Improved catalytic efficiency of P450 OleP for converting lithocholic acid into murideoxycholic and ursodeoxycholic acids through semi-rational and rational design

2.1 Materials

The primer STAR HS DNA polymerase and the Rapid Taq DNA polymerase were purchased from Takara (Dalian, China) and Vazyme Biotech Co., Ltd. (Nanjing, China), respectively. The genomic DNA extraction kits, plasmid miniprep purification kits, and DNA purification kits were purchased from Sangon Biotech (Shanghai, China), which also provided oligonucleotides synthesis and sequence analysis services. 5-Aminolevulinic acid (ALA), isopropyl β-d-1-thiogalactopyranoside (IPTG), and kanamycin sulfate were purchased from Aladdin (Shanghai, China), while NADH was sourced from BONTAC (Shenzhen, China). Unless otherwise specified, all other chemicals were purchased from Sinopharm Chemical Reagent Co., Ltd. (Shanghai, China).

2.2 Plasmids and strains

All the strains constructed, plasmids, and primers used in this study are listed in Table S1, Table S2, and Table S3, respectively. E. coli DH5α was used as the cloning host, E. coli BL21(DE3) and C41(DE3) as the expression hosts, and the plasmid pRSFDuet-1 as the expression vector. E. coli cells were routinely cultured in Luria-Bertani (LB) medium (10 g L⁻¹ tryptone, 5 g L⁻¹ yeast extract, and 10 g L⁻¹ NaCl) or LB agar medium (LB medium containing 2% (W/V) agar). Protein expression was carried out in Terrific Broth (TB) medium (12 g L⁻¹ tryptone, 24 g L⁻¹ yeast extract, 5 g L⁻¹ glycerol, 2.31 g L⁻¹ KH₂PO₄, and 12.54 g L⁻¹ K₂HPO₄). Kanamycin was added as a supplement at a final concentration of 50 μg mL⁻¹ when needed.

2.3 Production conditions for the P450 candidates and purification for OleP

The successfully engineered recombinant E. coli strains, each harboring a candidate P450 enzyme and redox partners, were initially activated by culturing on LB agar plates supplemented with 50 μg mL⁻¹ of kanamycin at 37 °C for 12 h. A single colony was picked and inoculated into 2 mL of LB broth containing 50 μg mL⁻¹ of kanamycin, followed by incubation at 37 °C with shaking at 220 rpm for 12 h to prepare the seed culture. Subsequently, 1 mL of the seed culture was transferred into 250 mL shaking flasks containing 50 mL of TB medium supplemented with 50 μg mL⁻¹ of kanamycin, and the cultures were incubated at 37 °C with shaking at 220 rpm. When the optical density at 600 nm (OD₆₀₀) reached 0.6–0.8 (usually after 3–5 h of incubation), the culture was cooled to 16 °C and supplemented with 0.4 mM ALA and 0.2 mM FeCl₃. Production of P450s proteins was induced by adding 0.2 mM IPTG, and the culture was further incubated for another 36 h. After cultivation, the cells were harvested by centrifugation at 18,400g for 10 min, washed twice with 100 mM potassium phosphate buffer (pH 7.4), and then prepared for subsequent purification and whole-cell catalysis experiments.

For the purification of OleP, the harvested pellet was resuspended in buffer A (20 mM Tris-HCl and 300 mM NaCl, pH 8.0). Phenylmethylsulfonyl fluoride (PMSF) was added to a final concentration of 1 mM to inhibit protease activity and prevent OleP degradation. Cell lysis was performed at 4 °C using a high-pressure homogenizer (Union Biotech, Shanghai, China). Subsequently, cell debris was removed by centrifugation at 18,400g for 30 min at 4 °C. The resulting soluble fraction containing OleP was loaded onto a 5 mL Ni-NTA His-Binding resin (PointBio, Shanghai, China) pre-equilibrated with buffer A. Non-target proteins were extensively washed off using a series of wash buffers (buffer A supplemented with 5 mM, 10 mM, 20 mM, and 50 mM imidazole, pH 8.0). Purified OleP was then eluted with an elution buffer consisting of buffer A with 0.15 M imidazole (pH 8.0). The eluted protein solution was subjected to buffer exchange and desalting into buffer B (20 mM Tris-HCl, pH 8.0) using a PD MidiTrap G-25 column. Further purification was carried out by ion-exchange chromatography on a HitrapQ column pre-equilibrated with buffer B, employing a gradient of increasing NaCl concentration. Following elution, the buffer of the purified OleP was exchanged to buffer C (20 mM Tris-HCl, 100 mM NaCl, pH 8.0) using an Amicon Ultra 10K centrifugal filter (Merck). The protein solution was subsequently loaded onto a Superdex 200 pg 26/600 GL column (Union-Biotech, Shanghai, China) equilibrated with a buffer C supplemented with 5% (v/v) glycerol. The quality and quantity of the purified OleP were assessed using SDS-PAGE and UV/visible spectroscopy.

2.4 Crystallization, X-ray data collection, and analysis

The OleP protein and its mutants were crystallized at a concentration of 14–18 mg/mL in SEC buffer (20 mM Tris-HCl, 100 mM NaCl, 5% (v/v) glycerol, pH 8.0) with reservoir solution (Table S4) by sitting drop vapor diffusion technique. The crystals were observed after 7 days in 0.8 μL sitting drops consisting of a 1:1 mixture of protein solution and reservoir solution at 16 °C. Then, the crystals were cryoprotected by transient soaking in corresponding reservoir solutions with additional 5% (v/v) glycerol and were transferred into liquid nitrogen. The data collection experiment was carried out on the BL19U1 beamline at Shanghai Synchrotron Radiation Facility (SSRF) (Shanghai, China), and the set was indexed and processed using the XDS1 program suite. The OleP holo crystal form diffracted to 1.98 Å resolution and belongs to the space group C121, with unit-cell parameters a = 183.60 Å, b = 61.21 Å, c = 99.07 Å, α = γ = 90.0°, β = 117.70° and two molecules in the asymmetric unit. The structures of OleP protein with substrate LCA and its mutants were phased by PHASER2 using predicted OleP holo coordinates as the molecular replacement model. Further manual model building was carried out using COOT4. The OleP model was refined using all data to 1.98 Å resolution, by the program REFMAC5 with TLS restraints. Phenix.table_one6 was used to generate the crystallographic statistics table (except for the value of Rwork/Rfree), indicated that 96.9% of the residues were in the most favored regions of the Ramachandran plot. Data collection, phasing and refinement statistics are given in Table S4. Structure figures were prepared with PyMOL2.3.4.

2.5 Whole-cell catalysis for the P450 candidates

To prepare P450s whole-cell biocatalysts, resting cell pellets with an optical density at 600 nm (OD₆₀₀) of 25 were resuspended in 100 mM potassium phosphate buffer (pH 7.4) supplemented with 5% (w/v) glucose and 5% (v/v) glycerol. For the bioconversion of LCA to MDCA or UDCA, 2 mL of the cell suspension was supplemented with 0.5 g L⁻¹ of the substrate LCA (1.33 mM), 5 g L⁻¹ of the cosolvent hydroxypropyl-beta-cyclodextrin (HP-β-CD), and 1 mM NADH. The whole-cell biocatalysis was conducted in 24-well plates at 25 °C and 220 rpm for 12 h. Subsequently, 1 mL of the reaction solution was collected and extracted twice with 1 mL of ethyl acetate. The resulting mixture was dried, redissolved in methanol, and then centrifuged at 18,400g for 20 min. The supernatant was filtered through a 0.22 μm filter for HPLC and UPLC-MS analysis. The conversion rate (%) was calculated using the following equation based on HPLC: Conversion (%) = (C_Actual)/(C_Theoretical) × 100%. C_Actual is the concentration of the actual amount of product generated, and C_Theoretical is the concentration of the theoretical amount of product that could be generated. The selectivity of MDCA was calculated as the ratio of the concentration of MDCA to the total concentration of all products. The selectivity of UDCA was calculated as the ratio of the concentration of UDCA to the total concentration of all products.

2.6 High-throughput screening method for UDCA and MDCA

The gene encoding 7β-HSDH (A4ECA9.1) from Collinsella aerofaciens was codon-optimized and synthesized by GenScript (Nanjing, China). The 7β-HSDH gene was subsequently subcloned into the C-terminal region of the OleP gene in plasmid pRSF-OleP-CamAB, with the incorporation of a ribosome binding site (RBS; GGATCCGAATTCGAGCTCATAAAAGGAGGAAAATAT), yielding the recombinant plasmid pRSF-OleP-CamAB-HSDH. After the whole-cell reaction, 50 μL of the reaction product was transferred to a 96-well microplate. Subsequently, 100 μL of a 2,4-dinitrophenylhydrazine solution (10 mM, dissolved in a 1:1 v/v mixture of dimethyl sulfoxide and ethanol) was added to the each well. The microplate was placed on a shaker and incubated at 25 °C with shaking at 800 rpm for 30 min. Following this, 100 μL of NaOH solution (4 M, dissolved in a 40% v/v ethanol aqueous solution) was added to the wells. Following the 30 min standing period at room temperature, the mixture was centrifuged at 18,400g for 5 min to collect the supernatant. A 100 μL aliquot of the supernatant was transferred to a new 96-well microplate, and the absorbance at 440 nm was measured. Samples with high absorbance values were selected for further HPLC analysis.

To obtain mutants capable of efficient MDCA synthesis, we utilized the HTS method previously established in our laboratory [11]. Specifically, following the whole-cell reaction, 50 μL of the reaction product was transferred into a 96-well microplate. Subsequently, 150 μL of a 75% H₂SO₄–EtOH solution was gradually added to each well. The reactions were carried out at room temperature, and the mixtures were centrifuged at 18,400g for 5 min to obtain the supernatant. The resulting mixtures exhibited a yellow coloration proportional to the MDCA concentration, which was quantitatively assessed by measuring absorbance at 370 nm. Samples exhibiting high absorbance values were selected for further HPLC analysis.

2.7 Molecular docking and molecular dynamics (MD) simulations

The substrate LCA was docked into the active site of the OleP_S240A crystal structure using Auto-Dock Vina in Chimera [23]. Molecular dynamics (MD) simulations of the mutant-substrate complex were performed to analyze the dynamic structural characteristics of the protein complexes. The protonation states of titratable residues (His, Glu, and Asp) were determined based on pKa values using PROPKA3.1 and through visual inspection of local H-bonding networks [24,25]. Histidine residues His 30, His 138, His 246, His 333, and His 354 were protonated at the δ position, while His 23, His 101, His 228, His 258, His 320, His 340, His 346, His 351, and His 355 were protonated at the ε position. Except for amino acids E89 and D91, which remain protonated, all other acidic residues are deprotonated. The substrate LCA was parameterized using the general AMBER Force Field (GAFF) [26], with atomic charge parameters fitted via the restrained electrostatic potential (RESP) method [27]. The RESP charges were calculated using Multiwfn software [28,29]. The force field parameters for Cpd I were sourced from a previous study [30]. The Amber ff14SB force field was employed for the protein residues [31]. The complex system was solvated in a 16 Å TIP3P water box, and Na⁺ was added to the protein surface to neutralize the total charge of the system.

All molecular dynamics simulations were performed using the GPU version of the Amber 18 software package. The detailed setting parameters were adopted from the previously reported [25]. After setup, the energy of the system was minimized using 5000 steepest descent and 5000 conjugate gradient steps. The system was then heated from 0 to 300 K under a canonical ensemble for 0.5 ns with a time step of 2 fs. Next, a 1 ns density equilibration was performed under the NPT ensemble at 300 K and 1.0 atm pressure to achieve uniform density following the heating dynamics. Subsequently, all restraints on the protein were removed, and the system underwent further equilibration for 4 ns under an NPT ensemble to stabilize pressure and temperature. A 50 ns production molecular simulation was then conducted. Analysis of the simulation data, including root mean square deviation (RMSD), root mean square fluctuation (RMSF), and hydrogen bond interactions, was performed using the CPPTRAJ module within Amber18.

2.8 Rational design of the P450 OleP

GEMS (GEMME + ESM_IF1 + MSA Transformer + SaProt) is a zero-shot prediction methodology designed to evaluate enzyme active site mutants by integrating structural, sequence, and evolutionary data of enzymes [32]. This approach assesses the existence likelihood of mutants relative to the wild type by incorporating structural data through ESM_IF1, evolutionary data via MSA Transformer and GEMME, and both sequence and structural data through SaProt, thus facilitating the creation of high-quality initial mutation libraries [32]. The detailed procedure for OleP design using GEMS is described as follows: substrate LCA was docked into the active site of the OleP_{S240A/F84Q/V291G} structure using Auto-Dock Vina in Chimera. Suitable reaction conformations were selected based on a comprehensive assessment of docking scores and binding poses. Subsequently, molecular dynamics simulations were conducted on the selected conformations. Upon stabilization, the conformations were clustered, and representative conformations were chosen from each cluster center. Virtual saturation mutagenesis was then performed on all amino acid residues within 6 Å of the non-hydrogen atoms of the substrate LCA to construct a mutant library. The virtual mutant library was evaluated using GEMS_greedy to optimize the balance between predicted scores and mutant diversity, followed by ranking of mutants according to their scores. Based on the semi-rational design results, candidate mutants were selected for further screening. The changes in binding energy before and after mutation were assessed using Rosetta Cartesian ΔΔG scoring, and mutants with significantly reduced binding energy were excluded. Thereafter, the protein-substrate co-folding tool Chai-1 was employed to determine the substrate-mutant binding conformation and evaluate whether the complex conformation was conducive to the reaction. Reasonable mutants were selected for 100 ns molecular dynamics simulations. If the simulated conformation proved stable, the mutant was constructed, and whole-cell catalytic experiment was performed to validate its functionality.

2.9 Analytical procedures

Cell density was assessed by measuring OD₆₀₀ using a spectrophotometer (UVmini-1240, Shimadzu Corporation, Japan). The quantification of MDCA and UDCA was performed using an HPLC instrument (Shimadzu, Kyoto, Japan), equipped with an evaporative light-scattering detector and a reverse-phase, ZORBAX Eclipse XDB-C18 column (5 μm, 4.6 × 250 mm, Agilent, USA). Detailed experimental procedures were performed as described in our previous report [11]. For the analysis of MDCA and UDCA using UPLC-MS, 1 μL of each sample was introduced onto an Agilent InfinityLab Poroshell 120 HILIZ-Z column (2.1 × 150 mm, 2.7 μm). The separation was conducted using an Agilent 1290 Infinity II LC system, and quantification was achieved with an Agilent 6495C triple quadrupole mass spectrometer (Agilent, USA) equipped with an electrospray ionization (ESI) interface operating in negative ion mode. Multiple reaction monitoring (MRM) mode was applied for data acquisition and target compound quantification. The elution process utilized a mobile phase A composed of 10 mM ammonium formate with 0.02% formic acid, and mobile phase B consisting of acetonitrile with 0.02% formic acid. The flow rate was maintained at 0.3 mL min⁻¹, and the solvent gradient was programed as follows over a total run time of 8.0 min: 0.1–4.0 min, 100% B; 4.0–4.5 min, 80% B; 4.5–8.0 min, 100% B. The mass-to-charge (m/z) of 375.0 was selected for monitoring LCA, while m/z 391.0 was used for monitoring MDCA and UDCA. All experiments were conducted in triplicate using cultures derived from three independent colonies (n = 3). Data analysis was carried out using GraphPad Prism 8.0 software, and results are presented as the mean ± standard deviation (SD) from the triplicate experiments.

3.1 Screening of efficient P450 enzymes for converting LCA to MDCA and UDCA

According to previous reports, the three-component P450 OleP can catalyze the production of 6β-OH deoxycholic acid from deoxycholic acid, as well as the synthesis of MDCA from LCA. However, when whole-cell catalysts were prepared using E. coli C43(DE3) strain harboring the pET28a-oleP and pACYCDuet-camA-camB plasmids, the conversion rate of LCA to MDCA was merely 1.6% [33]. To achieve efficient conversion of LCA to MDCA and UDCA, several P450s were selected to investigate their hydroxylation activity towards LCA, including the self-sufficient P450 BM3, two-component P450 enzymes (CYP5150AP3, P450_cur, and P450_FE), and other three-component P450 enzymes (CYP105A1, CYP105A3, CYP105AB3, CYP105D7, and CYP105P2).

An efficient redox partner SmCPR^I453V from Silybum marianum was employed to construct a whole-cell biocatalysis system for two-component P450s, while putidaredoxin reductase (CamA) and putidaredoxin (CamB) from Pseudomonas putida were used to establish the catalytic system for three-component P450s. The genes encoding theses redox partners were co-expressed with the corresponding P450 genes in the BL21(DE3) using a high-copy number plasmid pRSFDuet-1, resulting in ten recombinant E. coli strains harboring different P450s (Fig. 1A). Subsequently, the hydroxylation activity toward LCA was evaluated using these whole-cell biocatalysts.

The results showed that only the recombinant strain BL21-CamAB-OleP (harboring plasmid pRSF-CamAB-OleP) catalyzed the conversion of LCA to MDCA, with a conversion rate of 1.1% (Fig. 1B). Previous studies have reported that CYP5150AP3, P450_cur, and P450_FE exhibit steroidal 7β-hydroxylase activity towards 11-deoxycortisol, androstadienedione, and LCA, respectively, in the Pichia pastoris expression system [16,34,35]. However, none of them showed activity toward LCA in the E. coli expression system. To improve the catalytic performance of OleP, we reconstituted it with an enhanced electron transfer efficiency PetH/PetF redox partner and expressed it in E. coli C41(DE3), a strain that facilitates P450 folding via moderate T7 RNA polymerase expression [33]. This yielded the recombinant strain C41-PetHF-OleP (harboring plasmid pRSF-PetHF-OleP), which increased the conversion rate of LCA to MDCA to 32.5% (Fig. S1). Therefore, OleP was selected as the target enzyme for subsequent research.

3.2 Improving a high-throughput screening method for detecting UDCA

Despite advances in P450-mediated steroid hydroxylation via protein engineering [9,17], low steroid detection efficiency limits the screening of a large P450 mutation library. We previously established a high-throughput MDCA detection method using 75% H₂SO₄–EtOH, which outperforms the traditional NAD(P)H consumption assay in accuracy (Fig. 2A) [11]. For UDCA, a colorimetric assay based on the water-soluble tetrazolium salt, 2-(4-iodophenyl)-3-(4-nitrophenyl)-5-(2,4-disulfophenyl)-2H-tetrazolium (WST-1)/1-methoxy-5-methylphenazinium methyl sulfate (mPMS) system has been established, which couples NADPH production (from 7β-hydroxysteroid dehydrogenase catalyzed UDCA oxidation) [9]. Here, we improved the high-throughput detection method for UDCA by coupling two reactions (Fig. 2A): oxidation of UDCA to 7-keto-lithocholic acid (7-KLCA) via a NADP⁺-dependent 7β-hydroxysteroid dehydrogenase from Collinsella aerofaciens [19], followed by 2,4-dinitrophenylhydrazine (DNPH)-based ketone detection for 7-KLCA [36].

To optimize detection conditions, full-wavelength scanning (300-900 nm) was performed on DNPH-reacted solutions of UDCA, 7-KLCA, and their mixtures. The results showed that 7-KLCA peaked at 440 nm without interference from UDCA (Fig. 2B). The results showed that the absorbance detection of 7-KLCA at 440 nm was unaffected by the presence of UDCA. A standard curve of 7-KLCA concentration (0.125–2.0 mg/mL) versus absorbance at 440 nm exhibited a high correlation coefficient of 0.996 (Fig. 2C), and 7-KLCA concentration correlated positively with reaction solution color intensity (Fig. 2D). Validation with 48 samples via both this method and HPLC method showed 93.75% (45/48) of samples had a positive correlation between microplate absorbance and HPLC peak area (Fig. 2E). This high-throughput method enables the simultaneously detection of 96 samples in 60 min with 93% accuracy, realizing rapid and precise 7-KLCA quantification and improving UDCA detection efficiency for large-scale mutant screening.

3.3 Structural characteristics of the OleP-LCA complex

To provide structural information for the subsequent rational design of P450 OleP, the crystal structure of OleP was resolved. The purified OleP enzyme was obtained using a Hi-trap chelating HP column, gel filtration chromatography, and ion-exchange chromatography with a Mono Q column (GE Healthcare) (Fig. S2A–S2D). The purified OleP enzyme was crystallized by sitting-drop vapor-diffusion, producing OleP (PDB:21LA) and OleP-LCA (PDB:21GU) complex structures at 1.98 Å and 2.79 Å resolution (Fig. 3A, Fig. S2E, Fig. S3 and Table S4). The asymmetric unit contained two molecules, each adopting the canonical P450 fold with 14 α-helices, 15 β-strands (Fig. S4), and a central heme group coordinated by a thiolate ligand. Within each asymmetric unit, two OleP molecules (designated as chains A and B) adopt nearly identical conformations (RMSD = 0.31 Å), exhibiting high similarity to the previously reported crystal structure of OleP (PDB: 4XE3) with an RMSD of 1.1 Å (Fig. S5A) [37]. Structural alignment using the Dali Server confirmed that OleP shares high similarity with P450OleP (PDB: 4XE3, 8QYI, Z-score 61.4, 52.1, Fig. S5), including conserved catalytic motifs in the I- and L-helices. The iron in the heme group in the active site is axially coordinated with the conserved cysteine residue at position 356. The two propionate groups of the heme molecule form electrostatic interactions with two highly conserved histidine residues (at positions 101 and 354), two arginine residues (at positions 105 and 298), a leucine residue (at position 94), and a phenylalanine residue (at position 321).

Compared to 8QYI in PDB (RMSD = 2.5 Å, Fig. S5B), LCA in the OleP–LCA (PDB ID: 21GU) complex adopts similar conformations across all binding sites (Fig. 3B and Fig. S6A–S6C), indicating a conserved binding mechanism. The binding pocket of OleP is primarily composed of some hydrophobic residues, including Val93, Ser240, Ile243, Ala244, Val291, Gly294, Phe296, and Ile397 (Fig. 3C and Fig. S6C). Compared with the substrate-free structure of OleP (RMSD = 0.254), the B/C region is shifted outward, whereas the core substrate-binding cavity near the heme remains largely unchanged. In contrast, the overall active site cavity of 21GU is larger because the two helices in the G/H region adopt a more outward open conformation (Fig. S6D), resulting in a cavity volume of 2839 ų, approximately 1.3-fold greater than that of P450 OleP (8QYI) (Fig. 3D and Fig. S7). The distance between the C7 atom in LCA and the iron atom in the heme is approximately 4.3 Å, while the distance between the C6 atom in LCA and the iron atom in the heme is approximately 4.8 Å (Fig. 3B). This arrangement favors hydroxylation at the C6 position by the catalytic intermediate Compound I.

3.4 Semi-rational design of OleP to enhance the synthesis of MDCA and UDCA

To identify key residues affecting OleP-catalyzed LCA hydroxylation, a comprehensive mutant library with 60 amino acid sites in the substrate cavity was generated. This included ten heme-interacting residues, twenty-one residues within 5 Å of the heme, nine residues contacting the substrate LCA, and twenty residues within 5 Å of LCA (Fig. 4A–B, and S8A-B). Alanine residues at positions 95, 244, 293, 348, 353, and 359 were mutated to glycine, while the remaining 54 sites were substituted with alanine. The catalytic activity of the OleP variant was assessed by whole-cell catalysis, high-throughput screening, and HPLC analysis. Thirteen OleP mutants showed enhanced MDCA synthesis compared to the wild type, including P88A, A95G, H101A, R105A, L160A, S240A, S249A, S292A, A293G, V297A, H351A, H355A, and A359G. Notably, A95G, S240A, and S292A increased MDCA conversion rate from 32.5% to 89.4%, 94.5%, and 81.2%, respectively (Fig. 4C). Additionally, three mutants, including F84A, P88A, and S240A, enabled the 7β-hydroxylation of LCA to UDCA, with conversion rates of 1.5%, 1.8%, and 1.1%, respectively (Fig. 4D). The flexible B–C loop, positioned at the active pocket entrance, controls its opening and closing. Mutations F84A and P88A in this region modulate substrate specificity.

To enhance the catalytic activity of OleP for synthesizing MDCA and UDCA from LCA, F84, P88, and S240 were subjected to saturation mutagenesis. Mutations at these sites, except alanine substitutions, reduced UDCA conversion from LCA (Fig. S9). The S240A mutant (C41-PetHF-OleP_S240A) showed high catalytic efficiency, converting LCA to MDCA at 94.5% and to UDCA at 1.1% (Fig. S9). The previously identified PetF_F64D mutant, which enhanced electron transfer efficiency in the OleP catalytic reaction [11], was introduced to further improve the catalytic activity. The resulting strain (C41-PetHF_F64D-OleP_S240A) achieved conversion rates of 98.3% for MDCA (1.30 mM MDCA from 1.33 mM LCA at 12h; MDCA selectively:98.3%) and 1.7% for UDCA (22.6 μM UDCA from 1.33 mM LCA at 12h) (Fig. S1 and Fig. S10). Notably, the MDCA conversion yield was 1.22 times higher than the yield reported previously (80.9%) [11].

Reportedly, F84A, S240A, and V291A were identified as some of the most effective in catalyzing the conversion of LCA to UDCA [9]. Functionally, residue F84, situated on the BC loop, participates in substrate recognition [38]. S240, located on helix I, forms hydrogen bonds with the substrate [39]. V291, positioned on a β-sheet, contributes Pi-alkyl interactions that stabilize substrate binding [38,40]. In this study, we investigated the combined effect of mutations at F84 and V291 within an S240A background. The F84Q/S240A double mutant displayed a pronounced shift in regioselectivity, accompanied by a reduction in 6β-C-hydroxylation (Fig. S11). Notably, the triple mutant F84Q/S240A/V291G showed a dramatic decrease in MDCA conversion (conversion rate from 98.3% to 2.3%), while UDCA synthesis rose substantially from 1.7% to 6.5% (86.4 μM UDCA from 1.33 mM LCA at 12 h; UDCA selectively:73.9%) (Fig. S1 and Fig. S11). This corresponds to a 2.4-fold increase over the previously reported yield (2.68%) [9].

3.5 Optimization of OleP mutants based on rational design for enhanced UDCA synthesis

Through the application of semi-rational design, we successfully identified key residues involved in LCA hydroxylation and obtained the S240A mutant, along with its crystal structure, as well as the UDCA-enhanced triple mutant F84Q/S240A/V291G. This semi-rational design approach proved highly effective in the initial phase, enabling the construction of a focused mutant library for screening key residues without necessitating complex calculations. With these experimental advancements, our research objective evolved from merely identifying key residues to the precise optimization of UDCA synthesis efficiency and regioselectivity. Recognizing the limitations of semi-rational design in achieving multi-site precision, we employed rational design methodologies, specifically GEMS and Rosetta, to further enhance UDCA synthesis [32]. Building on the triple mutant F84Q/S240A/V291G, GEMS integrated structural, sequence, and evolutionary data to predict high-quality mutants, thereby guiding virtual saturation mutagenesis of 18 active site residues while ensuring diversity. The activity of the mutants was predicted with the GEME approach, and the top variants were reranked based on LCA binding affinity calculated by Rosetta Cartesian ΔΔG (Table S5 for ΔΔG, Table S6 for reranking results). From the top 150 GEMS-predicted variants, 31 variants were selected, considering the sequence diversity and active pocket volume (Fig. 5A). Additionally, Foldseek was used to identify homologous structures and guide further mutant designs based on sequence insertions or deletions (Fig. 5A). Of the 35 rationally designed mutants, three mutants, OleP_{F84Q/S240A/V291G/E89D}, OleP_{F84Q/S240A/V291G/E89N}, and OleP_{F84Q/S240A/V291G/V239I}, showed improved UDCA conversion (Fig. 5B). Among the variants, OleP_{F84Q/S240A/V291G/E89N} was the most effective, raising UDCA yield from 6.5% to 12.1% (160.9 μM UDCA from 1.33 mM LCA at 12 h; UDCA selectively:79.5%) (Fig. S1 and Fig. 5B).

To elucidate the catalytic mechanism by which the S240A mutant and the F84Q/S240A/V291G/E89N mutant efficiently catalyze LCA, the holo structure for the S240A mutant (PDB: 21NO) with a resolution of 2.85 Å was successfully obtained (Fig. S12 and 13 and Table S4). The crystal structure of mutant S240A exhibits high similarity to that of wild-type OleP, with an RMSD of only 0.4 Å (Fig. S13). Regrettably, the S240A complex structure containing LCA could not be obtained. Furthermore, the mutant OleP_S240A–LCA complex was modeled using molecular docking and molecular dynamics simulations. The results indicated the presence of two distinct conformations within the OleP-LCA complex structure (Fig. S14A and S14B). Configuration A represents the catalytic conformation, characterized by the orientation of the C6–H group towards the oxygen atom of Cpd I, which is conducive to hydroxylation (Fig. S14A and S14C). In contrast, configuration B arises from the interaction between residue S240 and the carboxyl group of LCA, causing the C6–H group to withdraw from Cpd I, thereby inhibiting C6 hydroxylation (Fig. S14B and S14D). In the OleP_S240A-LCA complex structure, the substitution of the polar residue serine (S) with alanine (A) eliminates the interaction between S240A and substrate LCA, resulting in a singular conformation, termed Configuration C. This stable conformation maintains a 3.0 Å distance between the C6 position and the central iron atom of the heme group (Fig. 5C and E). Further analysis of the MD simulation for S240A revealed that, in most conformations, the distance between C6–H and the oxygen atom of Cpd I is more favorable for hydrogen abstraction than that between C7–H and the oxygen atom of Cpd I (Figs. S15 and S16). This spatial arrangement enhances the activation rate of the iron atom for hydrogen abstraction from LCA and oxygen activation, thereby significantly increasing the C6 hydroxylation conversion rate from 32.5% to 94.5%.

In the OleP_{F84Q/S240A/V291G/E89N}-LCA complex, the active-site cavity is enlarged to 3363 ų, providing greater space for substrate accommodation and reorientation (Fig. S17A). Meanwhile, the quadruple mutations reshape the active-site interaction network (Fig. S17B). In particular, substitution of Glu89 with Asn introduces a hydrogen bond between E89 N and LCA, stabilizing a distinct binding mode (Configuration D). In this rearranged conformation, the distance between the C7 atom of LCA and the central iron atom of the heme group is reduced to 3.61 Å, whereas the distance between the C6 atom and the heme iron is approximately 4.23 Å (Fig. 5D and F, and S16), favoring attack by the catalytic intermediate Compound I at the C7 position. This more favorable catalytic geometry is consistent with the enhanced C7 hydroxylation activity, increasing the conversion rate from 6.5% to 12.1%. Together, these results suggest that the improved production of UDCA by OleP_{F84Q/S240A/V291G/E89N} arises from the combined effects of active-site expansion, substrate reorientation, and stabilization of a C7-favored binding configuration.