Production of diacetyl by metabolically engineered Enterobacter cloacae

Diacetyl, a high value product that can be extensively used as a food ingredient, could be produced from the non-enzymatic oxidative decarboxylation of α-acetolactate during 2,3-butanediol fermentation. In this study, the 2,3-butanediol biosynthetic pathway in Enterobacter cloacae subsp. dissolvens strain SDM, a good candidate for microbial 2,3-butanediol production, was reconstructed for diacetyl production. To enhance the accumulation of the precursor of diacetyl, the α-acetolactate decarboxylase encoding gene (budA) was knocked out in strain SDM. Subsequently, the two diacetyl reductases DR-I (gdh) and DR-II (budC) encoding genes were inactivated in strain SDM individually or in combination to decrease the reduction of diacetyl. Although the engineered strain E. cloacae SDM (ΔbudAΔbudC) was found to have a good ability for diacetyl production, more α-acetolactate than diacetyl was produced simultaneously. In order to enhance the nonenzymatic oxidative decarboxylation of α-acetolactate to diacetyl, 20 mM Fe3+ was added to the fermentation broth at the optimal time. In the end, by using the metabolically engineered strain E. cloacae SDM (ΔbudAΔbudC), diacetyl at a concentration of 1.45 g/L was obtained with a high productivity (0.13 g/(L·h)). The method developed here may be a promising process for biotechnological production of diacetyl.

diacetyl through metabolic engineering. In the present work, the ALDC encoding gene budA has been knocked out and the DRs encoding genes were also inactivated to construct a diacetyl producer ( Figure 1). Fe 31 was added to the medium to improve the nonenzymatic oxidative decarboxylation of a-acetolactate to produce diacetyl. Through the metabolic engineering approach described, 1.45 g/L diacetyl was synthesized within 11.3 h with a high yield of 0.21 mol/ mol using glucose as substrate.
Inactivation of DR-I in the ALDC mutant of E. cloacae SDM. Glycerol dehydrogenase (GDH) belongs to the medium-chain dehydrogenase family and accepts a broad range of substrates. Diacetyl could be reduced to (3R)-AC and (2R,3R)-2,3-BD by the GDH in K. pneumonia. A gdh gene (Gene bank: 13166340), which exhibits 59% sequence identity with that of K. pneumonia, was identified in the genome sequence of E. cloacae SDM. In this study, the protein encoded by gdh gene was renamed as DR-I due to its diacetyl reduction activity. As shown in Table 1, inactivation of DR-I would result in a lower DR activity of E. cloacae SDM (DbudADgdh) than that of the strain E. cloacae SDM and the mutant strain E. cloacae SDM (DbudA). However, the concentration of diacetyl increased modestly to only 326.7 mg/L ( Table 1) Inactivation of DR-II in the ALDC mutant of E. cloacae SDM. The genes that encode ALDC, ALS, and meso-2,3-BDH are sequentially clustered in one operon in E. cloacae SDM ( Figure S1). Our previously studied enzymatic reactions showed that meso-2,3-BDH can catalyze the conversion of diacetyl to (3S)-AC and further to (2S,3S)-2,3-BD as well as (3R)-AC to meso-2,3-BD. In this study, the meso-2,3-BDH (renamed as DR-II) encoding gene budC (Gene bank: 13167657) was knocked out through the allele exchange in E. cloacae SDM (DbudA) (Figure 2 IV-A).
As shown in Table 1, inactivation of DR-II would result in a sharp decrease of DR activity in E. cloacae SDM (DbudADbudC). The concentration of diacetyl increased to 416.1 mg/L after 36 h fermentation ( Figure 2 IV-C, Table 1). The budC mutant lost the ability to produce (2S,3S)-2,3-BD and meso-2,3-BD (Figure 2 IV-B). This  phenotype indicates that the formation of both (2S,3S)-2,3-BD and meso-2,3-BD depends on the activity of DR-II.
Then, DR-I and DR-II were both inactivated in the ALDC mutant of E. cloacae SDM ( Figure S2). As shown in Table 1, the DR activity would further decrease in the DR-I and DR-II double mutant. However, the glucose consumed, biomass, and concentration of diacetyl would also decrease in the mutant of E. cloacae SDM (DbudADbudCDgdh). Since the concentration (416.10 mg/L) of diacetyl obtained by E. cloacae SDM (DbudADbudC) was higher than that of other strains, E. cloacae SDM (DbudADbudC) was chosen for further investigation.
Diacetyl production by E. cloacae SDM (DbudADbudC). Diacetyl production using E. cloacae SDM (DbudADbudC) was conducted at 37uC in 300-mL shake flasks containing 50 mL medium. The medium was M9 medium supplemented with 18 g/L glucose and 5 g/L yeast extract 22 . The initial pH was 7.4. As shown in Figure 3, 59.8 mg/L diacetyl was obtained from 15 g/L glucose after 12 h of bioconversion. The yield of diacetyl was only at 0.83% of the theoretical value.
The concentration of a-acetolactate produced by E. cloacae SDM (DbudADbudC) was also analyzed during the 12 h of bioconversion. a-Acetolactate of 2.94 g/L was produced. This indicated that the strain E. cloacae SDM (DbudADbudC) showed an almost 3251 (mol/mol) co-production of a-acetolactate and diacetyl. Thus, diacetyl production could be further enhanced by the transformation of a-acetolactate accumulated in medium.
Optimization of the addition time of Fe 31 . In order to achieve higher diacetyl production, non-enzymatic oxidative decarboxyla-   Figure 4A and Figure 4B, although addition of Fe 31 at 12 h would result in a higher glucose utilization, the highest diacetyl concentration of 1.37 g/L was acquired when 20 mM Fe 31 was added at 10 h. Addition of Fe 31 at the beginning of the fermentation would inhibit the utilization of glucose and thus decrease the production of diacetyl by E. cloacae SDM (DbudADbudC).
Batch bioconversion under optimal conditions. Combining the results mentioned above, an optimal system for the production of diacetyl using E. cloacae SDM (DbudADbudC) was developed. Bioconversion was firstly conducted under the conditions mentioned above for 10 h ( Figure 5A). Then, 20 mM Fe 31 was added in the fermentation medium. As shown in Figure 5B, 1.45 g/L diacetyl was produced in 80 min after the addition of 20 mM Fe 31 . Glucose of 14.8 g/L was consumed during the bioconversion process. The yield of diacetyl was 0.21 mol/mol glucose. During the two-step bioconversion process, diacetyl was produced with a high productivity of 0.13 g/(L?h).

Discussion
Diacetyl has a strong buttery flavor and is mainly existed at low concentration in many dairy products, such as butter, beer, and fresh cheeses. Its formation in dairy products mainly results from the catabolism of a-acetolactate during 2,3-BD fermentation by certain species of lactic acid bacteria 14 . Due to the excellent performance of E. cloacae SDM as an efficient 2,3-BD producing strain, developing a metabolically engineered strain based on E. cloacae SDM through redirecting carbon flux toward the 2,3-BD pathways for the production of diacetyl is quite attractive and promising.
In the present study, the diacetyl production from glucose by E. cloacae SDM was firstly conducted through two genetic strategies: (i) inactivation of the ALDC gene (budA) to avoid enzymatic conversion of the diacetyl precursor a-acetolactate to (3R)-AC as described previously 14 and (ii) inactivation of the DR gene to avoid enzymatic reduction of diacetyl. Two DRs encoding genes (gdh and budC) were identified in the genome sequence of E. cloacae SDM. E. cloacae SDM (DbudADbudCDgdh) produced diacetyl at a concentration (318.31 mg/L) lower than that of E. cloacae SDM (DbudADbudC) (416.10 mg/L). This result indicates that DR might be important to strain SDM for glucose utilization and cell growth. On the other hand, when DR-I and DR-II were both inactivated in the ALDC mutant, (3R)-AC, (3S)-AC, and (2R,3R)-2,3-BD could still be detected ( Figure S2), indicating the presence of the third DR (DR-III) responsible for these chemical production in E. cloacae strain SDM (Figure 1). Although 2.94 g/L a-acetolactate was produced from 15 g/L glucose after 12 h of bioconversion, only 59.8 mg/L diacetyl was obtained and the final molar ratio of a-acetolactate and diacetyl was 3251 (Figure 3), implying an inefficient NOD of a-acetolactate to diacetyl. Thus, besides redirecting carbon flux toward production of a-acetolactate through genetic methods, more efficient chemical conversion of a-acetolactate into diacetyl should also be developed for optimal production of diacetyl. In the study by Gao et al. 12 , an efficient chemical conversion of a-acetolactate to diacetyl could be achieved by addition of Fe 31 . However, it was indicated that Fe 31 would also influence the glucose consumption ( Figure 4B) and hence might decrease the diacetyl production during the fermentation process. Thus, the addition time of 20 mM Fe 31 was also optimized in the present study. As shown in Figure 4A, when added at 10 h,   Several biotechnological routes have been used to produce diacetyl ( Table 2). Among all of the reported biotechnological processes, the group of Liu obtained the highest diacetyl concentration of 4.7 g/L with a metabolically engineered C. glabrata 12 . Efforts have been tried in order to increase the yield of diacetyl through inactivation of ALDC and overexpression of NADH oxidase in L. lactis. Using 5 g/L glucose as the substrate, the recombinant L. lactis produced 0.38 g/L diacetyl at a high yield of 0.16 mol/mol glucose 14 . In this study, metabolic engineering based on 2,3-BD pathway was used to reconstruct E. cloacae SDM as a novel biocatalyst for diacetyl production. Under optimal conditions, the recombinant E. cloacae SDM (DbudADbudC) could produce diacetyl with rather high concentration (1.45 g/L), productivity (0.13 g/(L?h)) and yield (0.21 mol/mol). Both the productivity and yield of diacetyl produced by the recombinant E. cloacae were new records for diacetyl production ( Table 2) Bacterial strains and plasmids. All the strains and plasmids used in this study are listed in Table 3. E. coli DH5a was used for general cloning procedures. The pKR6K was used for gene knock-out in E. cloacae strain SDM 24 . E. coli S17-1, which is able to host pKR6K and its derivatives, was used for conjugation with E. cloacae SDM. Lysogenic broth (LB) medium was used for the culture of E. coli and E. cloacae SDM. The selection medium in the conjugation experiments was M9 minimal medium supplemented with 1% sodium citrate as the carbon source and 0.05% ammonium chloride as the nitrogen source. Solid LB medium with 10% sucrose was used to select plasmid excision from the chromosome during the gene allelic exchange experiments. Kanamycin was used at a concentration of 50 mg/mL.
Knock out of the genes in E. cloacae SDM. Primers used in this study are listed in Table S1. Isolation of vectors, restriction enzyme digestion, agarose gel electrophoresis, and other DNA manipulations are carried out by standard protocols 25 . Mutants of E. cloacae strain SDM were generated by allele exchange using the suicide plasmid pKR6K 24 . The left and right flanking sequences were amplified from E. cloacae SDM and then ligated through PCR to get DbudA fragment using primer pairs PDbudA.f (EcoRI)/PDbudA.r (overlap), PDbudA.f (overlap)/PDbudA.r (BamHI). The gel-purified DbudA fragments were ligated to the pKR6K vector digested with the EcoRI and BamHI. The resulting plasmid was designated pKDbudA. For conjugation, donor and recipient strains were grown in LB to initial log phase (OD 600 nm 5 0.5), then collected and mixed at a ratio of 551 and spotted on LB plate. After 12 h of conjugation at 37uC, cells were recovered by washing the LB plate with normal saline and plated on the selection medium plates to eliminate the donor strain. The merodiploid (single-crossover) genotype was confirmed by PCR using primers PDbudA.f (EcoRI) and PDbudA.r (BamHI). Next, a single merodiploid colony was grown overnight in LB medium and appropriate dilutions were plated onto LB agar with 10% (w/v) sucrose, and then incubated overnight at 37uC. Colonies were screened by PCR using primers PDbudA.f (EcoRI) and PDbudA.r (BamHI). The budC and gdh mutants of strain SDM were generated by the same way of E. cloacae SDM (DbudA).  Batch fermentation. The batch fermentation was conducted in 300-mL shake flasks containing 50 mL medium. The medium consisted of M9 medium supplemented with 18 g/L glucose and 5 g/L yeast extract. The cultivation was carried out at 37uC and 180 rpm. The initial pH was adjusted to 7.4. Samples were collected periodically to determine the Cell density, concentrations of glucose, diacetyl, and a-acetolactate.
Enzyme activity assays. For the assays of the activities of ALDC and DR, cells of the strain were grown for 8 h, then centrifuged at 13,000 3 g for 5 min, and washed twice with 67 mM phosphate buffer (pH 7.4). Cells were finally resuspended with 67 mM phosphate buffer (pH 7.4) to an OD 600 nm of 20, and disrupted with an ultrasonic cell breaking apparatus (Xinzhi, Ningbo, China). Cell debris was removed through centrifugation at 13,000 3 g for 15 min. Enzyme activity was assayed in the resulting supernatant.
The activity of ALDC was assayed by detecting the production of AC from a-acetolactate 26 . a-Acetolactate was prepared immediately before use from ethyl 2-acetoxy-2-methyl-acetoacetate according to the protocol supplied by the manufacture. One unit of ALDC activity was defined as the amount of protein that produced 1 mmol of AC per min.
The activity of DR was assayed spectrophotometrically by measuring the change in absorbance at 340 nm corresponding to the oxidation of NADH (e 340 5 6,220 M 21 cm 21 ) at 30uC using a UV/visible spectrophotometer (Ultrospec 2100 pro, Amersham Biosciences, USA) 27,28 . The reaction solution for DR assay contained 5 mM of diacetyl and 0.2 mM of NADH in 67 mM phosphate buffer (pH 7.4). One unit of activity was defined as the amount of enzyme that consumed 1 mmol of NADH per min. The protein concentration was measured by the Lowry method, with bovine serum albumin as the standard 29 .
Analytical methods. Samples were withdrawn periodically and centrifuged at 12,000 3 g for 10 min. The Cell density was determined by monitoring the absorbance at 600 nm using a spectrophotometer (LENGGUANG-721, China) after an appropriate dilution. The concentration of glucose was measured enzymatically by a bio-analyzer (SBA-40D, Shandong Academy of Sciences, China) after diluting to an appropriate concentration. The concentrations of 2,3-BD and AC were analyzed by GC as described in Ma et al 6 . The concentrations of a-acetolactate and diacetyl were determined by the methods described in the previous reports 12, 30 . www.nature.com/scientificreports