Introduction

2,3-Butanediol (2,3-BD) can be efficiently produced by microbial fermentation as a platform and fuel bio-chemical1,2,3,4. Many microorganisms such as Enterobacter, Klebsiella, Bacillus and Serratia could be used to produce 2,3-BD from biomass5,6,7,8,9,10. Three key enzymes, including α-acetolactate synthase (ALS), α-acetolactate decarboxylase (ALDC) and 2,3-butanediol dehydrogenase (BDH), are involved in the biosynthesis of 2,3-BD from pyruvate2,4. Two molecules of pyruvate are condensed to α-acetolactate by ALS. Then, ALDC catalyzes the decarboxylation of α-acetolactate to produce (3R)-acetoin ((3R)-AC). (3R)-AC will be reduced to meso-2,3-BD and (2R,3R)-2,3-BD by meso-2,3-BDH and (2R,3R)-2,3-BDH, respectively11. Since meso-2,3-BDH and (2R,3R)-2,3-BDH could also catalyze the reduction of diacetyl to produce AC and 2,3-BD, these enzymes were also called as diacetyl reductase (DR).

α-Acetolactate produced in the 2,3-BD fermentation is unstable and can also be catalyzed through nonenzymatic oxidative decarboxylation to produce diacetyl12,13. Diacetyl is an important flavor compound responsible for the buttery aroma of many dairy products and is used as an ingredient in the food industry14. Diacetyl could also be produced through chemical methods. However, microbial production of diacetyl is preferred over chemical synthesis as a food or perfume additive due to the safety reasons15.

Various microorganisms such as lactic acid bacteria, Candida glabrata and Enterobacter aerogenes, have been used in the production of diacetyl12,16. For instance, a Lactococcus lactis mutant with low ALDC and lactate dehydrogenase activities was able to overproduce diacetyl with a final concentration of 0.52 g/L17. Studies by Guo et al. demonstrated that overexpression of NADH oxidase could efficiently tune the lactate and diacetyl production in L. lactis18. A combination of NADH oxidase overexpression with ALDC inactivation in L. lactis could produce diacetyl with a yield of 0.16 mol/mol of glucose, which was the highest yield of diacetyl synthesis till now14. Recently, C. glabrata CCTCC M202019 was metabolically engineered for diacetyl production. A high titer of 4.7 g/L of diacetyl was achieved with a yield of 0.10 mol/mol and a productivity of 0.07 g/(L·h)12. Diacetyl is a byproduct of the 2,3-BD fermentation, but most of the studies mentioned above have been based on strains with low 2,3-butanediol producing capabilities14,19,20. Due to their low efficiency of the glycolytic flux to diacetyl, the productivity and yield of diacetyl using these strains should be further enhanced for industrial production.

In our previous reports, Enterobacter cloacae subsp. dissolvens SDM can efficiently produce 2,3-BD with a high productivity and a high yield5. The key enzymes for 2,3-BD metabolism, including ALS, ALDC and the two DRs, were annotated based on the genome sequence of the strain21. Thus, it might be a good candidate for production of 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). Fe3+ was added to the medium to improve the nonenzymatic oxidative decarboxylation of α-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.

Figure 1
figure 1

Engineered pathway for diacetyl production in E. cloacae SDM.

NOD, nonenzymatic oxidative decarboxylation; ALS, α-acetolactate synthase, encoded by budB; ALDC, α-acetolactate decarboxylase, encoded by budA; DR-I, diacetyl reductase-I (glycerol dehydrogenase), encoded by gdh; DR-II, diacetyl reductase-II (meso-2,3-butanediol dehydrogenase), encoded by budC; DR-III, an undiscovered diacetyl reductase. Crosses represent the enzyme inactivation performed in this study.

Full size image

Results

Potential for diacetyl production by E. cloacae SDM

2,3-BD exists in three stereoisomeric forms: (2R,3R)-2,3-BD, (2S,3S)-2,3-BD and meso-2,3-BD. Recently, the mechanism of 2,3-BD stereoisomer formation was identified in 2,3-BD producing strains including K. pneumonia. (2R,3R)-2,3-BD and meso-2,3-BD are mainly produced from the (3R)-AC by the reactions catalyzed by (2R,3R)-2,3-BDH and meso-2,3-BDH, respectively11. (2S,3S)-2,3-BD could only be produced by the meso-2,3-BDH catalyzed reduction of (3S)-AC, which is not an enzymatic decarboxylation product of α-acetolactate but a reduction product of diacetyl11. As shown in Figure 2 I-B, the main metabolic products of E. cloacae SDM were meso-2,3-BD and (2S,3S)-2,3-BD when glucose was used as the carbon source. Low concentrations of (3R)-AC and (3S)-AC were also produced under aerobic conditions. (2S,3S)-2,3-BD could only be produced via the meso-2,3-BDH catalyzed two step reduction of diacetyl, the nonenzymatic oxidative decarboxylation product of α-acetolactate. Thus, diacetyl was produced as an intermediate of 2,3-BD biosynthesis in E. cloacae SDM. Redirecting more carbon flux toward the 2,3-BD into diacetyl through metabolic engineering of E. cloacae SDM might result in an efficient strain for the production of diacetyl.

Figure 2
figure 2

Molecular authentication and metabolic production analysis of E. cloacae SDM and its derivatives.

Panel I: E. cloacae SDM; Panel II: SDM (ΔbudA); Panel III: SDM (ΔbudAΔgdh); Panel IV: SDM (ΔbudAΔbudC). (A): Analysis of PCR fragments to confirm disruption of the gene. Lane M, molecular mass standard (Trans5K). (B): Fermentation products identified by GC. (C): Colorimetric detection of diacetyl (1, 2, 3 means all assays were performed by triplicate cultures). I-A: Lane 1–3: budA, gdh, budC products amplified with SDM genomic DNAs as the template. II-A: Lane 1–2, budA products amplified with SDM and SDM (ΔbudA) genomic DNAs as the templates, respectively. III-A: lane 1–2: gdh products amplified with SDM (ΔbudAΔgdh) and SDM genomic DNAs as the templates, respectively; lane 3–4: budA products amplified with SDM (ΔbudAΔgdh) and SDM genomic DNAs as the templates, respectively. IV-A: Lane 1–2: budC products amplified with SDM (ΔbudAΔbudC) and SDM genomic DNAs as the templates, respectively; lane 3–4: budA products amplified with SDM (ΔbudAΔbudC) and SDM genomic DNAs as the templates, respectively.

Full size image

Metabolic characteristics of the ALDC mutant of E. cloacae SDM

Although there was nonenzymatic oxidative decarboxylation of α-acetolactate in E. cloacae SDM, little diacetyl accumulated in the 2,3-BD fermentation process (Figure 2 I-A Figure 2 I-C and Table 1). Since α-acetolactate is mainly subjected to enzymatic conversion to (3R)-AC catalyzed by ALDC besides nonenzymatic oxidative conversion to diacetyl, lack of diacetyl accumulation might be due to the high degradation rate of α-acetolactate. Since the enzymatic irreversible reaction drains the available pool of α-acetolactate for diacetyl formation, knockout of the ALDC might be an effective method for the enhancement of diacetyl production in the E. cloacae strain SDM.

Table 1 Enzyme activities, glucose consumption and diacetyl production of E. cloacae SDM and its derivativesa
Full size table

In this study, E. cloacae SDM (ΔbudA) was constructed by knock-out of the budA gene (Gene bank: 13167655) through allele exchange (Figure 2 II-A). The effects of budA gene deletion on the ALDC activity and diacetyl formation of strain SDM are shown in Table 1 and Figure 2 II-C, respectively. In the native strain, the ALDC activity towards α-acetolactate was 3.81 ± 0.16 U/mg while little ALDC activity (0.04 ± 0.00 U/mg) was detected in E. cloacae SDM (ΔbudA). After 36 h fermentation, the concentration of diacetyl produced by E. cloacae SDM (ΔbudA) was 59.7 mg/L while only 2.85 mg/L diacetyl was obtained by the native strain SDM (Table 1). Besides diacetyl, (3R)-AC, (3S)-AC, (2R,3R)-2,3-BD, (2S,3S)-2,3-BD and meso-2,3-BD were also detected in the medium (Figure 2 II-B). These results indicate that the diacetyl would also be converted into those compounds in E. cloacae SDM (ΔbudA).

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 (ΔbudAΔgdh) than that of the strain E. cloacae SDM and the mutant strain E. cloacae SDM (ΔbudA). However, the concentration of diacetyl increased modestly to only 326.7 mg/L (Table 1). (3R)-AC, (3S)-AC, (2R,3R)-2,3-BD, (2S,3S)-2,3-BD and meso-2,3-BD would still accumulate during the fermentation (Figure 2 III-B).

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 (ΔbudA) (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 (ΔbudAΔbudC). 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 (ΔbudAΔbudCΔgdh). Since the concentration (416.10 mg/L) of diacetyl obtained by E. cloacae SDM (ΔbudAΔbudC) was higher than that of other strains, E. cloacae SDM (ΔbudAΔbudC) was chosen for further investigation.

Diacetyl production by E. cloacae SDM (ΔbudAΔbudC)

Diacetyl production using E. cloacae SDM (ΔbudAΔbudC) was conducted at 37°C 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 extract22. 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.

Figure 3
figure 3

Diacetyl production from glucose using E. cloacae SDM (ΔbudAΔbudC).

Glucose (), Cell density (), α-Acetolactate (), Diacetyl (). The bioconversion was carried out at 37°C in 300-mL shake flasks containing 50 mL of medium with pH adjusted to 7.4. The initial glucose concentration used was 18 g/L approximately. Error bars indicate standard deviation (n = 3).

Full size image

The concentration of α-acetolactate produced by E. cloacae SDM (ΔbudAΔbudC) was also analyzed during the 12 h of bioconversion. α-Acetolactate of 2.94 g/L was produced. This indicated that the strain E. cloacae SDM (ΔbudAΔbudC) showed an almost 32:1 (mol/mol) co-production of α-acetolactate and diacetyl. Thus, diacetyl production could be further enhanced by the transformation of α-acetolactate accumulated in medium.

Optimization of the addition time of Fe3+

In order to achieve higher diacetyl production, non-enzymatic oxidative decarboxylation of α-acetolactate should be enhanced. It was reported that α-acetolactate could be further converted to diacetyl by addition of Fe3+12,23. To study the effect of the addition time of Fe3+ on diacetyl production, 20 mM Fe3+ was added at 0, 2, 4, 6, 8, 10 and 12 h, respectively. The diacetyl production and glucose consumption were detected after 14 h fermentation.

As shown in Figure 4A and Figure 4B, although addition of Fe3+ at 12 h would result in a higher glucose utilization, the highest diacetyl concentration of 1.37 g/L was acquired when 20 mM Fe3+ was added at 10 h. Addition of Fe3+ at the beginning of the fermentation would inhibit the utilization of glucose and thus decrease the production of diacetyl by E. cloacae SDM (ΔbudAΔbudC).

Figure 4
figure 4

Effects of addition time of Fe3+ on diacetyl production (A) and glucose consumption (B).

The data were obtained after 14 h fermentation with Fe3+ addition at different time points. The bioconversion was carried out at 37°C in 300-mL shake flasks containing 50 mL of medium with pH adjusted to 7.4. The final concentration of Fe3+ added to medium was 20 mM. Error bars indicate standard deviation (n = 3).

Full size image

Batch bioconversion under optimal conditions

Combining the results mentioned above, an optimal system for the production of diacetyl using E. cloacae SDM (ΔbudAΔbudC) was developed. Bioconversion was firstly conducted under the conditions mentioned above for 10 h (Figure 5A). Then, 20 mM Fe3+ 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 Fe3+. 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).

Figure 5
figure 5

Time course of batch fermentation of diacetyl under the optimization condition.

Glucose (), Cell density (), α-Acetolactate (), Diacetyl (). (A) Time course of diacetyl and α-acetolactate production using E. cloacae SDM (ΔbudAΔbudC) before Fe3+ addition. (B) Time course of diacetyl production after 20 mM Fe3+ added at 10 h. The bioconversion was carried out at 37°C in 300-mL shake flasks containing 50 mL of medium with pH adjusted to 7.4. The initial glucose concentration used was 17.8 g/L. Error bars indicate standard deviation (n = 3).

Full size image

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 α-acetolactate during 2,3-BD fermentation by certain species of lactic acid bacteria14. 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 α-acetolactate to (3R)-AC as described previously14 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 (ΔbudAΔbudCΔgdh) produced diacetyl at a concentration (318.31 mg/L) lower than that of E. cloacae SDM (ΔbudAΔbudC) (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 α-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 α-acetolactate and diacetyl was 32:1 (Figure 3), implying an inefficient NOD of α-acetolactate to diacetyl. Thus, besides redirecting carbon flux toward production of α-acetolactate through genetic methods, more efficient chemical conversion of α-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 α-acetolactate to diacetyl could be achieved by addition of Fe3+. However, it was indicated that Fe3+ 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 Fe3+ was also optimized in the present study. As shown in Figure 4A, when added at 10 h, 20 mM Fe3+ could accelerate the NOD of α-acetolactate and accumulate the highest concentration of diacetyl.

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. glabrata12. 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 glucose14. 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 (ΔbudAΔbudC) 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). The carbon flux channeled into the diacetyl biosynthetic might be further enhanced since there were still (3R)-AC, (3S)-AC and (2R,3R)-2,3-BD accumulated during the fermentation (Figure 2 IV-C Figure 2 IV-B). This may be accomplished by searching the undiscovered diacetyl reductase, or overexpressing NADH oxidase, which could lead to prevention of NADH dependent reduction of diacetyl.

Table 2 Comparison of diacetyl production by different microorganisms
Full size table

Methods

Chemicals and biochemicals

(2R,3R)-2,3-BD (98.0%), (2S,3S)-2,3-BD (99.0%) and meso-2,3-BD (98.0%) were purchased from ACROS (The Kingdom of Belgium). Racemic AC, ethyl 2-acetoxy-2-methyl-acetoacetate and diacetyl were purchased from Sigma. NADH was purchased from Amresco. Restriction enzymes were purchased from TaKaRa Bio Inc. (China). PCR primers were prepared by Sangon (Shanghai, China). FastPfu DNA polymerase and T4 DNA ligase were purchased from Transgen Biotech (China) and MBI (USA), respectively. All other chemicals were of analytical grade and commercially available.

Bacterial strains and plasmids

All the strains and plasmids used in this study are listed in Table 3. E. coli DH5α was used for general cloning procedures. The pKR6K was used for gene knock-out in E. cloacae strain SDM24. 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 μg/mL.

Table 3 Bacterial strain and plasmid used in this studya
Full size table

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 protocols25. Mutants of E. cloacae strain SDM were generated by allele exchange using the suicide plasmid pKR6K24. The left and right flanking sequences were amplified from E. cloacae SDM and then ligated through PCR to get ΔbudA fragment using primer pairs PΔbudA.f (EcoRI)/PΔbudA.r (overlap), PΔbudA.f (overlap)/PΔbudA.r (BamHI). The gel-purified ΔbudA fragments were ligated to the pKR6K vector digested with the EcoRI and BamHI. The resulting plasmid was designated pKΔbudA. For conjugation, donor and recipient strains were grown in LB to initial log phase (OD600 nm = 0.5), then collected and mixed at a ratio of 5:1 and spotted on LB plate. After 12 h of conjugation at 37°C, 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 PΔbudA.f (EcoRI) and PΔbudA.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 37°C. Colonies were screened by PCR using primers PΔbudA.f (EcoRI) and PΔbudA.r (BamHI). The budC and gdh mutants of strain SDM were generated by the same way of E. cloacae SDM (ΔbudA).

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 37°C 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 α-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 × 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 OD600 nm of 20 and disrupted with an ultrasonic cell breaking apparatus (Xinzhi, Ningbo, China). Cell debris was removed through centrifugation at 13,000 × 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 α-acetolactate26. α-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 μmol 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 (ε340 = 6,220 M−1 cm−1) at 30°C 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 μmol of NADH per min. The protein concentration was measured by the Lowry method, with bovine serum albumin as the standard29.

Analytical methods

Samples were withdrawn periodically and centrifuged at 12,000 × 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 al6. The concentrations of α-acetolactate and diacetyl were determined by the methods described in the previous reports12,30.