Mechanism of 2,3-butanediol stereoisomers formation in a newly isolated Serratia sp. T241

Serratia sp. T241, a newly isolated xylose-utilizing strain, produced three 2,3-butanediol (2,3-BD) stereoisomers. In this study, three 2,3-butanediol dehydrogenases (BDH1-3) and one glycerol dehydrogenase (GDH) involved in 2,3-BD isomers formation by Serratia sp. T241 were identified. In vitro conversion showed BDH1 and BDH2 could catalyzed (3S)-acetoin and (3R)-acetoin into (2S,3S)-2,3-BD and meso-2,3-BD, while BDH3 and GDH exhibited the activities from (3S)-acetoin and (3R)-acetoin to meso-2,3-BD and (2R,3R)-2,3-BD. Four encoding genes were assembled into E. coli with budA (acetolactate decarboxylase) and budB (acetolactate synthase), responsible for converting pyruvate into acetoin. E. coli expressing budAB-bdh1/2 produced meso-2,3-BD and (2S,3S)-2,3-BD. Correspondingly, (2R,3R)-2,3-BD and meso-2,3-BD were obtained by E. coli expressing budAB-bdh3/gdh. These results suggested four enzymes might contribute to 2,3-BD isomers formation. Mutants of four genes were developed in Serratia sp. T241. Δbdh1 led to reduced concentration of meso-2,3-BD and (2S,3S)-2,3-BD by 97.7% and 87.9%. (2R,3R)-2,3-BD with a loss of 73.3% was produced by Δbdh3. Enzyme activity assays showed the decrease of 98.4% and 22.4% by Δbdh1 and Δbdh3 compared with the wild strain. It suggested BDH1 and BDH3 played important roles in 2,3-BD formation, BDH2 and GDH have small effects on 2,3-BD production by Serratia sp. T241.

and 2,3-BD stereoisomers from pyruvate [18][19][20][21] . First, two molecules of pyruvate condense to yield α -acetolactate and release one molecule of CO 2 by ALS, then α -acetolactate is decarboxylated into (3R)-AC by ALDC 22,23 . Meanwhile, α -acetolactate is also readily to undergo nonenzymatic oxidative decarboxylation and form diacetyl (DA) under oxygen supply conditions 24 . Finally, (3R)-AC and DA are reduced into 2,3-BD with the corresponding configurations by different BDHs. In addition, reports have shown that glycerol dehydrogenases (GDH) play an important role in the formation of 2,3-BD stereoisomers. A new model have been proposed for the formation of 2,3-BD stereoisomers in K. pneumoniae which includes two enzymes (meso-BDH and GDH). The meso-BDH enzyme catalyzes the stereospecific conversion of (3R)-AC to meso-2,3-BD and DA to (2S,3S)-2,3-BD via (3S)-AC as an intermediate, whereas GDH shows the abilities in the conversion from (3R)-AC to (2R,3R)-2,3-BD and DA to meso-2,3-BD via (3S)-AC as an intermediate. The two enzymes contribute to the formation of three 2,3-BD isomers in K. pneumoniae 25 . Therefore, the existence of multiple stereospecific dehydrogenase in natural strains is regarded as key factor for the mixed formation of AC and 2,3-BD stereoisomers 11,26 .
Serratia sp. T241, a newly isolated 2,3-BD producing strain in our lab, could utilize xylose and glucose as carbon sources to produce 2,3-BD, exhibiting a potential for 2,3-BD production based on lignocellulose-derived sugars. Interestingly, this strain T241 could produce three isomers of 2,3-BD simultaneously during the sugar fermentation process. Usually natural microorganisms produce a mixture of (2S,3S)-2,3-BD/meso-2,3-BD or (2R,3R)-2,3-BD/meso-2,3-BD. It suggested that there might be some differences for the mechanism of AC and 2,3-BD stereoisomers formation between strain T241 and other reported 2,3-BD producing strains. To clarify the differences of 2,3-BD isomers formation between Serratia sp. T241 and other 2,3-BD producing strains, a blast search based on Serratia sp. genome sequence was carried out using known functional BDHs (meso-BDH, (2S,3S)-BDH and (2R,3R)-BDH) and GDHs sequences. The obtained putative BDH/GDH genes were selected to perform in vitro and in vivo experiments. Heterogenous expression and pathway assembly in E. coli along with budA (acetolactate decarboxylase) and budB (acetolactate synthase), responsible for converting pyruvate into acetoin, confirmed that the putative BDH/GDH enzymes showed the abilities in the interconversion between AC and 2,3-BD, which contributed to all the AC and 2,3-BD isomers formation. Furthermore, gene deletion and enzyme assay verified the roles of the genes for AC and 2,3-BD isomers formation in Serratia sp. T241.
Stereospecific characteristics of BDH1, BDH2, BDH3 and GDH enzymes. Four BDH/GDH genes were cloned, expressed and purified as described in "Materials and methods" (Fig. S1). The purified enzymes were used to determine the kinetic parameters using AC and 2,3-BD as substrates under their optimal pH conditions. The comparative data of apparent K m and K cat values for BDH1-3 and GDH from Serratia sp. T241 were given in Table 1. BDH1 and BDH2 showed the activities for (3S/3R)-AC, meso-2,3-BD and (2S,3S)-2,3-BD as substrates. No activity for BDH1 and BDH2 could be measured when (2R,3R)-2,3-BD was used as a substrate. The meso-2,3-BD and (2R,3R)-2,3-BD could be reduced or oxidized by BDH3 and GDH with NADH/NAD + as coenzymes. (2S,3S)-2,3-BD was not a substrate for BDH3 and GDH at all. In addition, GDH also exhibited high catalytic efficiency for glycerol as a substrate (data no shown). So BDH3 and GDH from Serratia sp. T241 should be categorized as (2R,3R)-BDH and GDH.
Assembly of the BDH/GDH genes with AC operon in E. coli. The roles of BDH1, BDH2, BDH3 and GDH were investigated for the production of AC and 2,3-BD using E. coli BL21(DE3) as the host, which has no native AC and 2,3-BD production metabolism. As illustrated in Fig. 2, four genes encoding BDH1, BDH2, BDH3 and GDH from Serratia sp. T241 were cloned and assembled along with AC operon from S. marcescens H30 into pET28a vector, to generate the plasmids pET-budRAB-bdh1, pET-budRAB-bdh2, pET-budRAB-bdh3 and pET-budRAB-gdh. All of the plasmids harbored the AC operon (budRAB), which is used to produce acetoin from pyruvate. The genes bdh1, bdh2, bdh3 and gdh were expressed for evaluating the conversion of AC to 2,3-BD. The recombinant E. coli strains were subjected to batch fermentation using LB medium with 10 g/l glucose and E. coli carrying pET-budRAB was used as control. The results were given in Fig. 3. The control strain mainly produced 3.46 g/l of (3R)-AC and 0.11 g/l of (3S)-AC (Fig. 3a). In addition, a trace amount of (2R,3R)-2,3-BD and meso-2,3-BD could be detected in the broth by the control strain. As shown in Fig. 3b-e, all the strains carrying pET-budRAB-bdh1, pET-budRAB-bdh2, pET-budRAB-bdh3 and pET-budRAB-gdh produced (3R)-AC in high amounts. meso-2,3-BD was also produced by all the strains but in much lower amounts for the strains containing bdh3 and gdh genes. The strains containing bdh1 and bdh2 genes formed (2S,3S)-2,3BD which was absent from the strains containing bdh3 and gdh genes. Instead the strains containing bdh3 and gdh genes produced high amounts of (2R,3R)-2,3BD, which was totally absent in the other two strains.
T241 Δ bdh1, Serratia sp. T241 Δ bdh2, Serratia sp. T241 Δ bdh3 and Serratia sp. T241 Δ gdh were constructed as described in "Materials and methods". In vitro conversion showed that BDH1, BDH2, BDH3 and GDH have the abilities in the interconversion between AC and 2,3-BD. However, the roles of the four genes for 2,3-BD isomers formation in Serratia sp. T241 still remained unclear. Whether all the four enzymes regulated AC and 2,3-BD isomers formation by Serratia sp. T241 or not was unknown. So the four mutants and the wild strain were cultured in fermentation medium to investigate their metabolic characteristics. The cultures were carried out in 250 ml flask with 50 ml fresh fermentation medium, and the products in the broth were analyzed and quantified by GC system. The results were shown in Fig   were produced at concentrations of 1.73, 3.45, 2.14 and 3.67 g/l respectively (Fig. 4a-c). The cell growth of Serratia sp. T241 Δ bdh1 was slower than that of the wild strain, and the DCW of 7.7 g/l was obtained at 30h.   (Fig. 4h,i), which were higher in comparison to those of the wild strain. While the concentration of (2S,3S)-2,3-BD with 1.43 g/l showed somewhat lower than that of the wild strain (Fig. 4i). During the batch fermentation process, bdh3 gene deletion led to accumulation of (3R)-AC (6.60 g/l) and decrease of (2R,3R)-2,3-BD (0.98 g/l) obviously (Fig. 4k,l). Similar to Serratia sp. T241 Δ bdh3, 6.01 g/l of (3R)-AC could be produced by Serratia sp. T241 Δ gdh (Fig. 4n). However, Serratia sp. T241 Δ gdh still produced 4.40 g/l of (2R,3R)-2,3-BD (Fig. 4o). Correspondingly, the enzyme activities of the four mutants and the wild strain during the fermentation process also were measured, and the results were shown in

Discussion
2,3-BD and its dehydrogenation product, AC, could be produced by several natural strains such as K. pneumoniae 9 , K. oxytoca 10 , S. marcescens 5,11 and B. polymyxa 12 . However, these natural strains produce a mixture of 2,3-BD and AC, which limits their applications. So elucidation of the mechanism for 2,3-BD isomers formation and development of engineered strains for single configuration production of 2,3-BD are required. Previous studies showed that (3R)-AC as main intermediate product was produced from pyruvate by ALS and ALDC enzymes, whereas low level of (3S)-AC was obtained from DA, which was formed by a nonenzymatic oxidative decarboxylation of α -acetolactate 19 . BDH is a reversible enzyme involved in the last step from AC to 2,3-BD 14 .
Recent studies showed that multiple dehdyrogenases could carry out the conversion from AC to 2,3-BD such as (2S,3S)-BDH, (2R,3R)-BDH, meso-BDH and GDH, which resulted in different configuration formation of 2,3-BD 4,13,26 . Therefore, the catalytic efficiency of dehydrogenases and the existence of multiple dehydrogenases in 2,3-BD producing strain were regarded as the key factors for 2,3-BD isomers formation. Serratia sp. T241, a newly isolated xylose-utlizing strain, could produce high concentration of 2,3-BD with three configurations. In this study, four genes (bdh1, bdh2, bdh3 and gdh) involved in AC and 2,3-BD isomers formation in Serratia sp. T241 were identified. This is the first report that four enzymes played roles in 2,3-BD isomers production in one strain. The sequencing results showed the bdh1, bdh2, bdh3 and gdh genes from Serratia sp. T241 shared high identities of 88%, 64%, 64% and 86% with meso-BDH from S. marcescens H30, (2S,3S)-BDH from R. erythropolis, (2R,3R)-BDH from B. subtilis 168 and GDH from S. marcescens H30, implying that the four enzymes might contribute to 2,3-BD isomers formation in Serratia sp. T241. The four genes encoding the BDH1, BDH2, BDH3 and GDH enzymes was cloned and expressed in E. coli BL21(DE3), purified and characterized. All the four purified enzymes exhibited the activities for AC and 2,3-BD as substrates. The main differences occurred in 2,3-BD oxidation reaction. (2S,3S)-2,3-BD and meso-2,3-BD were the substrates of BDH1 and BDH2, which showed no activity for (2R,3R)-2,3-BD. While BDH3 and GDH showed the activities for (2R,3R)-2,3-BD and meso-2,3-BD as substrates, (2S,3S)-2,3-BD was not the substrate for BDH3 and GDH at all. According to their Furthermore, in vitro conversion and in vivo assemble of 2,3-BD pathway in E. coli revealed that the four enzymes contributed to meso-2,3-BD production in Serratia sp. T241. (2S,3S)-2,3-BD was obtained from (3S)-AC by BDH1 and BDH2, whereas BDH3 and GDH led to (2R,3R)-2,3-BD formation from (3R)-AC. The detailed model of AC and 2,3-BD isomers formation by Serratia sp. T241 could be inferred from the data of stereospecificity in catalytic reaction by the four purified enzymes ( Fig. 1 and Fig. S2). As shown in Fig. 5, (3R)-AC was produced from pyruvate via α -acetolacetate by ALS and ALDC, and DA was formed by a non-enzymatic oxidation decarboxylation of α -acetolactate. In Serratia sp. T241, (3S)-AC production from DA was afforded by BDH1, BDH2 and GDH. The features of BDH1 and BDH2 were similar to reported meso-BDH and (2S,3S)-BDH from K. pneumoniae and R. erythropolis respectively 27,29 . However, the GDH enzymes in previous studies showed no activity using DA as a substrate except that GDH from S. marcescens H30 which could converted DA into (3S)-AC and exhibited the same catalytic property with the GDH from Serratia sp. T241 25,32 . DA also was converted into (3R)-AC due to the existence of BDH3 in Serratia sp. T241. BDH1 and BDH2 exhibited high (S)-enantioselectivity for AC as substrate and catalyzed (3S)-AC and (3R)-AC into (2S,3S)-2,3-BD and meso-2,3-BD, respectively. In contrast, high (R)-enantioselectivity of BDH3 and GDH led to (2R,3R)-2,3-BD and meso-2,3-BD production from (3R)-AC and (3S)-AC. Similar to reported BDH/GDH from other strains, the four enzymes showed the abilities in the conversion from 2,3-BD to AC, and the corresponding configuration of AC could be obtained. It suggested that BDH1, BDH2, BDH3 and GDH were four reversible enzymes for the interconversion between AC and 2,3-BD. In addition, no DA could be detected from any form of 2,3-BD and AC, which showed that the conversions in catalytic reactions by the four enzymes were irreversible between DA and AC. In previous studies, the model of 2,3-BD stereoisomers formation in K. pneumoniae has been revealed 25 . In this model, two enzymes meso-BDH and GDH were identified, and contributed to three configurations of 2,3-BD production. The GDH enzyme in K. pneumoniae was responsible for (2R,3R)-2,3-BD and meso-2,3-BD formation from (3R)-AC and (3S)-AC, while (2S,3S)-2,3-BD and meso-2,3-BD were produced from (3S)-AC and (3R)-AC by meso-BDH. K. pneumoniae deficient in one of the two enzymes only produced two configuration of 2,3-BD 25,33 . For Serratia sp. T241, four enzymes including meso-BDH, (2S,3S)-BDH, (2R,3R)-BDH and GDH were identified and contributed to three configuration of 2,3-BD formation, which showed that the model of 2,3-BD stereoisomers formation in Serratia sp. T241 was more complicated than that of K. pneumoniae. According the catalytic properties of four enzymes, inactivating one of the genes encoding the four enzymes in Serratia sp. T241 still forms three configuration of 2,3-BD. The significance of such a complicated mechanism for physiological metabolism of Serratia sp. T241 still remain unknown, which needs further study in future.
Though in vitro conversion and in vivo assemble of 2,3-BD pathway in E. coli exhibited the catalytic properties and stereospecificity of four enzymes for DA, AC and 2,3-BD isomers as substrates, the expression level of four enzymes in Serratia sp. T241 during the fermentation process might be quite different. Therefore, four mutants deficient in BDH1, BDH2, BDH3 and GDH were constructed for investigating their roles in 2,3-BD isomers formation by Serratia sp. T241. As shown in Table 2, the enzyme activity loss of 98.4% could be observed in Serratia sp. T241 Δ bdh1, resulting in an obvious decrease of 2,3-BD yield, especially meso-2,3-BD with the loss of 97.7% when compared with the wild strain. A higher concentration of (3R)-AC (25.50 g/l) could be accumulated due to the deficiency of BDH1. In addition, deletion of bdh1 gene also led to low level of (2S,3S)-2,3-BD with the loss of 87.9%. So BDH1 played a vital role in 2,3-BD production in Serratia sp. T241. In contrast, slightly decrease of the enzyme activities was detected in the mutants of Δ bdh2, Δ bdh3 and Δ gdh, which still produced a large amount of meso-2,3-BD. However, the yields of meso-2,3-BD by the three mutants were slightly less than that of the wild strain, implying that BDH2, BDH3 and GDH might also contributed to meso-2,3-BD formation. During the fermentation process, low level of (2R,3R)-2,3-BD with the decrease of 73.3% was produced by Serratia sp. T241   Δ bdh3. Therefore, BDH3 might played an important role in (2R,3R)-2,3-BD production in Serratia sp. T241. Low concentration of (2R,3R)-2,3-BD in Serratia sp. T241 Δ bdh3 was produced partially due to the contribution of GDH. Similarly, BDH2 also contributed to (2S,3S)-2,3-BD production since a small amount of (2S,3S)-2,3-BD could be detected in Serratia sp. T241 Δ bdh1. These results showed that the four enzymes contributed to AC and 2,3-BD stereoisomers formation in Serratia sp. T241 during the fermentation process. The ratio of AC and 2,3-BD stereoisomers in the broth by Serratia sp. T241 partially depended on the catalytic efficiencies and expression level of the four enzymes. Considering the discovery of the four enzymes from the blast search based on the reported gene sequences, other dehydrogenase in Serratia sp. catalyzing the conversion from AC to 2,3-BD remained unknown. So it is necessary that development of the mutant deficient in all the four genes in future confirm 2,3-BD accumulation or not. In a conclusion, four genes involved in 2,3-BD isomers formation by Serratia sp. T241 were identified. BDH1, BDH2, BDH3 and GDH encoded by the four genes were categorized as meso-2,3-BD dehydrogenase, (2S,3S)-2,3-BD dehydrogenase, (2R,3R)-2,3-BD dehydrogenase and glycerol dehydrogenase, and contributed to 2,3-BD isomers formation. During the fermentation process, BDH1 and BDH3 played main roles in 2,3-BD production, while BDH2 and GDH only had small effect on 2,3-BD production in Serratia sp. T241. Bacterial strains, plasmids and bacterial growth condition. The strains and plasmids used in this study are listed in Table 4. Escherichia coli DH5α and BL21(DE3) as the cloning and expression hosts were grown at 37 °C. The pET28a vector was used for enzyme expression and the suicide vector pUTKm1 was used for gene knockout in Serratia sp. T241. Luria-Bertani (LB) medium was used for cultivation of E. coli and Serratia sp. T241. LB medium with 10 g/l glucose was used to cultivate E. coli containing the recombinant 2,3-BD pathway. Antibiotics were added in the following amounts (per ml) if necessary: 50 μ g kanamycin or 50 μ g ampicillin.

Methods
Serratia sp. T241 was maintained on agar slants in LB medium. The slants were incubated at 30 °C, and fully grown slants were stored at 4 °C. For seed preparation, a full loop of cells from a fully grown slant was inoculated into Identification of putative Serratia sp. T241 BDH/GDH genes. Blast searches of Serratia sp. AS12 local protein database developed by BioEditor software were carried out using known functional BDHs and GDHs. meso-BDH and GDH sequences are from S. marcescens H30 11,32 , while (2S,3S)-BDH and (2R,3R)-BDH sequences are from R. erythropolis 29 and B. subtilis 36 respectively. The putative genes with high scores were collected as candidate genes for further experiments.
Construction of E. coli BL21(DE3)/BDH and E. coli BL21(DE3)/GDH. The encoding sequences of the putative BDH and GDH genes were amplified by PCR with the genomic DNA of Serratia sp. T241 as template using the primers (Table 3), which contained the EcoRI and HindIII restriction sites respectively. The amplified products were ligated into the vector pET-28a at EcoRI and HindIII sites to generate the recombinant plasmids. The recombinant plasmids were transformed into E. coli BL21(DE3) for commercial sequencing and protein expression. Enzyme preparation, assay, and enzymatic reactions. The recombinant strains were cultured at 37 °C in a 250-ml flask containing 50 ml LB medium with kanamycin (50 μ g/ml), and expression was induced at a 2.5 h culture (about 0.6 OD 600 ) with 0.5 mM isopropyl-beta-D-thiogalactopyranoside (IPTG). Cells were harvested by centrifugation after 6 h, and cell lysate was prepared by sonication in an ice bath. The homogenate was centrifuged at 13,000 × g for 10 min to remove the debris 11 . The soluble fraction was subjected to purification under non-denaturing conditions with Ni-affinity chromatography using a Histrap HP column according the purification protocol (GE Healthcare, USA). The eluate from the column was pooled and desalted by a Hitrap desalting column (GE Healthcare, USA). Enzyme activity was determined spectrophotometrically by measuring the changes in absorbance at 340 nm and 40 °C corresponding to the oxidation of NADH or the reduction of NAD + . The reaction mixtures containing 50 mM potassium phosphate buffer, 0.2 mM NAD + for the oxidation reactions or 50 mM potassium phosphate buffer, 0.2 mM NADH for the reduction reactions were incubated at 40 °C for 5 min. After adding 10 μ l of approximately diluted purified enzyme solution, the reaction was started by the addition of the substrates. One unit of BDH activity was defined as the amount of enzyme required to reduce 1 μ mol of NAD(H) in one minute 11 . All enzyme activities were determined in triplicate. The protein concentrations of all samples were determined using the Bradford method, and bovine serum albumin served as the standard protein.
The enzymatic reactions were carried out similar to the assay method, except that the substrate and coenzyme concentrations were high. For the oxidation processes, a mixture containing 50 mM meso-2,3-BD/(2S,3S)-2,3-BD/(2R,3R)-2,3-BD, 4 mM NAD + , 50 mM potassium phosphate buffer and 20 μ g of purified enzyme in a final volume of 1 ml was incubated at 40 °C for 1 h. The reduction processes were carried out in 1 ml reaction system containing 50 mM DA or 100 mM (3S/3R)-AC, 4 mM NADH, 50 mM of potassium phosphate buffer and 20 μ g of purified enzyme at 40 °C for 1 h. The products in these reaction systems were extracted by ethyl acetate and then used to check the enzyme stereospecificity using a GC chromatograph system 32 . strain T241 and H30 as templates. The PCR products of the putative genes with RBS sequences were assembly into the downstream region of AC operon respectively by overlap extension PCR using the corresponding primers (Table 3). These overlapping PCR fragments were cloned into pET28a vector between the EcoRI and HindIII sites to generate the recombinant 2,3-BD pathway in E. coli 18 . Development of Serratia sp. mutant strains. Two DNA fragments (about 800 bp) from upstream sequence and downstream sequence of the putative gene with overlapping ends were amplified from Serratia sp. T241 using the primers listed in Table 3. The two fragments are then combined by overlapping PCR, generating an in-frame deletion construct of the putative gene. The overlapping PCR fragment was digested and cloned into the suicide vector pUTKm1 and transformed into E. coli S17-1 λ pir for conjugation with Serratia sp. T241. LB medium agar plate containing 50 μ g/ml kanamycin was used to screen the single crossover strains. The obtained single crossover strains were confirmed by PCR method. Then, one single crossover strain was grown in LB broth overnight, and plated onto LB medium agar plate. The double crossover strains were screened through kanamycin resistance phenotype 38 . The kanamycin-sensitive colonies were verified by PCR using the primers listed in Table 3.
The other putative genes mutants of Serratia sp. T241 were constructed using the same method. The obtained mutants of Serratia sp. T241 were stored in a glycerol suspension at − 80 °C for further experiments.
Analytical methods. The biomass concentration was determined by the optical density (OD) measured at 600 nm in a spectrophotometer (UV-1800, MAPADA) and correlated with dry cell weight (DCW) 11 .
The products in the broth were extracted by ethyl acetate with the addition of isopropanol as internal standard and then quantified using a gas chromatograph system (Agilent GC9860) with equipped with a chiral column (Supelco β -DEX ™ 120, 30-m length, 0.25-mm inner diameter). The operation conditions were as follows: N 2 was used as the carrier gas at flow rate of 1.2 ml/min; the injector temperature and the detector temperature were 215 and 245 °C, respectively; and the column temperature was maintained at 50 °C for 1.5 min, then raised to 180 °C at a rate of 15 °C/min. The concentration of the products was determined by calibration curves 11,32 .