Enhanced production of l-sorbose by systematic engineering of dehydrogenases in Gluconobacter oxydans

l-Sorbose is an essential intermediate for the industrial production of vitamin C (l-ascorbic acid). However, the formation of fructose and some unknown by-products significantly reduces the conversion ratio of D-sorbitol to l-sorbose. This study aimed to identify the key D-sorbitol dehydrogenases in Gluconobacter oxydans WSH-003 by gene knockout. Then, a total of 38 dehydrogenases were knocked out in G. oxydans WSH-003, and 23 dehydrogenase-deficient strains could increase l-sorbose production. G. oxydans-30, wherein a pyrroloquinoline quinone-dependent glucose dehydrogenase was deleted, showed a significant reduction of a by-product with the extension of fermentation time. In addition, the highest conversion ratio of 99.60% was achieved in G. oxydans MD-16, in which 16 different types of dehydrogenases were inactivated consecutively. Finally, the gene vhb encoding hemoglobin was introduced into the strain. The titer of l-sorbose was 298.61 g/L in a 5-L bioreactor. The results showed that the systematic engineering of dehydrogenase could significantly enhance the production of l-sorbose.


Introduction
L-sorbose is an essential precursor for the industrial fermentation production of 2-keto-L-gulonic acid , which is the direct precursor of L-ascorbic acid (vitamin C) [1,2]. The most used industrial method for producing L-sorbose is the conversion of D-sorbitol by Gluconobacter or Acetobacter species [3,4] in which membrane-bound D-sorbitol dehydrogenases are responsible for this enzymatic conversion. The catalytic activity center of membrane-bound D-sorbitol dehydrogenases is commonly exposed to the periplasmic space; therefore, the L-sorbose is generally produced in the periplasm and subsequently secreted into the culture broth [5,6]. In the classical vitamin C production process, the L-sorbose produced by Gluconobacter spp. was further converted into 2-KLG via a mixed fermentation of Ketogulonicigenium vulgare and Bacillus megaterium [7], and the resulting product 2-KLG was chemically esterified for vitamin C production [1,8].
Gluconobacter belongs to the family of Acetobacteraceae and is unsurpassed by other organisms in the capacity of incompletely oxidizing a great variety of carbohydrates, alcohols, and related compounds [4,9]. It can oxidize many substrates regio-selectively, hence is widely used for the industrial production of L-sorbose, 6-amino-L-sorbose, D-gluconic acid, and keto-gluconic acids [4,9]. L-sorbose is an essential substrate in vitamin C production, and mainly two types of membrane-bound D-sorbitol dehydrogenase are involved in the oxidation of D-sorbitol. One is pyrroloquinoline quinone (PQQ) dependent D-sorbitol dehydrogenase with two subunits (sldB and sldA), and the other one is FAD dependent D-sorbitol dehydrogenase with three subunits (sldS, sldL, and sldC) [10,11]. In addition, some cytoplasmic D-sorbitol dehydrogenases or xylitol dehydrogenases in G. oxydans mainly catalyze D-sorbitol to form D-fructose for cell growth [12,13]. Some special NADP + /NAD + -dependent D-sorbitol dehydrogenases from G. oxydans G624 could also catalyze D-sorbitol to form L-sorbose efficiently [14]. The membrane-bound D-sorbitol dehydrogenases are usually considered as the main enzymes responsible for catalyzing D-sorbitol into L-sorbose.
Genome deletion is expected to obtain genome-simplified chassis cells with rich application potential. Some studies on the simplification of the genomes of model microorganisms have been reported, including B. subtilis, Escherichia coli [15], Corynebacterium glutamicum [16], and Peer review under responsibility of KeAi Communications Co., Ltd.
Streptomyces avermitilis [17]. For G. oxydans, studies were mainly performed to identify the function of some genes by gene knockout, such as the gene cluster for the biosynthesis of PQQ [18] and the characterization of membrane-bound dehydrogenases from G. oxydans 621H [19]. Knockout of eight membrane-bound dehydrogenases in G. oxydans 621H could improve the titer and conversion rate of L-erythrulose to 242 g/L and 99%, respectively, in a dissolved oxygen (DO)-controlled stirred-tank bioreactor [20]. Thus, deleting enzymes unrelated to the L-sorbose formation may enhance the L-sorbose production by avoiding unnecessary energy consumption and providing limited membrane space to the most needed membrane-bound D-sorbitol dehydrogenase [21].
In this study, the critical membrane-bound D-sorbitol dehydrogenases in G. oxydans WSH-003 were identified by gene knockout. The genes of sldBA1 and sldSLC encode major membrane-bound D-sorbitol dehydrogenases that catalyze the conversion of D-sorbitol into Lsorbose. On the contrary, the gene of sldBA2 showed no catalytic activity for D-sorbitol, although it had high homology with sldBA1. After identifying the key D-sorbitol dehydrogenases of G. oxydans WSH-003, 38 predicted dehydrogenases unrelated to the synthesis of L-sorbose were knocked out. The results showed that 23 dehydrogenase-deficient strains could increase L-sorbose production and G. oxydans-30, a membrane-bound glucose dehydrogenase (mGDH)-deficient strain of them could significantly decrease a by-product in the fermentation broth with the extension of fermentation time. At the same time, G. oxydans MD-16, in which 16 kinds of dehydrogenases were knocked out consecutively, had the highest titer of L-sorbose (149.46 g/L) with a conversion rate of 99.60% in shake flasks. Besides, the overexpression of the gene vhb encoding Vitreoscilla hemoglobin (VHb) in G. oxydans MD-16, scaling up in a 5-L fermenter, the titer of L-sorbose reached 298.61 g/ L in 300 g/L D-sorbitol medium.

Strains, plasmids, and culture conditions
G. oxydans WSH-003 was obtained from Jiangsu Jiangshan Pharmaceutical Co., Ltd., and was sequenced in a previous study (GenBank Accession No. AHKI00000000.1) [22]. E. coli JM109 (E. coli JM109) was purchased from Novagen (Darmstadt, Germany) and used as a host for plasmid construction and amplification. The pMD19-T simple vector was purchased from TaKaRa (Dalian, China). The E. coli JM109 strains were cultivated in Luria-Bertani (LB) medium at 37 • C and 220 rpm. When required, tetracycline was added to a final concentration of 20 μg/mL. The G. oxydans strains were cultivated in D-sorbitol medium (80 g/L D-sorbitol and 10 g/L yeast extract) at 30 • C and 220 rpm. When required, fluorouracil (FU), kanamycin, cefoxitin, and tetracycline were added to a final concentration of 300 μg/mL, 50 μg/mL, 50 μg/mL, and 20 μg/mL, respectively.

Gene deletions and construction of recombinant strains
For knocking out dehydrogenases, the upstream and downstream homologous arms of dehydrogenases were amplified by PCR from the G. oxydans WSH-003 genome. The kanamycin resistance gene (kana) or kana-upp fusion segments was obtained by PCR amplification, and it was fused with the upstream and downstream homologous arms of dehydrogenases with resistance gene. Fragments (DHS-kana-DHX and DHS-kana-upp-DHX) corresponding to related dehydrogenases was obtained. Strains in which dehydrogenases was knocked out were generated by integrating the obtained homologous recombinant fragments into the G. oxydans WSH-003 parental strain. Membrane-bound D-sorbitol dehydrogenase genes of G. oxydans WSH-003 were deleted consecutively using the counter-selection marker of upp gene. D-sorbitol dehydrogenase knockout strains are listed in Table 1. The primers used to construct the D-sorbitol dehydrogenase knockout strains are given in Supplementary Table (Table S1). Dehydrogenases of unknown function of G. oxydans WSH-003 were deleted using the kana gene. Dehydrogenase of unknown function knockout strains are listed in Table 1. The primers used to construct the dehydrogenase knockout strains are given in Supplementary Table (Table S1). Dehydrogenase genes of G. oxydans WSH-003 were deleted consecutively using the counter-selection marker of the upp gene. Dehydrogenase combination knockout strains are listed in Table 1. The primers used to construct the combination knockout  Table (Table S1). All mutants were verified by polymerase chain reaction and sequencing.

Overexpression of vhb in G. oxydans
A plasmid containing the vhb gene driven by the P 2703 promoter or P vhb promoter was constructed to express vhb in G. oxydans [24]. The vhb gene with the promoter was synthesized by Sangon (Shanghai, China). The VHb was fused to the twin-arginine signal peptide to export it into the periplasm [25]. The plasmids and primers are listed in Table 1 and  Supplementary Table (Table S1), respectively. All plasmids constructed were transformed into competent E. coli JM109, and transformants were selected on LB medium plates containing 20 μg/mL tetracycline.
Transformation of G. oxydans was carried out using electroporation by the methods described previously [26].

Production of L-sorbose by G. oxydans in flasks
G. oxydans WSH-003 and dehydrogenase knockout strains preserved in − 80 • C ultra-low temperature refrigerator were streaked on D-sorbitol medium plates and cultured for 48 h at 30 • C. One single colony was picked up and inoculated into 250-mL flasks with 25 mL of D-sorbitol medium at 30 • C for 48 h, and then transferred to 25 mL of fermentation medium (D-sorbitol, 150 g/L; yeast extract, 10 g/L) in a 250-mL flask. The inoculation volume was 10% (v/v) and cultivations were performed at 30 • C. Tetracycline (20 mg/L), kanamycin (50 mg/L), or/and cephalosporins (50 mg/L) were added when necessary.

Production of L-sorbose by G. oxydans in a 5-L bioreactor
The batch fermentation process was performed to optimize the production of L-sorbose in a 5-L bioreactor system (T&J Bioengineering, Shanghai, China) filled with 3 L of fermentation medium (D-sorbitol, 300 g/L; yeast extract, 10 g/L). A 10% (v/v) inoculum of G. oxydans was added to the bioreactor and incubated at 30 • C at 400 rpm. The aeration rate was controlled at 1 vvm.

Analysis procedures
D-sorbitol and L-sorbose levels in the fermentation broth were determined by high-performance liquid chromatography (HPLC) (Shimadzu, Kyoto, Japan) using an Aminex HPX-87H column (Bio-Rad, CA, USA) at 35 • C with a flow rate of 0.5 mL/min and 5 mmol/L H 2 SO 4 as the eluent. The refractive index detector was selected for detection.

Identification of the key D-sorbitol dehydrogenases of G. oxydans WSH-003
The membrane-bound D-sorbitol dehydrogenase clusters are sldBA1, sldBA2, and sldSLC from G, oxydans WSH-003. However, the genes with the function of conversion of D-sorbitol into L-sorbose have not been studied. The knockout strains of different D-sorbitol dehydrogenases genes (sldBA1, sldBA2, and sldSLC) from G. oxydans WSH-003 are shown in Table 1. Verification of these strains with D-sorbitol dehydrogenase deficiency is shown in Fig. 1A. After cultivating these strains in D-sorbitol medium, the result showed no significant change in their growth compared with the wild-type strain (Fig. 1B). Among these, the growth of G. oxydans-C, G. oxydans-D, and G. oxydans-E decreased slightly; G. oxydans-G, which was deficient in all three D-sorbitol dehydrogenases, could also grow well in the D-sorbitol medium. G. oxydans-A, G. oxydans-B, and G. oxydans-C had no significant effects on the conversion of D-sorbitol into L-sorbose (Fig. 1C), and the L-sorbose titer were 70.17 g/L, 74.69 g/L and 72.94 g/L (Fig. 1G). However, G. oxydans-E that knocked out the genes of sldBA1 and sldSLC could not form L-sorbose from D-sorbitol (Fig. 1D). There was not L-sorbose produced in the fermentation broth, and still 74.16 g/L D-sorbitol left in the fermentation broth (Fig. 1G). The fermentation results of G. oxydans-G were the same as that of G. oxydans-E (Fig. 1E), and there was still 74.19 g/L D-sorbitol left in the fermentation broth (Fig. 1G). L-sorbose is an essential intermediate for manufacturing high valueadded products, such as vitamin C [2] and L-tagatose [27]. In the fermentation of L-sorbose, the highly active D-sorbitol dehydrogenase is essential. Although in vitro characterization of D-sorbitol dehydrogenase was conducted by exogenous expression of D-sorbitol dehydrogenase in E. coli [14] or direct isolation and identification of D-sorbitol dehydrogenase from G. oxydans [11], these in vitro results could not directly represent their in vivo production capacity in G. oxydans because membrane-bound dehydrogenases require the channeling of electrons into the respiratory chain [28]. Thus, in the present study, the critical D-sorbitol dehydrogenase of G. oxydans WSH-003 was verified by gene knockout, demonstrating that D-sorbitol dehydrogenases sldBA1 and sldSLC were key enzymes for L-sorbose formation, and one of them was sufficient to catalyze D-sorbitol to L-sorbose (Fig. 1G). Additionally, the resulting G. oxydans-G and G. oxydans-E, which lost the capacity lost the capacity to convert D-sorbitol to L-sorbose, may be used as a platform strain applied in the screening of high-activity D-sorbitol dehydrogenase in future studies.

Effects of dehydrogenase gene knockout on L-sorbose production
A total of 44 dehydrogenases of G. oxydans WSH-003 were selected to be knocked out, and dehydrogenase single knockout strain was used to verify the effect of dehydrogenase knockout on the production of Lsorbose. Among the 44 dehydrogenases, 6 dehydrogenases (PQQ-containing dehydrogenase 2, glucose dehydrogenase, NADH dehydrogenase type II2, NADH dehydrogenase type II, glucose dehydrogenase, and aldehyde dehydrogenase 2) could not be successfully knocked out in G. oxydans WSH-003, indicating that they might be essential genes. Finally, 38 dehydrogenase genes were knocked out successfully. Among the 38 dehydrogenase-deficient strains, 23 dehydrogenase-deficient strains could increase L-sorbose production. Compared with G. oxydans WSH-003 (136.95 g/L), the L-sorbose production with G. oxydans-12, G. oxydans-18, and G. oxydans-31 increased obviously, reaching 144.89 g/L, 144.14 g/L, and 144.77 g/L, respectively (Table 2). Noticeably, the fermentation broth color of G. oxydans-30 was significantly different from that of G. oxydans WSH-003 with the extension of fermentation time ( Fig. 2A). At the same time, an unknown by-product was obviously decreased in the culture broth of G. oxydans-30 using HPLC analysis (Fig. 2B). This by-product might be a reason why the strain could not reached a higher conversion rate. The dehydrogenase promiscuity may allow for the use of costeffective feedstock as substrates; however, these dehydrogenases may increase the formation of by-products and influence the yield and the downstream purification process. Therefore, reducing by-product formation through gene knockout is a valuable tool to demonstrate the function of a specific gene [29][30][31]. This study found that the number of by-products in the fermentation broth significantly reduced after membrane-bound glucose dehydrogenase was knocked out (Fig. 2). According to their protein sequences, these enzymes belonged to the quinoprotein GDH (EC 1.1.5.2) with PQQ as a cofactor in gram-negative bacteria [32][33][34]. The substrate spectra of GDH were broad, including D-galactose, D-mannose, D-xylose, L-arabinose, maltose, and D-allose [28]. The newly discovered membrane-bound glucose dehydrogenase with catalytic activity in L-sorbose conversion has expanded the substrate spectrum of the glucose dehydrogenase family. It can be used to produce other valuable products.

Effects of combinatorial knockout of dehydrogenase genes on Lsorbose production
The aforementioned results showed that some G. oxydans WSH-003 dehydrogenase− deficient strains increased the production of L-sorbose and reduced the formation of by-products. Therefore, a total of 27 genes of G. oxydans WSH-003 were knocked out by consecutively deleting one gene after the other using the upp gene as counter-selection markers [35]. Among the 27 G. oxydans WSH-003 dehydrogenase− deficient strains, the highest titer and yield of the strain G. oxydans MD-16 was 149.46 g/L and 99.60%, respectively, which increased by 6.54% compared with that of G. oxydans WSH-003 (93.06%) ( Table 3). Regarding the growth of G. oxydans WSH-003 dehydrogenase-deficient strains, G. oxydans MD-1, G. oxydans MD-2, G. oxydans MD-3, G. oxydans MD-4, G. oxydans MD-5, and G. oxydans MD-15 grew better than G. oxydans Δupp. However, the final OD 600 of G. oxydans MD-26 and G. oxydans MD-27 were just 1.14 and 1.10, respectively, which was reduced by more than 50% and could not be further engineered for gene knockout (Table 3).
In some previous studies, L-sorbose production was enhanced by the overexpression of sldBA1 with a newly identified strong promoter [3,36]. However, these methods did not increase the conversion rate of D-sorbitol to L-sorbose and reduce by-products in the fermentation  Table 3 Comparison of multi-deletion strains for L-sorbose production.
process. Thus, gene knockout was necessary to decrease or eliminate the formation of by-products and improve substrate conversion; for example, the cadaverine production increased threefold with the deletion of some genes [37]. The present study showed that deletion of dehydrogenases could indeed improve the conversion rate of D-sorbitol to L-sorbose and reduce the formation of by-products. The conversion rate of the strain G. oxydans MD-16 reached up to 99.60%, close to the theoretical conversion rate of D-sorbitol to L-sorbose.

Effects of vhb expression on L-sorbose production
During cell cultivation, the oxygen demand is extremely high because G. oxydans-mediated incomplete oxidation of D-sorbitol is closely coupled to the cellular respiration chain [38]. The intracellular and periplasmic vhb expression systems were constructed to improve the oxygen utilization efficiency. The results showed that both the final OD 600 values of G. oxydans WSH-003 and G. oxydans MD-16 expressing vhb were about 3, and the expression of hemoglobin had no obvious effect on cell growth. Besides, the G. oxydans MD-16 with p13-tet-P 2703 -Tat-vhb had the highest yield of L-sorbose (149.58 g/L), with the conversion ratio of 99.7% (Fig. 3A). The cell membrane imposed limits on the transfer of oxygen to intracellular VHb, and the twin-arginine translocase (Tat) pathway was to export active VHb into the periplasm [25]. The effect of VHb expression in the periplasm was further identified. The results showed that that overexpression of VHb in the periplasm space did not obviously affect the cell growth in shake flask fermentation and could slightly improve the titer of L-sorbose to 149.82 g/L with a conversion ratio of 99.88% (Fig. 3B).
Effective control of DO levels is essential to ensure that microbes maintain high cell densities during the fermentation process, which requires a constant increase in oxygen supply [39]. So far, various solutions have been investigated to increase the DO of the fermentation broth, including increasing agitation speed and oxygen partial pressure, using pure oxygen, or completely rebuilding fermentation vessels [40][41][42]. However, the aforementioned methods were usually expensive or led to cell damage. Alternatively, the overexpression of Vitreoscilla hemoglobin in microorganisms was used to improve cell growth and production [43][44][45]. In this study, the overexpression of hemoglobin VHb from Vitreoscilla increased oxygen availability and improved L-sorbose production without affecting cell growth (Fig. 3).

L-sorbose production in the 5-L bioreactor
Based on the effects of VHb expression in the shaking flask, the relatively high-production strain expressing VHb in the periplasm (G. oxydans MD-16 with p13-tet-P vhb -Tat-vhb) was chosen to scale-up in a 5-L bioreactor. The G. oxydans strain harboring p13-tet was used as the control during the fermentation process (Fig. 4). The production of L- Fig. 3. Effects of VHb expression on L-sorbose production in shaking flasks.

Conclusions
In this study, the role of G. oxydans WSH-003 dehydrogenases in Lsorbose production was systematically investigated. The results proved that the engineering of dehydrogenases could further improve the titer and conversion rate of L-sorbose. By knocking out some dehydrogenase genes, the amounts of by-products of G. oxydans could be significantly reduced. Meanwhile, both the titer and the conversion rate of L-sorbose were improved. Additionally, by systematically knocking out dehydrogenases of G. oxydans, a high-performing strain achieved 149.46 g/L L-sorbose titer and 99.60% theoretical yield in the flask. By overexpressing vhb in G. oxydans MD-16, the titer of L-sorbose reached 298.61 g/L in a 5-L bioreactor. The systematic characterization and engineering of dehydrogenases in G. oxydans could also provide references for more efficient biosynthesis of other compounds in this strain.

Declaration of competing interest
The authors declare that they have no competing interests.