Production of 14 α -Hydroxy Progesterone Using a Steroidal Hydroxylase from Cochliobolus lunatus Expressed in Escherichia coli

: Steroids with hydroxylation at C14 are drawing increased attention because of their diverse biological activities and applications. P -450 lun from Cochliobolus lunatus is the first fungal cytochrome P450 reported to have 14 α -hydroxylase activity. Studies have shown that P -450 lun catalyzes the hydroxylation of progesterone (PROG) at C14 α with low regiospecificity and activity. To improve its regiospecificity and activity for PROG, truncated forms of P -450 lun and its cognate redox partner CPR lun were functionally co-expressed in Escherichia coli . Then, a semi-rational protein engineering approach was applied to P -450 lun , resulting in a double-site mutant E109A/F297W with enhanced 14 α -position selectivity for PROG compared with the wild-type P -450 lun (97% vs. 28%). Protein structure analysis revealed that the F297W substitution can hinder the binding pose for 11 β -hydroxylation product formation. Finally, whole-cell catalysis was optimized, and the final titer of 14 α -OH-PROG reached 16.0 mg/L. This is the first report where a fungal 14 α -hydroxylase was functionally expressed in Escherichia coli . The steroid hydroxylation system obtained in this study can serve as a basis for the synthesis of 14 α -hydroxylated PROG and the rapid evolution of eukaryotic cytochrome P -450 lun .


Introduction
Steroids, pervasive terpene lipids in nature, have long served as therapeutic agents for a range of clinical diseases, including rheumatologic, autoimmune, and inflammatory disorders [1,2].The diverse physiological and pharmacological activities of steroids result from the strategic introduction of various functional groups onto the rigid gonane ring [3,4].A pivotal modification is the hydroxylation of the steroid backbone, a process that heightens the polarity of hydrophobic steroid molecules, thereby influencing their toxicity, cell membrane penetration ability, and the biological effects of steroid drugs ultimately [5].Of particular interest are 14-position hydroxylated steroids.14α-OH steroids exhibit specific biological activities, demonstrating potential antigonadotrophic and anticancer properties [6][7][8].In contrast, 14β-OH substituent is normally found in cardiotonic steroids, which are commonly used in treating congestive heart failure because of their cardiac effects, particularly the positive transducer effect [9][10][11].
Two approaches are currently used to incorporate 14-OH groups into steroids: chemical synthesis and biocatalysis.Nevertheless, the chemical synthesis route for steroid hydroxylation is beset by its complexity, low yields, and environmental unfriendliness [12,13].In contrast, biocatalysis is gaining prominence because of its high selectivity and atom utilization efficiency [14].Diverse microorganisms, including prokaryotes and fungi, can efficiently catalyze the production of steroids containing 14α-OH [15,16].Conversely, only certain plants and amphibians can yield 14β-hydroxylated sterols in small quantities [17].Fortunately, 14β-OH steroids can be obtained using the facile conversion of 14α-OH counterparts through chemical methods, enhancing the significance of investigating the biocatalytic synthesis of 14α-OH steroids.
Although many species can produce steroids with 14α-OH, research on the identification and characterization of these hydroxylation genes has been few reported.To our knowledge, 14α-hydroxylases have been obtained from Cochliobolus lunatus (anamorph Curvularia lunatus) [15,18,19], Bipolaris sp.[20], Thamnidum elegans [21], Crocus sativus, and Bufo toadstool [17].In contrast, the 14β-hydroxylases remain elusive.Although Crocus sativus and Bufo toadstool can produce 14β-hydroxylated steroids, the results of transcriptome data analysis and P450 gene screening experiments only proved the existence of 14α hydroxylation genes [17].This fact supports the hypothesis that nature exhibits a preference for generating 14α-OH, subsequently undergoing conformational conversion to 14β-OH through processes such as dehydration and hydration.
Two 14α-hydroxylases have been obtained from Cochliobolus lunatus: P-450 lun (from C. lunatus ATCCTM12017, equivalent to CYP103168 from C. lunatus CECT 2130) [15,19] and CYP14A (from C. lunatus CGMCC 3.3589 and C. lunatus JTU 2.406) [18].The corresponding electron transport protein, CPR lun , was identified using transcriptome analysis and genetic screening.P-450 lun and CYP14A, with approximately 82% sequence identity, all have low C14-hydroxylation specificity for C17-substituted steroids, such as progesterone (PROG), when co-expressed with CPR lun in Saccharomyces cerevisiae [15,18].Specifically, P-450 lun converts PROG into a mixture of products with 14α-OH and 11β-OH at a 1:1 ratio [18].PROG is the fundamental steroidal core for synthesizing C21 steroid derivatives of pharmaceutical significance.14-OH-PROG is a crucial intermediate in the chemical and biological synthesis of active steroids, such as cardenolides [10,22].Therefore, we performed engineering of P-450 lun to enhance its C14 regioselective hydroxylation of PROG in this study.
Yeasts are the primary choice for the heterologous expression of membrane-bound eukaryotic cytochrome P450 (CYP) because the composition and structure of the endoplasmic reticulum membrane of yeast cells allow the anchoring of membrane-bound proteins [23,24].The reported 14α-hydroxylases such as CYP11411, CYP44476, P-450 lun and CYP14A all employed S. cerevisiae as the host for investigating the transformation of steroid substrates [17].However, this expression system also comes with some drawbacks, including the unexpected formation of by-products, altered protein glycosylation, challenges in predicting plasma membrane permeability for specific compounds, and unpredictable targeting of recombinant proteins [25].Moreover, the burdensome operation and slow growth rate of yeasts limit the size and screening speed of mutant libraries for the directed evolution of enzymes with low activity and selectivity [23,26].Here, we propose using E. coli as an expression system because of its successful use in the heterologous expression of mammalian or fungal CYPs and operational convenience during enzyme directed evolution.
In this study, we report the functionally heterologous co-expression of N-terminally modified P-450 lun and CPR lun in E. coli.We also engineered the truncated P-450 lun , resulting in a double-site mutant (E81A/F269W) with enhanced regioselectivity and activity for the C14α hydroxylation of PROG.Finally, the whole-cell catalysis was optimized, and the production of 14α-OH-PROG was further increased.This is the first report of producing 14α-OH-PROG in E. coli with a fungal 14α-hydroxylase.

Identification of a Functional Truncated Form of P-450 lun (∆P-450 lun )
The P-450 lun and CPR lun were chemically synthesized and codon-optimized for expression in E. coli.As a redox partner, Flavin adenine dinucleotide (FAD) and flavin mononucleotide (FMN) containing CPR lun mediates electron transfer and is essential for P-450 lun catalytic activity [14,27,28].Class II P450 P-450 lun and its partners CPR lun can be typically co-expressed in a vector system using one or two "promoter-ORF(s)-terminator" expression cassettes [29].With the single cassette approach, P-450 lun and CPR lun can be spaced apart by a sequence containing the ribosome entry/binding site, and two distinct proteins will be expressed.The expression of the P-450 lun and CPR lun will be heterogeneous, with the upstream gene displaying higher expression [30,31].In addition, P-450 lun and CPR lun can also be linked by a nucleotide sequence (can be translated into a flexible or rigid linker) and expressed as a fusion protein.However, the nucleotide length and amino acid composition can severely affect the activity of the fusion protein [32][33][34].To avoid variability in expression efficiency and the selection of complex linkers, plasmid pRSFDuet with two expression cassettes was applied to express P-450 lun and CPR lun .
As membrane-bound proteins, eukaryotic P-450 lun and its cognate partner CPR lun possess N-terminal hydrophobic domains (residues 2-29 and 2-31, respectively) that anchor to the plasma membrane.However, E. coli lacks membrane-bound organelles, potentially leading to inclusion body formation or low protein expression due to its poor recognition ability of unique signaling sequences when expressing membrane-bound proteins [26,35].Fortunately, these limitations can be addressed by N-terminal modifications, including amino acid substitutions and truncations of hydrophilic sequences [26,36,37].
Firstly, we produced a recombinant E. coli BL21 (DE3) strain containing a modified P-450 lun (by replacing the second codon with alanine or substituting the N-terminal region with the hydrophilic sequences MAKKTSS or MALLLAVFL) [38][39][40] and ∆CPR lun (containing a truncated transmembrane region), while no detectable by-products of PROG were produced during whole-cell biotransformation.Then, transmembrane helices truncated P-450 lun and CPR lun (∆P-450 lun and ∆CPR lun , respectively) were introduced into E. coli BL21 (DE3), and detected products were generated (Figure 1A).High-performance liquid chromatography (HPLC) analyses revealed that PROG was converted into two major compounds and several uncharacterized hydroxylated by-products.Isolation, purification, and structural analysis using nuclear magnetic resonance (NMR) of the main compounds showed that they were 14α-OH-PROG and 11β-OH-PROG, respectively (with ratio of ~2:1) (Figure 1B; see NMR Data and spectrums in the Supporting Information).While the molar ratio of 14α-OH-PROG/11β-OH-PROG was different from that obtained from 1:1 in S. Cerevisiae, this could be the result of different culture conditions such as the pH, incubation time, and substrate dose [18].Therefore, the N-terminally modified P-450 lun and CPR lun from C. lunatus were functionally expressed in E. coli.However, the activity and regioselectivity of ∆P-450 lun for C14α position were still low (approximately 21% yield at a concentration of 3.5 mg/L).
The P-450lun and CPRlun were chemically synthesized and codon-optimized for expression in E. coli.As a redox partner, Flavin adenine dinucleotide (FAD) and flavin mononucleotide (FMN) containing CPRlun mediates electron transfer and is essential for P-450lun catalytic activity [14,27,28].Class II P450 P-450lun and its partners CPRlun can be typically co-expressed in a vector system using one or two "promoter-ORF(s)-terminator" expression cassettes [29].With the single cassette approach, P-450lun and CPRlun can be spaced apart by a sequence containing the ribosome entry/binding site, and two distinct proteins will be expressed.The expression of the P-450lun and CPRlun will be heterogeneous, with the upstream gene displaying higher expression [30,31].In addition, P-450lun and CPRlun can also be linked by a nucleotide sequence (can be translated into a flexible or rigid linker) and expressed as a fusion protein.However, the nucleotide length and amino acid composition can severely affect the activity of the fusion protein [32][33][34].To avoid variability in expression efficiency and the selection of complex linkers, plasmid pRSFDuet with two expression cassettes was applied to express P-450lun and CPRlun.
As membrane-bound proteins, eukaryotic P-450lun and its cognate partner CPRlun possess N-terminal hydrophobic domains (residues 2-29 and 2-31, respectively) that anchor to the plasma membrane.However, E. coli lacks membrane-bound organelles, potentially leading to inclusion body formation or low protein expression due to its poor recognition ability of unique signaling sequences when expressing membrane-bound proteins [26,35].Fortunately, these limitations can be addressed by N-terminal modifications, including amino acid substitutions and truncations of hydrophilic sequences [26,36,37].
Firstly, we produced a recombinant E. coli BL21 (DE3) strain containing a modified P-450lun (by replacing the second codon with alanine or substituting the N-terminal region with the hydrophilic sequences MAKKTSS or MALLLAVFL) [38][39][40] and ΔCPRlun (containing a truncated transmembrane region), while no detectable by-products of PROG were produced during whole-cell biotransformation.Then, transmembrane helices truncated P-450lun and CPRlun (ΔP-450lun and ΔCPRlun, respectively) were introduced into E. coli BL21 (DE3), and detected products were generated (Figure 1A).High-performance liquid chromatography (HPLC) analyses revealed that PROG was converted into two major compounds and several uncharacterized hydroxylated by-products.Isolation, purification, and structural analysis using nuclear magnetic resonance (NMR) of the main compounds showed that they were 14α-OH-PROG and 11β-OH-PROG, respectively (with ratio of ~2:1) (Figure 1B; see NMR Data and spectrums in the Supporting Information).While the molar ratio of 14α-OH-PROG/11β-OH-PROG was different from that obtained from 1:1 in S. Cerevisiae, this could be the result of different culture conditions such as the pH, incubation time, and substrate dose [18].Therefore, the N-terminally modified P-450lun and CPRlun from C. lunatus were functionally expressed in E. coli.However, the activity and regioselectivity of ΔP-450lun for C14α position were still low (approximately 21% yield at a concentration of 3.5 mg/L).

Engineering ∆P-450 lun to Improve Its Regiospecificity and Catalytic Activity
In pursuit of refining the regiospecificity of ∆P-450 lun at C-14α of PROG, a hybrid approach involving a semi-rational design and directed evolution was employed.Given that the crystal structure of P-450 lun is not available, a three-dimensional model was constructed using SWISS-MODEL [41] based on the crystal structure of CYP3A4 (PDB ID 5VCD, identity: 28%; similarity: 46%) [42].Although the sequence identity is not high, they possess a comparable overall fold and a conserved heme-binding core structure [43].To enhance model accuracy, GalaxyRefine [44,45] and Gromacs [46] were employed for side-chain repacking and global structural relaxation, resulting in an optimized model with an initial Z-score [47] of −8.25 (Figure S1) and more than 91.0% of residues located in the most favored regions of the Ramachandran plot [48,49] (Figure S2).
Molecular docking of PROG into the presumed structure of ∆P-450 lun was conducted to identify residues potentially influencing selectivity.The analysis of docking poses with the lowest binding free energy revealed two distinct binding modes (pose I and II), corresponding to 14α-and 11β-hydroxylation, respectively (Figure 2A).The lower free energy associated with mode II suggests a preference for 11β-OH.In this mode, the fourring structure of PROG assumed a slightly oblique orientation to the heme ring, and its β-side faced the heme group.The C17 acetyl group of PROG was close to the I-helix and formed hydrogen bonds with His121 and Ser294.While in pose I, the ligand rotated its plane by approximately 180 • , so the α-side faced the heme group.This binding mode was stabilized by a hydrogen bond between the C20-keto group of PROG and Thr369 and the hydrophobic interaction of the gonane ring structure with residues Phe297, Ala298, and Met365 (Figure 2B).
Molecular docking also showed the presence of eight residues (P108, E109, L122, S294.F297, T364, S367, and F368) within 5 Å of the ligands in both binding modes (Figure 2C).P108, E109, and L122 were located in substrate recognition site 1 (SRS-1).Residues S294 and F297 were located in the I-helix near the conserved sequence motif AGXXT (325-329), which was involved in selectivity.Residues T364, S367, and F368 were located at positions +5, +8, and +9 in the highly conserved EXXR motif in SRS-5.Given their proximity to the heme group, these three residues had a high potential for substrate binding and control of regioselectivity [50].We hypothesized that these eight residues contributed to the selectivity or activity of ∆P-450 lun and that replacing these amino acids could improve the activity or selectivity of hydroxylation at C14α.
Alanine scanning was performed across the eight residues to identify the key residues with a remarkable substantial effect on region-selectivity or activity.The substitutions E109A and F297A significantly improved the selectivity from 28% in the wild type to 57% and 73%, respectively.E109A also increased the PROG conversion from 75% to 83%.Although F297A decreased the conversion to 8%, only a trace amount of by-products was detected (Figure 2D).Motivated by these findings, E109 and F297 were selected for further analyses.We performed site saturation mutagenesis in these two residues using NNK codons in 24-well plates.To achieve a 95% coverage of all possible library variants, approximately 60 variants were screened in each single-site library.HPLC results showed that all screened E109 substitutions displayed similar or decreased PROG conversion and selectivity compared to E109A.Conversely, F297W substitution had the highest PROG conversion (41%) and selectivity (93%).To further enhance the regioselectivity and activity of ∆P-450 lun , we combined these two substitutions to obtain a double mutant E109A/F297W, which had higher selectivity (approximately 97%) and catalytic activity (Figure 3A,B; Table 1).conversion (41%) and selectivity (93%).To further enhance the regioselectivity and activity of ΔP-450lun, we combined these two substitutions to obtain a double mutant E109A/F297W, which had higher selectivity (approximately 97%) and catalytic activity (Figure 3A,B; Table 1).conversion (41%) and selectivity (93%).To further enhance the regioselectivity and activity of ΔP-450lun, we combined these two substitutions to obtain a double mutant E109A/F297W, which had higher selectivity (approximately 97%) and catalytic activity (Figure 3A,B; Table 1).

Computational Analysis of ∆P-450 lun and Its Variants
Structural and computational analyses were performed to understand the molecular basis for the selectivity and activity difference between ∆P-450 lun and its mutants for PROG.The binding modes of PROG in ∆P-450 lun were identical to PROG-bound CYP260A1, which can catalyze hydroxylation of PROG at C1α or C17α [51].Molecular docking revealed that the introduction of the larger indole group in F297W reduced the binding pocket volume of ∆P-450 lun by 168.5 Å 3 and brought in a significant steric hindrance against the substrate in pose I (Figure S7 and Figure 4C).The shift in selectivity might arise from the destabilization of the alternative binding mode due to the steric hindrance and hydrogen bond disruption caused by F297W.

Optimizing the Conditions of Whole-Cell Biocatalyst
To establish the optimal conditions for whole-cell biocatalysis, various parameters of the biotransformation process, including temperature, pH, and the additives, were systematically evaluated.A preliminary analysis showed the highest yield of 14α-OH-PROG obtained by ΔP-450lunE109A/F297W. Therefore, we used the E. coli strain containing pRS- Access tunnels facilitate the movement of ligands between the active site and solvent environment, particularly in enzymes with a buried active site [52,53].Using the Caver3.0server [54], we identified two distinctive access tunnels in ∆P-450 lun with features shared by cytochrome P450s [55] (Figure 4A).Tunnel 1, located between the F' helix, B-C loop, and β1 sheet, behaved as a 2a-like channel.In contrast, Tunnel 2, regressed through the B-C loop, could be merged and behaved as a 2e-like channel.Both channels were located in the P450 region involved in substrate specificity.
Detailed analysis revealed that the trajectories of these two channels converged at a node near the heme moiety.Residue E109, situated at the intersection of these two channels, might act as a gate controlling the spatial size of the node.E109 also could form an ionic bond with K489, serving as a critical bottleneck in access Tunnel 1.Therefore, substituting E109 with alanine widened both channels (bottleneck radius increased from 1.36 Å to 1.94 Å for Tunnel 1, from 1.23 Å to 1.50 Å for Tunnel 2) (Figure 4A,B).This modification might enhance the channel's suitability for substrate and water transport, consequently elevating the overall activity of ∆P-450 lun .Moreover, changes in the size, physicochemical properties, or dynamics of the access tunnels of an enzyme can affect its regioselectivity and catalytic activity.Cheng et al. found that a single residue mutation in the access tunnel of PpNHase or CtNHase could invert its regioselectivity toward aliphatic α,ωdinitriles [56].Meng et al. replaced the bottleneck residue F79 in the access tunnel of P450 Bsβ HI with relatively small amino acid alanine resulting the change of hydroxylation products' distribution toward myristic acid and pentadecanoic acid [57].So, substituting residue E109, which was located at the end of the access tunnel and the entrance to the binding pocket, with Alanine might contribute to the difference in the size and shape of the substrate access tunnel, and thus resulted in the diversity of reaction regioselectivity [56].

Optimizing the Conditions of Whole-Cell Biocatalyst
To establish the optimal conditions for whole-cell biocatalysis, various parameters of the biotransformation process, including temperature, pH, and the additives, were systematically evaluated.A preliminary analysis showed the highest yield of 14α-OH-PROG obtained by ∆P-450 lun E109A/F297W.Therefore, we used the E. coli strain containing pRSFDuet_∆P-450 lun E109A/F297W_∆CPR lun to analyze the effects of different reaction conditions on 14α-OH-PROG yield.Results showed that the optimal reaction temperature and pH were 30 • C and 7.4 (50 mM phosphate buffer), respectively (Figure 5A,B).Subsequent reaction variables were all evaluated under these conditions.
Since P450s are heme-containing enzymes, enhancing the intracellular concentrations of heme can potentially enhance the activity of whole-cell biocatalysts.We investigated the impact of adding different heme precursors, such as hemin, 5-ALA, and FeSO 4 •7H 2 O during the expression of P-450 lun [58].Given the limited import of heme into the BL21 (DE3) strain [59], adding heme did not significantly increase 14α-OH-PROG yield.However, supplementation with 5-ALA demonstrated an improvement, achieving the highest yield of 12.7 mg/L when 0.5 mM ALA was added (Figure 5C).In contrast, only trace amount of products were observed in the absence of 5-ALA during the induction of P-450 lun , indicating that E. coli relied mainly on exogenous 5-ALA for the biosynthesis of heme.Furthermore, the inclusion of iron might also contribute to the biosynthesis of heme.Adding 0.5 mM ALA to the medium and varying final concentrations of FeSO 4 •7H 2 O (5, 10, 20, 30, and 40 mg/L) were introduced, while no noticeable change in 14α-OH-PROG yield was observed.
As a member of class II P450s, P-450 lun requires two electrons from NADPH delivered by CPR lun to increase its catalytic activity [14,27].Thus, the nicotinamide cofactor may also be a limitation in current whole-cell systems.Various concentrations of NADPH (1, 1.5, or 2 equivalents) were introduced to the biotransformation system, and the production of 14α-OH-PROG was assessed at different time points (Figure 5D).The results were compared with the control reaction (without supplemented NADPH).The findings revealed that additional NADPH increased the initial rates and conversions, suggesting that the regeneration of the cofactor in cells was insufficient to sustain the catalytic reaction at the achieved rates and with the current coupling efficiency.Growth was interrupted when 1.5 equivalents of NADPH were added to the reaction system, possibly due to the low protein expression levels of P-450 lun or CPR lun .These results indicated that introducing the enzymatic regeneration systems (such as GDH/glucose) in the host as a substitute for nicotinamide cofactors may represent an economical approach to produce 14α-OH-PROG in this system.Following the analysis and optimization of reaction conditions, the highest conversion of PROG was achieved at 99%, corresponding to a 14α-OH-PROG yield of 16.0 mg/L.be a limitation in current whole-cell systems.Various concentrations of NADPH (1, 1.5, or 2 equivalents) were introduced to the biotransformation system, and the production of 14α-OH-PROG was assessed at different time points (Figure 5D).The results were compared with the control reaction (without supplemented NADPH).The findings revealed that additional NADPH increased the initial rates and conversions, suggesting that the regeneration of the cofactor in cells was insufficient to sustain the catalytic reaction at the achieved rates and with the current coupling efficiency.Growth was interrupted when 1.5 equivalents of NADPH were added to the reaction system, possibly due to the low protein expression levels of P-450lun or CPRlun.These results indicated that introducing the enzymatic regeneration systems (such as GDH/glucose) in the host as a substitute for nicotinamide cofactors may represent an economical approach to produce 14α-OH-PROG in this system.Following the analysis and optimization of reaction conditions, the highest conversion of PROG was achieved at 99%, corresponding to a 14α-OH-PROG yield of 16.0 mg/L.Thereafter, the effect of substrate concentration (ranging from 25 to 200 µM) was investigated using those optimized conditions mentioned above.The yield of 14α-OH-PROG was estimated after 8 h of transformation in the Erlenmeyer flasks.Results suggested that the maximum titer of 14α-OH-PROG was found at 50 µM substrate (Figure 6).At higher substrate concentrations (>50 µM), the accumulation of 14α-OH-PROG stopped or even showed a downward trend, which may result from the limited protein expression Thereafter, the effect of substrate concentration (ranging from 25 to 200 µM) was investigated using those optimized conditions mentioned above.The yield of 14α-OH-PROG was estimated after 8 h of transformation in the Erlenmeyer flasks.Results suggested that the maximum titer of 14α-OH-PROG was found at 50 µM substrate (Figure 6).At higher substrate concentrations (>50 µM), the accumulation of 14α-OH-PROG stopped or even showed a downward trend, which may result from the limited protein expression levels, poor substrates/products transport capacity of E. coli or substrates/products inhibition.
Catalysts 2024, 14, x FOR PEER REVIEW 9 of 14 levels, poor substrates/products transport capacity of E. coli or substrates/products inhibition.3. Materials and Methods.

Strains and Reagents
E. coli DH5α was used for DNA cloning and plasmid construction.E. coli BL21 (DE3) served as hosts for whole-cell biotransformation.Yeast extract and tryptone were purchased from Qxoid (Basingstoke, UK).Q5 Hot Start High-Fidelity DNA polymerase and restriction endonucleases were obtained from Thermo Scientific (Waltham, MA, USA).The plasmid purification kit was sourced from Omega Laboratories (Cleveland, OH, USA).Oligonucleotide synthesis and sequence analysis were performed by Novogene (Tianjin, China).PROG, 5-ALA and isopropy1 β-D-1-thiogalactopyranoside (IPTG) were purchased from Energy Chemical (Shanghai, China).Other chemicals were obtained from Genview (Beijing, China) with the highest available commercial grade.

Plasmid Construction and Protein Expression
The wild-type gene P-450lun (GenBank accession numbers: MN061487) and CPRlun (GenBank accession numbers: MN061485) were chemically synthesized and optimized considering the codon preferences of E. coli by Novogene (Tianjin, China).The expression construct for the normal or modified P-450lun and CPRlun were designed in pRSFDuet, respectively, using primers presented in Table S1.Sequences confirmation was performed by sequencing in Novegene (Tianjin, China).The recombinant cells transformed with expression vectors (pRSFDuet) harboring the normal or modified P-450lun gene and CPRlun gene were cultivated at 37 °C in Luria-Bertani (LB) medium with 50 µg/mL kanamycin.In total, 0.5 mM IPTG and 0.5 mM 5-ALA were added When cells' OD600 reached 0.6, and they continued to cultivate for 16 h at 28 °C.Then, cells were collected using centrifugation (10, 000g, 10 min, 4 °C) and resuspended (in 50 mM pH 7.4 potassium phosphate buffer) for the whole-cell biotransformation.

Whole-Cell Conversion of Progesterone Using E. coli in the Erlenmeyer flask
In total, 10 mL potassium phosphate buffer (50 mM, pH 7.4) with 50 g cell wet weight/L (cww/L) recombinant E. coli and 50 µM PROG was shaken at 250 rpm and 30 °C for 8 h.In total, 2 mL samples were taken and then extracted with equal volume ethyl acetate, dried with air-blowing, re-dissolved in 200 µL chromatography methanol, and

Strains and Reagents
E. coli DH5α was used for DNA cloning and plasmid construction.E. coli BL21 (DE3) served as hosts for whole-cell biotransformation.Yeast extract and tryptone were purchased from Qxoid (Basingstoke, UK).Q5 Hot Start High-Fidelity DNA polymerase and restriction endonucleases were obtained from Thermo Scientific (Waltham, MA, USA).The plasmid purification kit was sourced from Omega Laboratories (Cleveland, OH, USA).Oligonucleotide synthesis and sequence analysis were performed by Novogene (Tianjin, China).PROG, 5-ALA and isopropy1 β-D-1-thiogalactopyranoside (IPTG) were purchased from Energy Chemical (Shanghai, China).Other chemicals were obtained from Genview (Beijing, China) with the highest available commercial grade.

Plasmid Construction and Protein Expression
The wild-type gene P-450 lun (GenBank accession numbers: MN061487) and CPR lun (GenBank accession numbers: MN061485) were chemically synthesized and optimized considering the codon preferences of E. coli by Novogene (Tianjin, China).The expression construct for the normal or modified P-450 lun and CPR lun were designed in pRSFDuet, respectively, using primers presented in Table S1.Sequences confirmation was performed by sequencing in Novegene (Tianjin, China).The recombinant cells transformed with expression vectors (pRSFDuet) harboring the normal or modified P-450 lun gene and CPR lun gene were cultivated at 37 • C in Luria-Bertani (LB) medium with 50 µg/mL kanamycin.In total, 0.5 mM IPTG and 0.5 mM 5-ALA were added When cells' OD 600 reached 0.6, and they continued to cultivate for 16 h at 28 • C.Then, cells were collected using centrifugation (10,000× g, 10 min, 4 • C) and resuspended (in 50 mM pH 7.4 potassium phosphate buffer) for the whole-cell biotransformation.

Whole-Cell Conversion of Progesterone Using E. coli in the Erlenmeyer flask
In total, 10 mL potassium phosphate buffer (50 mM, pH 7.4) with 50 g cell wet weight/L (cww/L) recombinant E. coli and 50 µM PROG was shaken at 250 rpm and 30 • C for 8 h.In total, 2 mL samples were taken and then extracted with equal volume ethyl acetate, dried with air-blowing, re-dissolved in 200 µL chromatography methanol, and filtered through a 0.2 µm filter.Product formation was analyzed using reverse-phase HPLC to calculate the final conversions.The reaction mixture was separated by a Thermo Hypersil Gold C18 Column (4.6 × 250 mm, 5 µm) using solvent A (ddH 2 O with 0.1% v/v formic acid) and solvent B (acetonitrile).The flow rate was 1 mL/min.The gradient was set as 0-15 min 30% B-95% B, 15-20 min 95% B, 20-21 min 95% B-30% B, 25 min 30% B.

Engineering of P-450 lun for Improved Regioselectivity and Activity
Single-site mutation was introduced in P-450 lun with PCR using designed primers (Table S1) and Q5 hot-start high-fidelity DNA polymerase.The PCR reaction mixture (25 µL) was prepared, which included primers (1.25 µL, 10 µM), template plasmid (1 µL, 50 ng/µL), Q5 hot-start high-fidelity DNA polymerase (12.5 µL), and dd H 2 O (9 µL).The PCR product was digested by restriction enzyme Nhel/BspTI for 1.5 h at 37 • C and then connected into corresponding restriction sites of pRSFDuet_∆CPR lun using T4 DNA ligase (Thermo Scientific, Waltham, MA, USA).The final ligated products were transformed into E. coli BL21 (DE3) via electroporation and plated on LB agar plates containing 50 µg/mL kanamycin.Mutant colonies were inoculated in separate wells of 24-deep-well plates (0.2 mL, LB medium with 50 µg/mL kanamycin per well) and incubated at 300 rpm and 37 • C for 8 h.Then, 1.0 mL TB medium containing 50 µg/mL kanamycin, 0.5 mM IPTG, and 0.5 M 5-ALA was added to each well, and the plates continued to be incubated at 300 rpm and 28 • C for another 16 h.The cells were harvested by centrifuging the plates at 4000× g for 20 min, washed with 0.2 mL potassium phosphate buffer (50 mM, pH 7.4), and centrifuged again.Cells in each well were then resuspended in 0.5 mL of 50 mM pH 7.4 potassium phosphate buffer containing 50 µM PROG.The plates were shaken at 300 rpm and 30 • C for 12 h.Each sample was extracted with 500 µL ethyl acetate, dried with air-blowing, re-dissolved in 100 µL chromatography methanol, and product formation was analyzed using reverse-phase HPLC (described above).Desired mutations were confirmed using DNA sequencing.

Structural Modeling Analysis, Molecular Docking and Tunnel Analysis
Swiss-Model server (https://swissmodel.expasy.org/,accessed on 10 February 2023) was used to build a 3D structure model of ∆P-450 lun and its variants.To enhance the geometry and energy of the overall predicted model, energy minimization was conducted using the steepest gradient minimization algorithm, employing specific GROMACS routines.This involved performing 1500 minimization steps in a vacuum environment, with a maximum force convergence threshold set at 1.0 kJ/mol/nm.Additionally, a cut-off range of 1.4 nm was applied for both van der Waals and Coulomb interactions.Then, the optimized protein structure was refined using the GalaxyWeb server (https://galaxy.seoklab.org/index.html,accessed on 11 February 2023).The quality of the final generated model was assessed using a Ramachandran plot (SAVES server, https://saves.mbi.ucla.edu/,accessed on 12 February 2023) and Z-score (ProSA web, https://prosa.services.came.sbg.ac.at/prosa.php,accessed on 12 February 2023).Structural images were produced using PyMOL and Discovery studio visualizer 2019.
The molecular structure of the ligand PROG was constructed and optimized using ChemDraw.A grid box with dimensions of 5 Å surrounding the active site was generated, centered on the heme iron of P-450 lun or its variants.Molecular docking simulations were conducted using Autodock 4.2 and subsequently visualized using PyMOL.Each variant underwent 20 docking runs, and the resulting poses were clustered based on a Root Mean Square Deviation (RMSD) cutoff of 5 Å, employing default parameters.
Tunnel analysis was performed using Caver Analyst 2.0 (https://www.caver.cz/index.php?sid=121, accessed on 17 May 2023), with the following parameters: a probe radius of 1.4 Å, shell depth of 4 Å, shell radius of 3 Å, clustering threshold of 3.5, and the starting point set at the Fe atom of the heme co-factor.Following the tunnel calculations, the bottleneck radius and tunnel-lining residues for each tunnel were extracted from the tunnel statistics and residue graph functions.

Figure 2 .
Figure 2. (A) Superposition of the results from docking PROG in the homology model of ΔP-450lun.Yellow: docking pose I for the hydroxylation at C14α of PROG; purple: Docking pose II for the hydroxylation at C11β of PROG.(B) Interactions between PROG and ΔP-450lun in pose I and pose II.Analyzed using Discovery Studio Visualizer.Green dash: conventional hydrogen.Pink dash: hydrophobic (Pi-Alkyl).Residue numbering in bold corresponds to the long form of P-450lun and the circles to ΔP-450lun.(C) Superposition of selected amino acids for Alanine-scanning mutagenesis.(D) Whole-cell biocatalysis results of different Alanine substitution mutants in E. coli.Reaction conditions: 50 g cww/L recombinant E. coli cells were suspended in phosphate buffer (pH 7.4, 50 mM) containing 50 µM PROG; reactions were conducted at 28 °C, 250 rpm for 8 h.Error bars are standard deviation for three independent experiments.

Figure 3 .
Figure 3. Bioconversion of PROG using WT P-450lun and its mutants.(A) HPLC assay of PROG biotransformation using E. coli cells containing pRSFDuet-ΔP-450lun-ΔCPRlun and positive mutants of ΔP-450lun.(B) Distribution of the products of PROG using WT and corresponding mutants (E109A, F297A, F297W and E109A/F297W).Reaction conditions: 50 g cww/L recombinant E. coli cells were suspended in phosphate buffer (pH 7.4, 50 mM) containing 50 µM PROG; reactions were conducted at 28 °C, 250 rpm for 12 h.Error bars are standard deviation for three independent experiments.

Figure 2 .
Figure 2. (A) Superposition of the results from docking PROG in the homology model of ∆P-450 lun .Yellow: docking pose I for the hydroxylation at C14α of PROG; purple: Docking pose II for the hydroxylation at C11β of PROG.(B) Interactions between PROG and ∆P-450 lun in pose I and pose II.Analyzed using Discovery Studio Visualizer.Green dash: conventional hydrogen.Pink dash: hydrophobic (Pi-Alkyl).Residue numbering in bold corresponds to the long form of P-450 lun and the circles to ∆P-450 lun .(C) Superposition of selected amino acids for Alanine-scanning mutagenesis.(D) Whole-cell biocatalysis results of different Alanine substitution mutants in E. coli.Reaction conditions: 50 g cww/L recombinant E. coli cells were suspended in phosphate buffer (pH 7.4, 50 mM) containing 50 µM PROG; reactions were conducted at 28 • C, 250 rpm for 8 h.Error bars are standard deviation for three independent experiments.

Figure 2 .
Figure 2. (A) Superposition of the results from docking PROG in the homology model of ΔP-450lun.Yellow: docking pose I for the hydroxylation at C14α of PROG; purple: Docking pose II for the hydroxylation at C11β of PROG.(B) Interactions between PROG and ΔP-450lun in pose I and pose II.Analyzed using Discovery Studio Visualizer.Green dash: conventional hydrogen.Pink dash: hydrophobic (Pi-Alkyl).Residue numbering in bold corresponds to the long form of P-450lun and the circles to ΔP-450lun.(C) Superposition of selected amino acids for Alanine-scanning mutagenesis.(D) Whole-cell biocatalysis results of different Alanine substitution mutants in E. coli.Reaction conditions: 50 g cww/L recombinant E. coli cells were suspended in phosphate buffer (pH 7.4, 50 mM) containing 50 µM PROG; reactions were conducted at 28 °C, 250 rpm for 8 h.Error bars are standard deviation for three independent experiments.

Figure 3 .
Figure 3. Bioconversion of PROG using WT P-450lun and its mutants.(A) HPLC assay of PROG biotransformation using E. coli cells containing pRSFDuet-ΔP-450lun-ΔCPRlun and positive mutants of ΔP-450lun.(B) Distribution of the products of PROG using WT and corresponding mutants (E109A, F297A, F297W and E109A/F297W).Reaction conditions: 50 g cww/L recombinant E. coli cells were suspended in phosphate buffer (pH 7.4, 50 mM) containing 50 µM PROG; reactions were conducted at 28 °C, 250 rpm for 12 h.Error bars are standard deviation for three independent experiments.

Figure 3 .
Figure 3. Bioconversion of PROG using WT P-450 lun and its mutants.(A) HPLC assay of PROG biotransformation using E. coli cells containing pRSFDuet-∆P-450 lun -∆CPR lun and positive mutants of ∆P-450 lun .(B) Distribution of the products of PROG using WT and corresponding mutants (E109A, F297A, F297W and E109A/F297W).Reaction conditions: 50 g cww/L recombinant E. coli cells were suspended in phosphate buffer (pH 7.4, 50 mM) containing 50 µM PROG; reactions were conducted at 28 • C, 250 rpm for 12 h.Error bars are standard deviation for three independent experiments.

Figure 4 .
Figure 4. Channels and structural analyses of ΔP-450lun and its positive variants.(A) The trajectories of 2a-like channel and 2e-like channel in ΔP-450lun.The surface models of 2a-like channel and 2elike channel are colored in purple and yellow, respectively.The adjacent secondary structures are colored in grey.The important residues Glu109 and Lys489 near the node are colored in green, and heme is colored in rose.(B) Superposition of the residues 109 and Lys489 in the model of ΔP-450lun (grey) and E109A variant (orange).The E109A substitution broke the ionic bond between Glu109 and Lys489, resulting in the increase in distance between residues 109 and Lys489.(C) Structural comparison of ΔP-450lun and mutants F297W.Surface of Phe297 (green) and Trp297 (orange) are shown as yellow and pink, respectively.In the WT structure, PROG could be docked in the Pose II, resulting in hydroxylation at C11β.F269W substitution might hinder the formation of this binding pose.

Figure 4 .
Figure 4. Channels and structural analyses of ∆P-450 lun and its positive variants.(A) The trajectories of 2a-like channel and 2e-like channel in ∆P-450 lun .The surface models of 2a-like channel and 2e-like channel are colored in purple and yellow, respectively.The adjacent secondary structures are colored in grey.The important residues Glu109 and Lys489 near the node are colored in green, and heme is colored in rose.(B) Superposition of the residues 109 and Lys489 in the model of ∆P-450 lun (grey) and E109A variant (orange).The E109A substitution broke the ionic bond between Glu109 and Lys489, resulting in the increase in distance between residues 109 and Lys489.(C) Structural comparison of ∆P-450 lun and mutants F297W.Surface of Phe297 (green) and Trp297 (orange) are shown as yellow and pink, respectively.In the WT structure, PROG could be docked in the Pose II, resulting in hydroxylation at C11β.F269W substitution might hinder the formation of this binding pose.

Figure 5 .
Figure 5.Effect of reaction conditions on the production of 14α-OH PROG in the whole-cell biocatalysis.(A,B) The effects of reaction temperature and pH on the 14α-OH PROG production.(C) The effects of different final concentration of 5-ALA during the expression of P-450lun on the 14α-OH PROG production.(D) Time course of the biotransformation with different NADPH additions.Reaction conditions: 50 g cww/L recombinant E. coli cells were suspended in phosphate buffer (pH 7.4, 50 mM) containing 50 µM PROG; reactions were conducted at 30 °C, 250 rpm for 16 h.Error bars are standard deviation for three independent experiments.

Figure 5 .
Figure 5.Effect of reaction conditions on the production of 14α-OH PROG in the whole-cell biocatalysis.(A,B) The effects of reaction temperature and pH on the 14α-OH PROG production.(C) The effects of different final concentration of 5-ALA during the expression of P-450 lun on the 14α-OH PROG production.(D) Time course of the biotransformation with different NADPH additions.Reaction conditions: 50 g cww/L recombinant E. coli cells were suspended in phosphate buffer (pH 7.4, 50 mM) containing 50 µM PROG; reactions were conducted at 30 • C, 250 rpm for 16 h.Error bars are standard deviation for three independent experiments.

Figure 6 .
Figure 6.Effect of substrate concentration on the production of 14α-OH PROG in the whole-cell biocatalysis.Reaction conditions: 50 g cww/L recombinant E. coli cells were suspended in phosphate buffer (pH 7.4, 50 mM) containing 75 mM NADPH and varying concentration of PROG (dissolved in DMF, final 5%); reactions were conducted at 30 °C, 250 rpm for 8 h.Error bars are standard deviation for three independent experiments.

Figure 6 .
Figure 6.Effect of substrate concentration on the production of 14α-OH PROG in the whole-cell biocatalysis.Reaction conditions: 50 g cww/L recombinant E. coli cells were suspended in phosphate buffer (pH 7.4, 50 mM) containing 75 mM NADPH and varying concentration of PROG (dissolved in DMF, final 5%); reactions were conducted at 30 • C, 250 rpm for 8 h.Error bars are standard deviation for three independent experiments.

Table 1 .
Conversion rate of PROG hydroxylation and product distribution of P-450 lun and mutants from the bioconversion assay.