Identification of the enzyme catalyzing D-AC to D-2,3-BD in B. licheniformis

To produce meso-2,3-BD with high optical purity in B. licheniformis, we proposed to block the synthesis of D-2,3-BD by knocking out the gene encoding the enzyme catalyzing the conversion of D-AC to D-2,3-BD, so that the carbon flux from D-AC would be diverted into meso-2,3-BD (Fig. 1). It was hypothesized that the reversible conversion of D-AC to D-2,3-BD was catalyzed by a putative D-2,3-butanediol dehydrogenase (D-BDH) enzyme (Fig. 1). Although gene bdhA encoding the D-BDH enzyme had been identified in B. subtilis [12]), the analysis of the B. licheniformis genome did not reveal the existence of any homologs to bdhA. Alternatively, researchers have reported that the glycerol dehydrogenase (GDH) in B. licheniformis is capable of catalyzing the in vitro conversion of D-AC to D-2,3-BD [19]. Similar activities for GDHs (DhaD and GldA) were also observed in K. pneumonia [20]. Based on these results, the gdh gene in B. licheniformis was studied for its role in D-2,3-BD biosynthesis.

A 1104-bp putative GDH encoding gene (gdh) was identified in B. licheniformis WX-02 [21]. The deduced amino acid sequence of the protein (WX-02 GDH) was compared with GDHs reported in other strains via sequence alignments through ClustalW. Comparative analysis revealed that WX-02 GDH had nearly all the conserved residues of the iron-dependent alcohol dehydrogenase (Fe-ADH), including three glycine residues that are predicted to interact with the NAD(H) cofactor and three histidine residues that coordinate an iron cofactor. This suggests that the WX-02 GDH is a member of the Fe-ADH superfamily. In addition, a strong similarity (>48 %) was found between WX-02 GDH and GDHs, exhibiting D-BDH activity from other organisms, such as Serratia marcescens [22], Klebsiella pneumoniae (DhaD, KJ206474.1), Citrobacter freundii (DhaD, P45511.1), Klebsiella oxytoca (DhaD, YP_005016612), and B. licheniformis 10-1-A (GldA, AHW84748.1).

As shown in Fig. 2a, the His-tagged recombinant GDH protein was efficiently expressed in soluble form after IPTG induction. The purified GDH protein was observed as a single band on SDS-PAGE with an approximate molecular weight consistent with that predicted from its amino acid sequence (39.5 kDa). The activity of purified GDH from WX-02 was further investigated for its catalytic activities on various substrates with coenzymes. Figure 2b showed that WX-02 GDH accepted D-AC, D-2,3-BD, meso-2,3-BD, and glycerol as substrates. However, this enzyme exhibited the highest activity towards D-2,3-BD among various substrates. Its catalytic activity on glycerol was only around 5 % of that for D-2,3-BD. The preference on D-2,3-BD as the substrate by GDH was also reported in Hansenula polymorpha [23]. These results indicated that the GDH from B. licheniformis WX-02 possessed a substrate specific catalytic activity towards D-2,3-BD. Based on these observations, we concluded that the GDH enzyme was the best candidate enzyme for catalyzing D-2,3-BD synthesis in B. licheniformis WX-02. Consequently, we decided to delete the gdh gene, so the conversion of D-AC to 2,3-BD isomers could be diverted solely towards meso-2,3-BD production.

Deletion of gdh gene for production of meso-2,3-BD with high purity

To investigate the role of GDH in the biosynthesis of 2,3-BD isomers, we constructed WX-02Δgdh, a gdh-deficient strain of B. licheniformis WX-02. As shown in Fig. 3, WX-02 produced both meso-2,3-BD and D-2,3-BD, while no D-2,3-BD was produced by WX-02Δgdh. Complementation of gdh in WX-02Δgdh (strain WX-02Δgdh/pHY-gdh) restored its ability to produce D-2,3-BD. Figure 3 also showed dramatic differences in the production of 2,3-BD and D-AC between the wild-type and the engineered strain. Compared to WX-02, production of meso-2,3-BD and D-AC increased by 66.3 and 37.8 %, respectively, in WX-02Δgdh. The results indicate that the deletion of gdh gene from the wild-type strain eliminates the synthesis of D-2,3-BD, leading to an accumulation of its precursor D-AC, which in turn promoted the conversion of D-AC to meso-2,3-BD. Meso-2,3-BD was the only 2,3-BD isomer detected in the medium, resulting in a product of high purity. It was also found that the mutant WX-02Δgdh/pHY-gdh not only restored the production of D-2,3-BD but produced an even higher titer of total 2,3-BD isomers (D-2,3-BD and meso-2,3-BD) compared to that of wild-type. Correspondingly, the D-AC produced by the complementation strain was lower than the wild-type WX-02. The distribution profile of these metabolites indicates that the high expression of gdh gene is likely a result of the strong promoter P43 and multiple copies of pHY300PLK vector.

Deletion of acoR gene for accumulation of acetoin

In addition to serving as the precursor for 2,3-BD synthesis, acetoin has also been reported to be used as a carbon source by B. licheniformis. This is regulated by AoDH ES (Fig. 1) when glucose was depleted [17, 24]. Acetoin dissimilation would reduce its availability for meso-2,3-BD synthesis resulting in an adverse effect on 2,3-BD production. In this work, the acoR gene encoding transcriptional activator of AoDH ES was disrupted to block the degradation of acetoin. Two separate experiments were then performed to confirm the effect of the acoR knock-out on acetoin metabolism. First, the acoR-disrupted strain WX-02ΔacoR was grown in the medium containing acetoin as the sole carbon source. The wild-type strain WX-02 was also grown in this medium as a control. As shown in Fig. 4a, the mutant WX-02ΔacoR had a very poor growth compared to that of the wild-type WX-02. The poor growth caused by the acetoin utilization deficiency was also observed in B. subtilis and B. licheniformis strains with inactivated acoABCL operons [17, 25, 26]. The results of the acetoin consumption study presented in Fig. 4a also indicated the inability of utilizing acetoin by the acoR-disrupted mutant. Second, WX-02ΔacoR and WX-02 were grown in 2,3-BD production medium containing glucose as the carbon source. As shown in Fig. 4b, the mutant WX-02ΔacoR produced 15.7 and 4.3 % more D-AC and total 2,3-BD isomers than the wild-type WX-02, respectively. The results clearly demonstrate the role of acoR in D-AC metabolism, and the deficiency of acoR contributed to D-AC accumulation, thus favoring 2,3-BD production.

Double deletion of gdh and acoR for producing meso-2,3-BD with high purity and titer

Optimization of meso-2,3-BD production by the mutant WX-02ΔgdhΔacoR

To this point, the previously described results demonstrated the capability of the mutant WX-02ΔgdhΔacoR to produce meso-2,3-BD with a high purity and titer. A bench-scale fermentation was further developed in both batch and fed-batch modes to explore the potential of commercial production of meso-2,3-BD by this double-deletion mutant.

Figure 5 shows the batch fermentation profiles. The growth (Fig. 5a) and glucose consumption (Fig. 5b) of the mutant WX-02ΔgdhΔacoR were similar to the wild-type WX-02 strain. The WX-02ΔgdhΔacoR strain accumulated higher acetoin (Fig. 5c), but negligible D-2,3-BD isomer (Fig. 5d) compared to the wild-type. The meso-2,3-BD isomer produced from WX-02ΔgdhΔacoR strain was double that of the wild-type strain (Fig. 5e). The yield of meso-2,3-BD from glucose, Y _{meso-2,3-BD/glucose}, was 0.35 g/g for WX-02ΔgdhΔacoR, also much higher than wild-type (0.16 g/g). Finally, lactic acid was still produced as the main by-product (Fig. 5f), while acetic acid and ethanol were less than 1.0 g/L throughout the fermentation process (data not shown). It should be noted that the lactic acid produced by the mutant WX-02ΔgdhΔacoR was similar to that produced by the wide type strain under fermenter culture conditions (Fig. 5f). This trend is different from that in flask culture (Table 1), where WX-02ΔgdhΔacoR produced less lactic acid than WX-02. The reason may be due to the different operation conditions between the flask culture and the fermenter culture.

Compared to the shake-flask culture results (Table 1), both the wild-type and WX-02ΔgdhΔacoR strains produced more meso-2,3-BD with less accumulation of acetoin in the fermenter culture. This was due to the use of the two-stage agitation/aeration, and thus the controlled dissolved oxygen (DO) levels, during the fermenter operation. It has been reported that DO levels play a crucial role in the reversible conversion of acetoin to 2,3-BD [14–16, 19]. In this work, the relatively high DO level generated by high aeration and agitation during the initial fermentation stage produced an increased amount of acetoin, resulting in the favorable conversion to meso-2,3-BD during the second stage when DO was intentionally reduced to a lower level.

A fed-batch fermentation using WX-02ΔgdhΔacoR was further developed to improve the meso-2,3-BD isomer titer. As shown in Fig. 6, glucose was maintained at 10-20 g/L throughout the entire culture period through periodic feedings. Meso-2,3-BD titer reached up to 98.0 g/L with a yield Y _{meso-2,3-BD/glucose} of 0.40 g/g and productivity of 0.94 g/L–h, which was the highest meso-2,3-BD yield reported in Bacillus species (Table 2).

Strain	Substrate	2,3-BD isomer	Method	Concentration (g/L)	Yield (g/g)	Productivity (g/L–h)	References
Bacillus subtilis
AJ 1992 (WT)	Glucose	Mixture	Batch	2.5	0.38	0.33	[36]
RB03a	Glucose	Mixture	Batch	6.1	0.34	0.41	[37]
Bacillus amyloliquefaciens
F (WT)	Glucose	Mixture	Batch	33.0	0.33	–	[9]
B10-127 (WT)	Glucose	Mixture	Shake-flask	55.8	0.40	0.80	[38]
			Fed-batch	92.3	0.15	0.96
GARb	Crude glycerol	Mixture	Fed-batch	102.3	0.44	1.16	[39]
Bacillus licheniformis
Wild-type strain	Glucose	Mixture	Static flask	8.7	0.47	0.12	[40]
BL5c	Glucose	D-2,3-BD	Batch	12.2	0.45	2.65	[41]
WX-02ΔbudC	Glucose	D-2,3-BD	Shake-flask	30.8	0.27	1.28	[13]
X10 (WT)	Concentrated corn stover hydrolysate	Mixture	Fed-batch	74.0	0.47	2.06	[42]
DSM 8785 (WT)	Glucose	D-2,3-BD	Batch	72.6	0.42	0.86	[43]
			Fed-batch	144.7	0.40	1.14
			Immobilized	118.3	0.41	1.14
WX-02ΔgdhΔacoR	Glucose	Meso-2,3-BD	Shake-flask	28.2	0.24	1.18	This work
			Batch	35.1	0.35	1.30	This work
			Fed-batch	98.0	0.40	0.94	This work

^a B. subtilis 168 containing pAL10abA (P _alsSD -bdhA)

^b B. amyloliquefaciens B10-127 with pMA5-acr-HapII-dhaD-P_bdhA-alsR

^c B. licheniformis BL1Δldh

Strains and plasmids

Medium preparation and culture conditions

Escherichia coli and B. licheniformis seed cultures were performed in 250-mL flasks containing 50-mL LB medium at 37 °C on a rotary shaker (180 rpm). For 2,3-BD production, the seed culture with a OD₆₀₀ at 4.0 was inoculated (1 %, v/v) into either 250-mL flasks containing 50-mL medium or into a 5-L stir tank fermenter (Guo Qiang Bioengineering Equipment Company, Ltd. China) with a working volume of 3 L. The medium consisted of (per liter): glucose 120 g, corn steep liquor 33 g, (NH₄)₂SO₄ 9.0 g, K₂HPO₄ 1.0 g, MgSO₄ 1.5 g, NaCl 0.5 g, ZnCl₂ 0.12 g, FeCl₃ 0.001 g, and MnSO₄ 0.001 g, with an initial pH adjusted to 7.0 before autoclaving at 121 °C for 15 min [13]. Fermentations were carried out at 37 °C. The batch and fed-batch fermentation were operated under optimized conditions as follows: the pH was maintained at 6.0 with 6-M HCl or 7.5-M ammonia hydroxide; aeration, and agitation were, respectively, controlled at 3 L/min and 350 rpm for the first 16 h, and 1.5 L/min, and 200 rpm for the rest of fermentation. In the fed-batch operation, in addition to the above conditions used, a concentrated glucose solution (650 g/L) was added to the fermenter at a rate of 15 mL/h from 24–70 h to maintain the glucose concentration in the medium at a level of 10–20 g/L. As a result, the total glucose added to the fermenter was equivalent to 150 g/L.

Construction of the gene deletion mutants of B. licheniformis

To prepare the gdh-knock-out mutant, in-frame deletion of gdh gene in the genome of WX-02 was performed based on the method described previously [13]. Briefly, the 5′- and 3′-flanking sequences were amplified from the WX-02 genome using the primer pairs Δgdh-A-F/Δgdh-A-R and Δgdh-B-F/Δgdh-B-R, respectively. The two fragments were ligated by spliced overlap extension PCR (SOE-PCR) with the primers Δgdh-A-F and Δgdh-B-R. The fused fragment (1140 bp) was digested with XbaI and SacI, and inserted into T2(2)-ori, generating plasmid T2Δgdh. Transformation of B. licheniformis WX-02 with the knock-out plasmid T2Δgdh was performed via electroporation based on the previous protocols [33]. The kanamycin-resistant transformants were picked and verified by PCR with the primers T2-VF and T2-VR. A positive clone was incubated in LB medium containing 20-μg/mL kanamycin at 45 °C for 8 h, then streaked onto LB agar with kanamycin for another 8-h incubation to obtain single-crossover recombinants. The recombinants were grown in LB medium at 37 °C with serial subcultures to promote homologous recombination. The kanamycin sensitive colonies resulting from a double crossover event were confirmed by PCR with the primers Δgdh-F and Δgdh-R (Additional file 3: Figure S3). The mutant strain was designated as WX-02Δgdh. The acoR deletion mutant (WX-02ΔacoR) and gdh-acoR double-deletion mutant (WX-02ΔgdhΔacoR) were prepared using the similar procedures described above. The two mutants were confirmed through PCR with the primer pairs ΔacoR-F/R and Δgdh-F/R and ΔacoR-F/R, respectively (Additional file 3: Figure S3).

Construction of the gene complementation strain of B. licheniformis WX-02Δgdh

The P43 promoter was PCR-amplified from chromosomal DNA of B. subtilis 168 with the primers P43-gdh-TamyL-1 and 2 (Table 4). The gdh gene and terminator of the amyL gene were amplified from WX-02 genomic DNA with the primers P43-gdh-TamyL-3 to 6. The three fragments were fused by SOE-PCR with the primers P43-gdh-TamyL-1 and 6 (Table 4), and cloned into the XbaI and BamHI sites of pHY300PLK, forming the plasmid pHY-gdh. The expression vector pHY-gdh was then transformed into B. licheniformis WX-02Δgdh via electroporation based on a previously described method [34]. Positive transformants with tetracycline resistance were verified by PCR with the primers P43-gdh-TamyL-1 and 6 (Additional file 4: Figure S4), and was designated as WX-02Δgdh/pHY-gdh.

Cloning and expression of gdh gene in E. coli BL21(DE3)

The gdh gene was amplified from the genome of B. licheniformis WX-02 with the primers gdh-F and gdh-R (Table 4). The amplified product was ligated into the vector pET-28a(+) at NcoI and XhoI sites, resulting in the recombinant plasmid designated as pET-gdh. The plasmid was transformed into E. coli BL21(DE3) resulting in the recombinant strain E. coli BL21(DE3)/pET-gdh (Additional file 5: Figure S5) for protein expression. E. coli BL21(DE3)/pET-gdh was cultured in LB medium containing 20-μg/mL kanamycin at 37 °C to reach cell density (OD₆₀₀) of 1.0. IPTG was then added to the culture to a final concentration of 0.2 mM. Cells were grown for another 6 h and harvested by centrifugation at 8000 g for 15 min (4 °C). The cell pellets were disrupted with an Ultra-high Pressure Continuous Flow Cell Disrupter at 1000–1500 bar and 4 °C. The cell lysate was centrifuged at 12,000 g for 30 min (4 °C) to remove residual cell debris. The enzyme in the supernatant was purified using a Ni–NTA purification kits. The purity and identity of the protein were determined by sodium dodecyl sulfate polyacrylamide gel electrophoresis (SDS-PAGE).

Assay of activity and stereospecificity of glycerol dehydrogenase (GDH)

Here, one unit of activity was defined as the amount of enzyme required to consume or produce 1 μmol of NADH per minute. The specific activity of GDH was defined as the enzyme unit (U) divided by the amount of enzyme protein (mg). Protein concentrations were determined by the Bradford method [35].

Analyses of metabolites during fermentation

Cells were harvested from the fermentation broth via centrifugation at 12,000g for 10 min. The cell free media was kept for analysis. The cell pellets were washed twice with deionized water and re-suspended with deionized water for determining cell density at 600 nm (OD₆₀₀). Residual glucose concentration was measured enzymatically by a Bio-analyzer (SBA 40C, Shandong Academy of Sciences, China). Acetic acid, lactic acid, and ethanol were analyzed using an Agilent 1260 HPLC system. Analyte separation was achieved using Zorbax SB-Aq (4.6 mm ID × 25 0 mm, 5 μm) column with 99 % 20 mM Na₂HPO₄ and 1 % acetonitrile as the mobile phase (pH 2.0) at a flow rate of 0.5 mL/min. The column and detector temperatures were set at 35 °C; the injection volume was 10 μL and the detection wavelength was 210 nm. D-2,3-BD, meso-2,3-BD, and D-AC were analyzed by gas chromatography (Agilent Technologies 7890A) using the method described previously [13].