Flux Enforcement for Fermentative Production of 5-Aminovalerate and Glutarate by Corynebacterium glutamicum

: Bio-based plastics represent an increasing percentage of the plastics economy. The fermentative production of bioplastic monomer 5-aminovalerate (5AVA), which can be converted to polyamide 5 (PA 5), has been established in Corynebacterium glutamicum via two metabolic pathways. l -lysine can be converted to 5AVA by either oxidative decarboxylation and subsequent oxidative deamination or by decarboxylation to cadaverine followed by transamination and oxidation. Here, a new three-step pathway was established by using the monooxygenase putrescine oxidase (Puo), which catalyzes the oxidative deamination of cadaverine, instead of cadaverine transaminase. When the conversion of 5AVA to glutarate was eliminated and oxygen supply improved, a 5AVA titer of 3.7 ± 0.4 g / L was reached in microcultivation that was lower than when cadaverine transaminase was used. The elongation of the new pathway by 5AVA transamination by GABA / 5AVA aminotransferase (GabT) and oxidation by succinate / glutarate semialdehyde dehydrogenase (GabD) allowed for glutarate production. Flux enforcement by the disruption of the l -glutamic acid dehydrogenase-encoding gene gdh rendered a single transaminase (GabT) in glutarate production via the new pathway responsible for nitrogen assimilation, which increased the glutarate titer to 7.7 ± 0.7 g / L, i.e., 40% higher than with two transaminases operating in glutarate biosynthesis. Flux enforcement was more effective with one coupling site, thus highlighting requirements regarding the modularity and stoichiometry of pathway-specific flux enforcement for microbial production. the LdcC–Puo Rq –PatD variant with the putrescine oxidase from R. qingshengii to test its performance for glutarate production and ﬂux enforcement.


Introduction
Plastics are primarily synthesized chemically from petroleum and natural gas. However, the annual market volume of bioplastics is predicted to increase from 2.11 million tons in 2019 to 2.43 million tons in 2024 [1]. Among bioplastics, the biopolyamides find many applications ranging from the production of parachutes during World War II to filaments for 3D-printing at present [2,3]. Polyamides can be produced via two routes: either by the condensation of dicarboxylic acids with diamines or by the anionic ring-opening polymerization of lactams, which can be formed from ω-amino acids via cyclization [4]. Examples of these main building blocks range from succinate to sebacate for Figure 1. Fermentative production of 5-aminovalerate (5AVA) and glutarate. (A) A novel three-step metabolic pathway for production of 5AVA and glutarate was established by the exchange of a transaminase of an existing three-step pathway (PatA) with an oxidase (Puo). The anticipated shift of net nitrogen assimilation in the form of L-glutamic acid is counteracted by (B) endogenous Lglutamate synthesis by L-glutamic acid dehydrogenase(Gdh), which was deleted to enforce the flux towards glutarate. Pink areas depict decarboxylation reactions. Yellow areas depict transamination steps. Orange areas depict deamination steps. Blue areas depict oxidations. Red-shadowed genes (gabT, γ-aminobutyrate (GABA)/5AVA amino transferase; gabD, and succinate/glutaratesemialdehyde dehydrogenase) are native genes of Corynebacterium glutamicum. Additionally, copies of gabTD from Pseudomonas stutzeri were added by heterologous overexpression. The grey-shadowed gene (gdh, L-glutamic acid dehydrogenase) is native in C. glutamicum. The dark green-shadowed genes (ldcC, L-lysine decarboxylase; patA, putrescine transaminase; and patD, γ-aminobutyraldehyde dehydrogenase) are originally from Escherichia coli and were added by heterologous overexpression. The blue-shadowed gene (puo, putrescine oxidase; EC 1.4.3.10) was originally from Rhodococcus qingshengii and Paenarthrobacter aurescens, and it was heterologously expressed in C. glutamicum. Lglutamic acid is highlighted in bright green. OG: 2-oxoglutarate, Glu: L-glutamatic acid.
Flux enforcement has also been applied to improve glutarate production by C. glutamicum [15]. The transamination reactions catalyzed by cadaverine/putrescine transaminase PatA and GABA/5AVA amino transferase GabT during glutarate production were coupled to growth in the absence of the major ammonium assimilating enzyme L-glutamic acid dehydrogenase Gdh [15] ( Figure 1B). The resulting strain reached glutarate titers of about 5 g/L in shake flasks and about 25 g/L during fed-batch fermentation [15]. Here, the monooxyogenase enzyme putrescine oxidase encoded by puo from Rhodococcus qingshengii or Paenarthrobacter aurescens was tested for the Figure 1. Fermentative production of 5-aminovalerate (5AVA) and glutarate. (A) A novel three-step metabolic pathway for production of 5AVA and glutarate was established by the exchange of a transaminase of an existing three-step pathway (PatA) with an oxidase (Puo). The anticipated shift of net nitrogen assimilation in the form of l-glutamic acid is counteracted by (B) endogenous l-glutamate synthesis by l-glutamic acid dehydrogenase(Gdh), which was deleted to enforce the flux towards glutarate. Pink areas depict decarboxylation reactions. Yellow areas depict transamination steps. Orange areas depict deamination steps. Blue areas depict oxidations. Red-shadowed genes (gabT, γ-aminobutyrate (GABA)/5AVA amino transferase; gabD, and succinate/glutarate-semialdehyde dehydrogenase) are native genes of Corynebacterium glutamicum. Additionally, copies of gabTD from Pseudomonas stutzeri were added by heterologous overexpression. The grey-shadowed gene (gdh, l-glutamic acid dehydrogenase) is native in C. glutamicum. The dark green-shadowed genes (ldcC, l-lysine decarboxylase; patA, putrescine transaminase; and patD, γ-aminobutyraldehyde dehydrogenase) are originally from Escherichia coli and were added by heterologous overexpression. The blue-shadowed gene (puo, putrescine oxidase; EC 1.4.3.10) was originally from Rhodococcus qingshengii and Paenarthrobacter aurescens, and it was heterologously expressed in C. glutamicum.

Design of the Study: Comparing Flux Enforcement with Either a Single or Two Coupling Sites
The five-step LdcC-PatA-PatD-GabT-GabD route from l-lysine to glutarate conserves the energy available to the cell in two ways. First, the oxidoreductases PatD and GabD yield the reduced redox cofactor NADH. Second, during the conversion of cadaverine to 5-aminopentanal and 5AVA to glutarate semialdehyde, the transaminases PatA and GabT simultaneously yield l-glutamic acid from the TCA cycle intermediate 2-oxoglutarate. The operation of the reactions of PatA and/or GabT in the glutarate production pathway obviates the need to synthesize l-glutamic acid from 2-oxoglutarate and ammonium by NADPH-dependent l-glutamic acid dehydrogenase, which is encoded by gdh. Indeed, flux enforcement by the deletion of gdh was found to improve glutarate production via the five-step LdcC-PatA-PatD-GabT-GabD route [15] because the deletion of gdh was compensated for by the activities of the two transaminases PatA and GabT. This prompted us to analyze whether flux enforcement by gdh deletion coupled to a single transaminase reaction in the glutarate pathway was superior to coupling to the two transaminases PatA and GabT.
The fourth reaction in the five-step reaction sequence (transamination by GabT) was kept in order to maintain the "metabolic pull" of flux enforcement by gdh deletion, while the second reaction (transamination by PatA) was chosen to be replaced by an oxidative deamination reaction. Unlike the 2-oxoglutarate-dependent transamination catalyzed by PatA, oxidative deamination by an oxidase does not yield l-glutamic acid; thus, it cannot bypass NADPH-dependent l-glutamic acid dehydrogenase. A search for oxidases that accept substrates similar to cadaverine was conducted. Indeed, putrescine oxidase (EC 1.4.3.10) encoded by puo from Rhodococcus erythropolis accepts cadaverine as a substrate with a catalytic efficiency (k cat /K m ) of 220 s −1 mM −1 [20]. Another candidate from Paenarthrobacter aurescens also showed enzyme activity towards cadaverine, albeit with magnitudes lower k cat /K m (4.78 s −1 mM −1 ) [21]. Both enzymes contain non-covalently bound FAD. The putrescine oxidase gene puo Rq from Rhodococcus qingshengii was used due to the 98% identity similarity of its amino acid sequence to the characterized putrescine oxidase proteins from R. erythropolis [20] and the fact that genomic DNA from R. qingshengii was available from the strain collection Deutsche Sammlung von Mikroorganismen und Zellkulturen (DSMZ). The putrescine oxidases from P. aurescens (puo Pa ) and R. qingshengii (puo Rq )were tested in the second position of the five-step pathway for conversion of l-lysine to glutarate in order to compare the new routes (LdcC-Puo Rq -PatD-GabT-GabD and LdcC-Puo Pa -PatD-GabT-GabD) to the reference route of LdcC-PatA-PatD-GabT-GabD.

Proof of Principle: Putrescine Oxidases Oxidatively Deaminate Cadaverine in C. glutamicum
To test whether the putrescine oxidases from P. aurescens and R. qingshengii catalyzed the oxidative deamination of cadaverine to 5AVA in C. glutamicum, 5AVA production via the three-step pathway was analyzed after replacing transamination by PatA with a putrescine oxidase. The putrescine oxidase genes from P. aurescens and R. qingshengii, respectively, were heterologously expressed in the l-lysine-producing C. glutamicum strain GSLA2 to generate the LdcC-Puo Rq -PatD and LdcC-Puo Pa -PatD variants of the reference three-step LdcC-PatA-PatD route. GSLA2 is based on GRLys1, which was obtained by genome reduction and rational metabolic engineering of l-lysine production [22][23][24], but it also lacks the genes sugR, ldhA, snaA and cgmA to improve glucose consumption and to avoid N-acetylation and the export of cadaverine [8,23,[25][26][27]. The AVA1_patA reference strain produced 1.1 ± 0.0 g/L 5AVA in 3 mL microtiter plates ( Figure 2A). Notably, the strains with the new LdcC-Puo Rq -PatD and LdcC-Puo Pa -PatD routes produced 0.4 ± 0.0 g/L 5AVA (AVA1_puoRq) and 0.1 ± 0.0 g/L 5AVA (AVA1_puoPa), respectively. Thus, a proof of concept for the fermentative production of 5AVA with putrescine oxidases was achieved.
The spectrum of by-products formed indicated the incomplete conversion of L-lysine to 5AVA since 1.1 ± 0.1 and 3.6 ± 0.3 g/L of L-lysine and 0.3 ± 0.0 and 0.1 ± 0.0 g/L of cadaverine accumulated in the cultivations with AVA1_puoRq and AVA1_puoPa, respectively ( Figure 2A). This was not observed with the AVA1_patA reference strain. All strains produced glutarate (Figure 2A), which may be synthesized from 5AVA by endogenous, chromosomally encoded enzymes, namely GabT and GabD.

Deletion of Genes for Conversion of 5AVA to Glutarate and Change of the Microcultivation System Improved 5AVA Production
The accumulation of glutarate by the strains with the different three-step routes (LdcC-PatA-PatD, LdcC-PuoRq-PatD and LdcC-PuoPa-PatD) ( Figure 2A) prompted us to delete the GABA utilization operon. The deletion of the gabTDP operon has previously been shown to abrogate glutarate formation when using the LdcC-PatA-PatD reference route [10]. The assumption was that this would abolish glutarate formation as a by-product of the new AVA2_puoRq and AVA2_puoPa strains. It was, in fact, shown to be correct because glutarate was not detected after 48 h of the 3 mL microtiter plate cultivation of AVA2_puoRq, AVA2_puoPa, and the reference strain AVA2_patA ( Figure 2B). Compared to the strains possessing gabTDP operon (Figure 2A), the production of 5AVA increased two-to-five fold to final titers of 0.7 ± 0.0 g/L for AVA2_puoRq, 0.2 ± 0.0 g/L for AVA2_puoPa, and 5.2 ± 0.1 g/L for the reference strain AVA2_patA ( Figure 2B). While the accumulation of cadaverine was comparable for the three strains, the C. glutamicum strains Figure 2. Production titers of 5AVA and by-products by strains derived from (A) C. glutamicum GSLA2(pVWEx1-ldcC) and (B) GSLA2∆gabTDP(pVWEx1-ldcC). Three milliliter cultures were grown in the Duetz system with a 4% glucose minimal medium supplemented with 1 mM isopropyl βd-1-thiogalactopyranoside (IPTG) and harvested after 48 h. Values and error bars represent mean and standard deviation values (n = 3 cultivations).

Deletion of Genes for Conversion of 5AVA to Glutarate and Change of the Microcultivation System Improved 5AVA Production
The accumulation of glutarate by the strains with the different three-step routes (LdcC-PatA-PatD, LdcC-Puo Rq -PatD and LdcC-Puo Pa -PatD) ( Figure 2A) prompted us to delete the GABA utilization operon. The deletion of the gabTDP operon has previously been shown to abrogate glutarate formation when using the LdcC-PatA-PatD reference route [10]. The assumption was that this would abolish glutarate formation as a by-product of the new AVA2_puoRq and AVA2_puoPa strains. It was, in fact, shown to be correct because glutarate was not detected after 48 h of the 3 mL microtiter plate cultivation of AVA2_puoRq, AVA2_puoPa, and the reference strain AVA2_patA ( Figure 2B). Compared to the strains possessing gabTDP operon (Figure 2A), the production of 5AVA increased two-to-five fold to final titers of 0.7 ± 0.0 g/L for AVA2_puoRq, 0.2 ± 0.0 g/L for AVA2_puoPa, and 5.2 ± 0.1 g/L for the reference strain AVA2_patA ( Figure 2B). While the accumulation of cadaverine was comparable for the three strains, the C. glutamicum strains AVA2_puoRq and AVA2_puoPa accumulated l-lysine to high concentrations (6.9 ± 0.4 and 6.0 ± 0.3 g/L, respectively; Figure 2B). This indicated the incomplete conversion of l-lysine to 5AVA by the AVA2_puoRq and AVA2_puoPa strains.
Since the putrescine oxidase reaction requires molecular oxygen, we analyzed 5AVA production in another microcultivation system while allowing for a higher oxygen transfer rate than the current system (i.e., a Duetz system with a cultivation volume of 3 mL at 220 rpm). We used a BioLector with different filling volumes (800, 1000, and 1200 µL) at 1300 rpm, which are known to provide superior oxygen transfer rates [28,29]. Cultivation in the BioLector system lead to increased production of 5AVA by AVA2_puoRq about five-fold (3.5 ± 0.2, 3.5 ± 0.5, and 3.7 ± 0.4 g/L with filling volumes of 800, 1000, and 1200 µL) in comparison to the Duetz system. 5AVA production by AVA2_puoPa was increased by about two-fold (0.4 ± 0.0, 0.4 ± 0.0, and 0.5 ± 0.1 g/L with filling volumes of 800, 1000, and 1200 µL, respectively), as can be seen by comparing Figure 3 with Figure 2B. Notably, significantly less l-lysine was accumulated by the strain overexpressing puo Rq than the strain overexpressing puo Pa (4.3-4.7 g/L compared to 6.9 ± 0.4 g/L, respectively), as can be seen in Figure 3.
Taken together, cultivation with high oxygen transfer rates in the BioLector system improved 5AVA production by the new three step route and allowed us to identify the better performance of the LdcC-Puo Rq -PatD variant than of the LdcC-Puo Pa -PatD variant; however, the LdcC-Puo Rq -PatD variant was found to be inferior to the LdcC-PatA-PatD route in regards to 5AVA production. Nevertheless, we continued to use the LdcC-Puo Rq -PatD variant with the putrescine oxidase from R. qingshengii to test its performance for glutarate production and flux enforcement.
Catalysts 2020, 10, x FOR PEER REVIEW 6 of 15 AVA2_puoRq and AVA2_puoPa accumulated L-lysine to high concentrations (6.9 ± 0.4 and 6.0 ± 0.3 g/L, respectively; Figure 2B). This indicated the incomplete conversion of L-lysine to 5AVA by the AVA2_puoRq and AVA2_puoPa strains.
Since the putrescine oxidase reaction requires molecular oxygen, we analyzed 5AVA production in another microcultivation system while allowing for a higher oxygen transfer rate than the current system (i.e., a Duetz system with a cultivation volume of 3 mL at 220 rpm). We used a BioLector with different filling volumes (800, 1000, and 1200 μL) at 1300 rpm, which are known to provide superior oxygen transfer rates [28,29]. Cultivation in the BioLector system lead to increased production of 5AVA by AVA2_puoRq about five-fold (3.5 ± 0.2, 3.5 ± 0.5, and 3.7 ± 0.4 g/L with filling volumes of 800, 1000, and 1200 μL) in comparison to the Duetz system. 5AVA production by AVA2_puoPa was increased by about two-fold (0.4 ± 0.0, 0.4 ± 0.0, and 0.5 ± 0.1 g/L with filling volumes of 800, 1000, and 1200 μL, respectively), as can be seen by comparing Figure 3 with Figure 2B. Notably, significantly less L-lysine was accumulated by the strain overexpressing puoRq than the strain overexpressing puoPa (4.3-4.7 g/L compared to 6.9 ± 0.4 g/L, respectively), as can be seen in Figure 3. Taken together, cultivation with high oxygen transfer rates in the BioLector system improved 5AVA production by the new three step route and allowed us to identify the better performance of the LdcC-PuoRq-PatD variant than of the LdcC-PuoPa-PatD variant; however, the LdcC-PuoRq-PatD variant was found to be inferior to the LdcC-PatA-PatD route in regards to 5AVA production. Nevertheless, we continued to use the LdcC-PuoRq-PatD variant with the putrescine oxidase from R. qingshengii to test its performance for glutarate production and flux enforcement.

Flux Enforcement by Deletion of the L-Glutamic Acid Dehydrogenase Gene Improved Glutarate Production
Production can be coupled to growth requirements by flux enforcement. The flux enforcement of glutarate production has been achieved by the deletion of the L-glutamic acid dehydrogenase gene in C. glutamicum [15]. L-glutamic acid dehydrogenase is the major reactant that is necessary for the assimilation of ammonium into L-glutamic acid in C. glutamicum [30]. 2-Oxoglutarate-dependent transaminase reactions in synthetic cascades can substitute for L-glutamic acid dehydrogenase if they operate in the direction of L-glutamic acid formation. In the five-step pathway from L-lysine to glutarate with LdcC, PatA, PatD, GabT, and GabD, L-glutamic acid is synthesized from 2oxoglutarate by two reactions, namely those catalyzed by PatA and by GabT. Flux enforcement by

Flux Enforcement by Deletion of the l-Glutamic Acid Dehydrogenase Gene Improved Glutarate Production
Production can be coupled to growth requirements by flux enforcement. The flux enforcement of glutarate production has been achieved by the deletion of the l-glutamic acid dehydrogenase gene in C. glutamicum [15]. l-glutamic acid dehydrogenase is the major reactant that is necessary for the assimilation of ammonium into l-glutamic acid in C. glutamicum [30]. 2-Oxoglutarate-dependent transaminase reactions in synthetic cascades can substitute for l-glutamic acid dehydrogenase if they operate in the direction of l-glutamic acid formation. In the five-step pathway from l-lysine to glutarate with LdcC, PatA, PatD, GabT, and GabD, l-glutamic acid is synthesized from 2-oxoglutarate by two reactions, namely those catalyzed by PatA and by GabT. Flux enforcement by the deletion of the l-glutamic acid dehydrogenase gene gdh has been shown to improve glutarate production [15]. We reasoned that the beneficial effect of gdh deletion for the flux enforcement of glutarate production may be more pronounced in a five-step pathway from l-lysine to glutarate that only involves a single transaminase reaction instead of two. Therefore, we constructed a strain that expressed the oxidase gene puo Rq instead of the transaminase gene patA. The GSLA2G strain, which lacked gdh coding for l-glutamic acid dehydrogenase, was transformed with plasmids for the expression of either the LdcC-Puo Rq -PatD-GabT-GabD variant with a single transaminase (GLUT_puoRq) or, as a reference, the LdcC-PatA-PatD-GabT-GabD variant with two transaminases (GLUT_patA). The GLUT_puoRq strain grew slower (0.08 ± 0.00 compared to 0.11 ± 0.00 1/h, respectively) and to an about 20% lower maximal biomass concentration (3.6 ± 0.1 compared to 4.5 ± 0.2 g/L cell dry weight, respectively) than the GLUT_patA reference strain. Importantly, while the by-product formation was similar for both strains, GLUT_puoRq produced about 40% more glutarate (7.7 ± 0.7 compared to 5.5 ± 0.2 g/L, respectively; Figure 4) than the reference strain GLUT_patA. This confirmed our hypothesis that flux enforcement by gdh deletion coupled to a production pathway with a single transaminase was superior to a route using two transaminase reactions as metabolic coupling sites.
Catalysts 2020, 10, x FOR PEER REVIEW 7 of 15 involves a single transaminase reaction instead of two. Therefore, we constructed a strain that expressed the oxidase gene puoRq instead of the transaminase gene patA. The GSLA2G strain, which lacked gdh coding for L-glutamic acid dehydrogenase, was transformed with plasmids for the expression of either the LdcC-PuoRq-PatD-GabT-GabD variant with a single transaminase (GLUT_puoRq) or, as a reference, the LdcC-PatA-PatD-GabT-GabD variant with two transaminases (GLUT_patA). The GLUT_puoRq strain grew slower (0.08 ± 0.00 compared to 0.11 ± 0.00 1/h, respectively) and to an about 20% lower maximal biomass concentration (3.6 ± 0.1 compared to 4.5 ± 0.2 g/L cell dry weight, respectively) than the GLUT_patA reference strain. Importantly, while the by-product formation was similar for both strains, GLUT_puoRq produced about 40% more glutarate (7.7 ± 0.7 compared to 5.5 ± 0.2 g/L, respectively; Figure 4) than the reference strain GLUT_patA. This confirmed our hypothesis that flux enforcement by gdh deletion coupled to a production pathway with a single transaminase was superior to a route using two transaminase reactions as metabolic coupling sites.

Discussion
This study showed that the oxidative deamination of cadaverine by putrescine oxidase Puo can operate in C. glutamicum to replace cadaverine transaminase PatA in synthetic pathways that convert cadaverine to 5AVA or glutarate. Studying glutarate biosynthesis pathways with either Puo or PatA allowed us to determine the effectiveness of flux enforcement (Figure 4). We showed how L-glutamic acid synthesis (which is required for growth) that is conducted by either one or two transamination reactions in the synthetic glutarate biosynthesis pathways can compensate for the absence of Lglutamic acid dehydrogenase caused by the deletion of its gene gdh. Glutarate production was higher when it was coupled to L-glutamic acid synthesis by one transamination (GabT) compared to a 1:2 stoichiometry, i.e., when L-glutamic acid was formed in reactions catalyzed by two transaminases (PatA and GabT) in the glutarate production pathway. Puo is not more efficient than PatA per se, because 5AVA production via the LdcC-PuoRq-PatD route was found to be lower than production via the LdcC-PatA-PatD route ( Figure 2B). Thus, the better glutarate production via the LdcC-PuoRq-PatD-GabT-GabD route compared to the LdcC-PatA-PatD-GabT-GabD route (Figure 4) resulted from the better efficiency of flux enforcement with a single coupling site as compared to two coupling sites (Figure 1).
Putrescine oxidase belongs to the group of diamine oxidases found in bacteria, archaea and eukaryotes that are usually active with short aliphatic diamines in order to utilize them as carbon and/or nitrogen sources [31]. Applications in enzyme catalysis benefit from the fact that these flavoproteins utilize molecular oxygen as an electron acceptor instead of costly redox cofactors such as NAD(P)H [20,21,31,32], e.g., for the enzymatic transformation of cadaverine to 5-aminopentanal by a diamine oxidase [33]. One constraint to their application is the concomitant formation of hydrogen peroxide, an oxidative stressor that causes significant damage to the enzymes used in biocatalysis and to DNA and proteins in vivo [34]. As C. glutamicum, the production host chosen here, shows a high natural resistance to hydrogen peroxide due to its very active catalase (commercialized as CAS Number 9001-05-2 [34]), pathways operating with putrescine oxidase functioned well.

Discussion
This study showed that the oxidative deamination of cadaverine by putrescine oxidase Puo can operate in C. glutamicum to replace cadaverine transaminase PatA in synthetic pathways that convert cadaverine to 5AVA or glutarate. Studying glutarate biosynthesis pathways with either Puo or PatA allowed us to determine the effectiveness of flux enforcement (Figure 4). We showed how l-glutamic acid synthesis (which is required for growth) that is conducted by either one or two transamination reactions in the synthetic glutarate biosynthesis pathways can compensate for the absence of l-glutamic acid dehydrogenase caused by the deletion of its gene gdh. Glutarate production was higher when it was coupled to l-glutamic acid synthesis by one transamination (GabT) compared to a 1:2 stoichiometry, i.e., when l-glutamic acid was formed in reactions catalyzed by two transaminases (PatA and GabT) in the glutarate production pathway. Puo is not more efficient than PatA per se, because 5AVA production via the LdcC-PuoRq-PatD route was found to be lower than production via the LdcC-PatA-PatD route ( Figure 2B). Thus, the better glutarate production via the LdcC-PuoRq-PatD-GabT-GabD route compared to the LdcC-PatA-PatD-GabT-GabD route (Figure 4) resulted from the better efficiency of flux enforcement with a single coupling site as compared to two coupling sites (Figure 1).
Putrescine oxidase belongs to the group of diamine oxidases found in bacteria, archaea and eukaryotes that are usually active with short aliphatic diamines in order to utilize them as carbon and/or nitrogen sources [31]. Applications in enzyme catalysis benefit from the fact that these flavoproteins utilize molecular oxygen as an electron acceptor instead of costly redox cofactors such as NAD(P)H [20,21,31,32], e.g., for the enzymatic transformation of cadaverine to 5-aminopentanal by a diamine oxidase [33]. One constraint to their application is the concomitant formation of hydrogen peroxide, an oxidative stressor that causes significant damage to the enzymes used in biocatalysis and to DNA and proteins in vivo [34]. As C. glutamicum, the production host chosen here, shows a high natural resistance to hydrogen peroxide due to its very active catalase (commercialized as CAS Number 9001-05-2 [34]), pathways operating with putrescine oxidase functioned well. However, the decrease in the growth rate of the strains with Puo-supported glutarate production in comparison to strain with PatA-supported glutarate production may be attributed to H 2 O 2 accumulation.
Aerobic fermentation processes are often limited by low oxygen transfer rates, which may be detrimental to growth and production, in particular for biosynthesis routes requiring molecular oxygen. Since putrescine oxidase requires molecular oxygen [31,35], cultivation platforms supporting different oxygen transfer rates (Duetz microtiter plates (MTPs) and BioLector MTPs) were compared [28,29]. Production was higher in the platform with the higher oxygen transfer rates (as can be seen by comparing Figures 2B and 3) in the strain where 5AVA production was based on oxidase activity. The observed benefit may not only have been due to the better provision of molecular oxygen for putrescine oxidase, since even when assessing only the production of precursor l-lysine, C. glutamicum requires a minimum oxygen transfer rate of 14 mmol L −1 h −1 for optimal biomass and l-lysine production, which cannot be reached in Duetz plates [29,36]. This oxygen bottleneck was also observed in closed tubes for an enzyme activity assay that employed the DavA-DavB route from l-lysine via 5-aminovaleramide to 5AVA, where davB encodes a l-lysine monooxygenase [12]. Thus, the better performance of the three-step LdcC-PatA-PatD pathway for the conversion of l-lysine to 5AVA compared to the variants with putrescine oxidase Puo Rq or Puo Pa instead of transaminase PatA (LdcC-Puo Rq -PatD and LdcC-Puo Pa -PatD) may also have been due the fact that transaminase PatA does not require molecular oxygen in oxygen-restricted conditions (Duetz MTPs).
The substrate spectrum of putrescine oxidase is relatively broad because, besides the C4 diamine putrescine and the C5 cadaverine, longer diamines (hexamethylenediamine, spermine, spermidine) are also accepted-though the shorter C3 diaminopropane is not [20]. This feature is not limited to the oxidases, as it is also true for the transaminases PatA and GabT (accepting substrates of different chain lengths: diamines such as putrescine and cadaverine and ω-amino acids such as GABA and 5AVA, respectively,). Transaminases, in particular, are known for their promiscuity towards a broad range of substrates [37,38], which proves advantageous in the evolutionary sense but can translate to a drawback for short-term production experiments [39]. Nevertheless, the broad substrate spectra have been employed for the metabolic engineering of diamines, ω-amino acids, and dicarboxylic acids [14], though it has had to be ensured that only a precursor of one chain length was synthesized. This specificity has been achieved by narrow spectrum decarboxylases such as l-ornithine decarboxylase yielding putrescine from l-ornithine for the C4 products or, as also used here, l-lysine decarboxylase for C5 products [10,14,40].
The cascading of enzymes in biocatalysis [41] helps to overcome the need for the purification of intermediates, to regenerate redox cofactors [42], or to displace the reaction equilibrium towards product formation when coproducts tautomerize, dimerize, cyclize, or polymerize, as has been shown for self-sufficient transamination reactions [43]. Cascaded enzymes in metabolic pathways typically allow for coupling to cellular respiration, e.g., for redox cofactor regeneration in whole-cell biotransformations [44]. For fermentation processes, flux enforcement is a powerful tool to link production to an essential metabolic pathway. In C. glutamicum, l-lysine and glutarate production titers have benefitted from this technique [6,15,19]. Similarly, the metabolic cutoffs of genes in the pentose phosphate pathway have been found to enforce the co-utilization of pentoses with methanol by complementation with genes for methanol and formaldehyde assimilation [45,46].
Here, we compared flux enforcements with two or one coupling sites, specifically to see whether the synthesis of l-glutamic acid in the absence of l-glutamic acid dehydrogenase was enabled by GabT alone or by PatA and GabT if glutarate was produced. With one coupling site (GabT alone), the titer of glutarate was increased by 40% compared to two coupling sites (Figure 4). Theoretically, a change from a 2:1 stoichiometry to a 1:1 stoichiometry in flux enforcement should have increased the glutarate titer by 100%. Similar observations were published for the flux enforcement of l-lysine production by the deletion of the genes for the TCA cycle enzyme succinyl-CoA synthetase genes, and it was hypothesized that bypassing reactions such as the glyoxylate shunt compensated, to some extent, for the lack of succinyl-CoA synthetase [19]. Similarly, bypassing reactions compensating for the lack of l-glutamic acid dehydrogenase may have reduced the beneficial effect of flux enforcement. In addition to l-glutamic acid dehydrogenase, C. glutamicum possesses the glutamine synthetase/glutamine-oxoglutarate amidotransferase (GS/GOGAT) system for l-glutamic acid synthesis and ammonium assimilation [47]. The combined reactions of glutamine synthetase and glutamine-oxoglutarate amidotransferase differ from l-glutamic acid dehydrogenase with the requirement for ATP; thus, the GS/GOGAT-encoding genes are only upon induced nitrogen starvation with basal expression at the nitrogen concentrations used here for growth and production [48,49]. Thus, it is conceivable that GS/GOGAT activity reduced the beneficial effect of flux enforcement by the gdh deletion observed here (Figure 4). The additional deletion of the GS/GOGAT genes in the gdh deletion strain is not helpful because l-glutamic acid dehydrogenase and the GS/GOGAT system are the only pathways for net ammonium assimilation, whereas transaminases only transfer ammonium from an existing donor to an oxoacid acceptor.
Glutarate production may be rationally improved through metabolic engineering targeting, e.g., precursor supply, redox cofactor regeneration, or by-product formation [50], as well as by enzyme engineering, e.g., targeting catalytic efficiency or the product inhibition of the synthetic glutarate biosynthesis pathway [51]. Recently, the application of adaptive laboratory evolution (ALE), which has revealed answers to fundamental evolutionary questions in E. coli [52,53], gained attention in strain development for growth-associated fermentative processes [54]. In C. glutamicum, ALE has been exploited for increased methanol-dependent growth and tolerance, followed by whole-genome sequencing and mutational analysis, respectively [45,55,56]. Notably, flux enforcement by gdh deletion in the glutarate-producing strains described here and previously [15] reduced the growth rate. Thus, it may be possible to accelerate glutarate production by this strain through ALE and to rationalize the selected genomic changes by reverse genetics. Importantly, reaching the industrial maturity of glutarate production by recombinant C. glutamicum genetic and metabolic robustness, as well the scaling-up of such production, remains to be achieved.

Molecular Biology Methods
The isolation of the genomic DNA of C. glutamicum and classical methods that include the plasmid isolation, molecular cloning, and heat-shock transformation of E. coli and the electroporation of C. glutamicum were performed as described previously [58,59]. ALLin HiFi DNA Polymerase (HighQu, Kraichtal, Germany) was used to amplify DNA sequences, with plasmid or genomic DNA used as the template. The oligonucleotides that were used as primers in this study are listed in Table 3.
The genomic DNA (gDNA) of Rhodococcus qingshengii (DSM 45257) and Paenarthrobacter aurescens (DSM 20116) was obtained from the German Collection of Microorganisms and Cell Cultures (DSMZ), and the gDNA of E. coli MG1655 was isolated in [58]. The pBV2xp plasmid was digested with the restriction enzymes BamHI and SacI (New England Biolabs, Ipswich, MA, USA). The DNA fragments were joined by the means of isothermal DNA assembly [60]. To overexpress puo Rq , puo Pa , and patD in C. glutamicum, their genes were amplified from pBV2xp-puo Rq -patD, pBV2xp-puo Pa -patD, and the genomic DNA of C. glutamicum, respectively, and they were assembled into BamHI-linearized pEKEx3 and pEC-XT99A by Gibson Assembly using the respective primers. C. glutamicum was transformed with the constructed plasmids and empty vectors.

HPLC Analysis
The quantification of amino acids and cadaverine and glutarate in the cultivation medium was performed with a high-pressure liquid chromatography system (1200 series, Agilent Technologies Deutschland GmbH, Böblingen, Germany), as described previously [61]. After the centrifugation of 1 mL cell cultures at 14000 rpm for 10 min, the supernatant was stored at −20 • C prior to analysis.
The amino acids l-lysine and 5AVA and the diamine cadaverine were detected with a fluorescence detector (FLD G1321A, 1200 series, Agilent Technologies) after the derivatization of the samples with OPA (ortho-phthaldialdehyde). The detection of glutarate was done with a refractive index detector (RID G1362A, 1200 series, Agilent Technologies) and a diode array detector (DAD G1315B, 1200 series, Agilent Technologies).

Conclusions
By exchanging a putrescine transaminase with an oxidase in a synthetic glutarate pathway in C. glutamicum, a proof of concept for the production of 5AVA and glutarate was achieved. The heterologous expression of putrescine oxidase from two strains, R. qingshengii and P. aurescens, showed 5AVA production titers of 0.4 ± 0.0 and 0.1 ± 0.0 g/L 5AVA, respectively, and could be improved by the deletion of the glutarate module gabTDP and improved oxygen availability to a maximum of 3.7 ± 0.4 g/L 5AVA with the LdcC-Puo Rq -PatA-GabT-GabD route. Upon the disruption of gdh, which encodes the enzyme that is responsible for the major nitrogen assimilation reaction in C. glutamicum, the focus of the flux enforcement towards glutarate production tightened on a single transaminase reaction, as compared to the two transaminases PatA and GabT, resulting in a product titer increase of 40% (7.7 ± 0.7 vs. 5.5 ± 0.2 g/L). These results highlight requirements regarding the modularity and stoichiometry of the pathway-specific flux enforcement for microbial production.