Engineering a Novel Cofactor‐Versatile Short‐Chain Dehydrogenase for Efficient Synthesis of Chiral Alcohols

ABSTRACT

Short‐chain dehydrogenases/reductases (SDRs) are widely used for the asymmetric synthesis of chiral alcohols, which are critical building blocks in pharmaceutical manufacturing. In this study, a novel SDR from Bacillus subtilis (BsSDR) was identified and engineered for the asymmetric synthesis of (S)−1‐(2,6‐dichloro‐3‐fluorophenyl)ethanol ((S)‐CFL). By integrating in silico molecular docking with FuncLib based design, a focused mutant library of BsSDR targeting key hotpots was constructed. Four of the constructed 31 variants exhibited markedly enhanced activity. Subsequent combinatorial mutagenesis generated a “best” variant BsSDR_M2 (I178L/T179L), which displayed a 107.4‐fold increase in catalytic activity, along with excellent enantioselectivity (> 99.9% e.e.) and broad substrate scope. Notably, unlike most SDRs that are strictly dependent on NAD(H) or NADP(H), the BsSDR also possesses a remarkable capacity to utilize the noncanonical cofactor NMNH. Moreover, in addition to robust enantioselective reductive activity toward CFA, the BsSDR also exhibits glucose‐oxidizing activity, driving cofactors regeneration. In the presence of NMNH, BsSDR_M2 displayed a 33.3‐fold improvement in catalytic efficiency (k _cat/K _M) toward substrate 1‐(2,6‐dichloro‐3‐fluorophenyl)ethanone (CFA) compared to the wild‐type (WT), along with an extended half‐life (t _1/2) of 52.1 h at 45°C. Using Escherichia coli expressing BsSDR_M2 as the only biocatalyst and glucose as the cosubstrate, 100 g/L CFA was reduced within 6 h in 92.1% conversion and > 99.9% e.e., achieving a space‐time yield (STY) of 368.2 g/(L·d). This work highlights the potential of BsSDR as a cost‐effective and sustainable biocatalyst for the efficient synthesis of (S)‐CFL and other high‐value chiral alcohols.

1. Introduction

Chiral alcohols represent an important class of optically active compounds in which hydroxyl groups are directly bound to stereogenic carbon centers (Breuer et al. 2004; Kuriyama et al. 2010). Owing to structural versatility, chiral alcohols with diverse substituents and functional groups serve as indispensable intermediates in asymmetric synthesis (Khan et al. 2025; Li et al. 2024; Matsuyama et al. 2002). Notably, more than 20% of pharmaceutical intermediates contain chiral alcohol motifs (Khan et al. 2025; Li et al. 2024; Patel 2011; Patel 2001). The synthetic pathways for a range of representative active pharmaceutical ingredients (APIs) (de Gonzalo et al. 2022; Hoyos et al. 2019; Jiang and Fang 2020), such as atorvastatin (Gong et al. 2018; Ma et al. 2010), rosuvastatin (Dai et al. 2023; Liu et al. 2018), ezetimibe (Zhang et al. 2021), montelukast (Shafiee et al. 1998), ticagrelor (Liu et al. 2022), and crizotinib (Zong et al. 2019), necessarily feature at least one asymmetric reduction.

To date, many catalysts have been explored for the asymmetric reduction of ketones, including chiral oxazaborolidine catalysts (Corey and Helal 1998; Shende et al. 2018), chiral metal catalysts (Li et al. 2015; Mallat et al. 2007), and biocatalysts (Naim et al. 2024; Nie et al. 2018; Xie et al. 2009; Zhang et al. 2024; Zhou et al. 2020). The most common chemical methods for synthesizing chiral alcohols, which include the addition reaction to carbonyl compounds and the hydrogenation reduction of ketones, rely on chiral reagents (Shaikh et al. 2008; Yang et al. 2023). However, these reagents are not only complex and costly to prepare, but also harmful to the environment.

So far, biocatalytic asymmetric reduction catalyzed by short‐chain dehydrogenases/reductases (SDRs) has emerged as a pivotal strategy for the synthesis of chiral alcohol, owing to its exceptional enantioselectivity, mild reaction conditions, environmental sustainability, and broad substrate versatility (Chadha et al. 2024; Liu et al. 2019; Yao et al. 2025). For instance, Zong et al. (2019) developed a quadruple mutant (ADH‐LKM, F147L/Y190P/V196L/A202W) that could completely convert 150 g/L 1‐(2,6‐dichloro‐3‐fluorophenyl) ethanone into the crizotinib precursor (S)−1‐(2,6‐dichloro‐3‐fluorophenyl)ethanol with 99.9% e.e. within 12 h at 40°C (Zong et al. 2019). We engineered a carbonyl reductase (CR) from Levilactobacillus suantsaii for the generation of a robust mutant LsCR_M4, which efficiently reduced 600 g/L 2‐chloro‐1‐(3,4‐difluorophenyl)ethanone (CFPO) to (1S)−2‐chloro‐1‐(3,4‐difluorophenyl)ethanol ((S)‐CFPL), a key chiral building block for ticagrelor, in > 99% yield and 99.5% e.e. (Liu et al. 2023). Despite tremendous progress in the biocatalysis of chiral alcohols, discovering novel catalytic properties and developing industrial biocatalysts for reductases remain fundamental to industrial biocatalysis.

In this study, an SDR from Bacillus subtilis (BsSDR) was identified from our group's enzyme library, exhibiting significant reductive activity toward 1‐(2,6‐dichloro‐3‐fluorophenyl)ethenone (CFA). Notably, BsSDR can utilize both NAD(P)H and the noncanonical cofactor NMNH, exhibiting remarkable cofactor versatility. Besides reductive activity toward CFA, BsSDR also exhibits oxidative activity toward glucose, making cofactors regeneration (Scheme 1). Furthermore, engineering BsSDR was implemented, yielding a beneficial variant BsSDR_M2. Molecular docking and molecular dynamics (MD) stimulations are conducted to elucidate the underlying mechanisms for catalytic activity and thermostability enhancement of BsSDR_M2.

Scheme 1.

BsSDR‐catalyzed CFA biotransformation into (S)‐CFL, a precursor of crizotinib. In this reaction system, glucose serves as the cosubstrate for cofactors regeneration, and BsSDR acts as the only biocatalyst for both reduction and oxidation reactions.

2. Materials and Methods

2.1. Strains, Plasmids, and General Chemicals

Strains of Bacillus subtilis, Thermus thermophilus HB27, and Clostridium beijerinckii CGMCC 1.335 were stored in our laboratory. Target genes encoding SDRs were amplified by polymerase chain reaction (PCR) from genomic DNA of the donor strains using gene‐specific primers. Amplified fragments were cloned into the pET‐28a(+) expression vector, and the sequence‐verified recombinant plasmids were transformed into competent E. coli BL21(DE3) cells (lab‐constructed). The gene encoding BsSDR (GenBank accession No. WP_003245662.1) used in this study was cloned from Bacillus subtilis (GenBank accession No. AL009126.3). All PCR primers, synthesized by Jiutian Gene Technology (Tianjin) Co. Ltd., are listed in Supporting Information S1: Table S1. Phanta Super‐Fidelity DNA polymerase and the restriction enzyme DpnI were purchased from Vazyme Biotech Co. Ltd. (Nanjing, China). The Plasmid Miniprep Kit and PCR Clean‐up Kit were sourced from Hangzhou Axygen Biotechnology Co. Ltd. Substrates CFA and the product (S)‐CFL were supplied by Shanghai Titan Scientific Co. Ltd. Other reagents and chemicals, all of analytical grade, were used without additional purification.

2.2. Site‐Directed Mutagenesis and Mutant Library Construction

The PCR was performed using the gene bssdr as the template, with primers for mutants listed in Supporting Information S1: Table S2. The reaction mixture (25 µL) consisted of 12.5 µL 2× Phanta Max Buffer, 1 µL dNTP Mix (10 mM), 2 µL each primer (10 µM), 1 µL template plasmid, and 1 µL Phanta Max Super‐Fidelity DNA Polymerase, with the remaining volume (7.5 µL) made up with deionized water. PCR was performed under the following conditions: initial denaturation at 95°C for 3 min; 30 cycles of denaturation (95°C, 15 s), annealing (60°C, 15 s), and extension (72°C, 4 min); followed by a final extension at 72°C for 10 min. Following PCR, the products were treated with 0.5 μL Dpn I at 37°C for 1 h to digest template DNA, and then transformed into competent E. coli BL21(DE3) cells for protein expression.

2.3. Cosubstrate Screening

Cosubstrate screening was conducted using crude recombinant WT‐BsSDR enzyme derived from cell lysate. Enzyme activity was determined in a 500 μL reaction system containing 5 mM CFA, 450 μL crude enzyme, 10% (w/v) sugars (xylose, lactose, maltose, glucose, sucrose), 10% (v/v) alcohols (methanol, ethanol, isopropanol, glycerol), and volume adjusted to 500 μL with 100 mM sodium phosphate buffer (pH 7.0). Reactions were carried out at 50°C, 800 rpm for 8 h, and the reaction mixtures were subsequently extracted with 1 mL of ethyl acetate. The organic phase was separated, dried over anhydrous sodium sulfate, and filtered through a 0.22 μm hydrophilic membrane. The product (S)‐CFL was then analyzed by gas chromatography (Agilent 8860) equipped with a CycloSil‐B chiral capillary column (30 m × 0.25 mm i.d., 0.25 μm film thickness). Both the injector and detector temperatures were set at 250°C, while the column temperature was maintained at 155°C for 15 min.

2.4. Screening for Beneficial Mutants

The WT‐BsSDR and its mutants were expressed in E. coli, and their catalytic activities were evaluated by using a whole‐cell biocatalysis system. Reactions were carried out in 500 μL of 100 mM sodium phosphate buffer (pH 6.0) containing 10 mM CFA, 10 g/L DCW of WT‐BsSDR or its variants, and 8% (w/v) glucose at 60°C and 800 rpm. The reaction time was set to 4 h for screening single‐point mutants, and 0.5 h for screening combined mutants. All reactions were determined by gas chromatography (details in Section 2.3).

2.5. Expression and Purification of BsSDRs

E. coli BL21(DE3)/pET‐28a(+)‐bssdr and E. coli BL21(DE3)/pET‐28a(+)‐esgdh were constructed in our group for producing WT‐BsSDR and Exiguobacterium sibiricum glucose dehydrogenase (EsGDH). Recombinant E. coli BL21(DE3)/pET‐28a (+)‐bssdr, E. coli BL21(DE3)/pET‐28a(+)‐bssdr‐mutants, and E. coli BL21(DE3)/pET‐28a(+)‐esgdh were grown in Luria‐Bertani (LB) supplemented with kanamycin (50 μg/mL) under standard conditions (37°C, 220 rpm). Protein expression was induced with 0.1 mM IPTG when the culture reached an OD₆₀₀ of 0.6–0.8, followed by incubation at 16°C with shaking at 180 rpm for 20 h. Cells were collected by centrifugation at 8000 rpm for 10 min at 4°C, resuspended in 100 mM sodium phosphate buffer (pH 7.0), and then disrupted via sonication. The soluble proteins were purified from the lysate by nickel‐affinity chromatography. Purified proteins were verified by SDS‐PAGE (Supporting Information S1: Figure S1) and quantified using the Bradford protein assay.

2.6. Cofactor Screening

Cofactor screening was conducted using purified WT‐BsSDR enzyme. Enzyme activity was determined in a 500 μL reaction mixture containing 5 mM CFA, 0.1 mg purified WT‐BsSDR enzyme, and 5 or 10 mM cofactors (NADH, NADPH, NMNH), with the final volume adjusted using 100 mM sodium phosphate buffer (pH 7.0). Reactions were carried out at 50°C, 800 rpm, for 4 h. All reactions were determined by gas chromatography (details in Section 2.3).

2.7. Determination of Kinetic Parameters for WT‐BsSDR and Mutant BsSDR_M2

To determine the apparent kinetic parameters for the cofactors (NAD⁺, NADP⁺, and NMN⁺), cofactor concentrations were varied while glucose was maintained at a constant 8% (w/v). The kinetic parameters of WT‐BsSDR and its variant BsSDR_M2 for the substrate CFA were determined by measuring initial reaction rates at varying CFA concentrations with constant NMNH (2 mM). The absorbance variation at 340 nm was monitored by a microplate reader (Thermo Scientific) at 50°C. All measurements were performed in triplicate, and kinetic parameters (v _max and K _M) were derived from nonlinear regression analysis of Michaelis‐Menten plots using Origin 2021 software.

2.8. Thermostability Characterization for WT‐BsSDR and Mutant BsSDR_M2

Thermostability is characterized by three critical parameters: half‐inactivation temperature (T ₅₀ ¹⁵ and T ₅₀ ³⁰), half‐life (t _1/2), and the melting temperature (T _m). To determine T ₅₀ ¹⁵ and T ₅₀ ³⁰, purified WT‐BsSDR and the mutant BsSDR_M2 were incubated over a temperature gradient ranging from 30°C to 80°C for 15 and 30 min, respectively, followed by rapid cooling in an ice‐water bath for 5 min. The residual activities were determined under the following conditions: 50 μg of purified enzyme, 2 mM substrate (CFA), 5 mM coenzyme (NMNH), 60°C, and pH 7.0. The original activity before heat treatment was defined as 100%.

T _m of WT‐BsSDR and BsSDR_M2 were analyzed using a circular dichroism (CD) spectrometer (Applied Photophysics Ltd.). Purified WT‐BsSDR and BsSDR_M2 were adjusted to a concentration of 0.1–0.2 mg/mL by dilution and then transferred into a 10 mm quartz cuvette. Samples were scanned continuously over a wavelength range of 160–280 nm, with the temperature ramped from 20°C to 90°C at a gradient of 5°C/min.

To determine the half‐life (t _1/2) of purified WT‐BsSDR and BsSDR_M2, each purified protein (50 μg) was incubated at 45°C for appropriate durations, after which residual activities were measured under the conditions of 50°C and pH 7.0. The t _1/2 values were calculated based on the first‐order deactivation functions:

$$ln(A/A_0)=‐k_{d}t,$$ (1)

$$t_{1/2}=\frac{ln2}{k_d}.$$ (2)

A ₀ represents the initial activity, U, and A the residual activity after incubation time t, U; and k _d is the deactivation rate constant, h⁻¹.

2.9. Computational Modeling, Molecular Docking, and MD Simulations

For the WT‐BsSDR, a three‐dimensional (3D) model of its complex with NMNH was acquired using AlphaFold3. The structure of the mutant protein was determined via the PyMOL (v 2.5.6) (http://www.pymol.org) program, which incorporates an amino acid mutation module. The substrate CFA was docked into WT‐BsSDR and its mutant using the YASARA (version 16.4.6). We selected optimal docking conformations that captured the precise spatial orientation of the substrate, then optimized the substrate and each BsSDR variant using the AMBER GAFF and AMBER 99SB force fields, respectively. The entire model was solvated in a rectangular TIP3P water box, and counterions were added to neutralize the system charge. AMBER 18 software was employed to prepare the initial coordinates and topology parameters for MD simulations (Chen et al. 2023; Qiu et al. 2024; Yao et al. 2025). To eliminate unfavorable interatomic contacts, the initial models were subjected to a two‐stage energy minimization protocol: constrained minimization of the solvent with the protein and substrate fixed, followed by unrestrained full‐system relaxation and gradual heating from 0 to 303 K over 300 ps. Following a 100 ps density equilibration in the NPT ensemble (303 K, 1.0 atm) to facilitate system relaxation, all simulations were carried out for 100 ns in the same ensemble with a 1 fs time step (Fan et al. 2022; Qiu et al. 2024; Yao et al. 2025). Simulation results and intermolecular interactions were analyzed and visualized with PyMOL (v 2.5.6) and Discovery Studio 2019 Client, respectively.

2.10. Binding Free Energy Calculation via MM/GBSA and Dynamic Cross‐Correlation Map (DCCM) Analysis

The binding free energy (ΔG_bind) of the substrate to each BsSDR was computed employing the Molecular Mechanics Generalized Born Surface Area (MM‐GBSA) approach. The DCCMs were computed from the equilibrated portions of the MD simulations for both WT‐BsSDR and BsSDR_M2 using the Bio3D package in R package (Grant et al. 2006). The final 10 ns trajectory was saved at 10 ps intervals and converted to Bio3D input format using the CatDCD trajectory processing tool, and correlation matrices were generated based on the Cα atoms of all amino acid residues (Yu and Dalby 2018). DCCM matrices were generated using Origin 2021.

2.11. Substrate Scope

Ketones, aldehydes, and keto‐esters were used to measure the substrate scope of WT‐BsSDR and BsSDR_M2. An assay was performed using a 500 μL reaction system consisting of 100 mM phosphate buffer (pH 7.0), 20 mM substrates (1a–16a), 10 g/L DCW CR cells, and 8% (w/v) glucose. After stirring at 60°C and 1000 rpm for 30 min, the reaction was terminated by adding 1 mL of ethyl acetate. The conversion and e.e. values were determined by chiral HPLC or GC analyses, with the corresponding detection conditions described in Supporting Information S1: Table S3.

2.12. Reaction Condition Optimization for (S)‐CFL Biosynthesis

The effects of temperature, pH, and substrate to glucose molar ratio on the catalytic activities of WT‐BsSDR and BsSDR_M2 were investigated. For temperature optimization, reactions were carried out in a 500 μL system containing 100 mM sodium phosphate buffer (pH 6.0), 13.8 g/L CFA, 10 g/L DCW BsSDR_M2, and 8% (w/v) glucose. The mixture was incubated at 30°C–80°C with shaking at 1000 rpm for 30 min. The pH dependence was investigated in a pH range of 4.0–9.0 at 60°C using various buffering systems: 100 mM sodium acetate buffer (pH 4.0–6.0), 100 mM sodium phosphate buffer (pH 6.0–8.0), and 100 mM Tris‐HCl buffer (pH 7.0–9.0). Reactions were performed in a 500 μL system containing 13.8 g/L CFA, 10 g/L DCW BsSDR_M2, and 8% (w/v) glucose, and incubated at 60°C with shaking at 1000 rpm for 30 min in buffers of varying pH (4.0–9.0). The substrate to glucose molar ratio was optimized within the range of 1:1 to 1:5 (mol/mol). Reactions were performed in a 500 μL system containing 100 mM sodium phosphate buffer (pH 6.0), 10 g/L DCW BsSDR_M2, and 13.8 g/L CFA at varying substrate to glucose molar ratios, followed by incubation at 60°C with shaking at 1000 rpm for 30 min.

To enhance the solubility of substrate CFA, cosolvent was screened among dimethylformamide (DMF), dimethyl sulfoxide (DMSO), methanol, ethanol, and isopropanol (IPA). Reactions were performed in a 500 μL system containing 100 mM sodium phosphate buffer (pH 6.0), 10 g/L DCW BsSDR_M2, 8% (w/v) glucose, and 13.8 g/L CFA, incubated at 60°C and 1000 rpm for 30 min with various organic solvents. Impact of the optimal solvent concentration on activity was determined in the range of from 10% to 60% (v/v). Reactions were performed in a 500 μL system containing 100 mM sodium phosphate buffer (pH 6.0), 10 g/L DCW of BsSDR_M2, 13.8 g/L CFA, and various concentrations of DMSO, incubated at 60°C and 1000 rpm for 30 min. Catalytic activity was determined by gas chromatography.

Reactions (50 mL) were conducted under the following conditions: BsSDR_M2 was incubated at 60°C in 100 mM sodium phosphate buffer (pH 6.0). All reactions were performed with agitation at 800 rpm in a mixture containing 50% (v/v) DMSO, a substrate to glucose molar ratio of 1:3 (mol/mol), 50 or 100 g/L CFA, and 10 g/L DCW BsSDR_M2. A batch‐wise feeding strategy was adopted for substrate addition: substrate CFA was initially set at 14 g/L, with 14 g/L further supplemented every 30 min until a final concentration of either 50 or 100 g/L was reached. A 6 M NaOH solution was used to maintain the pH within the range of 5.95–6.05 by using an automatic titrator. At specified intervals during the reaction, 100 μL aliquots of the reaction mixture were withdrawn, then extracted with 4.9 mL ethyl acetate. The organic phase was used to quantify (S)‐CFL and CFA following the method above‐mentioned.

2.13. Statistical Analysis

Each experiment was independently conducted in triplicate, and data are presented as the mean ± standard deviation (SD). Statistical significance was determined using an unpaired two‐sided Student's t‐test, and significance levels were annotated as follows: *p < 0.05; **p < 0.01; ***p < 0.001.

3. Results and Discussion

3.1. Parental SDR Mining and Identification

Enzyme discovery for chiral biomanufacturing remains a critical research direction in modern biocatalysis, primarily driven by the increasing demand for high stereoselectivity in the synthesis of pharmaceutical intermediates, fine chemicals, and agrochemicals (Gu et al. 2024; Li et al. 2021; Reetz et al. 2024). Here, we established a reductase library comprising 19 SDRs originating from Bacillus subtilis, Thermus thermophilus HB27, and Clostridium beijerinckii CGMCC 1.335. The catalytic activities were evaluated by using recombinant E. coli‐expressed SDRs as biocatalysts and CFA as the model substrate, together with glucose and Exiguobacterium sibiricum glucose dehydrogenase (EsGDH) for cofactors regeneration (Figure 1A). Among them, three SDRs, namely BsSDR, TtSDR1, and TtSDR4, exhibited considerable catalytic activity toward substrate CFA (Supporting Information S1: Table S4). Notably, TtSDR1 and TtSDR4 from Thermus thermophilus showed undesirable reverse R‐stereoselectivity toward CFA despite detectable catalytic activity. In contrast, the BsSDR from Bacillus subtilis showed the highest catalytic activity and excellent S‐selectivity, affording (S)‐CFL in 11.59% conversion and > 99.5% e.e. Hence, BsSDR was chosen for subsequent studies.

Figure 1.

(A) Strategies for the discovery of novel SDRs. (B) The optimal cosubstrate screening for WT‐BsSDR. CK: in the absence of any cosubstrates. Data are presented as the mean ± SD from three independent measurements (n = 3). ND indicates not detected.

3.2. Cosubstrate Screening for BsSDR

Conventional single‐substrate biocatalytic systems suffer from inherent limitations, including irreversible redox cofactor depletion and poor long‐term catalytic stability (Wolfe et al. 2024; Liu et al. 2025). In contrast, the one‐enzyme dual‐substrate strategy allows SDRs to catalyze both target substrates and cosubstrates, achieving efficient in situ cofactors regeneration, thus reducing cofactors cost, and improving overall process robustness (Sharma et al. 2022). To evaluate the feasibility and practical potential of BsSDR, a variety of sugars and alcohols were systematically screened as alternative cosubstrates. Using BsSDR‐expressing E. coli whole‐cells as the biocatalyst, glucose exhibited the highest catalytic activity among the test cosubstrates (Figure 1B). Then, the optimal reaction conditions of BsSDR were investigated, focusing on three key parameters: glucose concentration, temperature, and pH. As shown in Supporting Information S1: Figure S2A, the optimal glucose concentration was identified as 8% (w/v). The optimal pH and temperature for BsSDR were determined to be 7.0°C and 60°C, respectively (Supporting Information S1: Figure S2B,C).

3.3. Engineering BsSDR for Improved Thermostability and Activity

Protein engineering is a transformative approach for industrial biotechnology, facilitating the optimization of enzymes to satisfy stringent industrial criteria for stability, specificity, and efficiency (Dai et al. 2023; Gonlepa et al. 2025; Gu et al. 2021; Liu et al. 2023; Reetz et al. 2024; Shi et al. 2025). Thus, protein engineering was performed to enhance the catalytic activity of BsSDR, particularly toward bulky ketone substrates. The three‐dimensional (3D) model of BsSDR was constructed by employing AlphaFold 3, and amino acid sequence alignments reveal a catalytic triad comprising residues S177, Y190, and K194, as well as a typical (α/β)₈‐barrel structure (Figure 2A and Supporting Information S1: Figure S3). According to Ramachandran plot analysis, the 3D model quality is satisfactory, as most residues are located in allowed regions (Supporting Information S1: Figure S4). According to previous studies (Filling et al. 2002; Guo et al. 2014), the basic residue S177 stabilizes the substrate, and Y190 is the catalytic residue and participates in proton transfer.

Figure 2.

(A) Cartoon diagram of the BsSDR and substrate CFA docking complex, modeled using YASARA. (B) Close‐up view of the catalytic site (S177‐Y190‐K194) and CFA. (C) Schematic illustration of FuncLib, a computational pipeline for the automated design of multipoint mutations at enzyme active sites through phylogenetic analysis. Seven residues (Q130, I178, T179, K182, P222, A234, E238) were chosen as mutation sites. (D) Mutation sites were screened by computing the sequence space of hotspot residues. (E) Catalytic activity and enantioselectivity of WT‐BsSDR and its variants. One unit of enzyme activity (U) was defined as the amount of BsSDRs required for 1 μmol (S)‐CFL formation per hour under the reaction conditions of 60°C and pH 6.0. The specific activities of WT‐BsSDR and the variants Q130L, I178L, T179L, and K182E (U/mg DCW) are marked as red circles, green inverted triangles, blue triangles, pink squares, and purple rhombus, respectively. The other variants are represented by black squares. The e.e. values are represented by blue pentagrams. Errors indicate the SD of three replicates.

Subsequently, the substrate CFA was docked into the predicted active center, and the optimal binding conformation was selected for further analysis (Figure 2A,B). Molecular docking revealed 11 residues (I178, T179, Q130, K182, N184, L187, P220, G221, P222, A234, and E238) located within 5 Å of the substrate CFA, which were considered potential hotspots for mutagenesis. These candidate residues were further analyzed using FuncLib (http://FuncLib.weizmann.ac.il), an automated platform for designing multipoint active‐site mutations based on phylogenetic and structural considerations. Through this analysis, 7 residues (Q130, I178, T179, K182, P222, A234, and E238) were prioritized for mutagenesis (Figure 2C,D). A total of 31 mutants were constructed and experimentally validated, among which 13 exhibited improved catalytic activity. The beneficial mutations were primarily clustered around positions Q130, I178, T179, and K182. Notably, variants Q130L, I178L, T179L, and K182E exhibited significantly enhanced catalytic activity toward CFA compared to WT‐BsSDR, increased by 12.2‐, 7.5‐, 5.2‐, and 2.3‐folds, respectively (Figure 2E).

3.4. Combinatorial Engineering of Beneficial Variants for Synergistic Catalytic Enhancement

Synergetic mutation effects are often exploited in protein engineering to simultaneously improve multiple catalytic properties via combinatorial mutation strategies (Gong et al. 2018). Accordingly, combinatorial engineering was conducted on the most promising single‐site variants (Q130L, I178L, T179L, and K182E) to further improve catalytic performance. Six double mutants (Q130L/I178L, Q130L/T179L, Q130L/K182E, I178L/T179L, I178L/K182E, and T179L/K182E) and four triple mutants (Q130L/I178L/T179L, Q130L/I178L/K182E, I178L/T179L/K182E, and Q130L/T179L/K182E) were generated. Among them, the mutant I178L/T179L exhibited the most pronounced enhancement in catalytic activity, with a 107.4‐fold increase relative to WT‐BsSDR. In contrast, the quadruple mutant Q130L/I178L/T179L/K182E displayed a complete loss of catalytic activity toward CFA (Figure 3).

Figure 3.

Activity and enantioselectivity of combinatorial mutants. Asterisks indicate statistically significant differences compared with the WT (*p < 0.05; **p < 0.01; ***p < 0.001).

3.5. Cofactor Preference of BsSDR

As is known, most native SDRs are nicotinamide cofactor‐dependent enzymes, with a strong preference for either NADP(H) or NAD(H) (Kallberg et al. 2002; Persson et al. 2003; Black et al. 2020; King et al. 2020; Zhou et al. 2025). Given the significant differences in stability, bioavailability, and cost of these cofactors, understanding an enzyme's cofactor utilization is critical to its practical biocatalytic applications. Thus, to clarify the cofactor utilization of BsSDR, a systematic evaluation was performed using nicotinamide cofactors, including the native cofactor NADP(H), NAD(H), and the noncanonical cofactor NMNH. As shown in Figure 4, purified BsSDR exhibited substantial catalytic activity in the presence of both NADP(H), NAD(H), and NMNH. Notably, BsSDR showed the highest catalytic activity when NMNH was used as the cofactor, producing 0.16 and 0.18 g/L (S)‐CFL at 5 and 10 mM NMNH, respectively. Therefore, NMNH was identified as the preferred cofactor for BsSDR, differing from most SDRs that typically rely on NADP(H) or NAD(H). This finding not only expands the known cofactor versatility within the SDR family but also highlights BsSDR as a promising biocatalyst for industrial chiral alcohol synthesis.

Figure 4.

The optimal cofactor screening for WT‐BsSDR. Errors indicate the standard deviation (SD) of three replicates. CK: No cofactor was added. Errors indicate the SD of three replicates.

To elucidate the underlying mechanism by which BsSDR functions as an NMNH‐preferred enzyme, amino acid sequence alignments were performed against 10 different NADPH‐dependent SDRs randomly selected from the NCBI database (Supporting Information S1: Figure S3). The cofactor‐binding sites of SDRs are highly conserved, and typically contain Thr and Gly residues, forming the characteristic TG‐X₃‐G‐X‐G motif (Jörnvall et al. 2010). Furthermore, the 20 amino acids downstream of this motif play a key role in determining the cofactor specificity of SDRs (Lesk 1995). As shown in Supporting Information S1: Figure S3, the canonical TG‐X₃‐G‐X‐G motif was identified at residues 47–54. Analysis of these 20 downstream residues revealed that BsSDR possessed unique hydrophobic residues at positions 59 (Val), 60 (Leu), 71 (Val), and 73 (Leu) compared with other SDRs, whereas residue 58 (Ser) exhibited an opposite hydrophilic pattern. More strikingly, BsSDR has an alkaline Lys at position 63 instead of the acidic Glu that exists in the other 8 aligned SDRs, a neutral Asn at position 67 in place of the alkaline Lys conserved in all other SDRs, and an alkaline Arg at position 55 rather than the neutral Leu found in the aligned counterparts. These substitutions alter both the charge distribution and hydrophobicity in the region downstream of the TG‐X₃‐G‐X‐G motif, which underlies the intrinsic preference of BsSDR for NMNH and clearly distinguishes it from other SDRs.

3.6. Kinetic Parameters Determination for WT‐BsSDR and Variant BsSDR_M2

As shown in Table 1, kinetic parameters of WT‐BsSDR and its variant BsSDR_M2 toward the natural cofactors NAD⁺ and NADP⁺, as well as the noncanonical cofactor NMN⁺, were determined to further elucidate their catalytic efficiency. Intriguingly, besides activity toward natural cofactors, both WT‐BsSDR and mutant BsSDR_M2 exhibited substantial activity with the noncanonical cofactor NMN⁺. Moreover, compared with WT‐BsSDR, BsSDR_M2 exhibited 1.3‐ and 5.6‐fold improvements in k _cat and k _cat/K _M toward NMN⁺, respectively (Table 1 and Supporting Information S1: Figure S5). Concurrently, the kinetic parameters of WT‐BsSDR and its variant BsSDR_M2 toward substrate CFA were determined to evaluate their catalytic efficiency. BsSDR_M2 exhibited a significantly higher turnover number (k _cat) toward CFA, with a 4.9‐fold increase compared to WT‐BsSDR. In addition, BsSDR_M2 exhibited 83.6% lower K _M for CFA than that of WT‐BsSDR, confirming its higher substrate affinity (Table 2). Remarkably, BsSDR_M2 demonstrated an exceptional 33.3‐fold improvement in catalytic efficiency (k _cat/K _M) toward CFA.

Table 1.

Kinetic parameters of WT‐BsSDR and its variant BsSDR_M2 toward cofactors.

Enzyme	Cofactor	K M (mM)	k cat (s−1)	k cat/K M (s−1 mM−1)
WT‐BsSDR	NAD+	4.58 ± 0.87	0.81 ± 0.04	0.18 ± 0.03
	NADP+	3.15 ± 0.38	1.17 ± 0.07	0.37 ± 0.1
	NMN+	3.52 ± 0.50	1.25 ± 0.2	0.36 ± 0.02
BsSDRM2	NAD+	2.12 ± 0.22	1.94 ± 0.4	0.92 ± 0.1
	NADP+	1.22 ± 0.17	2.67 ± 0.3	2.19 ± 0.2
	NMN+	1.19 ± 0.17	2.85 ± 0.5	2.39 ± 0.11

Note: Reactions were conducted in pH 7.0, 100 mM sodium phosphate buffer, at a constant glucose concentration of 8% (w/v). NAD⁺, NADP⁺, and NMN⁺ represent oxidative cofactors.

Table 2.

Kinetic parameters of WT‐BsSDR and its variant BsSDR_M2 toward CFA.

Varianta	K M (mM)	k cat (s−1)	k cat/K M (s−1 mM−1)
WT‐BsSDR	15.13 ± 0.99	0.10 ± 0.01	0.007 ± 0.001
BsSDRM2	2.48 ± 0.20	0.59 ± 0.04	0.24 ± 0.02

Note: NMNH was supplemented as the only reduced cofactor.

^a Purified WT‐BsSDR and BsSDR_M2 were used.

3.7. Catalytic Activity and Thermostability Characterization for WT‐BsSDR and BsSDR_M2

To characterize the thermostabilities of WT‐BsSDR and BsSDR_M2, Tₘ, T ₅₀ ¹⁵, T ₅₀ ³⁰, and t _1/2 were determined. As listed in Table 3, the T _m values of WT‐BsSDR and BsSDR_M2 were 52.0°C and 75.4°C, respectively. The thermal melting curves are shown in Supporting Information S1: Figure S6. At 45°C, the t _1/2 of BsSDR_M2 reached 52.1 h, longer than that of WT‐BsSDR. Compared with WT‐BsSDR, the T ₅₀ ¹⁵ of BsSDR_M2 was increased by 9.8°C, and the T ₅₀ ³⁰ of BsSDR_M2 reached 62.4°C (Table 3). Collectively, these results demonstrate that BsSDR_M2 has improved thermostability.

Table 3.

Thermal stability and activity of WT‐BsSDR and its variant BsSDR_M2.

Enzyme	T m (°C)	ΔT m (°C)	T 50 15 (°C)	T 50 30 (°C)	45°C
					k d (×10−2 h−1)	t 1/2 (h)
WT‐BsSDR	52.0 ± 0.3	23.4	56.7 ± 1.8	55.6 ± 1.1	1.41	49.2 ± 2.5
BsSDRM2 (I178L/T179L)	75.4 ± 0.2		66.5 ± 2.7	62.4 ± 2.4	1.33	52.1 ± 2.1

3.8. Mechanistic Insights Into Activity and Stability Enhancement

In order to investigate the relative affinity of WT‐BsSDR and BsSDR_M2 for substrate CFA, we evaluated the binding free energy (ΔG_bind) using the MM‐GBSA method. As summarized in Supporting Information S1: Table S5, the calculated ΔG_bind between CFA and BsSDR_M2 was lower than that of WT‐BsSDR, decreasing from −12.99 kcal/mol (WT‐BsSDR) to −17.38 kcal/mol (BsSDR_M2). Consistent with the kinetic analysis, BsSDR_M2 displays a higher affinity for the substrate CFA. To further elucidate the affinity changes of WT‐BsSDR and BsSDR_M2 with CFA, binding free energy decomposition was performed. As shown in Figure 5A,B, the contributions of residues L178 and L179 in BsSDR_M2 to substrate CFA binding were significantly lower than those of residues I178 and T179 in WT‐BsSDR. This indicates that intramolecular interactions further stabilize the binding of the substrate CFA by enhancing the stability among CFA, the surrounding residues in the central binding region, and the two mutated residues.

Figure 5.

Free energy decomposition per residue for WT‐BsSDR (A) and BsSDR_M2 (B), respectively. Gibbs free energy landscape at 303 K for WT‐BsSDR (C) and BsSDR_M2 (D). (E) Secondary structure comparison between WT‐BsSDR and BsSDR_M2 during 100 ns MD simulations. Locations of residues 252–263 and 188–211 on the surface of WT‐BsSDR (F, G).

To explore the underlying mechanisms of enhanced catalytic performance, MD simulations (100 ns) of WT‐BsSDR and BsSDR_M2 were conducted at 303 K. Root‐mean‐square deviation (RMSD) analysis of MD trajectories revealed that the WT‐BsSDR reached a stable conformation at approximately 40 ns, whereas the BsSDR_M2 mutant achieved structural stability within only 10 ns, suggesting increased structural rigidity of the mutant (Supporting Information S1: Figure S7A). Meanwhile, root‐mean‐square fluctuation (RMSF) values were analyzed to validate the accuracy of the analysis. Near the mutation site, the RMSF values of BsSDR_M2 were notably different from WT‐BsSDR. Specifically, RMSF decreased at sites 160–180, with no significant changes observed at other sites (Supporting Information S1: Figure S7B). Furthermore, the Gibbs free energy landscape corresponding to the conformation was calculated using the radius of gyration (Rg, Supporting Information S1: Figure S7C) and RMSD to unravel the molecular mechanisms by which mutation (I178L/T179L) enhances thermal stability. Compared to WT‐BsSDR, BsSDR_M2 displayed reduced Rg values, reflecting its more compact and stable global conformation. At 303 K, the Gibbs free energy landscape of WT‐BsSDR (Figure 5C) revealed intensified red regions and an expanded high free energy region, reflecting lower conformational stability, while BsSDR_M2 (Figure 5D) presented a broadened blue distribution region, indicating increased conformational stability. BsSDR_M2 exhibited lower minimum Rg and RMSD values than WT‐BsSDR, suggesting a lower Gibbs free energy for the mutant. Simulations showed that I178L and T179L mutations significantly affected conformational dynamics, thus enhancing thermostability.

To identify structural differences that may underlie their functional disparities, secondary structural analyses were also conducted to compare the WT‐BsSDR and the mutant BsSDR_M2. Generally, protein secondary structures are categorized as helices (α‐helix, 3–10 helix, and π‐helix), sheets (β‐sheet and antiparallel β‐sheet), and irregular structures (β‐bends and β‐turns). The BsSDR_M2 mutations showed no significant changes in the overall secondary structure motifs. However, as shown in Figure 5E, two specific regions (residues 188–211 and 252–263) exhibited notable structural rearrangement, where the original loose and flexible β‐turns (purple) were transformed to more rigid and stable α‐helices (green). The detailed probability data for helix, sheet, and coil structures in residues 188–211 and 252–263 are shown in Supporting Information S1: Table S6. The secondary structure analysis revealed that, in BsSDR_M2, the proportion of the stable α‐helix increased, whereas that of the flexible β‐turn decreased. Residues 188–211 and 252–263 were located in the surface region of BsSDR (Figure 5F,G), and the enhanced rigidity of these surface residues presumably contributes to the thermostability of BsSDR.

3.9. Attacking Distance and Molecular Interaction Analysis of WT‐BsSDR and BsSDR_M2

To gain deeper insight into the mechanism underlying the enhanced catalytic performance of BsSDR_M2, the docking results of WT‐BsSDR and BsSDR_M2 were analyzed. Docking result comparisons between WT‐BsSDR and BsSDR_M2 revealed two key differences: (1) the distance between the hydrogen atom of Y156 and the carbonyl oxygen of CFA decreased from 2.8 to 1.8 Å, and (2) the distance between the C4 hydrogen of the NMNH amide ring and the carbonyl carbon of CFA was reduced from 3.4 to 2.8 Å (Figure 6A,B). These shortened catalytic distances in BsSDR_M2 are expected to facilitate more efficient proton and hydride transfer, thereby enhancing CFA binding affinity and improving catalytic efficiency. Additionally, we analyzed the interactions of the I178 and T179 mutation sites with the substrate CFA as well as with the surrounding residues. Compared to WT‐BsSDR, BsSDR_M2 exhibits stronger interactions with substrate CFA, as an additional π–π T‐shaped interaction forms between F240 and CFA. Structural analysis revealed that the I178L and T179L mutations introduced new hydrogen bonds (Figure 6C,D) and a remodeling of hydrophobic interactions with CFA and surrounding residues (Figure 6E,F). In our opinion, the double mutation I178L/I179L induces alterations in intramolecular interactions, thereby influencing the structural and functional properties of BsSDR.

Figure 6.

(A) Analysis of the distance between substrate and Y190 and coenzyme NMNH in WT‐BsSDR. (B) Analysis of the distance between substrate and Y190 and coenzyme NMNH in BsSDR_M2. Hydrogen bond network of WT‐BsSDR (C) and BsSDR_M2 (D). Interactions of mutation sites with surrounding residues in WT‐BsSDR (E) and BsSDR_M2 (F). Red dashed lines represent H‐bonds, green dashed lines represent hydrophobic interactions, purple dashed lines represent π–π T‐shaped interactions, and blue dashed lines represent cation‐π interactions.

3.10. Dynamics Cross‐Correlation Map (DCCM) and Pocket Dimensions Analysis

DCCM is a powerful tool for assessing the motion within enzymatic systems during MD simulations, particularly focusing on the Cα atom, a key indicator of protein rigidity (Gao et al. 2024; Ma et al. 2022). The DCCMs of WT‐BsSDR and BsSDR_M2 were generated from the final 20 ns segments of their respective MD simulations, with correlation coefficients (C _ij) visualized in distinct colors. Compared with WT‐BsSDR, the overall protein consistency of BsSDR_M2 showed no significant change. However, a marked reduction in negative cross‐correlation was observed in the substrate‐binding domain of mutant BsSDR_M2, specifically between residues 50–125 and 175–250. While DCCM analysis revealed that the mutations strengthened the correlated motions between L178/L179 and their neighboring residues. Therefore, BsSDR_M2 exhibited more cohesive motions than WT‐BsSDR, either positively or negatively cross‐correlated, which is hypothesized to positively influence BsSDR thermostability (Figure 7A,B). In addition, the substrate‐binding pocket dimensions of WT‐BsSDR and BsSDR_M2 were analyzed using Caver Analyst 2.0. The calculated substrate‐binding pocket volume of BsSDR_M2 (1534.8 ų) was 390.1 ų larger than that of WT‐BsSDR (1144.7 ų), indicating a more spacious and flexible catalytic cavity. This expansion of the catalytic pocket volume likely facilitates the accommodation of the sterically bulky substrate CFA, thereby improving substrate accessibility and enhancing the catalytic performance of BsSDR_M2 (Figure 7C,D).

Figure 7.

Dynamic cross‐correlation matrix (DCCM) analysis of WT‐BsSDR (A) and BsSDR_M2 (B). Correlation coefficients (C _ij) for the dynamic cross‐correlation map (DCCM) are displayed in a color‐coded manner. C _ij values ranging from 0 to 1 indicate a positive correlation, whereas values ranging from −1 to 0 indicate a negative correlation. Substrate‐binding pocket volumes of WT‐BsSDR (C) and BsSDR_M2 (D) were calculated by using Caver Analyst 2.0.

3.11. Substrate Specificity Characterization

To explore the substrate scope and broad applicability of BsSDR_M2 toward structurally diverse ketones, whole‐cell biotransformation was performed at pH 7.0, 100 mM sodium phosphate buffer, using 16 additional ketone substrates. Compared with WT‐BsSDR, the mutant BsSDR_M2 exhibited notably higher activities toward aromatic ketone substrates. Among 8 test aromatic ketones (9a–16a), 6 substrates (9a–12a, 14a, and 16a) displayed significantly enhanced activity, while maintaining strict stereoselectivity. This phenomenon can be attributed to the structural modification of the catalytic pocket in BsSDR_M2, where the increased cavity volume facilitates the access and release of aryl ketone substrates. In contrast, BsSDR_M2 exhibited lower catalytic activity toward smaller aliphatic ketones (1a–4a, 6a), likely due to suboptimal interactions within its expanded active site. Both BsSDR_M2 and WT‐BsSDR retained high catalytic activity toward aldehyde substrates (7a and 8a), indicating that the catalytic mechanism remains conserved despite the introduced mutations. Overall, the expanded substrate scope of BsSDR_M2 includes more industrially relevant building blocks, rendering it a versatile biocatalyst for producing key chiral alcohols used in fine chemicals and complex pharmaceuticals (Table 4).

Table 4.

Activity and stereoselectivity of WT‐BsSDR and BsSDR_M2 toward diverse substrates.

Substrates	Conv. (%)a		e.e.p (%)b
	WT‐BsSDR	BsSDRM2	WT‐BsSDR	BsSDRM2
1a	52.39 ± 2.12	26.54 ± 1.33	21.5 (S) ± 1.08	58.4 (R) ± 1.12
2a	38.87 ± 1.33	24.81 ± 1.24	66.6 (S) ± 1.33	72.9 (S) ± 1.51
3a	64.47 ± 1.01	70.28 ± 2.11	> 99.5 (S) ± 0.05	> 99.5 (S) ± 0.04
4a	57.53 ± 1.99	36.23 ± 1.82	51.0 (R) ± 1.35	60.3 (R) ± 1.29
5a	77.72 ± 1.60	91.08 ± 2.25	> 99.5 (R) ± 0.02	> 99.5 (R) ± 0.03
6a	97.35 ± 2.59	92.62 ± 1.63	23.66 (S) ± 0.32	4.82 (S) ± 0.41
7a	92.70 ± 1.87	94.51 ± 1.51	/	/
8a	66.13 ± 1.31	78.15 ± 1.91	/	/
9a	57.69 ± 1.12	71.92 ± 2.60	> 99.5 (R) ± 0.03	> 99.5 (R) ± 0.03
10a	88.08 ± 1.35	88.67 ± 2.43	> 99.5 (R) ± 0.05	> 99.5 (R) ± 0.06
11a	95.27 ± 2.52	100.00 ± 1.01	> 99.5 (S) ± 0.02	> 99.5 (S) ± 0.03
12a	27.04 ± 1.36	100.00 ± 1.31	> 99.5 (S) ± 0.02	> 99.5 (S) ± 0.02
13a	95.38 ± 2.57	91.02 ± 1.55	> 99.5 (S) ± 0.05	> 99.5 (S) ± 0.06
14a	27.57 ± 1.65	37.91 ± 1.90	> 99.5 (R) ± 0.06	> 99.5 (R) ± 0.05
15a	ND	ND	ND	ND
16a	47.26 ± 1.12	57.33 ± 2.13	> 99.5 (R) ± 0.15	> 99.5 (R) ± 0.10

Note: Keto‐esters (1a–6a), aldehydes (7a–8a), ketones (9a–16a).

Abbreviation: ND, not detectable.

^a Experiments were conducted in triplicate.

^b The configurations were determined by comparing the retention times with authentic samples or reported data (Supporting Information S1: Figure S8).

Substrate specificity analysis revealed that ester steric hindrance strongly affects the enantioselectivity toward β‐keto esters (1a‐4a). Substrate 3a gave the highest enantioselectivity, while 1a and 2a showed reduced selectivity with decreasing steric bulk. For α‐keto esters (5a, 6a), 5a displayed high enantioselectivity, while 6a showed poor selectivity, likely because its flexible aromatic substituent disrupts ordered binding. In addition, 3a and 5a differ in ester position, steric hindrance, and polarity, leading to distinct binding orientations that generate S‐ and R‐configured alcohols, respectively. For substrate 1a, WT‐BsSDR produced the S‐isomer, whereas BsSDR_M2 yielded the R‐isomer, suggesting that mutation I178L/T179L altered pocket size, polarity, or steric properties and thereby reversed stereoselectivity.

Aromatic ketone substrates (9a–14a, 16a) generally afforded excellent enantioselectivity because their bulky, rigid substituents restrict binding to a single favorable conformation. Substrates 9a and 14a, lacking ortho‐substitution and having low steric hindrance, preferentially bind with the aryl group in the large hydrophobic pocket and the alkyl chain in the small pocket, yielding the R‐product. In contrast, ortho‐substitution in 11a increases steric hindrance, reverses binding orientation, giving the S‐product. Likewise, the bulkier alkyl chains of 12a and 13a alter the steric balance between aryl and alkyl groups, switching binding preference and thus yielding the S‐product. Polar substituents also modulate pocket interactions. For example, para‐fluoro substitution in 14a has little effect on hydrophobic binding preference, retaining the R‐configuration identical to 9a, whereas chlorinated alkyl substituents in 10a, 13a, and 16a may enhance polarity and additional interactions with polar residues, thereby influencing carbonyl orientation and product configuration (Table 4).

3.12. BsSDR_M2‐Catalyzed (S)‐CFL Synthesis

To evaluate the industrial applicability of BsSDR_M2, a gram‐scale biosynthesis of (S)‐CFL was performed using E. coli BL21(DE3) whole cells expressing BsSDR_M2. Given that the substrate CFA is nearly water‐insoluble and tends to crystallize or precipitate at low temperatures or in aqueous media, the use of an appropriate cosolvent is crucial for enhancing its solubility and facilitating substrate diffusion into the active site of BsSDR_M2. Among the test cosolvents, DMSO was the most effective for BsSDR_M2‐catalyzed reactions (Figure 8D). Subsequent optimization of biocatalysis parameters, including temperature, pH, DMSO concentration, and molar ratio of substrate to glucose, was performed to maximize the catalytic efficiency of BsSDR_M2 in (S)‐CFL biosynthesis. As shown in Figure 8A,B, BsSDR_M2 exhibited optimal catalytic activity at 60°C and pH 6.0. The optimal molar ratio of substrate to glucose and DMSO concentration was determined to be 1:3 (mol/mol) and 50% (v/v), respectively (Figure 8C,E). Under the optimized conditions, whole cells expressing BsSDR_M2 (10 g DCW/L) achieved nearly complete conversion (> 94.53%) of 50 g/L CFA within 3 h. When the substrate concentration was increased to 100 g/L, complete conversion (> 92.06%) was achieved within 6 h, corresponding to a high space‐time yield (STY) up to 368.24 g/L/day (Figure 8F).

Figure 8.

(A) Effect of temperature on the catalytic activity of BsSDR_M2. (B) Effect of pH on the catalytic activity of BsSDR_M2. (C) Effect of molar ratio of CFA to glucose on the catalytic activity of BsSDR_M2. (D) Effect of organic cosolvents on the catalytic activity of BsSDR_M2. CK: No other organic solvent was added. (E) Effect of DMSO concentration on the catalytic activity of BsSDR_M2. (F) Reduction reaction processes of BsSDR_M2 toward substrate CFA under varying substrate concentration conditions (50, 100 g/L). CK: Whole‐cell reaction without BsSDR_M2. Errors indicate the SD of three replicates.

Conventional industrial synthesis of crizotinib intermediates relies heavily on chemical asymmetric reduction with expensive catalysts, such as Ir‐Spiro‐PAP (Iridium‐pyridine amino phosphine hydrogen) and chiral ligands (Qian et al. 2014). Limitations such as low substrate loading, stringent equipment requirements, and high costs restrict their large‐scale industrial application. Recently, ADH‐based biocatalysis has been widely applied for ketone CPE reduction and asymmetric synthesis of (S)‐CFL, addressing the limitations of chemical methods with high specificity and mild reaction conditions (Zong et al. 2019). In this study, a novel cofactor‐versatile SDR (WT‐BsSDR) was identified and was further engineered into BsSDR_M2, which exhibits good catalytic performance in the synthesis of (S)‐CFL, thus facilitating the industrial manufacture of crizotinib.

4. Conclusions

We identify the BsSDR from B. subtilis as an efficient stereoselective reductase capable of converting CFA to (S)‐CFL. Cofactor screening demonstrated that BsSDR exhibits remarkable cofactor versatility by utilizing both the noncanonical cofactor NMNH and the native NAD(P)H. To further improve its catalytic performance, in silico molecular docking combined with FuncLib‐guided rational design was employed to engineer BsSDR, yielding the “best” variant BsSDR_M2 (I178L/T179L). BsSDR_M2 exhibited significant enhancements in both the activity and thermostability toward substrate CFA, as well as higher conversion toward other sterically bulky aromatic ketones compared with WT‐BsSDR. In addition to the CFA‐reductase activity, BsSDR also possesses glucose‐oxidizing activity, enabling in situ cofactors regeneration. This allows us to establish a “One enzyme, two reactions” system, eliminating the need for an additional enzyme for cofactor regeneration. Based on this system, BsSDR_M2 achieved complete conversion of 100 g/L CFA within 6 h, with a STY of 368.24 g/(L·d), demonstrating great potential for industrial application in crizotinib biomanufacturing. Overall, the robust catalytic properties of BsSDR_M2 underscore its potential as a cost‐effective biocatalyst for (S)‐CFL synthesis, laying a solid foundation for the industrial production of crizotinib.

Author Contributions

Ya‐Jun Wang, Li‐Li Yao, Bin Xue, and Xian‐Heng Song conceived and designed the study. Li‐Li Yao, Yuan‐Fan Ye, Kai‐Cheng Huang, and Hui‐Qi Mao performed the experiments. Li‐Li Yao, Ya‐Jun Wang, Xian‐Heng Song, Yan‐Jun Wan, and Feng Xue analyzed the results. Li‐Li Yao, Bin Xue, and Ya‐Jun Wang wrote this manuscript.

Conflicts of Interest

The authors declare no conflicts of interest.

Supporting information

Figure S1: SDS‐PAGE analysis of BsSDR and its mutants. Lane M: Protein marker; Lane 1: Control; Lane 2: Crude enzyme of WT‐BsSDR; Lane 3: Crude enzyme of BsSDRM1 (T179L); Lan 4: Crude enzyme of BsSDRM2 (I178L/T179L); Lan 5: Purified enzyme of WT‐BsSDR; Lan 6: Purified enzyme of BsSDRM1 (T179L); Lane 7: Purified enzyme of BsSDRM2 (I178L/T179L).

Figure S2: (A) Effect of glucose concentration on WT‐BsSDR activity. Reaction conditions: 500 μL system containing 10 mM CFA, 10 g/L DCW, pH 7.0 sodium phosphate buffer (100 mM), and glucose at varying concentrations (0‐10% w/v); incubated at 50°C, 1000 rpm for 4 h. (B) Effect of pH on WT‐BsSDR activity. Reaction conditions: 500 μL reaction system containing 10 mM CFA, 10 g/L DCW, 8% (w/v) glucose; incubated at 60°C, 1000 rpm for 4 h in different buffers (100 mM sodium citrate at pHs 4.0, 5.0, 6.0; 100 mM sodium phosphate at pHs 6.0, 7.0, 8.0; 100 mM Tris‐HCl at pHs 7.0, 8.0, 9.0). (C) Effects of temperature on WT‐BsSDR activity. Reaction condition: 500 μL reaction system containing 10 mM CFA, 10 g/L DCW, 8% (w/v) glucose, and pH 7.0 sodium phosphate buffer (100 mM); incubated at 1000 rpm for 4 h.

Figure S3: Multiple‐squence alignment of WT‐BsSDR with ten different NADPH‐dependent short‐chain dehydrogenases (SDRs). The amino acid sequence of WT‐BsSDR was aligned with those of other SDRs; SDR1 (PDB ID: 4RF2), SDR2 (PDB ID: 8KCN), SDR3 (PDB ID: 7P36), SDR4 (PDB ID: 4RF5), SDR5 (PDB ID: 4RF3), SDR6 (PDB ID: 7P7Y), SDR7 (PDB ID: 1ZK4), SDR8 (GenBank accession No: WP_057878290.1), SDR9 (GenBank accession No: WP_129033283.1), SDR10 (GenBank accession No: KRL08675.1).

Figure S4: Evaluation of the WT‐BsSDR model using procheck. Procheck analysis showed 91.8% of residues in most favored regions, 8.2% in additional allowed regions, and no residues in generously allowed or disallowed regions.

Figure S5: Kinetic parameters determination for WT‐BsSDR and BsSDRM2 toward CFA and various cofactors. (A) Fitting of WT‐BsSDR initial reaction rate (v) at varying CFA concentrations. (B) Fitting of BsSDRM2 initial reaction rate (v) at varying CFA concentrations. (C) Fitting of WT‐BsSDR initial reaction rate (v) at varying NMN+ concentrations. (D) Fitting of BsSDRM2 initial reaction rate (v) at varying NMN+ concentrations. (E) Fitting of WT‐BsSDR initial reaction rate (v) at varying NAD+ concentrations. (F) Fitting of BsSDRM2 initial reaction rate (v) at varying NAD+ concentrations. (G) Fitting of WT‐BsSDR initial reaction rate (v) at varying NADP+ concentrations. (H) Fitting of BsSDRM2 initial reaction rate (v) at varying NADP+ concentrations.

Figure S6: Thermal unfolding curves of WT‐BsSDR and BsSDRM2.

Figure S7: The RMSD (A), RMSF (B) and Rg (C) of WT‐BsSDR and BsSDRM2. MD simulations were performed at 303 K during 100 ns.

Figure S8: HPLC/GC analysis of 16 substrates (1a‐16a) and their corresponding chiral alcohols synthesized by using WT‐BsSDR and BsSDRM2.

Table S1: Primers used in this study.

Table S2: Primer design for site‐directed mutagenesis.

Table S3: GC/HPLC detection condition for chiral alcohols detection.

Table S4: Screening for CFA‐reducing SDRs.

Table S5: Binding free energy of WT‐BsSDR and BsSDRM2 calculated by MM‐GBSA.

Table S6: Probability of helix, sheet, and coil in the region of residues 188‐211 and 252‐263 of WT‐BsSDR and BsSDRM2.

Yao, L.‐L. , Song X.‐H., Huang K.‐C., et al. 2026. “Engineering a Novel Cofactor‐Versatile Short‐Chain Dehydrogenase for Efficient Synthesis of Chiral Alcohols.” Biotechnology and Bioengineering 123: 2062–2079. 10.1002/bit.70239.

Data Availability Statement

The data that support the findings of this study are available in the supporting material of this article.