 h of shakeflask cultivation (n = 3, error bars = SD). The C. glutamicum BE strain containing no plasmids was used as a non-engineered control. a The final OD 600 at the end of cultivation is shown for all strains tested, and an experimentally determined correlation factor (0.28) was used to determine the biomass yield (Y X/S ). b The production characteristics include the final titers for l-lysine (light-grey bars), 5AVA (dark-grey bars), and glutaric acid (black bars)

Examination of 5AVA production by engineered C. glutamicum expressing the davAB genes under the control of the strong H36 promoter in fed-batch fermentation
Having achieved improved 5AVA production in flask culture, fed-batch culture of C. glutamicum BE (p36davAB3) was performed next in a 5-L fermentor. C. glutamicum BE (p36davAB3) produced 19.7 g/L 5AVA in 157 h, with the overall yield and productivity of 0.08 g/g and 0.16 g/L/h, respectively (Fig. 3d). This strain also accumulated 13.4 g/L glutaric acid as a byproduct at the end of the cultivation. On the other hand, l-lysine accumulation decreased significantly compared to that observed with C. glutamicum BE (pJS38). l-Lysine accumulated to 3.7 g/L in 25 h, but production remained between 1 and 2 g/L over the entire cultivation period (Fig. 3d). Citric acid was another major byproduct, but its concentration remained at ~1 g/L throughout cultivation. Notably, the production patterns observed during fed-batch fermentation were different from those observed during flask cultivation. Although large portions of l-lysine remained unconverted at the end of the flask cultivation, very little l-lysine remained in fed-batch fermentation. This indicates that control of pH and provision of sufficient air streams were beneficial for 5AVA production and provided better results during fed-batch fermentation. A sufficient air supply is important for cultivation because lack of sufficient air can result in accumulation of substantial concentrations of lactic and acetic acids in the fermentation broth [30,41]. These byproducts were not observed in our cultivation conditions, contrary to previous reports. These results suggest that the strategy combining improved expression of the davA gene fused with His 6 -Tag at its N-Terminal and the davB gene as an operon under control of the strong synthetic H 36 promoter was successful in directing most of the l-lysine pool toward 5AVA, resulting in efficient production of 5AVA.
Construction of an engineered C. glutamicum gabT mutant for enhanced production of 5AVA with greatly reduced glutaric acid production While 19.7 g/L of 5AVA could be produced by fed-batch cultivation of C. glutamicum BE (p36davAB3), glutaric acid, a major byproduct, was still produced to a relatively high concentration (up to 13.4 g/L). In order to further enhance 5AVA production, conversion of 5AVA into glutaric acid should be minimized. However, no enzyme responsible for converting 5AVA into glutaric acid is known in C. glutamicum. Thus, we performed moleculardocking simulations, which suggested possible interactions between endogenous GabT and 5AVA (Additional file 2: Figure S2, Additional file 3: Figure S3). GabT shares homology (60 % by primary peptide structure) with 4-aminobutyrate aminotransferase (SGR_1829) in Streptomyces griseus, which exhibits 60 % relative aminotransfer activity for 5AVA [42]. Although the pyridoxal phosphate moiety was in the correct orientation and position, the orientation of bound 5AVA was twisted, possibly because of the larger size of the substrate being accommodated in the active site. Additionally, the enzyme also shared high homology (Additional file 3: Figure S3) with P. putida DavT, which binds 5AVA as its natural substrate (Additional file 4: Figure S4). The major difference in active sites between GabT and DavT is that DavT contains a glutamine residue (Gln80) rather than a methionine residue, enabling accommodation of the ω-amino group in the binding pocket (Additional file 3: Figure S3). However, GabT from S. griseus, even with the methionine residue at this position, is sufficiently promiscuous to accept 5AVA as a substrate [42]. Thus, we could conclude from the docking simulations that the endogenous C. glutamicum GabT might accommodate 5AVA as a substrate for aminotransfer reactions, leading to the formation of glutaric acid.
Based on the above results, the gabT gene (E.C. 2.6.1.19, encoding 4-aminobutyrate aminotransferase, Ncgl0462) was deleted from the chromosome of C. glutamicum BE to construct C. glutamicum AVA2. C. glutamicum AVA2 produced 17.5 g/L of l-lysine by flask cultivation in 44 h, with no residual 5AVA detected (Fig. 6). This result suggests that deletion of the gabT gene did not inhibit cell growth and l-lysine production. Plasmid p36davAB3 was then transformed into C. glutamicum AVA2 to assess 5AVA production. Fed-batch cultivation of engineered C. glutamicum AVA2 (p36davAB3) in a 5-L fermenter resulted in production of 33.1 g/L 5AVA with greatly reduced glutaric acid (2.0 g/L) and l-lysine (648.3 mg/L) at the end of cultivation. The overall yield and productivity obtained were 0.1 g/g glucose (163.1 mmol/mol) and 0.22 g/L/h, respectively (Fig. 7). Cells grew to an OD 600 of 134 in 153 h, with a measured dry cell weight of 36.1 g/L, with the maximum specific growth rate was 0.4 h −1 . The maximum specific 5AVA productivity was 65.8 mg/g/h, which gradually decreased to 22.1 mg/g/h at the end of cultivation. Additionally, the l-lysine concentration peaked at 28 h, but remained as low as 1 g/L for the remainder of the cultivation. As expected, gabT deletion resulted in a significant decrease in glutaric acid production compared to that observed in the parent strain. However, the continued presence of glutaric acid in the culture broth suggests that unknown aminotransferases still remain in C. glutamicum that are capable of converting 5AVA to glutaric acid, although at lower efficiencies than GabT.

Conclusions
In this study, we report development of engineered C. glutamicum strains for the production of 5AVA from glucose. Expression of two key enzymes, lysine 2-monooxygenase and delta-aminovaleramidase, was systematically optimized by examining different promoters, origins of replication, codon usage of the davAB genes, and even 5′ modification of the davA gene with a His-tag, all of which Fed-batch cultivation of C. glutamicum AVA2 harboring p36davAB3 for the production of 5AVA in laboratory-scale bioreactor from glucose. a Characteristics of the fed-batch cultivation profile including growth (filled circles, OD 600 ), residual sugar (empty circles; g/L) and b production titers of products including l-lysine (dark diamonds), 5AVA (magenta diamonds), and glutaric acid (green triangles) are plotted against the cultivation time were found to be important for determining the optimal and stable plasmid-based expression of the davAB genes in C. glutamicum. In addition, production of a major byproduct, glutaric acid, could be significantly reduced by identifying previously unknown enzyme GabT responsible for converting 5AVA to glutaric acid and deleting the corresponding gene from the chromosome. Fed-batch cultivation of the final engineered C. glutamicum AVA2 strain harboring p36davAB3 produced 33.1 g/L 5AVA with greatly reduced glutaric acid (2.0 g/L). The metabolically engineered C. glutamicum strains developed in this study should be useful for enhanced fermentative production of the novel C5 platform chemical, 5AVA, from renewable resources such as glucose.

Strains and plasmids
All bacterial strains and plasmids used in this study are listed in Additional file 5: Table S1. All DNA manipulations were performed following standard procedures [43]. Primers used in this study (Additional file 6: Table  S2) were synthesized at Bioneer (Daejeon, Korea). C. glutamicum BE (KCTC 12390BP) was used as the base strain for 5AVA production. Polymerase chain reaction (PCR) was performed with the C1000 Thermal Cycler (Bio-Rad, Hercules, CA, USA). The general PCR condition for amplifications of target genes using primer sets listed in Additional file 6: Table S2 is as follows: 1 cycle of 95 °C for 5 min; 30 cycles of 94 °C for 30 s, 52 °C for 30 s, 72 °C for 1 min 30 s; and a final extension of 72 °C for 5 min. The final reaction volume is 20 μL. The in-house-developed C. glutamicum shuttle vector pKCA212-MCS was constructed by cloning the origin of replication of the cryptic plasmid pCC1 [33] into pKA212-MCS at the AatII and XhoI sites. The origin of replication of pCC1 was synthesized by GenScript (http://www.genscript.com) based on the reported sequence. Plasmid pKA212-MCS was constructed by replacing the chloramphenicol-resistance gene of pKA312-MCS [11] with a kanamycin-resistance gene obtained from pZA21-MCS (http://www.expressys. com) by restriction digest with AatII and SpeI. The davAB genes from pKE112-DavAB [20,21] were restrictiondigested and ligated into pKCA212-MCS to construct pKCA212davAB using the same restriction enzyme sites (EcoRI/KpnI, KpnI/BamHI). A 16-bp untranslated region (tttcacacaggaaaca) containing a ribosome-binding site was present between the two genes for davB expression. The same genes were also cloned into pEKEx1 to construct pJS30. The codon-optimized versions of davAB genes (Additional file 7: Table S3) with preferred codon usage in C. glutamicum were synthesized by Bioneer (Daejeon, Korea) and cloned into the EcoRI/BamHI restriction enzyme sites in pEKEx1 to yield pJS38.
To construct promoter variants of pEKEx1, promoterless pEKEx1 was created by removing the tac promoter and the initial 778 bp of the coding sequence of the LacI Q gene from pEKEx1 by restriction digestion with EcoRV/EcoRI. The desired promoters were similarly designed as previously described [5,34,44] and inserted into the promoterless pEKEx1 vector. The region 250 bp upstream of the start codon for the superoxide dismutase gene (NCgl2826, E.C. 1.15.1.1) was amplified by polymerase chain reaction (PCR) from C. glutamicum ATCC 13032 chromosome using primers Psod_F_EcoRV and Psod_R_EcoRI, and then digested and cloned into the EcoRV/EcoRI sites of the promoterless pEKEx1 vector to construct pJS57. The 248-bp sequence upstream of the start codon for the gene encoding the elongation factor Tu (Ncgl0480, E.C. 3.6.5.3) was amplified by PCR from C. glutamicum ATCC 13032 chromosome using the primers Ptuf_F_EcoRV (v2) and Ptuf_R_EcoRI, and then digested and cloned into the promoterless pEKEx1 vector to yield pJS58. The codon-optimized davAB genes from pJS38 were restriction-digested with EcoRI/BamHI and cloned into the pJS57 and pJS58 vectors at the EcoRI/BamHI restriction enzyme sites to make pJS59 and pJS60, respectively.
Plasmid p36davAB2 was constructed from pEKEx1 by cloning the codon-optimized davAB genes. Promoterless pEKEx1 was constructed by methods similar to those described in the previous paragraph, except that the genes were cloned into the EcoRV/PstI restriction sites of the vector. The P H36 promoter was amplified by PCR using the JW02H-F and JW02H-R primers from pCES208H36GFP, and the fragments were restrictiondigested with EcoRV/EcoRI. A second round of PCR using primers JW02AB-F and JW02AB-R from pJS38 generated codon-optimized davAB gene fragments that were restriction-digested with EcoRI/PstI. The resulting products were then ligated into the EcoRV/PstI restriction sites of the promoterless pEKEx1 vector to yield p36davAB2.
Plasmid p36davAB1 was constructed from pCES208 by cloning the codon-optimized davAB genes. Products from the first round of PCR using primers JW01A-F and JW01A-R were used for amplification of the davA gene, which was then restriction-digested with BamHI/SfiI. The second round of PCR used primers JW01B-F and JW01B-R to amplify the davB gene, which was then restriction-digested with NotI. These fragments were cloned into the pCES208H36GFP vector [36] by replacing the egfp gene to yield p36davAB1.
Plasmid p36davAB3 was constructed from the pCES208H36EGFP vector [36]. The codon-optimized davA gene fused with His 6 -Tag at its N-Terminal was amplified using primers JW03A-F and JW01A-R and restriction-digested with BamHI and SfiI. The codonoptimized davB gene was amplified using primers JW01B-F and JW03B-R and restriction-digested with NotI. The two products were then cloned into the pCES208H36EGFP vector by replacing the egfp gene to construct p36davAB3.
Plasmid pJS113 beta was constructed from the pK19mobsacB vector [45]. Primers 113 i1F beta and 113 i1R beta were used to PCR-amplify the upstream region and a portion of the gabT gene from C. glutamicum. Primers 113 i2F beta and 113 i2R beta were then used to PCR-amplify the downstream region and a portion of the gabT gene of C. glutamicum. The two PCR products were joined by a third PCR using primers 113 i1F beta and 113 i2R beta. The final PCR product was cloned into the PstIdigested pK19mobsacB to make pJS113 beta. pJS113 beta was subsequently used to disrupt the gabT gene in the C. glutamicum BE chromosome, resulting in the strain C. glutamicum AVA2. This in-frame deletion left a 330-bp deletion in the 280-609 region of the 1347-bp gabT gene.
E. coli DH5α and TOP10 strains (Additional file 5: Table S1) were used for general cloning purposes. All constructed plasmids introduced into C. glutamicum, except for pJS113 beta, were prepared in unmethylated form using the methylation-deficient E. coli JM110 strain (Stratagene; Agilent Technologies, Santa Clara, CA, USA). pJS113 beta was propagated in C. glutamicum by bacterial conjugation using E. coli S17-1 as a donor [45]. Plasmids were introduced via electroporation as previously described [46]. Cells were transferred to a microcuvette and electroporated using a micropulser. Cells were transformed with about 2 μg of DNA by electroporation (1.8 V and 400 Ω). Pre-chilled preculture medium (900 μL) was added and the transformed cells were allowed for growth recovery for 2 h without shaking in a 30 °C incubator. The transformed cells were then spread onto the agar plates containing kanamycin as a selective marker.

Flask cultivation
Stock cells stored in glycerol were used to inoculate 5-mL pre-cultures, which were grown at 30 °C with shaking at 200 rpm in an incubator (JSSI-300C; JS Research Inc., Gongju, Korea) for 17-18 h. Cells suspended in 250-μL aliquots of pre-culture were harvested by centrifugation (Centrifuge 5415 D; Eppendorf, Hamburg, Germany) and transferred to a 25-mL primary culture in autoclaved 300-mL baffled Erlenmeyer flasks, each containing 1.5 g of CaCO 3 to maintain the pH at ~7.0 during cultivation. Primary cultures were grown with shaking in an incubator for 44 h. When appropriate, isopropyl-β-d-thiogalactopyranoside (IPTG) at a final concentration of 0.5 mM was used to induce gene expression during the early log phase (OD 600 = 0.5-0.6), with 25.0 μg/mL kanamycin added for selective pressure (Ravasi et al. [35]).

Fed-batch fermentation
Stock cells stored in glycerol were used to inoculate 5.0-mL pre-cultures, which were grown at 30 °C with shaking in an incubator for 17-18 h. Two 1-mL samples of the pre-culture were transferred to two 1-L Erlenmeyer flasks, each containing 100 mL of seed medium, and grown with shaking (200 rpm) in a 30 °C incubator for 19-20 h. The entire seed culture (200 mL) was added as the inoculum to the 1.8-L primary culture in a fermenter (initial OD 600 = 1.5-2.0 in 2 L). IPTG at a final concentration of 1 mM and kanamycin (25.0 μg/L) were also added during fermentation inoculation. A NBS Bio-Flo 3000 fermenter system (New Brunswick Scientific, Edison, NJ, USA) equipped with a 6.6-L jar was used for all fed-batch cultivation experiments. The pH was maintained at 7.0 by addition of 28 % (v/v) ammonia solution (Junsei Chemical Co., Ltd., Tokyo, Japan). Temperature and agitation were maintained at 30 °C and 600 rpm, respectively, by a proportional-integral-derivative controller throughout the entire cultivation period. The aeration rate was maintained at 1 L/L/min. Foaming was suppressed by addition of 1:10 diluted antifoam 204 (Sigma-Aldrich). Feeding solution (200 mL) was manually added each time the residual glucose level decreased to <20 g/L.

Analytical procedures
Two high-performance liquid chromatography (HPLC) systems, Agilent 1100 (Agilent Technologies) and Waters Breeze 2 (Waters Corporation, Milford, MA, USA), were used to determine the metabolite concentration in the culture broth. For detection of amino compounds, the supernatant of the culture samples was reacted with o-phthaldehyde as previously described [13] prior to injection into the Eclipse Zorbax-AAA column (Agilent Technologies). Linear gradients of mobile phase A [10 mM Na 2 HPO 4 , 10 mM Na 2 B 4 O 7 ·10H 2 O, and 8 mg/L NaN 3 (pH 7.2)] and mobile phase B (methanol, acetonitrile, and water at a volumetric ratio of 45:45:10) were used to separate the amino acids in the column. Borate buffer (0.4 M; pH 10.2) was used as a buffering agent rather than pH 9.0 buffer as previously described [13]. The derivatized compounds were detected using a diodearray detector at 338 nm. The column temperature was set to 25 °C, and the flow rate of the pump was set to 0.640 mL/min. The following gradient was applied for resolving the compounds: 0-0.5 min, 0 % B; 0.5-18 min, a linear gradient of B from 0 to 57 %; 18-26 min, a linear gradient of B from 57 to 100 %; 26-31.8 min, 100 % B; 31.8-31.9 min, a linear gradient of B from 100 to 0 %; 31.9-32 min, 0 % by volume. Glutaric acid was detected using the Waters Breeze 2 HPLC system (Waters Corporation) with a MetaCarb 87H column (Varian; Crawford Scientific, Strathaven, UK) and a constant flow of sulfuric acid solution at 0.5 mL/min. The Waters Breeze 2 system included an isocratic pump (Waters 1515; Waters Corporation), a refractive index detector (Waters 2414; Waters Corporation), and an autosampler (Waters 2707; Waters Corporation).
Cell growth was monitored by measuring the OD 600 with an Ultrospec 3000 spectrophotometer (Amersham Biosciences, Uppsala, Sweden). The correlation factor (0.28 g of dry weight of cells per L per OD 600 of 1) was experimentally determined and used for biomass concentration calculation of flask-cultivated cells. This correlation factor was in agreement with a previously reported value [47]. Glucose concentration was measured using a 2700 biochemistry analyzer (YSI, Yellow Springs, OH, USA). When necessary, diluted HCl solution was used to neutralize CaCO 3 in the cultivation media.

Molecular docking simulation
Molecular docking simulations were performed using Autogrid and Autodock 4.2.5.1 software [48]. Gasteiger charges and hydrogen atoms were added using AutoDockTools 1.5.6. A Lamarckian genetic algorithm with default parameters was used, and no peptide residues were kept flexible. The docking grid was set to encompass the catalytic pocket, but not the entire enzyme. For docking of the natural substrate of 4-aminobutyrate aminotransferase, the substrate molecules were separately saved from a known structure (PDB ID: 4ATQ) [49] as a single molecule in the form of an external aldimine and used as a ligand. For docking with 5AVA aminotransferase, pyridoxal phosphate and 5AVA in the form of an external aldimine were used as the ligand. Torsion about the bond between the pyridine moiety of the pyridoxal phosphate and the Schiff base was not allowed during docking. The docking results were visualized using PyMol 1.6 (https:// www.pymol.org/pymol) without additional hydrogen atoms.