Combinatorial metabolic engineering of industrial Gluconobacter oxydans DSM2343 for boosting 5-keto-D-gluconic acid accumulation

Background L-(+)-tartaric acid (L-TA) is an important organic acid, which is produced from the cream of tartar or stereospecific hydrolysis of the cis-epoxysuccinate. The former method is limited by the availability of raw material and the latter is dependent on the petrochemical material. Thus, new processes for the economical preparation of L-TA from carbohydrate or renewable resource would be much more attractive. Production of 5-keto-D-gluconate (5-KGA) from glucose by Gluconobacter oxydans is the first step to produce L-TA. The aim of this work is to enhance 5-KGA accumulation using combinatorial metabolic engineering strategies in G. oxydans. The sldAB gene, encoding sorbitol dehydrogenase, was overexpressed in an industrial strain G. oxydans ZJU2 under a carefully selected promoter, P0169. To enhance the efficiency of the oxidation by sldAB, the coenzyme pyrroloquinoline quinone (PQQ) and respiratory chain were engineered. Besides, the role in sldAB overexpression, coenzyme and respiratory chain engineering and their subsequent effects on 5-KGA production were investigated. Results An efficient, stable recombinant strain was constructed, whereas the 5-KGA production could be enhanced. By self-overexpressing the sldAB gene in G. oxydans ZJU2 under the constitutive promoter P0169, the resulting strain, G. oxydans ZJU3, produced 122.48 ± 0.41 g/L of 5-KGA. Furthermore, through the coenzyme and respiratory chain engineering, the titer and productivity of 5-KGA reached 144.52 ± 2.94 g/L and 2.26 g/(L · h), respectively, in a 15 L fermenter. It could be further improved the 5-KGA titer by 12.10 % through the fed-batch fermentation under the pH shift and dissolved oxygen tension (DOT) control condition, obtained 162 ± 2.12 g/L with the productivity of 2.53 g/(L · h) within 64 h. Conclusions The 5-KGA production could be significantly enhanced with the combinatorial metabolic engineering strategy in Gluconobacter strain, including sldAB overexpression, coenzyme and respiratory chain engineering. Fed-batch fermentation could further enlarge the positive effect and increase the 5-KGA production. All of these demonstrated that the robust recombinant strain can efficiently produce 5-KGA in larger scale to fulfill the industrial production of L-TA from 5-KGA.

Background L-(+)-tartaric acid (L-TA), an important naturally existing hydroxyl carboxylic acid, is mainly used as antioxidant in food industry, as a chiral reagent in organic synthesis, as an acidic reducing agent in the textile industry and in galvanochemistry [1][2][3]. It is also an alternative to citric acid as an acidulant in food additives for its superior organoleptic properties [4]. Currently, the L-TA that is commercially available is produced exclusively through the stereospecific hydrolysis of cis-epoxysuccinate [5][6][7]. In this process, the reaction is catalyzed by the cis-epoxysuccinate hydrolase from Rhodococcus rhodochrous, Nocardia tartaricans, Corynebacterium sp. or Pseudomonas sp. [7]. However, cis-epoxysuccinate is derived from petrochemical-based precursor maleic anhydride, and this limits the production of L-TA. Therefore, development of the sustainable alternative solution for L-TA manufacture has recently attracted increasing attention.
A promising route has been employed for the production of L-TA by sequential whole-cell catalyzed oxidation and chemical catalysis, in which the glucose was first biologically converted to 5-keto-D-gluconate (5-KGA) by Gluconobacter oxydans, then to L-TA in the presence of ammonium vanadate as a trace element [4,8]. Thus, the strategy towards an efficient synthetic route to L-TA was to optimize the enzymatic production of 5-KGA [4,9].
So far several attempts have been made to increase 5-KGA production, such as overexpression of gluconate: NADP 5-oxidoreductase [9], inactivation of GA2DH [3,17] and overexpression of SLDH [18,19] and optimization of the media conditions [20]. The highest yield achieved of 5-KGA was 240-295 mM (about 60 g/L) in a batch fermentation over periods as long as 72 h with a productivity of 0.83 g/(L · h) [18]. However, the methods used are Fig. 1 Scheme of glucose metabolism and the respiratory chain in Gluconobacter oxydans DSM2343. Abbreviations: mGDH, membrane-bound PQQ-dependent glucose dehydrogenase; SldAB, membrane-bound PQQ-dependent sorbitol dehydrogenase; GA2DH, membrane-bound FADdependent gluconate 2-dehydrogenase; CydAB, cytochrome bd oxidase; CyoBACD, cytochrome bo 3 oxidase; QrcABC, cytochrome bc 1 complex; PntAB, transhydrogenase. the red-cross means the metabolic pathway cut off individual engineering approaches and the efficiency of 5-KGA production was unsatisfactory, especially there was a large amount of residual GA left at the end of the biotransformation process.
In early work, we constructed the recombinant strain G. oxydans ZJU2, in which the GOX1231 and GOX1081 genes were markerless deleted [21]. The engineering of 2-KGA and acetic acid metabolism pathway could enhanced the 5-KGA production, but the titer and productivity of 5-KGA was unsatisfied. Hence, we set out to develop an efficient Gluconobacter cell factory to facilitate 5-KGA biosynthesis using combinatorial strategies. The sldAB gene, encoding the SLDH, was plasmid-based overexpressed with a strong promoter, P 0169 [22]. Inspired by the mechanism from most of the Gluconobacter, it is proposed that the cofactor PQQ and respiratory chain engineering could be enhanced efficiency of the membrane-bound quinoaproteins. As a result, the genes involved with the PQQ cluster [23,24] and terminal ubiquinol cytochrome bo 3 oxidase [25] were fused expression. The related specific enzyme activity, H + /O ratio and 5-KGA titer were investigated. Base on the pH shift and dissolved oxygen tension (DOT) control, the robust G. oxydans cells were more facilitated 5-KGA accumulation by the fed-batch fermentation. This study represents combinatorial engineering approaches collectively increased the titer of 5-KGA in the G. oxydans, which can provide insights into devising engineering strategies to improve the object production.

Results and discussion
Promoter selection in G. oxydans strain Increasing interest in Gluconobacter has led to investigations on strain improvement. The well characterized promoter is a prerequisite to the understanding of gene expression. So far, a few studies concerning the isolation and characterization of G. oxydans promoters, such as P 0169 [22], P tufB [26], P 0264 [27], and P 0452 [27], have been reported. However, which one is more suit for gene expression in G. oxydans is inconclusive. To select the appropriate promoter for gene expression in G. oxydans, we generated the different promoters in front of a gfp report gene in a pBBR1MCS5 vector. The promoter activity was indirectly determined by measuring the whole cell fluorescence intensity (RFU/OD 600 ). As shown in Fig. 2, the RFU/OD 600 presented a linear increase with the cell growth until 24 h. Green fluorescent protein (GFP) contains a fluorescent cyclic tri-peptide, and oxygen is required for the final oxidation of the mature, cyclized fluorophore of GFP [22]. For this reason, the RFU was impacted by the level of the DO, which explained that the RFU/OD 600 remained stable after 24 h. However, among this four different promoters, the cells expressing gfp under regulation of P 0169 promoter exhibited outstanding effect in the G. oxydans. The observation elucidated that the P 0169 promoter was evidently recognized by G. oxydans and could reliably drive heterologous gene expression in G. oxydans. 14  The glucose metabolism of the mutant G. oxydans ZJU2 strain showed the positive effect on the 5-KGA production [21], but it remained low productivity of 5-KGA. To boost the 5-KGA production, the sldAB gene was plasmid-based overexpressed in G. oxydans ZJU2 under the control of the selected P 0169 promoter, generating G. oxydans ZJU3. The enzyme activity of SLDH of G. oxydans ZJU3 toward GA and the mRNA abundance of sldAB were investigated ( Table 1). The SLDH activity was 2.55 ± 0.04 U/mg protein, which was more 3-fold higher than that of the reference strain. The sldAB expression data obtained were normalized to the transcriptional level of the 16S RNA. G. oxydans ZJU3 achieved the relative transcriptional level of 4.12 ± 0.04, which was 4-fold higher than those of the control. The results showed that the specific enzyme activities and transcription of the sldAB gene in G. oxydans could be markedly enhanced by sldAB overexpressed under the control of the selected P 0169 promoter. In the literature [26], the stability and transcriptional level of the mRNA by adding poly (A/T) tails at the 3'-UTR of the sldAB were discussed. It was revealed that an artificial poly (A/T) tail was proposed to slow down the mRNA degradation process in bacteria and the high sldAB expression levels were achieved. This well demonstrated the importance of the mRNA stability on the gene expression, which should be considered in our late study.

Batch fermentation by sldAB overexpression strain
The batch fermentation by the recombinant G. oxydans ZJU3 strain was performed on a 15 L agitation tank under DOT rich condition. The results demonstrated that all tested strains had the similar trend of glucose consumption rate and cell growth (Fig. 3). The glucose was rapidly oxidized and exhausted at 30 h, while about 130 ± 1.76 g/L of GA was accumulated over the same time frame. The reference strain, G. oxydans ZJU2, accumulated 82.48 ± 1.10 g/L 5-KGA in the broth, but the residual GA was quite high at about 62.07 ± 1.04 g/L (Fig. 3a). In the fermentation process using recombinant G. oxydans ZJU3, 122.48 ± 0.41 g/L of 5-KGA with the productivity of 1.92 g/(L · h) were obtained at 64 h ( Fig. 3b). Compared with the reference strain, the titer of 5-KGA was increased 48.99 % and the residual GA was 24.95 ± 0.76 g/L, decreased 59.80 %. However, the DCW of the tested strains was about 3.58 ± 0.13 g/L, which was lower than other bacteria. In recent study, DNA microarray analysis and 13 C metabolic flux analysis ( 13 C -MFA) are used to characterize the two growth phases of G. oxydans in the presence of glucose [25]. In the first growth phase, 90 % of the glucose is oxidized by the mGDH to GA, accompanied by reasonable growth, high demand for oxygen and a low formation of CO 2 . The onset of phase II results in reduced biomass and demand for oxygen while the GA was oxidized in the periplasm to 5-KGA (Fig. 3). Thus, only a small percentage of the carbon source is taken up by cell, resulting in modest cell growth and poor cell yield. G. oxydans IFO3293 and G. oxydans 621H cells cultivated on glucose medium provids a cell yield of 0.09 g cdw /g glucose [28,29]. In comparison to these values, E. coli reaches a value of 0.49 g cdw /g glucose [30], and Bacillus subtilis reaches a yield of 0.32 g cdw /g glucose [31]. In phase I and II, the cytoplasmic sugar catabolism proceeded predominantly via the PPP, particularly in phase II [28].
During the glucose metabolism by G. oxydans, the mGDH and SLDH serve as the main enzymes, and PQQ is the cofactor. The PQQ supplementation experiments showed that the control strain G. oxydans ZJU2 was not significantly influenced by the exogenous PQQ (Fig. 3c), but the G. oxydans ZJU3 strain showed the respond to PQQ supplementation in a similar manner, achieving the highest 5-KGA production when the added PQQ was up to 500 μg/L (Fig. 3d). A total of 131.92 ± 2.11 g/L of 5-KGA was produced by G. oxydans ZJU3. These results indicated that the addition of PQQ could enhance the production of 5-KGA in the SLDH overexpression strain, which was in agreement with a recent report that the overexpression of PQQ-dependent dehydrogenases could lead to imbalance between coenzyme PQQ level and the corresponding quinoproteins [24].

Enhanced PQQ biosynthesis to improve 5-KGA production
To further augment the 5-KGA production, the PQQ biosynthesis gene, pqqABCDE, was engineered and the biomass growth, PQQ concentration and 5-KGA production were shown in Table 2. In G. oxydans ZJU4, overexpression of pqqABCDE gene cluster under the control of P 0169 promoter led to the excretion of 674.82 ± 4.12 μg/L of PQQ into the supernatant, which was enhanced by 383.53 % compared with that by the parent strain while the 5-KGA concentration was 131.76 ± 1.89 g/L, an increase of 7.58 %. This also confirmed the results from the PQQ supplement experiment (Fig. 3) that the coenzyme PQQ was the key factor driving the 5-KGA production. In G. oxydans ZJU5, simultaneous expression of the PQQ gene cluster and an associated tldD gene produced 757.83 ± 2.43 μg/L of PQQ, which was increased by 12.30 % compared with that from G. oxydans ZJU4. These results were consistent with the literature finding that the tldD gene was related to the PQQ biosynthesis [32]. The 5-KGA production by G. oxydans ZJU5 was 134.88 ± 2.16 g/L. In addition, the cell growth of G. oxydans ZJU5 was comparable to that of G. oxydans ZJU4. This finding implied that tldD gene expression did not affect the cell growth.
Overexpression of metabolic pathways involving redox reactions may lead to cofactor imbalances, thus impairing the yield of products. Cofactor engineering approaches are often adopted to compensate for imbalance of cofactors to improve product biosynthesis [33], for example, manipulating the availability of intracellular NADH [34] and NADPH [35]. In G. oxydans,  quinoproteins and their cofactor PQQ can catalyze the oxidation of substrates by PQQ regulated electron transfer in the respiratory chain. However, the biosynthesis of quinoproteins and their cofactor PQQ are usually independent [24]. Therefore, the PQQ gene cluster was overexpressed in Gluconobacter oxydans WSH-003 [23] and Ketogulonigenium vulgare [24], which demonstrated to improve 2-keto-L-gulonic acid (2-KLG) production by cofactor engineering. These advance significantly facilitated the development of the efficient strains to produce 5-KGA.
Notably, it was reported that disruption of tldD gene in G. oxydans led to a drop of PQQ excretion below the detection limit and a decrease in cell growth, indicating that the related tldD gene was essential for PQQ biosynthesis [32]. The TldD protein of G. oxydans 621H, related to the E. coli TldD, is a peptidase involved in processing of small peptides. In other PQQ-producing bacteria, the peptidase-like protein PqqF is required for PQQ synthesis and has a similar function as the TldD in G. oxydans [32]. Therefore, our study demonstrated that overexpression of the tldD gene could increase PQQ level by 12.30 % in G. oxydans ZJU5 compared with that of G. oxydans ZJU4 (Table 2). This cofactor engineering was adopted to compensate the shortage of cofactors in the PQQ-dependent SLDH overexpression strain and to improve the 5-KGA production.

The engineering of the respiratory chain
In an earlier study, the cytochrome bo 3 oxidase (cyoBACD, GOX1911-1914) was found to be the main component for proton extrusion via the respiratory chain in G. oxydans 621H [25]. To enhance the respiratory chain activity and the 5-KGA production, the cytochrome bo 3 oxidase was engineered. The cyoBACD genes were fused into the PQQ overexpression plasmid (pUCpr-T1 and pUCpr-T2) under the P 0169 promoter control, generating pUCpr-T3 and pUCpr-T4. The corresponding plasmids were electrotransferred into G. oxydans ZJU3, resulting in G. oxydans ZJU6 and G. oxydans ZJU7, respectively.
The batch fermentations of G. oxydans ZJU4, G. oxydans ZJU5, G. oxydans ZJU6, and G. oxydans ZJU7, under an excess of oxygen, showed that all recombinant G. oxydans strains reached their maximal OTR. The time points of maximal OTR correlated with the time points at which about 60 % glucose had been predominantly oxidized to GA in the periplasm (Fig. 4). G. oxydans ZJU4 and G. oxydans ZJU5 reached their maximal OTR of 26.52 mmol/L · h and 29.85 mmol/L · h, respectively at 19 h. After complete consumption of glucose, the OTR decreased to 1.5 to 1.8 mmol/L · h. The specific maximal CO 2 production rate (CTR) of both strains was 7.5 mmol/L · h at 32 h, then it began to decrease (Fig. 4a,  b). The CTR decreased below 5.0 mmol/L · h at 57 h for   (Fig. 4c, d). This indicated that the oxidation of GA to 5-KGA by G. oxydans ZJU7 had completely finished at 47 h. Using the constructed G. oxydans strains, the H + /O ratio and terminal ubiquinol bo 3 oxidase activity were also measured when the cells were in the logarithmic growth phase (DCW of 1.5). The results were shown in Table 3. The average H + /O ratio and terminal ubiquinol oxidase activity of the control strain G. oxydans ZJU3 were 1.24 ± 0.11 and 0.31 ± 0.04 μmol/min · mg, respectively. The G. oxydans ZJU4 and G. oxydans ZJU5 had an H + /O ratio and terminal ubiquinol oxidase activity comparable to those of the control strain, implying the mutant growth was not impaired under conditions of PQQ cluster overexpression. However, the recombinant G. oxydans ZJU6 and G. oxydans ZJU7 showed a 64 % increase of the H + /O ratio (1.95 ± 0.23 and 2.01 ± 0.16) and 1.5-fold of the ubiquinol oxidase activity (0.78 ± 0.05 μmol/min · mg and 0.80 ± 0.06 μmol/min · mg) compared with the control strain, respectively. A total of 141.86 ± 2.89 g/L and 144.52 ± 2.94 g/L 5-KGA was accumulated by G. oxydans ZJU6 and G. oxydans ZJU7, respectively. This indicated the expression of ubiquinol bo 3 oxidase could enhance respiratory proton translocation and increase the 5-KGA production in G. oxydans.
G. oxydans possesses a branched respiratory chain consisting of two terminal ubiquinol oxidases, cytochrome bo 3 oxidase and cytochrome bd oxidase (cydAB, GOX0278-0279) (Fig. 1). The genome sequence also revealed genes for a cytochrome bc 1 complex (qrcABC, GOX0565-0567) and a soluble cytochrome c (GOX0258) [10]. The absence of cytochrome bd oxidase did not affect the cell growth or proton extrusion via the respiratory chain, whereas absence of the genes encoding cytochrome bo 3 oxidase caused a severe growth defect [25]. Plasmidbased overproduction of cytochrome bo 3 oxidase under the P 0169 promoter control increased the respirationdriven proton extrusion by 66.7 % in G. oxydans ZJU6 and G. oxydans ZJU7, compared with the control strains ( Table 3). The increase in the H + /O ratio might be owing to the presence of more bo 3 quinol oxidase and its proposed high oxygen affinity, which should favor oxygen reduction under the experimental conditions of an oxygen pulse [25]. The H + /O ratio for the reference strain, G. oxydans ZJU3, was measured as 1.21 ± 0.11. In the literature, H + /O ratios reported for E. coli vary between 3.4 and 4.5 [36,37], which was more than 2-fold higher than that of the G. oxydans ZJU3 strain. A major difference of the respiratory chain between the two species is the lack of the multisubunit proton-pumping NADH dehydrogenase I (NDH-I) in the G. oxydans [10,25]. However, NDH-I is preferentially synthesized during anaerobic growth in the presence of alternate electron acceptors [38]. Therefore, what extent NDH-I contributes to the H + /O ratio is unclear. As the bd oxidase was not relevant for proton translocation, an H + /O ration of 4 might be assumed for G. oxydans [39]. However, our results were much lower than a value of 4. Richhardt explained that cytochrome bo 3 oxidase might not function as a primary proton pump but as a Na + pump and the respiratory chain could involve a reverse electron transfer coupled to an influx of protons [25].
Interestingly, an increased 15.48 % cell yield was observed when the bo 3 oxidase gene was overexpressed (Fig. 4 g, h). The possibility of an increased bo 3 oxidase level in G. oxydans could cause a shift of the electron flux from the non-proton pumping bd oxidase to the proton pumping bo 3 oxidase [25]. Another explanation for the increased cell growth is the assumption that the activity of the bo 3 oxidase is limited and can be increased by its overproduction. This was confirmed in the experiments that the ubiquinol oxidase activity increased 2.5-fold (Table 3) and the OTR enhanced 9.12 mmol/L · h (Fig. 4) by overexpressed the cyoBACD genes, indicating that cytochrome bo 3 quinol oxidase played the positive role in the cell growth and substrate oxidation. Recently, a limitation of oxygen consumption by the activity of the terminal oxidase was reported after overexpressing the membrane-bound glucose dehydrogenase (mGDH, GOX0265) [40]. This implied that the activity of bo 3 quinol oxidase might be high enough to satisfy the increased capacity for oxygen reduction.

Process optimization and fed-batch fermentation
The constructed G. oxydans ZJU7 showed good potential for 5-KGA production. To fulfill the requirement of industrial production of 5-KGA, the fermentation process was optimized, including the two-stage pH control, DO control and glucose feed. As shown in Fig. 5a, the glucose was oxidized faster and exhausted at 14 h when the pH was controlled at 5.5. This result was consistent with the previous results where a pH of 5.5 was suitable for the activity of membrane-bound glucose dehydrogenase [41]. However, a pH of 4.5 was more suitable for synthesis of 5-KGA by SLDH. In the early study, the pH of the culture medium was an important factor c The 5-KGA production and conversion rate between G. oxydans ZJU7 and wild-type strains. black-bar, 5-KGA, red-bar, conversion rate for selective production of 5-KGA by G. suboxydans IFO 12528, in which an 87 % glucose conversion rate could be achieved by controlling the medium pH in a range of 3.5-4.0 [20]. According to the experimental results, we have established a two-stage pH control strategy, whereas the pH is controlled as 5.5 in the first stage to let the glucose oxidation and cell growth, then it is shifted to 4.5, facilitating 5-KGA formation.
In its natural habitat, G. oxydans is likely subjected to low-oxygen stress conditions because of the rapid oxygen consumption as a result of its own metabolism. The OTR and CTR were previously investigated during the 5-KGA production, in which the high OTR was observed when the glucose was oxidized to GA in the periplasm (Fig. 4a, b, c, d). It was reported that oxygen limitation could cause expression changes of 486 genes, representing 20 % of the chromosomal genes [42]. Therefore, during this stage, it must ensure the adequate DOT above 20 % by increasing the agitation speed. The onset of 5-KGA formation decreased the demand for DOT but increased the CTR (Fig. 4a, b, c, d). Hence, to improve the cost performance of the industrial production of 5-KGA, DOT was continuously controlled by adjusting the agitation speed according to OTR and CTR.
Additionally, to achieve the hyper-production of 5-KGA by the recombinant G. oxydans ZJU7 strain, the glucose fed-batch fermentation was carried out under the DOT and pH control. A glucose feed of about 1400 g was started at 12 h when the glucose concentration was below 40 g/L and was used to maintain the glucose at 40~50 g/L (Fig. 5b). The GA quickly accumulated during this stage. When the glucose concentration dropped to 14 g/L at 28 h, the 5-KGA began to quickly be formed and the pH shifted from 5.5 to 4.5. Concurrently, the DOT was controlled above 20 %. Using glucose fed-batch fermentation, DOT control, and a pH shift strategy, the 5-KGA titer was increased to 162 ± 2.12 g/L with the 2.53 g/(L · h) productivity by the recombinant G. oxydans ZJU7 strain, which was increased by 10-fold compared to the wild-type strain (Fig. 5c). This indicated that the PQQ overexpression and respiratory chain modification could efficiently enhance 5-KGA accumulation. Furthermore, the problem of residual GA was well solved, and the final concentration of GA was reduced to 15.38 g/L, which should decrease the separation difficulty. These results illustrated that the supplement of co-enzyme to the membrane-bound dehydrogenases was increased by overexpression of the PQQ cluster genes. The efficiency of electron transfer to O 2 was enhanced by overexpression of the genes for the bo 3 oxidase, which increased the activity of respiratory chain.

Conclusion
Bacteria of the genus Gluconobacter exhibit so-called oxidative fermentation, or incomplete oxidation, a highly unusual metabolic feature that has been exploited industrially for decades in the production of vitamin C, GA and dihydroxyacetone [43]. In this study, the role of the combinatorial metabolic engineering of the industrial G. oxydans for the boosting of 5-KGA accumulation was investigated. In summary, the sldAB was overexpressed under the selected P 0169 promoter. In addition, the PQQ gene cluster and terminal ubiquinol bo 3 oxidase were fused overexpression to strengthen the glucose oxidation. Under the optimized culture conditions of the fedbatch process, the combinatorial approaches collectively increased the 5-KGA titer 10-fold, reached 162 ± 2.12 g/ L. The results showed great potential for optimizing the current producer strain of G. oxydans used in industrial biotechnology, which presents the first report of 5-KGA production by combinatorial metabolic engineering approaches in G. oxydans. We envision that these approaches could provide framework for devising engineering strategies to improve the production of biochemicals in G. oxydans. Nevertheless, some issues are still worth further study, for example, how many copies of pqq cluster and bo 3 oxidase genes overexpression can be matched the quinoproteins requirements, especially in the quinoproteins overexpression strains. The genome sequence has revealed that the TCA cycle is incomplete as genes for succinate dehydrogenase and succinyl CoA synthetase were absent [10]. To develop more robust strains, the strategy involving direct repair of this defective metabolic pathway by genomic integration of heterologous genes should be investigated, while considering the G. oxydans as a broadly applicable host for oxidative industrial bioconversions.

Bacterial strains, plasmids, and media
The bacterial strains, and plasmids used in this study are listed in Table 4. Escherichia coli strains were cultivated in Luria-Bertani (LB) medium at 37°C. The 50 μg/mL gentamicin or 100 μg/mL ampicillin were used whenever required. Agar (1.5 %) was added to obtain solid media. Gluconobacter oxydans DSM2343 strains were cultivated on mannitol medium (MP) containing 5 g/L yeast extract, 3 g/L peptone, and 25 g/L mannitol at 30°C. For G. oxydans possesses a natural resistance against cefoxitin, thus, as a precaution to prevent bacterial contamination, 50 μg/mL cefoxitin was added.

Construction of shuttle vector
A shuttle vector pUCpr, compatible to the broad-hostplasmid pBBR1MCS5, was constructed. The 2446-bp par-rep gene fragment of cryptic plasmid pGOX3 from G. oxydans DSM2343 was amplified with primers pr_PstI_F / pr_SalI_R ( Table 5). The sequenced PCR product was digested and inserted into the PstI/SalI site of the pUC19, resulting in pUCpr [44].

Promoter selection
Four different promoters from G. oxydans DSM2343 were carefully selected. A putative promoter sequence, P 0169 [22], the promoter of elongation factor Tu, P tufB [26], and ribosomal proteins L35 and L13, P 0264 and P 0452 [27], were amplified by PCR with primers as listed in Table 5. The resulting DNA fragments P tufB , P 0264 , P 0452 and P 0169 were digested with restriction enzymes SacI and XbaI, then were ligated into pBBR1MCS-5 [45], generated the vector pBB5-P tufB , pBB5-P 0264 , pBB5-P 0452 and pBB5-P 0169 , respectively. The report gene gfp was amplified from the cloning vector pET28 (a)-GFP. The resulting product was digested with XbaI and HindIII, and then cloned into the XbaI and HindIII site of pBB5-P tufB , pBB5-P 0264 , pBB5-P 0452 and pBB5-P 0169 to generate the corresponding promoter strength reporter plasmids.
The reporter plasmids were first transformed into E. coli DH5α, analyzed for the correct insert by DNA sequencing and then transformed into G. oxydans DSM2343 by electroporation (2000 V, 200 Ω and 25 μF) in a 2-mm cuvette using a Gene Pulser II (Bio-Rad, München, Germany) as described previously [46]. The transformants were selected by cefoxitin and gentamicin. The whole cell fluorescence intensity (RFU/OD, the relative fluorescence unit divided by the corresponding cell density) was measured [22]. Cells were harvested and washed twice with KPB buffer (pH 7.0), and then photographed by using a confocal laser scanning microscope. The cell density was measured by the absorbance at 600 nm using the spectrophotometer (UVmini-1240, SHI-MADZU ® ), which determined the strengths of the different promoters. Overexpression of sldAB gene in G. oxydans ZJU2 Based on the bioinformatics analysis of G. oxydans 621H genome sequence [10], SLDH_XbaI_F / SLDH_E-coRI_R primers were designed and the open reading frames (ORFs) of SLDH (sldAB, GOX0854-0855) was PCR-amplified with the primers. The genomic DNA of G. oxydans DSM 2343 was used as a template. The sequenced amplicon was digested with the restriction endonucleases XbaI and EcoRI and cloned into pBB5-P 0169 vector restricted with the same enzymes, resulting in plasmids pBB5-P 0169 -sldAB. This plasmid and as a control pBB5-P 0169 vector were transferred into the desired G. oxydans ZJU2 strain by electroporation and were selected for a gentamycin-resistant phenotype. The correct strain G. oxydans ZJU2/pBB5-P 0169 -sldAB was named G. oxydans ZJU3.
Co-expression of the PQQ biosynthesis genes and the terminal ubiquinol cytochrome bo 3 oxidase genes in G. oxydans ZJU3 In the oxidation process by the membrane-bound SLDH, the PQQ serves as prosthetic groups [10,15,16], which transfer electrons to the respiratory chain [10]. To enhance the 5-KGA production, the genes encoding the cofactor PQQ and the terminal ubiquinol cytochrome bo 3 oxidase of the respiratory chain were reinforced. To fulfill the experiments, the PQQ gene cluster (GOX0983-0987) and the related tldD gene (GOX1104) were cloned and the generated plasmids pUCpr-T1 and pUCpr-T2. The constructed plasmids were transferred into G. oxydans ZJU3, resulting the recombinant strains G. oxydans ZJU4 and G. oxydans ZJU5, respectively. The PQQ and terminal ubiquinol cytochrome bo 3 oxidase fused expression The capital and underlined sequences indicate the restriction enzyme sites plasmid was constructed by the pEASY-Uni seamless cloning and assembly kit CU101 (TransGen, China). The PCR-amplified promoter P 0169 , pqqABCDE, and tldD were cloned into the XbaI site of the shuttle vector pUCpr, which resulted in the plasmid pUCpr-T1-P 0169 and pUCpr-T2-P 0169 . Then the cyoBACD (GOX1911-1914) was inserted the XbaI and KpnI site of pUCpr-T1-P 0169 and pUCpr-T2-P 0169 to generate the fusion expressed plasmid pUCpr-T3 and pUCpr-T4, respectively. After verification of the accuracy of the plasmid by the sequencing, the fusion plasmid was transferred in the recombinant strain G. oxydans ZJU3 by electroporation [46] and selected for a gentamycin-kanamycin resistant phenotype, generated G. oxydans ZJU6 and G. oxydans ZJU7.

Measurements of enzyme activity and protein concentration
For the purpose of preparation of the cell crude extract and membrane fraction, a single colony of the G. oxydans strain was pre-incubated in MP medium. The G. oxydans cells were harvested by centrifugation (10,000 × g, 5 min, 4°C) and resuspended in 20 mL 50 mM sodium phosphate buffer (pH 6.0). The cells were disrupted with an ultrasonifier (JY92-2D, Xinzhi, NingBo) for 50 cycles (2000 w, 3 s sonication, 5 s pause) on ice.
The cell debris were removed by centrifugation at 5500 × g for 20 min at 4°C and the supernatant was used as the crude extract. For the preparation of membranes, the supernatant was centrifuged for 60 min at 180,000 × g at 4°C. The resulting sediments were collected and resuspended into 50 mM sodium phosphate buffer (pH 6.0), and used as the membrane fraction. Enzyme activities were determined using a spectrophotometer (Uvmini-1240 SHIMADZU ® ). Substratedependent changes of redox states of artificial electron acceptors (2, 6-dichlorophenolindophenol, DCPIP, Sigma) were determined at 600 nm and 30°C. The basal reaction mixture contained 50 mM PBS pH 6.0, 0.25 mM DCPIP, and 0.325 mM phenazine methosulphate (PMS, Sigma), which was prepared and pre-warmed to 30°C before the assay. Measurements were performed in a cuvette with a 1-cm light path containing a 0.8-mL basal reaction mixture and 10 μL enzyme pre-incubated at 30°C for 5 min. The reaction was started by adding 20 μL of a 2.0 M gluconate solution. One unit of enzyme activity (U) was defined as the amount of enzyme that can catalyze the conversion of 1 μM DCPIP per min at 30°C. The concentration of proteins was determined with the Pierce™ BCA assay kit.

Quantitative real-time PCR (RT-PCR)
Cells were harvested at an OD 600 of 2.5 at room temperature and immediately frozen in liquid nitrogen. Cells were then stored at −80°C until RNA extraction.
Total RNA was extracted with RNAiso™ Plus from Takara (Dalian, China) and treated with RNase-free DNase. Following chloroform extraction step, RNA was precipitated with isopropanol and the pellet washed twice in 75 % ethanol. After air-drying, RNA was resuspended in RNase-free water. The quantity of total RNA was verified using an Eppendorf Biophotometer (Eppendorf, Hamburg, Germany). The cDNA was synthesized from the total RNA using a PrimeScript RT Reagent Kit (Perfect Real Time) (Takara) according to the manufacturer's protocol. The products were quantified via realtime PCR with StepOnePlus TM Real-Time PCR System (Applied Biosystems, USA) using primer RTsldh_F/ RTsldh_R. The 16S rRNA gene was used as internal standard based on the primer RT16S_F/RT16S_R.

Characterization of respiration activity and H + /O measurements
Cultivation of G. oxydans strains was performed and the respiration activity was measured by the exhaust gas analysis system, such as the oxygen transfer rate (OTR), carbon dioxide transfer rate (CTR). The number of H + moved upon respiration (H + /O ratio) is principally important for the efficiency of cellular ATP production. Hence the H + /O ratio of bacterial cells have been measured to evaluate the efficiency of the respiratory chain. The H + /O and terminal ubiquinol oxidase activity were measured using a method previously reported in the literature [36].

Batch fermentations
Batch and fed-batch fermentations were conducted in a 15 L stirred tank bioreactor (Fus-D; Guoqiang Bioengineering Equipment Co., Ltd, Shanghai, China) with 9 L of the initial medium, which was composed of 0.41 g/L (NH 4 ) 2 SO 4 , 0.1 g/L (NH 4 ) 2 HPO 4 , 0.01 g/L MgSO 4 · 7H 2 O, 3.0 g/L corn steep liquor paste and 100-150 g/L glucose (depending on the experiment) [19]. The CaCO 3 (20 g/L) was sterilized separately and then added to the medium. The seed culture was prepared by inoculation of a single colony into a 5-mL MP medium tube, then into 200 mL of fresh seed medium in 500-mL flasks and cultivated on a rotary shaker at 220 rpm for 16 h. The seed culture (5 %, v/v) was then inoculated into the fermentation medium and the fermentation was carried out at 30°C and pH was controlled at 5.0 by automatic addition of 5 M NaOH. In order to examine whether PQQ could enhance the 5-KGA production in engineered strains, different amount of PQQ (0 μg/L, 100 μg/L, 200 μg/L, 500 μg/L) was exogenously added to the mixed cultures at the beginning of the fermentation process, and the 5-KGA concentrations were measured after 64 h.