Introduction

The fluorination of organic compounds is a prevalent practice1,2,3, particularly in the field of medicine4,5,6, where fluorine-containing drugs surpass 50% by 20237,8,9. The incorporation of gem-difluoro/monofluoro groups can alter the physical and chemical properties as well as the biological activity of compounds due to the favorable biocompatibility, high electronegativity, and stability of fluorine atoms10,11. Notably, fluorinated amide compounds, such as aryldifluoroamides and monofluoroamides12,13,14, are frequently found in natural products, pharmaceuticals, and biologically active molecules (Fig. 1a). Various strategies have been employed for the synthesis of fluoroamide compounds (Fig. 1b). The utilization of an oxidant in conjunction with for oxidative desulfurization-difluorination presents a method for the synthesis of fluorinated amide compounds15. Nonetheless, these reactions exhibit a restricted scope, and the fluorinating reagents utilized are characterized by high sensitivity, rendering them susceptible to instability and potential explosion reactions16. Subsequently, a variety of transition metal catalysts have been developed for the synthesis of fluorinated amide compounds. These include palladium-catalyzed arylation of α,α-difluoroacetamides16, nickel-catalyzed difluoroalkylation of arylboronic acids17, and nickel-catalyzed tandem difluoroalkylation-arylation of amides18. Additionally, photocatalysis technology has emerged as a crucial component in the synthesis of fluorinated amide compounds. Specifically, under the catalysis of fac-Ir(ppy)3, unactivated (hetero)aromatics and hydrazones can engage in cross-coupling reactions with bromodifluoroacetamide compounds, leading to the formation of difluoroamide compounds19,20. Moreover, the synergistic interaction between thiols and the photocatalyst offers an effective route for the synthesis of difluoroamide compounds21,22. Recently, the introduction of the 4-dimethylaminopyridine-boron radical has opened a novel pathway for the synthesis of fluoroamides by activating the carbonyl oxygen and facilitating a spin-center shift, which promotes the cleavage of the C-F bond in trifluoromethyl groups, thereby enabling selective functionalization23. However, the synthesis of fluoroamide compounds continues to face significant challenges, particularly concerning the generation of by-products and the attainment of both chemical and stereoselectivity. Given the importance of fluorinated amide structures, it is crucial to continue advancing efficient and sustainable methods for their synthesis.

Fig. 1: The significance of fluorinated amide compounds, the development of their synthetic strategies and experimental design in this study.
figure 1

a The fluorinated amide compounds are commonly structural units found in pharmaceuticals. b Strategies for synthesizing fluorinated amides by chemical methods. c Fluorinated reagent F-1 used in this study and enzymatic synthesis of fluoroamide compounds. Under blue light irradiation, fluorinated reagents undergo coupling with alkenes in an enzymatic chiral environment, resulting in fluorinated amide compounds with pronounced distal chirality.

Full size image

Enzyme catalysis, known for its remarkable stereoselectivity and evolvability, has found widespread application in the synthesis of complex chemical compounds that are typically difficult to access using conventional chemical approaches24,25. In nature, biocatalysts capable of catalyzing fluorination reactions are relatively limited11. Fluorinase from Streptomyces cattleya is the only currently known enzyme for incorporating fluorine into organic molecules, which demonstrates the ability to catalyze the formation of 5’-fluorodeoxy adenosine (5’-FDA) from S-adenosyl-l-methionine (SAM)26. To broaden the scope of biocatalytic fluorine chemistry, researchers have successfully employed non-heme iron enzymes to catalyze the synthesis of fluorinated compounds, including radical-mediated fluorine atom transfer for the stereoselective fluorination of C(sp3)-H bond27,28, and the catalytic enantioselective difunctionalization of alkenes via trifluoromethyl radical intermediates29,30. Additionally, enzymes capable of incorporating fluorinated motifs into organic molecules have been utilized in the synthesis of various fluorinated products, such as the synthesis of fluorinated β-amino acids by threonine aldolase31, the synthesis of chiral fluorides by aldolase32, the synthesis of trifluoromethyl-substituted cyclopropane compounds by myoglobin/protoglobin-catalyzed carbene transfer reaction33,34, and the synthesis of α-amino C−H fluoroalkylation compounds by P450 variant35. Furthermore, the advent of photoenzymes has considerably broadened the catalytic repertoire of enzymes36,37, encompassing Csp3-Csp3 cross-electrophile couplings38, oxidative cross-coupling39, and radical acylation40. These reactions indicated the potential for utilizing the activity of photoenzyme-catalyzed intermolecular coupling to introduce fluorine groups into specific sites. Recently, Zhao group has developed a photoenzymatic hydrofluoroalkylation that integrates fluorinated motifs into olefins41. This work advances enzymatic strategies for the introduction of fluorinated groups at the terminal positions of compounds by photoenzymes. It is noteworthy that, besides introducing fluorine groups at the termini of molecules, accurately introducing fluorine groups at specific internal locations within a molecule is also challenging42,43,44.

In this study, we synthesized fluorinated amide compounds with distal chirality (γ-to F) by photoenzymes. Ene-reductases demonstrated new catalytic capabilities under blue light irradiation conditions. A suitable fluorinated reagent (F-1) was designed for the efficient incorporation of fluorine groups at specific internal positions. Five flavin-dependent ene-reductases were screened with F-1 and electron-rich alkenes for coupling reactions (Fig. 1c), with OYE1 selected for its catalytic activity and stereoselectivity. By optimizing the reaction system and performing enzyme engineering, we successfully obtained fluorinated amide compounds with excellent remote chirality (up to 97%). Additionally, the study showed the influence of key amino acids Y196 and Y375 on stereoselectivity and reaction activity. This study not only presents a potential enzymatic approach for synthesizing fluorinated amides but also contributes to the advancement of remote chiral compound synthesis through photoenzymes.

Results

Experimental design

Flavoproteins demonstrated their capability to catalyze a variety of reactions, including hydroalkylation45,46,47,48,49, asymmetric hydroamination50, enantioselective reduction of enamides51, radical cyclization52, hydrosulfonylation53, and lactonization54. The cofactor flavin mononucleotide (FMN) plays a significant role in hydrogen atom transfer mechanisms involved in these reactions. Despite the significant contributions of flavin-dependent enzymes to the field of enzymatic catalysis, their potential applications in the complex domain of fluorine chemistry remain largely underexplored. Current literature indicates a scarcity of research focused on the incorporation of fluorine groups at specific positions within compounds via flavoproteins. We propose that a radical containing fluorine at a designated site can be generated under light excitation. Drawing on prior studies and the presence of fluoroamide structures in certain pharmaceuticals, we have identified a precursor, designated as F-1, which possesses a leaving group and is anticipated to yield a fluoramide radical upon excitation with visible light (Fig. 1c). Five flavin-dependent enzymes, OYE1/2/3 and XenA/B were examined under blue light irradiation (Supplementary Fig. 1 and Fig. 2). Glucose dehydrogenase (GDH), d-(+)-glucose, and NADP+ were added to the reaction system to provide NADPH. The ERs successfully catalyzed the synthesis of the target compounds under blue light, with no product formation observed in the absence of ERs or light conditions (Supplementary Fig. 3). We selected OYE1 for the subsequent experiments.

Fig. 2: Reaction condition optimization and molecular docking for semi-rational design to increase yield and ee values of OYE1.
figure 2

a The optimization of the reaction system, including light intensity, reaction buffer, pH and amount of enzyme. Reaction system: F-1 (10 mM), E-1 (5 mM), NADP+(500 μM), Glucose (50 mM), GDH (5 U/mL), ER (0.5-2 mol%), DMSO and glycerol (10% v/v, respectively). The total volume is 800 μL. #blue LED: 12 W 455 nm. *chiral carbon. b The protein structure was obtained from PDB: 3TX9. The docking results were analyzed using PyMOL. Yellow represents the cofactor FMN. Blue represented the product. White represented the key amino acids. The cartoon of the protein was shown in cyan. The catalytic effects of mutants were investigated. Bold formatting represents mutants with significantly enhanced effects. c Topographic steric maps near the active pockets of OYE1-WT and OYE1-Y375F. Topographic steric maps were generated using the online tool SambVca 2.1, with amino acids near the active site and substrate channel selected for analysis. Source data are provided as a Source Data file.

Full size image

Interestingly, we observed that the inclusion of the amide group yielded a favorable effect on the reaction, leading to increased efficiency and reactivity. Upon conducting the reaction with ethyl bromodifluoroacetate (F-0) under blue light irradiation, no products corresponding to the ERs were identified. Nevertheless, upon substituting the ester group with an amide group, the product formation was successfully confirmed through GC-MS analysis (Supplementary Fig. 4). Our computational studies revealed that the amide substrate displayed stronger binding to the enzyme than the ester substrate (Supplementary Fig. 5), which might provide an explanation for the lack of reactivity or the relatively diminished reactivity of ester substrates.

Reaction optimization

The feasibility of the experimental design was confirmed as the initial stage for the photocatalytic enzymatic synthesis of fluorinated amides. Nevertheless, the results obtained revealed the need for further enhancement of both the reaction yield and stereoselectivity (Fig. 2a).

In the initial stage, precise control over chemical reactions was achieved by adjusting light parameters, transitioning from the original 12 W 455 nm to 45 W 465 nm wavelength range. These optimized conditions led to an increase in yield. Specifically, the yield increased from 15% to ~21%, while the enantiomeric excess (ee) value remained unchanged (93%) (Fig. 2a, entries 1 and 3). We speculated that the slight increase in yield with higher light power might have been attributed to the enhanced excitation of the substrate and the more thorough illumination of the entire reaction system. Subsequently, the impact of sodium phosphate buffer (NaPi) and Tris-HCl reaction buffers at various pH levels was investigated. The results indicated that the reaction performed more effectively in an alkaline buffer, resulting in a yield increase of ~1.7 times (Fig. 2a, entries 3 and 4). Additionally, adjustments were made to the amount of enzyme ER added. As anticipated, elevating the enzyme loading to 2 mol% led to a reaction yield of 51% and an ee value of 93%. Ultimately, the alkaline buffer (Tris-HCl pH = 8.5) was selected, and the enzyme amount (2 mol% ER) in the reaction system was increased for subsequent investigations.

To enhance the reaction yield, compound 1 was employed as the ligand for molecular docking in order to establish a complex with OYE1 for the purpose of identifying crucial residues (Fig. 2b). Subsequently, we performed the site-directed mutagenesis targeting seven key residues in proximity to the ligands: Y375/Y196/F123/W116/Y82/M39/T37. Taking into account steric hindrance and potential interactions, these specific sites were predominantly selected for mutation utilizing the amino acids: alanine (A), serine (S), tyrosine (Y), phenylalanine (F), and tryptophan (W) (Fig. 2b and Supplementary Table 1). OYE1 was rapidly evolved by creating a smaller, high-quality mutant library.

The results revealed that the Y375F mutation led to a yield increase (>90%), while maintaining the stereoselectivity essentially unchanged ( >90%) (Fig. 2b, entry 13 and Supplementary Fig. 6). The computational analysis revealed the Y375F mutant demonstrated an enhanced capacity to bind the radical F-1’ compared to the wild type (WT) (Supplementary Fig. 7c). Furthermore, the SambVca55 tool indicated that the Y375F mutant might experience diminished steric hindrance relative to the WT, which could promote a more efficient transport of the substrate passage to the active site located above the flavin mononucleotide (FMN) (Fig. 2c and Supplementary Fig. 8). Specifically, the T37S mutation significantly improved stereoselectivity (97%). However, this mutation led to a notable decrease in yield (13%) (Fig. 2b, entry 16). Subsequently, saturation mutagenesis was performed at the T37 site in an attempt to enhance the yields. Nevertheless, T37 mutants exhibited either a marginal increase or a loss of catalytic activity (Supplementary Fig. 9). Through a comparative analysis of the yields and stereoselectivities of the mutants, the OYE1-Y375F was identified as the optimal variant.

Substrate scope investigations

Subsequently, we investigated the substrate spectrum of the enzyme. We assessed the impact of substrate size by employing fluorinated reagents with varying chain lengths (F-1-F-6) and a range of alkenes featuring diverse substituents and steric hindrance (E-1-E-10) (Supplementary Fig. 10a, b). Our research compared these substrates using both the OYE1-WT and the OYE1-Y375F mutant (Fig. 3c). Overall, the Y375F mutation proved beneficial in enhancing yields while keeping the ee values relatively stable in comparison to the wild type.

Fig. 3: Substrate scope of photoenzyme catalysis.
figure 3

Reaction conditions: fluorinated reagent: 10 mM, alkene: 5 mM, NADP+: 0.5 mM, glucose: 50 mM, GDH: 0.5 μg/μL, ene-reductase: 2 mol% based on fluorinated reagents, DMSO: 10% (v/v), glycerol: 10% (v/v), Tris-HCl (pH = 8.5): add to 800 μL. Reaction at room temperature for 16 h. †(Z)-But-2-en-2-ylbenzene was used as the alkene. *chiral carbon. The yields were determined by 1H NMR with 1,3,5-trimethoxybenzene as an internal standard. The ee values were calculated using the product peak area from chiral HPLC.

Full size image

The impact of fluorinated radical precursor substrates with varying chain lengths on the reaction was investigated. As the result showed in Fig. 3c, the yield of compounds 1-3 ranges from 91% to 67%, accompanied by slightly reduced enantioselectivities from 95% to 86% ee. Furthermore, the influence of non-aromatic fluorinated reagents in the reaction was also investigated (4). revealing inferior yield for non-aromatic substrates (35%), and a simultaneous reduced stereoselectivity (76% ee).

For alkene substrates, a variety of alkenes with different substituents at various positions were examined (5-14). The experimental results revealed that the substituent groups on the benzene ring (5-9) had minimal effects on the yields (83-99%) and stereoselectivities (90-97% ee). The ortho-substitution reaction of alkene substrates might have a certain impact on product yield for 10 and 11 (60% and 49%), with high stereoselectivity (96% and 89%). Compound 12, with a low yield of 37%, exhibited a notably high stereoselectivity (90%). The excellent diastereoselectivity (dr > 20:1) was observed for 13 with low yield and ee. In addition, the utilization of α-ethyl styrene in the reaction led to decreased yield and stereoselectivity (14). A mutant OYE1-Y375A, showing enhanced stereoselectivity (from 36% to 69%) despite a slight decrease in yield, was identified (Supplementary Fig. 11). We speculated that the improved stereoselectivity of mutant OYE1-Y375A may be attributed to reduced steric hindrance, allowing more space for accommodating the ethyl group.

In this study, we also explored the impact of monofluorinated reagents on the reaction (15, 16). Despite the enzyme’s stereoselectivity in terminal olefins (>90%), it did not exhibit stereoselectivity in specific positions of fluorine atoms (dr ≈ 1:1). To determine the absolute configuration of the products catalyzed by the enzyme, X-ray diffraction experiments were conducted, revealing that product 8 was predominantly the S-isomer (Fig. 3, Supplementary Fig. 12 and Supplementary Table 2 and Table 3). These results suggested that the enzyme had a wide broad substrate spectrum, enabling it to catalyze various fluorinated reagents and alkenes.

Mechanistic investigations

We first investigated whether an EDA complex was formed during the reaction. UV-Vis spectroscopy indicated that no EDA complex was formed in this process (Supplementary Fig. 13). Based on previous research findings, the cofactor FMN demonstrated the capability to generate FMNH through the collaborative influence of GDH, NADP+, and glucose. When exposed to light excitation, FMNH undergoes a single-electron transfer (SET) process to form FMNH•, which in turn facilitates hydrogen atom transfer (HAT)45. Furthermore, specific amino acids may play a crucial role in providing hydrogen atoms during the catalytic process. Therefore, we conducted an investigation on amino acids (Y82/Y196/Y375) that could potentially be involved in the HAT process (Fig. 4a).

Fig. 4: Investigation of the enzyme’s catalytic mechanism through mutagenesis and deuteration experiment.
figure 4

a Investigation of the impact of three tyrosine residues (Y82/Y196/Y375) on the reaction process. b Deuterium labeling experiments. Track the source of chiral hydrogen atoms in the product and provide the evidence for the reaction mechanism (D-Tris: Tris+DCl+D2O, pH = 8.5).

Full size image

Y82A/Y196A/Y375A exhibited varying degrees of reduced activity. In particular, the yield (9%) and ee values (52%) of Y196A were significantly diminished. In contrast, while the yields of Y82A (24%) and Y375A (23%) decreased, the ee values did not show significant changes (Y82A: 92%; Y375A: 88%, respectively). This disparity suggested that Y196 might play a more crucial role in both binding and stabilizing the substrate. The reduction in yield could be attributed to the replacement of Y with A, leading to substantial modifications in the enzyme’s active site, which adversely affected substrate binding. Consequently, we mutated tyrosine residue to structurally similar phenylalanine, which was unable to facilitate the HAT process. The introduction of the Y196F mutation consistently resulted in a decrease in ee values, including Y196F (ee: 73%), Y82F/Y196F (ee: 59%), Y196F/Y375F (ee: 73%), Y82F/Y196F/Y375F (ee: 56%). However, compared to the wild type, the mutants containing Y196F exhibited alterations only in stereoselectivity, with nearly unchanged or even increased yields (WT: 53%; Y196F: 56%; Y82F/Y196F: 53%; Y196F/Y375F: 95%; Y82F/Y196F/Y375F: 99%). Based on the outcomes of these mutants, we speculated that Y196 had a role in stabilizing the product through hydrogen bond interactions (Fig. 4a and Supplementary Fig. 14). Subsequently, we performed deuterium labeling experiments to further investigate the source of the hydrogen atoms (Fig. 4b and Supplementary Fig. 15). For mutants containing the Y196F, the proportion of deuterated products generated from deuterated solvent (D-Tris) was relatively low ( < 25%), suggesting that the solvent was not the primary hydrogen source. Conversely, the introduction of d-Glu-1,2,3,4,5,6,6-d7 significantly increased the proportion of deuterated products (Y196F: 38%; Y82F/Y196F: 40%; Y196F/Y375F: 57%, Y82F/Y196F/Y375F: 70%), an effect further amplified when both deuterated solvent and deuterated glucose were present (Y196F: 60%; Y82F/Y196F: 59%; Y196F/Y375F: 79%, Y82F/Y196F/Y375F: 77%). Furthermore, under standard conditions, the mutants did not exhibit a significant reduction in yields compared to the wild type (WT), implying that the hydrogen atom source has not been disrupted. These results suggested that the primary source of hydrogen atoms originated from the deuterated flavin cofactor (FMNH). It is also noteworthy that in the absence of the Y196F mutation, the proportion of deuterated glucose was relatively low. We hypothesized that this phenomenon might be attributed to proton exchange interactions between the Y196 residue, FMNH, and the solvent, which could modulate the generation and distribution of deuterated products.

To better understand the reaction mechanism and its enantioselectivity, we performed molecular dynamics (MD) simulations and quantum mechanics/molecular mechanics (QM/MM) calculations. Based on the initial conformation obtained from docking, a 200-ns MD simulation was performed. The results demonstrated that the conformation of the FMN cofactor and amino acid residues exhibited no significant displacement during the simulation (Supplementary Fig. 16). In addition, Our MD simulations have shown that both F-1 radical and E-1 can exchange the position during MD simulations, leading to two representative binding conformers of RC1a and RC1b (Fig. 5a). As both conformers can undergo fast exchange, both conformers are assumed to be in the same energy level. Starting from RC1a, the C-C coupling between F-1 radical and E-1 forms radical intermediate (IC1a), with a barrier of 4.5 kcal/mol. The subsequent HAT from FMNH• to the C-radical center affords (R)-1 (IC1a → PCa). As for RC1b, the C-C coupling between F-1 radical and E-1 experiences a barrier of 2.8 kcal/mol, which can eventually afford the (S)-1 (RC1b → IC1b → PCb). Since the C-C coupling step is highly exothermic and irreversible, the stereoselectivity of the reaction is dominated by the C-C coupling step (TS1a vs. TS1b). Thus, our QM/MM calculations suggest the reaction would be dominated by the formation of (S)-1, which agrees with our experiments. In accordance with previous studies56,57,58, a catalytic mechanism was proposed in Fig. 5b. Initially, FMN was reduced to FMNH in the presence of GDH/glucose/NADP+. Subsequently, the binding of the fluorinated reagent F-1 to the active site of the ene-reductase formed a complex with FMNH (Int A). Upon excitation by visible light, the excited FMNH• underwent SET, followed by the cleavage of the C-Br bond of F-1, resulting in a radical intermediate in Int B. Due to the thermodynamic driving force, the hydrophobic α-methyl styrene E-1 would bind to the active site after the release of the by-product Br- (Int C), as previously proposed45,48. Subsequently, a radical C-C bond coupling could readily occur between the radical intermediate and E-159, leading to the formation of the prochiral radical intermediate (Int D). Finally, product 1 was formed with the transfer of the hydrogen atom, while FMNH• was converted to oxidized flavin (FMNOX) (Int E). The enzymatic cofactor FMNOX could be converted back to FMNH by the regeneration system (GDH/glucose/NADP+), initiating the subsequent catalytic cycle.

Fig. 5: Computational studies and proposed reaction mechanism.
figure 5

a QM/MM calculations revealed the changes in energy barriers and configurational preference during the product formation process. Source data are provided as a Source Data file. b Proposed mechanism. Radical formation and chiral hydrogen atom transfer process during the reaction.

Full size image

Discussion

In this study, we developed a photocatalytic enzymatic strategy for synthesizing fluorinated amide compounds. In contrast to conventional chemical synthesis methods, enzyme catalysis demonstrated gentle and environmentally friendly reaction conditions. Our results indicated that photoenzymes exhibited notable catalytic efficiency, compatibility with various substrates, and precise regulation of distal chirality. Additionally, our mechanism study identified the significant contributions of Y196 and Y375 in controlling chirality and enhancing catalytic efficiency. This study extended the enzymatic synthesis methods for fluorinated amide compounds, demonstrating great potential for the production of biologically active molecules incorporating fluorinated amides. Additionally, it notably diversified the range of applications for photoenzymes. The developed photoenzymes are anticipated to have extensive utility in the production of valuable fluorinated molecules.

Methods

Photoenzymatic reactions experiments

In this study, the reactions involved were all performed in 4 mL glass vials. Under nitrogen, the substrates were added to the reaction flask according to their final concentrations: fluorinated reagent: 10 mM, alkene: 5 mM, NADP+: 0.5 mM, d-(+)-glucose: 50 mM, DMSO: 80 μL, glycerol: 80 μL, glucose dehydrogenase (GDH): 0.5 μg/μL, and ene reductase (ER): 1–2 mol%. Tris-HCl (pH = 8.5) was added to the reaction system in a total volume of 800 μL. Four-microliter glass vial was sealed with a threaded cap. It was sealed with a Parafilm sealing film wrapped around the bottle cap again. Reactions were placed in a 12 W 455 nm parallel light reactometer or 45 W 465 nm LED blue light with two cooling fans (Supplementary Fig. 2). After 16 h of reaction, the reaction was terminated by adding an equal volume of acetonitrile. The enzyme was completely inactivated after 15 min at room temperature.

After the termination of the reaction, an equal volume of ethyl acetate was added for extraction, which was repeated three times. The reactions were detected by HRMS, 1H NMR (400 MHz, CDCl3, 64 scans), and chiral HPLC.

Molecular docking

Molecular docking was conducted using AutoDock Vina60. The protein structure was obtained from the PDB database (PDB code: 3TX9). Water molecules and ligands were removed using PyMOL, while FMN was retained as the protein structure for docking. The ligand underwent energy minimization before conducting docking. The docking pocket included FMN. After docking, the results were analyzed using PyMOL.

Yield calculation

Yield determination experiment: Reaction system: fluorinated reagent: 10 mM, alkene: 5 mM, NADP+: 0.5 mM, glucose: 50 mM, GDH: 0.5 μg/μL, ene-reductase: 2 mol%, DMSO: 10% (v/v), glycerol: 10% (v/v), Tris-HCl (pH = 8.5): add to 800 μL. Reaction at room temperature for 16 h. An equal volume of acetonitrile was added to terminate the reaction. An equal volume of ethyl acetate was added for extraction, repeated three times. Distillation was done under reduced pressure until ethyl acetate was completely evaporated. Six hundred microliters of CDCl3 were added. 5.6 mg of 1,3,5-trimethoxybenzene was added (as an internal standard) to the deuterated reagent.

Stereoselectivity evaluation

Stereoselectivity experiments were conducted using chiral HPLC. Initially, conditions for resolving racemates were optimized. After the conditions were established, enantiomeric separations of enzymatic reaction products were performed. The Daicel chiral columns used were IA/IC/OJ-H (4.6 mm × 250 mm, 5 μm) with a mobile phase of hexane/isopropanol. The ee values were calculated by recording peak areas in the HPLC chromatograms.

System setup

The initial structure was constructed based on the crystal structures of OYEl (PDB code.3TX9), and mutant structures of OYE1_Y375F were built based on SWISS-MODEL homology modeling61. The substrates were docked into the active site using AutoDock Vina60. We assigned the protonation status of titratable residues (His, Glu, Asp) based on the pKa values from the PROPKA software in combination with a careful visual inspection of local hydrogen-bonded networks62. His29, His105, His159, His191, His205, and His237 were protonated at the ε position, His43, His286, His329, and His380 were protonated at the δ position, while all the Glu and Asp residues were deprotonated. The Amber ff14SB force field was employed for the protein residues63, while the general AMBER GAFF was used for substrates64. The partial atomic charges of substrates were obtained from the RESP at the B3LYP/6-31G(d) level of theory65. Sodium ions were added to the protein surface to neutralize the total charge of the systems. Finally, the resulting system was solvated in a rectangular box of TIP3P waters extending up to a minimum distance of 16 Å from the protein surface.

Molecular dynamics (MD) simulations

The energy minimization procedure was initially conducted using the steepest descent method, followed by conjugate gradient optimization. The system was gradually heated from 0 K to 300 K over 50 ps in the NVT ensemble with 25 kcal/(mol·Å²) harmonic restraints applied to protein backbone atoms. Subsequently, a 1 ns NPT density equilibration was performed at 300 K to establish uniform system density, followed by a 4 ns NPT equilibration to stabilize temperature and pressure fluctuations. Finally, a 200 ns NPT production simulation was conducted. Notably, when dual substrates were present in the active pocket, two distinct binding modes were observed during the initial 4 ns NPT equilibration. To systematically investigate these conformations, separate 200 ns production simulations were executed for each binding pattern, with additional 5 kcal/(mol·Å²) harmonic restraints applied to substrates to compensate for missing force field parameters of radical species. All simulations employed the SHAKE algorithm to constrain hydrogen-containing bonds (2fs integration time step)66. All MD simulations used the GPU-accelerated Amber 20 package67. The trajectory analyses were performed using the Cpptraj module68.

QM/MM calculations for enzymatic reactions

We performed conformational sampling on a 200 ns MD trajectory using the K-means clustering algorithm and selected representative structures for quantum mechanics/molecular mechanics (QM/MM) calculations. Enzyme-catalyzed reaction simulations were carried out using the ChemShell 3.7 software package, with the QM region computed by Turbomole 7.5 and the MM region handled by DL_POLY69,70,71,72. The polarizing influence of the enzymatic environment was incorporated through an electronic embedding scheme, while the QM/MM boundary was effectively addressed using hydrogen link atoms implemented with the charge-shift model. For whole reactions, the QM region consists of the FMN, two substrates, and the sidechain of Y196/Y375F. During QM/MM geometry optimizations, the QM region was studied with the B3LYP density functional with two levels of theory. For geometry optimization, the double-ζ basis set def2-SVP was used. The energies were further corrected with the larger basis set def2-TZVP for all atoms. Dispersion corrections computed with Grimme’s D3 method were included in all calculations.

Reporting summary

Further information on research design is available in the Nature Portfolio Reporting Summary linked to this article.