A short‐chain carbonyl reductase mutant is an efficient catalyst in the production of (R)‐1,3‐butanediol

Abstract

R‐1,3‐butanediol (R‐1,3‐BDO) is an important chiral intermediate of penem and carbapenem synthesis. Among the different synthesis methods to obtain pure enantiomer R‐1,3‐BDO, oxidation–reduction cascades catalysed by enzymes are promising strategies for its production. Dehydrogenases have been used for the reduction step, but the enantio‐selectivity is not high enough for further organic synthesis efforts. Here, a short‐chain carbonyl reductase (LnRCR) was evaluated for the reduction step and developed via protein engineering. After docking result analysis with the substrate 4‐hydroxy‐2‐butanone (4H2B), residues were selected for virtual mutagenesis, their substrate‐binding energies were compared, and four sites were selected for saturation mutagenesis. High‐throughput screening helped identify a Ser154Lys mutant which increased the catalytic efficiency by 115% compared to the parent enzyme. Computer‐aided simulations indicated that after single residue replacement, movements in two flexible areas (VTDPAF and SVGFANK) facilitated the volumetric compression of the 4H2B‐binding pocket. The number of hydrogen bonds between the stabilized 4H2B‐binding pocket of the mutant enzyme and substrate was higher (from four to six) than the wild‐type enzyme, while the substrate‐binding energy was decreased (from −17.0 kJ/mol to −29.1 kJ/mol). Consequently, the catalytic efficiency increased by approximately 115% and enantio‐selectivity increased from 95% to 99%. Our findings indicate that compact and stable substrate‐binding pockets are critical for enzyme catalysis. Lastly, the utilization of a microbe expressing the Ser154Lys mutant enzyme was proven to be a robust process to conduct the oxidation–reduction cascade at larger scales.

The utilization of a microbe expressing the CpSADH and LnRCRSer154Lys mutant enzymes were proven to be a robust process to conduct the oxidation‐reduction cascade at larger scales.

INTRODUCTION

Enantiopure alcohols represent a major category of chiral, versatile compounds that are widely used for the synthesis of drugs, foods and pesticides (Patel, 2013). R‐1,3‐butanediol (R‐1,3‐BDO) is an important chiral alcohol with a short‐length carbon chain. It has been used for the synthesis of pheromones, fragrances and insecticides (Zu et al., 2020). Moreover, in the early nineties, Japanese researchers found that R‐1,3‐BDO could be a precursor for the synthesis of β‐lactam antibiotics such as penem and carbapenem (Iwata et al., 1990). Carbapenem showed a broad spectrum of activity and improved resistance to most β‐lactamases produced by bacteria. Therefore, it has previously been regarded as the last resort against antibiotic‐resistant bacteria (Papp‐Wallace et al., 2011). Due to its application in the synthesis of multiple compounds, the demand for R‐1,3‐BDO has increased consistently.

Early attempts to achieve the chemical synthesis of R‐1,3‐BDO involved the use of L‐threonine as the starting substrate. Though L‐threonine is cost‐effective and abundantly available, the method involving its use is relatively inefficient. The synthetic route involved four steps and involved the use of sulfuric acid, hydrogen, Pd catalyst and methanol. The final yield associated with this method was only approximately 70% (Larchevêque et al., 1991).

Because of the advancements in synthetic biology, researchers were able to finely adjust microbial metabolic pathways and obtain R‐1,3‐BDO through fermentation using a low‐cost substrate such as glucose from the culture medium. However, these methods either resulted in a moderate yield (around 13%) when using glucose as the starting material (Nemr et al., 2018), or an unsatisfactory product titre compared to that observed during enzymatic transformation. So far, the highest R‐1,3‐BDO titre obtained using the synthetic biological method has been approximately 1.5% (Kataoka et al., 2014).

In comparison to the chemical and synthetic biology‐based methods for the synthesis of R‐1,3‐BDO (Table S1), the enzymatic method displayed great advantages with regard to the simplicity of procedure, conversion ratio, substrate tolerance, and product yield. Three enzymatic methods can be used for R‐1,3‐BDO preparation. The first involves the oxidation of the S‐1,3‐BDO enantiomer in racemate by dehydrogenase, while ensuring that the R‐1,3‐BDO enantiomer is preserved. However, the maximum yield of R‐1,3‐BDO was theoretically limited to 50%, and the actual enantiomeric excess (e.e) of the R enantiomer was approximately 95% (Matsuyama et al., 1993). The low product e.e associated with the method limits its further application because this value should exceed 99% for the product to be an eligible chiral intermediate in organic synthesis. The second method involved the use of dehydrogenases for the enantio‐selective reduction of 4‐hydroxy‐2‐butanone (4H2B) to R‐1,3‐BDO. Zheng et al. reported that they used a C. krusei cell catalyst to transform 45 g/L of 4H2B into R‐1,3‐BDO with 99% e.e., and obtained a yield of 83.9% in approximately 50 hours (Zheng et al., 2012). Recombinant dehydrogenase was also used in a continuous bed reactor in which 4H2B was used as the substrate. The substrate concentration was 250 g/L, the product yield was 99%, and the e.e. was 99% when the transformation time was 500 h (Itoh et al., 2007). Upon analysing the product yield and e.e. value, the method involving the use of dehydrogenase seemed to be feasible for industrial use. However, compared to racemic 1,3‐BDO (around 6$/kg), 4H2B is not cost‐effective (around 60$/kg). Meanwhile, the reduction process is time‐consuming, which indicates that the current use of dehydrogenases was not efficient. Recently, a new enzymatic method that involved oxidase‐dehydrogenase catalysis (oxidation–reduction cascade) was developed for R‐1,3‐BDO synthesis (Zu et al., 2020). This method reduced the substrate cost significantly and resulted in the generation of R‐1,3‐BDO with a high e.e. (99.5%). The substrate in the method is racemic 1,3‐BDO. However, the enzyme catalysts used in the process were not robust because the catalysis duration time was around 100 h.

Based on the above analysis, the most promising enzymatic method for R‐1,3‐BDO production is the oxidation–reduction cascade. The major challenge associated with these methods is the inferior catalytic efficiency of dehydrogenases. In our previous studies, we developed an oxidation–reduction enzyme cascade (Scheme 1). A robust dehydrogenase CpSADH that oxidized the S‐1,3‐BDO enantiomer in racemic BDO (k _cat/K _m = 1.8 × 10⁶M⁻¹·min⁻¹) was cloned and expressed (Yamamoto et al., 1995). In 8 h, E. coli expressing CpSADH transformed 80 g/L of racemic BDO into 4H2B and R‐1,3‐BDO with 96% e.e. and 45% conversion. However, the catalytic turnover of a dehydrogenase LnRCR (the wild‐type dehydrogenase in this study) used for the reduction step was much lower. E. coli expressing LnRCR converted 60 g/L of 4H2B into R‐1,3‐BDO and helped achieve 98% e.e. and 90% conversion in 30 h.

Scheme 1.

The oxidation–reduction enzyme cascade for preparation of R‐1,3‐BDO Rac‐1, 3‐BDO, racemic 1,3‐butanediol; R‐1,3‐BDO, R‐1,3‐butanediol 4H2B, 4‐hydroxy‐2‐butanone; CpSADH, E. coli expressing CpSADH, LnRCR, E. coli expressing LnRCR

As annotated in GenBank, LnRCR belongs to the short‐chain carbonyl reductase (SDR) family. The members of the family show low sequence identity. However, they share similar spatial protein structures and have a highly conserved nicotinamide cofactor‐binding site for NADH or NADPH (Graff et al., 2019). Our previous studies have shown that LnRCR is a dehydrogenase that shows a preference towards NADH. Native or engineered SDRs are employed as efficient enzymes for the reduction of a spectrum of ketones to generate enantiopure alcohols (Li et al., 2022). Therefore, we believe that LnRCR could act as a robust catalyst after its modification using protein engineering techniques. The engineered catalyst would facilitate the refinement of the enzymatic cascade (Scheme 1) and enhance the efficiency of the enzymatic method. This could further reduce the price of enantio‐pure R‐1,3‐BDO (now around 350$/kg).

RESULTS AND DISCUSSION

Molecular docking

According to the catalytic mechanism of SDR (Filling et al., 2002; Li et al., 2022), we proposed the catalytic process by which LnRCR facilitated the reduction of 4H2B. In summary, the hydroxyl group of Tyr donates a proton to the carbonyl oxygen of 4H2B, polarizing the carbonyl carbon (C2) of 4H2B to facilitate nucleophilic hydride transfer. This results in the formation of a proton relay system, consisting of a hydrogen on the C4 of NADH, hydroxyl hydrogen of Tyr157, 2OH and 3OH of ribose, Lys155 side chains, and water bound to the backbone carbonyl of Asn111. The system shuttles protons from the bulk solvent to the Asn111 residue. Based on these assumptions, CDOCKER docking was conducted and 30 enzyme‐substrate configurations were generated. In these configurations, the distances between the donor and acceptor of the proton relay system ranged from 1.8 to 3.5 Å. Thus, these configurations were putative rational poses.

The configuration with the lowest docking energy was chosen for further analysis. In this configuration, the distances between the hydrogen on C4 (NADH) and O6 (4H2B), hydroxyl hydrogen of Tyr157 and O6 (4H2B), hydroxyl hydrogen of Tyr157 and oxygen of 2OH (ribose), hydrogen of Lys155 side chains and oxygen of 3OH (ribose) are 3.2, 3.3, 1.8 and 1.9 Å, respectively. As shown in Figure 1, the activity of LnRCR was in accordance with Prelog's rule (Prelog, 1964) for the reduction of 4H2B, which occurred with the addition of the R‐hydride of NADPH to the re‐faced ketone. Therefore, only R‐1,3‐BDO was produced during the reduction process. The docking results were in agreement with the catalysis process, in which the e.e value of R‐1,3‐BDO was approximately 95%.

FIGURE 1.

Docking 4H2B in LnRCR‐binding pocket hydrogen bonds are indicated by blue dashes.

Virtual mutagenesis and NNK saturation mutagenesis

Though LnRCR showed acceptable levels of enantioselectivity for R‐1,3‐BDO production, the catalytic activity was relatively low (53.6 U/g dry cell weight) compared to other reductases. For example, a robust SDR from Lentilactobacillus parabuchneri showed a specific activity of 15750.0 U/g dry cell weight in reduction of tert‐butyl (S)‐6‐chloro‐5‐hydroxy‐3‐oxohexanoate (Liu et al., 2017). The moderate activity hampered the further applicability for LnRCR. A low catalytic efficiency was usually correlated with a low substrate‐binding ability, which was suggestive of high levels of binding energy (Londhe et al., 2019). To obtain a mutant with a better substrate‐binding ability, we first performed virtual saturation mutagenesis. The residues that were within 8 Å of the substrate (Figure 2A) were chosen for virtual saturation mutagenesis and the changes in free energy (ΔG) levels were compared. Altogether, 500 mutants were obtained. A mutation in the Gly95 site (from Gly to Pro) resulted in the maximum level of reduction in binding energy (−0.95 kJ/mol). Mutations at the Gly95, Ser144, Ser154 and Val195 sites resulted in ΔG values ranging from −0.79 to 0.52 kJ/mol, which implied that these mutations were beneficial. Other mutants were associated with ΔG values ranging from −0.50 to 2.50 kJ/mol, which indicated that they resulted in neutral or deleterious effects (Spassov & Yan, 2013; Table 1).

FIGURE 2.

The residues representing the LnRCR‐binding pocket (A) and protein purification of the mutants (B) (A), Residues constituting the LnRCR‐binding pocket are represented by lines; NADH and 4H2B are displayed as sticks.

TABLE 1.

Beneficial mutants identified using virtual mutagenesis.

Mutation	ΔG	Mutation	ΔG	Mutation	ΔG	Mutation	ΔG
Gly95Pro	−0.95	Gly95Leu	−0.80	Ser154Lys	−0.79	Ser144Cys	−0.74
Gly95Val	−0.67	Val195Trp	−0.63	Val195Arg	−0.62	Gly95Asp	−0.61
Gly95Lys	−0.60	Gly95Cys	−0.56	Gly95Gln	−0.55	Gly95Glu	−0.53
Gly95Cys	−0.52

Note: ΔG unit: kJ/mol.

Four sites (Gly95, Ser144, Ser154 and Val195) with ΔG values less than −0.50 kJ/mol were selected for NNK saturation mutagenesis (Kretz et al., 2004). The high‐throughput screening of the resultant single mutant libraries facilitated the identification of three mutants (Ser154Lys, Ser154Glu and Ser154Asp) with higher levels of activity than that of the wild‐type LnRCR. The proteins derived from these three mutants were purified (Figure 2B) and their catalytic constants were determined using 4H2B as the substrate (Table 2). Among these, Ser154Lys showed the highest turnover and catalytic efficiency. The Ser154Lys mutant was further optimized by combinatorial saturation mutations at the 95th, 144th and 195th positions. However, no mutant with further improved activity was detected.

TABLE 2.

Kinetic parameters for the reduction of 4H2B by purified enzymes.

Enzyme	K m	k cat	k cat / K m
	[mM]	[min−1]	[M−1·min−1]
LnRCR	28.0 ± 3	1.28 × 104	4.6 × 105
LnRCRSer154Lys	20.1 ± 2	1.98 × 104	9.9 × 105
LnRCRSer154Glu	30.1 ± 4	1.30 × 104	3.3 × 105
LnRCRSer154Asp	35.6 ± 2	1.33 × 104	3.7 × 105

Mechanism underlying the enhanced catalytic efficiency of LnRCRS154K

An analysis of molecular dynamics (MD) simulation results revealed that the RMSF values for several sites (52EAAG55, 66VTDPAF71, 98AATVGDY104, 124MQ125, 128LK129, 148SVGFANK154 and 168TQN170) in the LnRCRS154K protein were moderately changed, as compared to those in LnRCR (Figure 3, A). All sites except for 66VTDPAF71, 98AATVGDY104 and 148SVGFANK154 were distant from the catalytic site. Therefore, we did not analyse further. With regard to the three sites surrounding the catalytic site, the VTDPAF and SVGFANK sites became more flexible, while the AATVGDY site became more inflexible (Figure 3A and Figure 3B). The three sites all occurred in the NADH‐binding pocket. The 4H2B‐binding pocket occurred at a site deep inside the NADH‐binding pocket. As depicted in Figure 4, when the MD simulation was stabilized (after 100 ns), the 4H2B BPV of LnRCR fluctuated between 200 and 600 ų. The 4H2B BPV of LnRCRS154K was reduced from approximately 200 to 150 ų and finally became stabilized. The volume of 4H2B was estimated to be 90 ų using the surface mode of Discovery Studio software. For 4H2B, the binding pocket of LnRCR was capacious. As proposed previously, compact binding pockets provided a conducive environment for small substrate catalysis (Guo et al., 2022). From this point of view, the binding pocket of LnRCRS154K seemed tight and energetically stable.

FIGURE 3.

Variations in the RMSF values of residues in MD (A) and the three sites surrounding the catalytic site (B). (B), blue, stable area; red, flexible area; yellow, residue Lys154.

FIGURE 4.

Variations in the binding pocket volumes (A) and HBond numbers (B) in MD A, Variations in the binding pocket volumes; LnRCR is indicated by purple lines, and LnRCRS154K is indicated by blue lines B, HBond numbers in MD; the maximum number of hydrogen bonds formed in a specific time frame (ns) have been shaded. LnRCR is shown in dark grey shade, and LnRCRS154K is shown in grey shade.

Because the hydrogen bond (HBond) was considered to be the most important factor in protein–ligand interactions (Wade & Goodford, 1989), the number of HBonds formed between 4H2B and the enzymes were analysed. As depicted in Figure 4B, during MD, the number of HBonds between 4H2B and the two proteins was significantly different. For example, the maximum number of HBonds in the 4H2B‐LnRCRS154K complex was 6 and appeared four times, while that in the 4H2B‐LnRCR complex was 5 and was observed only once. More HBonds were formed in the 4H2B‐LnRCRS154K complex than in the 4H2B‐LnRCR complex. This phenomenon was more pronounced in the later stage of MD.

The total binding energy, which is an important measure of binding affinity, was calculated using ‘gmx_mmpbsa’. A decrease in binding energy indicates that a more stable protein–ligand complex was formed, and is usually associated with increased enzyme activity. The total binding energy of the protein–4H2B complex decreased from −17.0 kJ/mol to −29.1 kJ/mol when there was a mutation of Ser to Lys in residue 154 (Table 3). This decrease is mainly attributable to the polar solvation energy (PB).

TABLE 3.

Calculation of binding energy using MMPBSA (energy unit: kJ/mol).

Complex	dG	MM	TdS	PB	SA
		COU + VDW
LnRCRWT	−17.005	−506.1	59.2	464.6	−34.7
		−249.6 + −256.4
LnRCRS154K	−29.111	−502.9	54.0	455.9	−36.2
		−217.9 + −284.9

Abbreviations: COU, electrostatic energy; dG, protein‐–ligand ‐binding energy; MM, molecular mechanics vacuum energy; PB, polar solvation energy; SA, non‐polar solvation energy; VDW, Van der Waal''s energy.

The 154th residue of LnRCR is vital for 4H2B catalysis because the replacement of a single residue at the LnRCRS154K site resulted in an approximately 115% increase in catalytic efficiency (Table 2). The e.e. value of the product was also increased from 94.5% to 99.0%. As shown in Figure 3B, the catalytic site of LnRCRS154K was composed of two binding pockets that showed a significant difference in volume. The bigger pocket facilitated NADH binding, while the other pocket facilitated 4H2B binding. The 154th residue occurred at the corner of the NADH‐binding pocket and close to the 4H2B‐binding pocket. The mutated lysine was in an area (SVGFANK) that eventually became more flexible. The 4H2B‐binding pocket was located in the middle of two flexible areas (VTDPAF and SVGFANK). After residue replacement, the two flexible areas seemed to ‘clamp’ the movements of the 4H2B‐binding pocket, which facilitated the formation of a smaller and more stable 4H2B‐binding pocket. The shrunken binding pocket of the mutant resulted in higher levels of interactions between the protein and ligand such as those resulting from the formation of hydrogen bonds, which stabilized the substrate and intermediates during catalysis. A direct effect was the obvious reduction in the binding energy of the LnRCRS154K‐4H2B complex. Taken together, in LnRCRS154K, the substrate was easier to bind and transform than in LnRCR; therefore, the productivity of R‐1,3‐BDO was observed to be increased. The value of the Michaelis–Menten constant was decreased for the LnRCRS154K protein, which also indicated its higher affinity to 4H2B, as compared to that of LnRCR to 4H2B (Table 2). Because LnRCRS154K also followed Prelog's rule (Prelog, 1964), the shrunken binding pocket resulted in increased steric hindrance for S‐1,3‐butanediol. Therefore, the trace activity of the enantiomer was reduced further. The changes in the activity of the two 1,3‐butanediol enantiomers enhanced the enantio‐selectivity of LnRCRS154K (Table 3).

Reduction and oxidation–reduction cascade transformation

Escherichia coli expressing LnRCRS154K were subjected to a 1‐L scale‐up transformation. Sixty grams of 4H2B substrate were used for transformation and reactions were completed in 16 h (conversion 90%, Figure 5A). After performing recovery procedures, 45.9 g of products were recovered (85.0% yield). The R‐1,3‐BDO product had a purity of 99.5% and an e.e. value of 99.9% (Figure 5D), [α]22 D = ‐29.9 (C = 1.00, ethanol). In comparison, the R‐1,3‐BDO product of LnRCR transformation had an e.e. value of 95.0% (Figure 5C); [α]22 D = ‐28.5 (C = 1.00, ethanol). The specific production rate (SPR) was 475.8 mg/h·g, while the SPR of C. krusei cells was estimated to be 37.8 mg/h·g (Zheng et al., 2012). Immobilized E. coli cells containing recombinant LsADH dehydrogenase were applied in the same manner in a packed bed reactor, and the SPR of the immobilized cells was estimated to be 50.0 mg/h·g (Itoh et al., 2007). LsADH protein shared 98.8% sequence identity with LnRCR and had variations in only three amino acid residues, but both of them had a serine residue at the 154th position. Thus, the LnRCRS154K mutant is considered to be an efficient catalyst for R‐1,3‐BDO production. Recently, an SDR LxCAR S154Y mutant was proven to be a robust catalyst for the production of (R)‐[3,5‐bis(trifluoromethyl) phenyl] ethanol ((R)‐BTPE) (Li et al., 2022). LnRCR and LxCAR exhibit a protein sequence identity of 89.2% and both have a serine residue at the 154th position. Furthermore, the two proteins showed similar domain arrangements and overall structures with an RMSD of 0.41 Å. Therefore, the 154th position is a key target site that needs to be taken into consideration while engineering proteins with a structure similar to that of LnRCR.

FIGURE 5.

Time course of E. coli expressing LnRCRS154K transformation (A) and chiral GC analysis (B), racemic 1,3‐BDO; (C), E. coli expressing LnRCR transformation; (D), E. coli expressing LnRCRS154K transformation 1, R‐1,3‐BDO; 2, S‐1,3‐BDO.

To further reduce the substrate cost, a refined oxidation–reduction cascade was employed to prepare R‐1,3‐BDO. In this cascade, the previous 4H2B substrate (around 60$/kg) was replaced with racemic 1,3‐BDO (around 6$/kg). The two‐step cascade included the first oxidation of racemic 1,3‐BDO by CpSADH. E. coli expressing CpSADH converted 80 g/L of racemic BDO into 4H2B and R‐1,3‐BDO with 96% e.e. and 45% conversion in 8 h. After discarding CpSADH cells, the supernatant was further transformed by E. coli expressing LnRCRS154K for 15 h. The recovered 66.4 g R‐1,3‐BDO product had a purity of 99.5% and an e.e. value of 99.8%. The SPR of the cascade was 366.0 mg/h·g and the total yield of the cascade was 83%. The results showed that the substrate cost of the oxidation–reduction cascade was reduced significantly.

In this study, we proved that compact and stable substrate‐binding pockets are vital for enzyme catalysis. Our findings could guide other researchers focused on the use of protein engineering techniques. However, only a beneficial single mutant was identified in this study. This could be attributable to the fact that either the mutant sample size was not large enough or high‐throughput screening was ineffective. We are conducting further studies on both counts to explore these limitations, in order to identify a more robust mutant.

EXPERIMENTAL PROCEDURES

Chemicals, bacterial strains, plasmids and culture media

4‐hydroxy‐2‐butanone (4H2B), Racemic 1,3‐BDO and R‐1,3‐BDO were purchased from Acros (Beijing, China). Other chemicals used were analytical reagents and commercially available. E. coli str. K‐12 substr. MG1655 (ATCC47076) cultures were grown in either Luria Broth (LB) medium at 37°C. E. coli TOP10 (Tsingke Biotech) cells were used for cloning and expression studies. The plasmid pET30a (Invitrogen) was used for standard cloning and expression in E. coli. To isolate bacterial strains carrying the appropriate recombinant plasmids, 30 μg/mL kanamycin was added to the medium.

Extraction and purification of DNA

Plasmids were isolated using the TsMini Kit (Tsingke, Beijing, China), and DNA was purified using the Tsgel Midi Kit (Tsingke, Beijing, China). Each of these procedures was performed according to the manufacturer's protocols.

Gene cloning and expression in Escherichia coli

The LnRCR gene (GenBank access number: BAP47553.1) was synthesized by Genscript Co. (Nanjing, China). The gene was cloned into the NdeI and XhoI sites of pET30a plasmid, and the resulting plasmid was named as pETLnRCR. The construct was transformed into E. coli BL21(DE3) cells and grown overnight in LB broth at 37°C. Five milliliter of the overnight culture was transferred into 500 mL of LB broth and allowed to grow to an optical density of 0.8 (OD 580). Isopropyl‐β‐D‐thiogalactoside (IPTG) was then added at a final concentration of 1.0 mM, and after induced for 12 h, the cells were harvested by centrifugation (4000 × g for 10 min).

The CpSADH gene (GenBank access number: JABWAB010000004.1) was synthesized by Genscript Co. (Nanjing, China). The gene was cloned into the NdeI and XhoI sites of pET30a plasmid, and the resulting plasmid was named as pETCpSADH. The expression procedure for pETCpSADH was same as pETLnRCR.

Purification of His‐tagged proteins on nickel (Ni)‐chelating columns

Purification was performed using the buffer supplied by Novagen (New Jersey, USA), in accordance with the manufacturer's protocol. Cells grown in 500 mL expression culture were suspended in 40 mL of binding buffer (50 Mm NaH2PO4, 300 mM NaCl and 10 mM imidazole, at pH 8.0) and disrupted via ultrasonication in an ice bath for 20 min at 200 W. The supernatant was applied to a 1 mL Novagen His∙Band gravity flow column equilibrated with 20 mL Ni‐NTA‐binding buffer. The column was then washed with 20 mL wash buffer (50 mM NaH₂PO₄, 300 mM NaCl and 20 mM imidazole, at pH 8.0). The His‐tagged protein was eluted with 10 mL elution buffer (50 mM NaH₂PO₄, 300 mM NaCl and 200 mM imidazole, at pH 8.0). To remove the imidazole from the elution buffer, the eluate was applied to a 5 mL HiTrap desalting column (GE Healthcare, USA), and elution was performed using 50 mM of Tris–HCl buffer (150 mM NaCl, at pH 7.5). Protein samples were collected and dialysed against 5000 mL of 20 mM phosphate buffer (pH 7.2). The buffer was changed three times in 24 h.

Dehydrogenase activity assay and measurements of catalytic constants

The activity was assayed by adding 30 μg of purified enzymes to 200 μL of the 50 mM 4H2B solution (pH 8.0, 2 mM NADH, 50 mM Tris–HCl with 10% isopropanol). The reaction mixture was incubated at 30°C for 6 hours. Following transformation, the reaction mixture was extracted with 400 μL of ethyl acetate, and conversions were analysed via an Agilent 7890 GC system (Agilent) equipped with an HP‐5 column (30 m × 0.25 μm; Agilent, CA, USA). GC analysis was performed with an FID (flame ionization detector) under the following conditions: injection and detection temperature of 250°C, column temperature of 50°C for 1 min, raised to 100°C at 5°C /min, split ratio of 40, and a flow rate of 1 mL/min of nitrogen. One unit (U) of enzyme activity is defined as the amount of enzyme that catalyses the conversion of one nanomole of substrate per minute.

Chiral GC and specific optical rotation assays for enantiomeric excess (e.e) of 1,3‐BDO

Chiral GC

Twenty‐five microlitres of the reaction solution were mixed with 50 μL N,N‐dimethylformamide. Subsequently, an excess (375 μL) of N,N‐diethyltrimethylsilylamine (DETMSA) was added at room temperature. The mixture was then quickly heated to 80°C until it became homogeneous. Only the 1,3‐BDO in the solution was derivatized, while 4H2B was not. The derivatized solution was extracted with 400 μL ethyl acetate and then used for chiral GC analysis. The analysis was conducted using an Agilent 7890 GC system (Agilent, CA, USA) equipped with a CP‐ChiraSil‐DEX CB column (30 m × 0.25 μm; Agilent). The conditions used for GC were as follows: injection and detection temperature of 220°C, column temperature of 63°C for 30 min, a split ratio of 20 and a flow rate of 1 mL/min of nitrogen. The e.e is defined as 100 × (A − B)/(A + B), where A and B are the amounts of two enantiomers of 1,3‐BDO.

Specific optical rotation

Specific optical rotation ([α]22 D) measurements were performed using a JASCO P‐1010 polarimeter in a 100 × 2 mm cell at room temperature (22°C). The specific optical rotation of standard R‐1,3‐BDO was determined to be −30.0 (C = 1.00, ethanol).

Determination of protein concentration and SDS‐PAGE

The protein concentration was determined using the BCA Protein Assay Kit (Thermo Fisher Scientific) using bovine serum albumin (BSA) as a standard.

SDS‐PAGE was performed as described previously (Laemmli, 1970) with a 6% polyacrylamide stacking gel and a 12% polyacrylamide separating gel. Protein purity was estimated using Glyko BandScan software (Glyko, Novato, USA).

Homology modelling and energy minimization

Homology modelling of LnRCR was performed with SWISS‐MODEL (Waterhouse et al., 2018), using 6XNB (a short‐chain alcohol dehydrogenase from Leifsonia xyli) as a template. The nicotinamide adenine dinucleotide (NAD) molecule was inserted into the LnRCR NAD‐binding site, and then the energy minimization of the protein‐NAD complex was performed in Rosetta energy terms with a Rosetta relax script (Leaver‐Fay et al., 2011).

Molecular docking and in silico mutagenesis

Enantiomers of 1,3‐BDO were docked into the catalytic site of LnRCR using Discovery Studio 4.5 (Accelrys, San Diego, CA). Briefly, CDOCKER docking and interaction energies were measured, and proteins and ligands were parameterized using CHARMm force fields. Top ligand positions were clustered within a root‐mean‐square deviation of 2.0 Å and scored against the CDOCKER Interaction Energy. The resulting global structure with the lowest energy was selected for further analysis.

To investigate the role of binding pocket residues in complex stabilization, we performed computational site‐directed mutagenesis according to the ‘Calculate Mutation Energy (Binding)’ protocol available in Biovia Discovery Studio 4.5. The energy minimization of final docked complexes was performed with CHARMm using the Smart Minimizer algorithm in Discovery Studio 4.0 (Accelrys, San Diego, CA, USA). In silico mutagenesis was performed by calculating the free binding energy of the docked complex. Thus, the residues of the binding pocket were mutated into 19 different amino acids, to estimate the impact of each mutation on the binding within the complex. The mutation binding energy was finally calculated as ΔΔG_mut = ΔΔG_bind(mutant) ‐ ΔΔG_bind (wild type), where ΔΔG_mut is the mutation energy and ΔΔG_bind is the difference in the free energy between the complex and unbound states.

Construction of NNK saturation mutant libraries and high‐throughput screening of mutants

LnRCR NNK saturation mutant libraries for four residues (Gly95, Ser144, Ser154 and Val195) were generated through NNK saturation mutagenesis method (Kretz et al., 2004) using primers listed in Table S2. The PCR reaction mix was performed in a 25 μL final volume containing: 10 × KOD buffer (2.5 μL), MgCl2 (1 μL, 25 mM), dNTP (5 μL, 2 mM each), primers (0.5 μL of each pair of primers, 2.5 μM), template plasmid (pETLnRCR, 1 μL, 10 ng/μL) and 1 unit of KOD DNA polymerase. The PCR protocol comprised two rounds of amplification cycles (five cycles in the first round and 25 cycles in the second round) with an initial denaturation at 94°C for 10 min. The first round of the PCR programme consisted of five cycles of 40 s of denaturation, 40 s of annealing at 53°C and 50 s of elongation at 68°C. The second round consisted of 25 cycles of 1 min of denaturation, 1 min of annealing at 60°C and 6 min of elongation at 68°C. The final elongation step was for 30 min at 72°C. Subsequently, the template plasmid was removed from the 10 μL reaction mix by digesting it with 1 unit of DpnI (New England Biolabs, Beijing, PRC). Finally, the mixture was transformed into E. coli BL21(DE3) and mutations in the clones were confirmed through DNA sequencing.

The library screening of mutant was performed using a high‐throughput screening method that involved the chromogenic reaction of 4H2B and 2,4‐dinitrophenylhydrazine (Li et al., 2010). Specifically, 30 OD of library E. coli cells was added to 150 μL of 50 mM 4H2B solution (pH 8.0, 50 mM Tris–HCl with 10% isopropanol) in 96‐well plates. The reaction mixture was incubated at 30°C for 6 h. The 2,4‐dinitrophenylhydrazine solution was prepared by adding 1 g of 2,4‐dinitrophenylhydrazine to 7.5 mL of concentrated sulfuric acid, and this solution was slowly added to 75 mL of 95% ethanol. Finally, the solution was diluted to a volume of 250 mL using distilled water. Twenty‐five microlitres of library transformation suspensions were mixed in 96‐well plates with 100 μL of 2,4‐dinitrophenylhydrazine solution. Next, 5 mL of 0.8 M sodium hydroxide was added and mixed in the wells. Subsequently, the mixture was centrifuged (3500 × g, 30 min), and the supernatant was transferred into a new microplate. Finally, the absorbance was measured using a BioTeK Synergy 2 Multi‐Mode Microplate Reader (BioTeK) at λ = 520 nm. Reaction mixtures without resting cells served as the controls.

Molecular dynamics simulations

Molecular dynamics (MD) simulations targeting protein–ligand complexes were generated using GROMACS software (Van der Spoel et al., 2005). The protein–ligand complex was placed in a 12‐surface polyhedron simulation box containing a water solvent (TIP3P model). To attain overall system neutrality, ions (Na⁺ and Cl⁻) were added to the solvent. Next, energy was minimized to eliminate unfavourable system interactions. The system was then balanced under isothermal‐isobaric conditions. The temperature and pressure were maintained at 300 K and 1 bar using the Berendsen thermostat and Parrinello–Rahman barostat models, respectively, each of which was simulated for 100 ns. The results were recorded every 0.01 ns. Next, the root means square fluctuation (RMSF) values were calculated using the GROMACS software and mapped using Sigmaplot (Inpixon HQ, Palo Alto, CA, USA). Finally, the binding pocket volumes were calculated using POVME version 3.0 software (Wagner et al., 2017).

Molecular mechanics Poisson–Boltzmann surface area determination

The molecular mechanics Poisson–Boltzmann surface area (MMPBSA), a strategy used to appraise interaction‐free energies, was used to analyse biomolecular interactions. First, a protein–ligand snapshot of MD trajectories was extracted from the stable region of each complex to incorporate the MD simulations and their binding energy calculations. Finally, the binding energy was calculated using the MMPBSA protocol in the g_mmpbsa package (Kumari, et al., 2014).

Escherichia coli expressing LnRCRS154K transformation and R‐1,3‐BDO recovery

The scaled‐up transformation was conducted in a 2 L reaction tank. Sixty grams of BDO were dissolved in 1 L of 50 mM Tris–HCl (pH 8.0) containing 10% isopropanol. Then, 6 g (dry weight) of E. coli expressing LnRCRS154K were applied. The parameters of the reaction tank were set as follows: temperature, 30°C; RPM, 200, air sparging rate, 200 mL/min. The pH was adjusted to 8.0 with an ammonia solution (0.5 M). The specific production rate (SPR) is defined as the amount of product catalysed using 1 g (dry weight) of cells per hour.

The transformation process was monitored via GC. After the reaction is completed, the mixture was first concentrated via rotator evaporation at 40°C, to yield 120 g of viscous liquid. Then, 500 mL ethanol was added and R‐1,3‐BDO was extracted via distillation (95°C/10 mm Hg).

Transformation in the oxidation–reduction cascade

The scaled‐up transformation process was conducted in a 2 L reaction tank. Eighty grams of BDO were dissolved in 1 L of 50 mM Tris–HCl (pH 8.0). Then, 6 g (dry weight) of E. coli expressing CpSADH were applied. The parameters of the reaction tank were as follows: temperature, 30°C; RPM, 200 and air sparging rate, 300 mL/min. The pH was adjusted to 8.0 with an ammonia solution (0.5 M). The transformation was monitored by GC analysis. When the conversion of 45% of the reaction mixture was completed, the transformation suspension was centrifuged and supernatant was collected. The supernatant was added to 10% isopropanol. Then, 6 g (dry weight) of E. coli expressing LnRCRS154K were applied. The parameters of the reaction tank were set to ensure that the E. coli expressing LnRCRS154K transformation occurred effectively. The transformation process was monitored by GC. After the reaction was completed, the R‐1,3‐BDO recovery process was the same as above.

Software and online service

The program BLASTX (Altschul et al., 1990) was used for protein homology searching, and Vector NTI 8.0 (Informax) was used for sequence alignments. Figures of computer‐aided modelling were prepared by PyMOL software (DeLano, 2002).

AUTHOR CONTRIBUTIONS

Xiaoyan Guo: Formal analysis (lead); methodology (lead). Yunfang Gao: Formal analysis (equal); methodology (equal). Fangzheng Liu: Data curation (equal). Yong Tao: Investigation (lead). Haibo Jin: Funding acquisition (equal); investigation (equal). Jianjun Wang: Project administration (lead); writing – original draft (lead); writing – review and editing (lead). Wu Sheng: Funding acquisition (lead).

CONFLICT OF INTERESTS STATEMENT

The authors declare that they have no competing interests.

Supporting information

Appendix S1.

Guo, X. , Gao, Y. , Liu, F. , Tao, Y. , Jin, H. , Wang, J. et al. (2023) A short‐chain carbonyl reductase mutant is an efficient catalyst in the production of (R)‐1,3‐butanediol. Microbial Biotechnology, 16, 1333–1343. Available from: 10.1111/1751-7915.14249