Overcoming NADPH product inhibition improves D-sorbitol conversion to L-sorbose

Gluconobacter oxydans sorbitol dehydrogenase (GoSLDH) exhibits a higher catalytic efficiency than other l-sorbose producing enzymes. During the reaction catalysed by GoSLDH, NADP+ is reduced to NADPH and d-sorbitol is oxidized to l-sorbose. However, GoSLDH activity is inhibited by the NADPH (Ki = 100 μM) formed during the enzymatic reaction. Therefore, Escherichia coligosldh-lrenox producing both GoSLDH for d-sorbitol oxidation and LreNOX (NAD(P)H oxidase from Lactobacillus reuteri) for NADP+ regeneration was generated and used for l-sorbose production. Whole cell biocatalysts with the LreNOX cofactor recycling system showed a high conversion rate (92%) of d-sorbitol to l-sorbose in the presence of low concentration of NADP+ (0.5 mM). By alleviating NADPH accumulation during the catalytic reactions, E. coligosldh-lrenox exhibited 23-fold higher conversion rate of d-sorbitol than E. coligosldh. l-Sorbose production by E. coligosldh-lrenox reached 4.1 g/L after 40 min, which was 20.5-fold higher than that of E. coligosldh. We also constructed G. oxydansgosldh and G. oxydansgosldh-lrenox strains, and they exhibited 1.2- and 2.9-fold higher conversion rates than the wild-type G. oxydans KCTC 1091. The results indicate that overcoming NADPH product inhibition using LreNOX improves chemical production in NADP+-dependent enzymatic reactions.

In the current study, we characterized a novel water-forming LreNOX from L. reuteri showing a high cofactor preference towards NADPH unlike most other NAD(P)H oxidases (NOXs) which exclusively use NADH as a substrate. We developed a co-expression system in which GoSLDH encoded by the G. oxydans G624 gosldh was used as the l-sorbose producing enzyme and LreNOX was used as the cofactor-regenerating enzyme. Further, we demonstrated a simple, highly efficient, and economical whole-cell biocatalysis system comprised of GoSLDH coupled with LreNOX to regenerate the cofactor NADP + from NADPH. This system reduces the NADPH inhibition effect in the GoSLDH reaction and enables high production of l-sorbose from d-sorbitol.

Results and Discussion
GoSLDH inhibition by NADPH. Escherichia coli pET28-gosldh carrying GoSLDH, which encodes SLDH from G. oxydans G624, was constructed in our previous study 17 . During the catalytic process, NADP + was reduced to NADPH, and d-sorbitol was oxidized to l-sorbose. Purified GoSLDH from the induced E. coli pET28-gosldh showed high catalytic activity of 3570 U mg protein −1 (in the direction of l-sorbose production). During whole cell biocatalysis by E. coli gosldh in the presence of 0.5 mM NADP + , the NADPH concentration reached 147 μM, showing a conversion rate of d-sorbitol to l-sorbose of only 2.6% (Fig. 1a,b). To gain insight into the inhibitory effects of NADPH, the inhibition constant (K i ) of NADPH for GoSLDH was determined. Double reciprocal plots revealed competitive inhibition of GoSLDH by NADPH (Fig. 2a). The K i value of NADPH was determined to be 100 µM. Inhibition kinetic analysis, performed with NADPH concentrations ranging from 0 to 170 µM, showed that the apparent K m increased as the NADPH concentration increased. K m increased from 3.9-to 8.9-fold compared to the values without supplementary NADPH as NADPH concentration increased from 70 to 150 µM. After 20 s of E. coli gosldh whole cell biocatalysis, up to 160 μM NADPH was accumulated, resulting in 80% inhibition of GoSLDH activity (Fig. 1b). The reduction in NADPH formation would likely increase l-sorbose production by reducing GoSLDH inhibition. Therefore, we established a cofactor regeneration system using a low initial concentration (0.5 mM) of NADP + .  Binding thermodynamics studies. We performed isothermal titration calorimetry (ITC) experiments to characterize the binding thermodynamics of NADP + to GoSLDH (100 μM) in 100 mM glycine-NaOH buffer (pH 10) at 25 °C. As shown in Fig. 3a, the binding of NADP + to GoSLDH in the absence of d-sorbitol was exothermic (ΔH =-754 cal mol −1 ) with a K a value of 1.12 × 10 4 M −1 or K d (dissociation constant) value of 89.3 μM. This is equal to a ΔG value of −5.5 kJ mol −1 , showing that the binding is mostly entropy-driven (with a calculated ΔS of 16.0 cal mol −1 deg −1 ). Next, binding of NADPH to GoSLDH (100 μM) was found to be exothermic (ΔH = −578 cal mol −1 ) with K a (affinity constant) value of 1.49 × 10 4 M −1 or a K d value of 67.1 μM, indicating the higher binding affinity of GoSLDH for NADP + (22 μM lower K d ) (Fig. 3b). This equals a ΔG value of −5.6 kJ mol −1 , showing the binding is mostly entropy-driven (with a calculated ΔS of 17.1 cal mol −1 deg −1 ). NADP + was also titrated into GoSLDH (100 μM) in the presence of NADPH (300 μM). As shown in Fig. 3c, binding of NADP + to the protein was not detected, suggesting that NADPH and NADP + compete with each other because these two molecules bind to the same site in GoSLDH. Because GoSLDH does not reduce reaction activity at 25 °C in the presence of 200 mM l-sorbose and 0.5 mM NADPH, NADPH solely inhibits the oxidation reaction of GoSLDH.
Characterization of NADH oxidase. NOXs catalyse the oxidation of NAD(P)H by simultaneously reducing molecular O 2 to water or hydrogen peroxide [24][25][26] . NOXs play important roles in regenerating oxidized cofactors required for numerous enzymatic reactions. Because nearly all proteins are deactivated upon exposure to H 2 O 2 , H 2 O-forming NOXs are considered as more effective than H 2 O 2 -forming NOXs for NAD(P) + regeneration 24 . Most NOXs show NADH oxidation activity rather than NADPH oxidation activity. In this study, we characterized a novel NOX from L. reuteri (LreNOX) which efficiently oxidized both NADPH and NADH. The NOX gene from L. reuteri was isolated from genomic DNA using specific primers derived from the coding sequence ( Table 1). The gene was cloned into the T7 promoter-based plasmid pET28a to produce pET28a-lrenox and then heterologously expressed in E. coli BL21 (DE3) cells. The enzyme was purified by nickel-nitrilotriacetic acid affinity chromatography; purity was assessed by polyacrylamide gel electrophoresis in the presence of sodium dodecyl sulphate (Fig. S1). The optimal pH and temperature for the oxidation of NADPH by purified LreNOX were 6.0 and 60 °C, respectively. Investigation of the kinetic parameters with NADPH as a substrate was performed using the purified enzyme LreNOX under the optimal assay conditions. Figure S2 shows that LreNOX oxidized NADPH with a high specific activity (V max = 161 U mg −1 ) and binding affinity (K m = 35 µM). Compared to L. san-Nox2, LreNOX showed 14.6-fold higher activity towards NADPH. This suggests that LreNOX is a good candidate for enzymatic regeneration of the oxidized cofactor NADP + .

Strategy for overcoming NADPH inhibition in l-sorbose production. GoSLDH catalyses strict
NADP + -dependent conversion of d-sorbitol to l-sorbose and exhibits NADPH product inhibition. Accumulation of NADPH during the SLDH reaction is a bottleneck in the conversion of d-sorbitol to l-sorbose. Therefore, it is important to reduce NADPH accumulation to increase l-sorbose production. To make the reaction more efficient and economical, it is necessary to regenerate the cofactor NADP + from NADPH. In the present study, we used highly active LreNOX to regenerate NADP + . In the coupled system of GoSLDH and LreNOX, LreNOX recycled the reduced cofactor NADPH in the reaction mixture to the oxidized cofactor NADP + . Cofactor regeneration may provide an alternative tool for not only reducing the product NADPH concentration but also minimizing the reaction cost because of the lower concentration (0.5 mM NADP + ) of the initial cofactor (Fig. 1a). When we used E. coli BL21-Codon Plus (DE3)-RIL cells, no conversion of d-sorbitol to l-sorbose was observed. Therefore, we used E. coli BL21-Codon Plus (DE3)-RIL cells as an expression host for lrenox and gosldh. The biocatalytic activities of the purified enzymes prepared from E. coli pET28-lrenox and E. coli pET28-gosldh cells were assayed using d-sorbitol as a substrate. First, the enzyme was purified from induced E. coli pET28-gosldh ( Table 1). The purified GoSLDH showed a high SLDH activity of 3570 U mg −1 in the direction of l-sorbose production. Second, E. coli pet28-lrenox harbouring lrenox, which encodes LreNOX from L. reuteri, was constructed. Most recombinant wild-type NOXs showed activity with NADH, but less or no activity with NADPH. To regenerate NADP + for l-sorbose production, we screened NOXs from different organisms and found that LreNOX can oxidise NADPH. LreNOX exhibited higher activity towards NADPH (160 U mg −1 ) than previously reported NOXs.
Because whole cells contain NAD + and NADP + reservoirs as a continuous source of cofactors 21 , E. coli-pET28-lrenox was used for cofactor NADP + regeneration in the current study. When GoSLDH and LreNOX are expressed individually in two different cells, the cellular membranes of the cells may limit the exchange of pyridine cofactors between the two cells 27 . To avoid retardation of cofactor exchange, E. coli gosldh-lrenox cells co-expressing LreNOX and GoSLDH were constructed as shown in Fig. 4. LreNOX oxidized the NADPH produced by GoSLDH and provided the oxidized form the cofactor NADP + for the biocatalytic process of GoSLDH, without the limitation of a transmembrane process. Efficient production of l-sorbose was possible by removing NADPH inhibition using E. coli gosldh-lrenox as the biocatalyst and d-sorbitol as the substrate (Fig. 4).

Effects of pH and cofactor concentration on l-sorbose production.
To achieve high product conversion, the effect of pH on l-sorbose yield from d-sorbitol was investigated. The reaction used 50 mM of d-sorbitol as the initial substrate and 3.0 mg dry cell weight (DCW) mL −1 of E. coli gosldh or E. coli gosldh-lrenox as the biocatalyst under different pH conditions. After a 1-h reaction with E. coli gosldh or E. coli gosldh-lrenox , the conversion rates were measured by monitoring the product concentration. The conversion rate was highest at pH 8.0 for E. coli gosldh-lrenox (32.4%) and pH 9.0 for E. coli gosldh (5.8%), respectively. The optimal pH values of purified GoSLDH and LreNOX were pH 10.0 and 5.0, respectively (Fig. 5a). Cofactor concentration is another important parameter in the catalytic process and requires careful monitoring because high cofactor concentrations may lead to enzyme inhibition and high production costs 18,19,28 . The effect of cofactor concentration on l-sorbose production was investigated (Fig. 1). During whole cell biocatalysis with E. coli gosldh in the presence of 0.7 mM NADP + , the NADPH concentration reached 147 μM, which is higher than the K i,NADPH value (100 μM) of GoSLDH. In contrast, biocatalysis with E. coli gosldh-lrenox produced only 32.2 μM NADPH (Fig. 1a), showing a 42.7% improvement in l-sorbitol conversion. As a result, E. coli gosldh-lrenox exhibited a 17.7-fold increase in d-sorbitol conversion to l-sorbose compared to E. coli gosldh (Fig. 1a). The conversion rates of E. coli gosldh-lrenox and E. coli gosldh were 46% and 3.4%, respectively, under the optimal NADP + concentration. LreNOX co-expression was beneficial for producing l-sorbose because of the reduction in NADPH accumulation (Fig. 1). The optimal initial concentration of NADP + was 0.7 mM for producing l-sorbose in the absence of the NADP + regeneration system. Incorporation of LreNOX lowered the NADP + requirement to 0.05 mM for the conversion of sorbitol to l-sorbose. In the current study, therefore, a 14-fold lower NADP + concentration was used with the NADP + regeneration system.

Effects of cell density and substrate concentration on l-sorbose production. The effect of E.
coli gosldh and E. coli gosldh-lrenox cell density on l-sorbose production was investigated (Fig. 5b). The conversion of d-sorbitol to l-sorbose increased in the range of 1.5-3 mg DCW mL −1 (E. coli gosldh-lrenox ). When the E. coli gosldh-lrenox DCW was increased to greater than 3.0 mg mL −1 , the production of l-sorbose remained nearly unchanged,  indicating that the optimal cell amount was 3.0 mg mL −1 . Under the optimized conditions, E. coli gosldh-lrenox exhibited a 47.4% conversion rate when we used 50 mM d-sorbitol, 0.5 mM NADP + , and 3.0 mg DCW mL −1 . For E. coli gosldh , the conversion rate was only 2-5.6%, although 1.5-15 mg DCW mL −1 was used. Substrate concentration is an important parameter determining the rate of biocatalysis. Therefore, we investigated the effect of substrate concentration on the conversion rate by varying the d-sorbitol concentration over 10-90 mM. The maximum conversion was 92.0% and 23.7% when 10 mM d-sorbitol was used with whole cell biocatalysts E. coli gosldh-lrenox and E. coli gosldh , respectively. The maximum l-sorbose concentrations were 4.1 and 0.23 g/L, respectively, when whole cell biocatalysts E. coli gosldh-lrenox and E. coli gosldh were used in presence of 50 mM d-sorbitol. Because of the oxidizing activities of membrane-bound dehydrogenases, G. oxydans has been used for industrial production of l-sorbose for decades [29][30][31] . Heterogeneous expression of sldh in Pseudomonas putida IF03738 was reported previously 32 , and purified SLDH showed 61.5% conversion with 1.12 g/L of l-sorbose at 25 °C for 24 h. In the current study, we obtained 3.7-fold higher production of l-sorbose with the whole cell biocatalyst E. coli gosldh-lrenox than that obtained with the SLDH system expressed in P. putida IF03738 32 .

Effects of lrenox expression in G. oxydans.
To further confirm the effect of lrenox expression on L-sorbose production, recombinant G. oxydans strains harboring the pBBR-adh-gosldh and pBBR-adh-gosldh/ lrenox plasmids were constructed (Fig. S3). Gene expression of gosldh and lrenox in G. oxydans was confirmed by real time PCR using the appropriate primers (Table S1, Fig. S3). G. oxydans (KCTC 1091), G. oxydans gosldh , and G. oxydans gosldh-lrenox showed conversion rates of 33.3%, 40.0%, and 96.6%, respectively, after a 12 h reaction (Fig. S4). The conversion rates of G. oxydans gosldh and G. oxydans gosldh-lrenox were 1.2 and 2.9 times higher, respectively, than that of wild-type G. oxydans. We also compared productivity of the strains. The L-sorbose productivity of G. oxydans gosldh-lrenox (145 g/L/h of L-sorbose from 150 g/L of D-sorbitol during a 12 h reaction) was 2.5 and 2.4 times higher than that of G. oxydans KCTC 1091 and G. oxydans gosldh strains, respectively. Results suggest that enhanced productivity of the G. oxydans gosldh-lrenox strain is attributable to NADP + regeneration, leading to minimization of NADPH-mediated GoSLDH inhibition. We have compared the present conversion rate achieved by E. coli with those achieved by Gluconobacter species (Table 2). Although we did not optimize the process for G. oxydans gosldh-lrenox culture in this study, the production rate of L-sorbose (12.1 g/L/h) was similar to the highest level reported previously (13.6 g/L/h). In this manuscript, we report for the first time that NADPH inhibits GoSLDH, and lrenox expression minimizes NADPH inhibition, thereby enhancing L-sorbose production by E. coli and G. oxydans.

Conclusions
GoSLDH catalyses strict NADP + -dependent interconversion between d-sorbitol and l-sorbose but shows NADPH product inhibition. To avoid inhibition by NADPH, we coupled a cofactor regenerating LreNOX with GoSLDH to perform bioconversion using whole cells of both E. coli gosldh-lrenox and E. coli gosldh as biocatalysts. In the coupled system, LreNOX was used to remove the reduced cofactor NADPH from the reaction mixture and recycle it to the oxidized cofactor NADP + . Biocatalytic synthesis of l-sorbose was successfully performed using the E. coli gosldh-lrenox whole-cell system. Under optimal conditions, the maximum l-sorbose conversion with E. coli gosldh-lrenox reached 92%, which is much higher than that obtained using the SLDH heterogeneous expression system (61.5%) 32 . l-Sorbose production with E. coli gosldh-lrenox was 4.1 g/L after a 40-min reaction, which was 20.5-fold higher than that using E. coli gosldh . Strategy for cofactor regeneration provides an efficient tool for preventing NADPH accumulation and product inhibition, and provides a new biocatalytic method for chemical production using the NADP + -dependent enzymatic reaction. Additionally, production costs can be reduced because of the lower NADP + amounts required.

Materials and Methods
Plasmids and reagents. Ex-Taq DNA polymerase, a genomic DNA extraction kit, pGEM-T easy vector, and reagents for polymerase chain reaction were purchased from Promega (Madison, WI, USA). T4 DNA ligase and restriction enzymes were purchased from New England Biolabs (Ipswich, MA, USA). The pET28a expression vector, plasmid isolation kit, and nickel-nitrilotriacetic acid superflow column for His-tag protein purification were obtained from Qiagen (Hilden, Germany) 20 . Oligonucleotide primers were obtained from Bioneer (Daejeon, Constructions of E. coli gosldh and E. coli gosldh-lrenox . The synthesized sldh sequence was based on the DNA sequence of the Polyol specific long-chain dehydrogenase from G. oxydans G624 (GenBank accession number AB028937.1). Codon optimization was performed as previously reported 17 . The optimized sldh sequences were named as gosldh. The gosldh fragments, flanked by NdeI and XhoI restriction enzymes sites at the 5′ and 3′ ends, respectively, were synthesized by GeneScript (Piscataway, NJ, USA) and cloned into the pUC57 vector. Codon-optimized gosldh was digested by NdeI-XhoI and then ligated into the expression vector pETDuet-1 from Novagen (Madison, WI, USA) to express gosldh. Additionally, the lrenox fragment was obtained from the genome of L. reuteri using primers Nox1 and Nox2 (Table 1). To construct a co-expression system carrying gosldh and lrenox, pEasy-Blunt-nox was digested with BamHI and NotI, and the gel-purified nox fragment was ligated to pETDuet-gosldh digested with the same restriction enzymes.  [7][8] or Glycine-NaOH buffer (pH 9-10) and maintained at 4 °C for further studies. To optimize the parameters of pH and cell density to produce l-sorbose from E. coli gosldh , the cell pellets described above were resuspended in 100 mM Tris-HCl buffer (pH 7-8) or Glycine-NaOH buffer of varying pH (9-10), at a cell density of 1.5-18 mg DCW mL −1 . The effect of d-sorbitol and NADP + concentrations on the rate of l-sorbose production from E. coli gosldh was evaluated by varying the initial d-sorbitol concentration from 10 to 90 mM and NADP + concentration from 0.1 to 0.8 mM. The reaction conditions for the coupled system were the same as those for producing l-sorbose with whole-cell E. coli gosldh alone. Escherichia coli carrying the expression plasmid pACYCDuet-1 with the gosldh genes was used as the control strain. For all combinations of pH, NADP + concentration, cell density, and substrate concentration, biotransformation of d-sorbitol to l-sorbose was conducted at 30 °C with shaking at 150 rpm. Samples of the cell suspension (1 mL) were obtained periodically and centrifuged at 13,000 rpm for 5 min; the supernatants were used to analyse l-sorbose concentration.