Introduction

Selective functionalization of carbon‒hydrogen (C-H) bonds is a powerful strategy in installing new functional groups into organic molecules, whereby the synthetic routes are potentially largely simplified (e.g., avoiding tedious and costly prefunctionalization steps) for more sustainable chemical synthesis1,2. Over the last decade, organic chemistry has witnessed significant advances in transition metal-based complexes3,4,5 and organocatalysts6 used for C-H bond functionalization reactions. On the other hand, enzymes are the method of choice for catalyzing challenging reactions such as C-H bond activation and functionalization7,8,9. When a metal ion or organic cofactor is embedded into the well-evolved supermolecular structure of a protein, the regio- and stereoselectivity are granted in a catalytic process. In particular, oxidoreductases such as oxidases, monooxygenases, and peroxygenases have been widely investigated for selective C-H functionalization7,8,9,10,11,12. Using the technology of directed evolution, C-H functionalization can be expanded to non-natural reactions such as transforming C-H bonds into C-Si13, C-B14, and C-N15 bonds. Among these enzymes, heme-thiolate-containing enzymes such as P450 monooxygenases and unspecific peroxygenases (EC 1.11.2.1, UPOs) relying on oxoferryl-heme as the oxygenating species (so-called compound I) have been investigated thoroughly for C-H bond oxyfunctionalization reactions12,16. P450s are typically dependent on nicotinamide cofactors (NADP(H)) coupled with a regeneration system to reduce O2 as the oxygen source, and the resultant sophisticated electron transport chains impose low efficiencies on the catalytic transformations. In contrast, the UPOs directly use partially reduced H2O2 as an oxygen- and electron-source to catalyze C-H bond oxyfunctionalization. In principle, UPOs have substrate and product scopes similar to those of P450s but with higher efficiencies (e.g., higher turnover numbers)12.

To date, the UPOs have shown promiscuous activity and enable a wide variety of reactions, such as hydroxylation, epoxidation, demethylation, halogenation, sulfoxidation, and alcohol oxidation17. In particular, UPOs-catalyzed hydroxylation could accept benzylic, aromatic, allylic or propargylic C-H bonds in a rather regio- and/or stereoselective manner, giving access to various precursors18 and drug analogs that are industrially relevant19. Very recently, we also demonstrated that peroxygenase enabled shifting the reactivity of aromatic C-H hydroxylation towards a dearomatization reaction by capturing the aromatic oxide, an instant intermediate for the NIH shift to phenols20.

Of note, the established demethylation is a less explored reaction for UPOs, which convert ethers to cleaved products via a hydrogen abstraction of the alpha C-H bond and an oxygen rebound mechanism21,22. The demethylation reaction thus far, as suggested by its name, only resulted in the formation of cleaved products due to the instability of the hemiacetals towards spontaneous hydrolysis in aqueous conditions (e.g., sole formation of 4-hydroxybutanal via tetrahydrofuran oxidation shown in Fig. 1a). Interestingly, peroxygenases have shown unusual stability in pure organic solvents upon immobilization23,24. This inspired us to assume that the hydrolysis of hemiacetals obtained from immobilized peroxygenases can be circumvented under neat reaction conditions. Hemiacetal compounds are valuable synthetic intermediates in chemical synthesis25. For example, cyclic hemiacetals frequently act as vital functional moieties in pharmaceutical development (Supplementary Fig. 1). Traditionally, the synthesis of hemiacetal compounds is primarily restricted to the addition reaction between an alcohol and an aldehyde/ketone, or obtained by reducing lactone with metal catalysts26,27. However, hemiacetals have not been deemed significant for UPOs or enzymes in general due to their instability in aqueous conditions28.

Fig. 1: Peroxygenase-catalyzed conversion of ether compounds.
figure 1

a The established cleavage of ethers and b the proposed oxyfunctionalization for cyclic hemiacetal synthesis.

Full size image

Herein, we demonstrate the enzymatic synthesis of cyclic hemiacetals by peroxygenase-catalyzed C-H bond oxyfunctionalization under neat reaction conditions (Fig. 1b). The reaction engineering allowed us to switch the peroxygenase-catalyzed ether cleavage to hemiacetals synthesis. This work will not only expand the scope of the known C-H oxyfunctionalization chemistry enabled by peroxygenases but also contribute a catalytic method for hemiacetal synthesis in the synthetic community.

Results and discussion

Proof-of-concept experiments

We began to evaluate the envisioned hemiacetal synthesis by using UPO under neat reaction conditions. The hydroxylation of tetrahydrofuran (THF, 1) to tetrahydrofuran-2-ol (1a) was chosen as a model reaction (Fig. 2a). THF serves a dual role as both the substrate and reaction medium. The recombinant peroxygenase from Agrocybe aegerita (rAaeUPO, PaDa-I variant) was prepared according to a previous protocol29. To obtain immobilized enzyme, we first screened seven commercial resin carriers that have abundant amino groups (LX 700, 703, 704), epoxy groups (LX 600, 603, 609) or strong adsorption capacity (LX 1000). The amino resin was activated with glutaraldehyde before immobilization, while the other resins were used without any pretreatment. Although a number of oxidative in situ generations of H2O2 have been reported for UPOs30,31,32,33,34,35,36,37, we decided to use a syringe pump for the supply of H2O2 to the enzyme to reduce the complexity of the overall reaction system. Under the reaction conditions chosen, the amino resin LX 700 excelled over the other six in terms of the desired product concentration (Fig. 2b) and the amount of the enzymes immobilized (Supplementary Table 1, Supplementary Fig. 2). A steady increase in product formation was observed, with a significant amount of 1a (24.6 mM) obtained in 24 h. 2.8 mM of the overoxidized product γ-butyrolactone (1b) was observed (Fig. 2c). The formation of products 1a and 1b is contradictory to the established THF oxidation catalyzed by peroxygenase, in which only hydroxylated 4-hydroxybutanal was observed under aqueous condition21. We attribute this effect to the improved stability of cyclic hemiacetal product 1a under near-neat reaction conditions.

Fig. 2: Peroxygenase-catalyzed hydroxylation of tetrahydrofuran.
figure 2

a Reaction scheme. b Screening of immobilized enzymes under neat conditions. Conditions: [THF] = 0.5 mL (12.3 M), [immobilized rAaeUPO] = 100 mg (corresponding to 1.45 μM), [H2O2] = 6 mM h−1, 30 °C, 7 h. c Time course of the hydroxylation of tetrahydrofuran yielding tetrahydrofuran-2-ol (squares, 1a) and γ-butyrolactone (diamonds, 1b). Conditions: [THF] = 0.5 mL (12.3 M), [immobilized rAaeUPO] = 125 mg (corresponding to 1.81 μM), [H2O2] = 6 mM h−1, 30 °C, 24 h. The reported value is based on the mean value of two distinct experiments (n = 2).

Full size image

Characterization of the enzymatic oxyfunctionalization reaction

The robustness of the synthesis of 1a was determined by the performance of the immobilized peroxygenase, the combinatorial supply of H2O2 and its consumption by the enzyme under neat reaction conditions. In a next step, we systematically varied these parameters to optimize the reaction conditions by using the rAaeUPO immobilized on LX 700 (Table 1). In terms of the enzyme concentration, a value of approximately 100 mg (corresponding to 1.45 μM) was found to be optimal with respect to the desired product concentration and the initial reaction rate of THF hydroxylation (Table 1, entries 1–3). This observation is reasonable, as at lower loading, the enzymes were inactivated by H2O2 feeding with an applied rate of 6 mM h−1, while at higher enzyme loading, a mass transfer limitation could slow down the reactions (Supplementary Fig. 3). When the varied H2O2 was investigated, it turned out that at a rate of 6 mM h−1 yielded the highest production formation (Table 1, entries 2, 4, and 5). Lowering the feeding rate to 3 mM h−1 resulted in an almost linear increase in the product concentration. In contrast, a higher feeding rate could inactivate the enzymes quickly (Supplementary Fig. 4). With the experiments performed, only a minor concentration of the overoxidized lactone 1b was observed, corresponding to an overall selectivity of approximately 90% (Table 1). A turnover number (TON) of 27380 was achieved for 1a synthesis, which is superior to the natural enzyme activity of peroxygenases in C-H oxyfunctionalization reactions38. It is of interest to note that the efficiency of immobilized rAaeUPO in using H2O2 is significantly higher than that of tBuOOH in terms of product concentration and selectivity (Table 1, entries 2, 6 and 7, and Supplementary Fig. 5). This is rationalized that the kinetically less effective tBuOOH has a good affinity with rAaeUPO in apolar solvents, whereas H2O2 is expected to have a higher affinity with the enzyme in slightly polar solvent (i.e. THF)39. Additionally, a possible hydroxylation of the C-H bond leading to tetrahydrofuran-3-ol was not observed in all experiments as judged by gas chromatography using a commercial standard compound, suggesting excellent regioselectivity of the hydroxylation of THF by rAaeUPO. The control reactions using rAaeUPO or H2O2 alone, or using the resin and H2O2 together did not yield any oxidized products of THF (Table 1, entries 8–10).

Table 1 Optimization of the envisioned enzymatic hemiacetal synthesis
Full size table

Substrate scope of the enzymatic cyclic hemiacetal synthesis

To show the applicability of the envisioned enzymatic approach for hemiacetal synthesis, we next investigated a range of cyclic ether substrates with electronic and structural diversity (Fig. 3, and Supplementary Figs. 6–18). The six- and seven-membered cyclic ethers were readily oxidized into the corresponding hemiacetal products (Fig. 3, 2a and 3a), while the conversion of oxetane was not observed. The highest enzyme activity with a TON of 95172 was achieved with tetrahydropyran (THP), which is comparable to the activity of peroxygenases in cycloalkane oxidation35,36,40. 1,4-Dioxane and benzo-1,4-dioxane were solely converted into hemiacetals with lower enzyme activity (Fig. 3, 5a, 12a). The substituted THF and THP were also accepted by rAaeUPO (Fig. 3, 6a-11a). In particular, the dihydrobenzo-furan and dihydrobenzo-pyran compounds, which are potentially more interesting in the pharmaceutical industry, were oxidized into the desired hemiacetals albeit with varied enzyme activities. The synthesis of some cyclic hemiacetals was also performed on a preparative scale at 50 mL. The solvent was evaporated and recycled, and the products were purified via flash chromatography after 24 hours of oxyfunctionalization reactions. Isolated quantities of 0.28, 0.69, 0.21 and 0.51 g were obtained for 1a, 2a, 5a and 10a (Supplementary Fig. 19), respectively. The substrates bearing moderate deactivating groups (Fig. 3, 13a-16a) were not converted. Additionally, a few noncyclic ethers were tested (Supplementary Table 2) and only C-O bond cleaved products were obtained. Interestingly, although the hydroxylation products of the two piperidine derivatives were not observed (Fig. 3, 17a and 18a), the corresponding lactam products via possible overoxidation of hemiaminals were obtained with 4.3 and 2.1 mM (Supplementary Figs. 17 and 18), respectively. This could be due to the low reactivity of rAaeUPO with piperidine compounds and the poor stability of hemiaminals41.

Fig. 3: Substrate scope of cyclic ether oxidation reaction.
figure 3

Conditions: [substrates] = 0.5 mL, [immobilized rAaeUPO] = 100 mg (corresponding to 1.45 μM), [H2O2] = 6 mM h−1, 30 °C, 24 h. The selectivity was determined by GC. Selectivity = [1-18a]/([1-18a]+[1-18b]) × 100%. TON = [1-18a]/[immobilized rAaeUPO]. N.D. = not determined. The ee value was determined by chiral HPLC after the derivatization of the product.

Full size image

Meanwhile, we also determined the enantiomeric excess (ee) of the obtained hemiacetal products. To determine the absolute configuration, (R)-selective alcohol dehydrogenase from Lactobacillus kefir was used to prepare the standard compounds with complementary chirality starting from lactones, respectively (Supplementary Information). As shown in Fig. 3, while the model substrate did not show a chirality, >99% ee in (R)-configuration was achieved with 2a, 3a, 5a, 7a, 8a, 10a, 11a and 12a (Supplementary Figs. 28–36).

Synthesis of lactones and lactams

Although a minor amount of lactones were formed along with the hemiacetal products, we wondered whether it is possible to switch the product distribution towards the formation of lactones by carrying out the reaction engineering, i.e. switching to an aqueous condition. The lactone compounds are versatile building blocks in flavors, fragrances, polymers, etc42. The enzymatic synthesis of lactones represents a green alternative to current chemical methods43. To show the enzymatic synthesis of lactones starting directly from the C-H bond hydroxylation of cyclic ethers via the in situ use of hemiacetal (Fig. 4a), we then investigated the reaction pH, a crucial parameter determining the stability of the hemiacetals. As shown in Fig. 4b, a change of the pH from acidic to neutral and alkaline conditions remarkably influenced the formation of lactone 1b. Below pH 5, only hemiacetal was obtained, while at pH 9 the lactone product formed exclusively. Figure 4c shows the reaction course starting from the oxidation of THF catalyzed by rAaeUPO in the presence of H2O2. Clearly, with the consumption of THF, hemiacetal 1a formed first, and then a stepwise oxidation of the intermediate 1a into lactone 1b occurred. 5.6 mM of 1b was obtained, corresponding to a conversion of 56% and a TON of 3733 for the enzyme. A gap in the mass balance (ca. 20%) is attributed to the hydrolysis of hemiacetal 1a into 4-hydroxybutanal.

Fig. 4: Hydroxylation of tetrahydrofuran for γ-butyrolactone synthesis by rAaeUPO.
figure 4

a Reaction scheme. b Influence of the reaction pH on the product distribution of hemiacetal 1a (light green rectangles) and lactone 1b (dark green rectangles). c The representative time course showing THF conversion (triangles, 1), and tetrahydrofuran-2-ol (squares, 1a) and γ-butyrolactone (circles, 1b) formation. Reaction conditions for (b), [THF] = 8 mM, [rAaeUPO] = 1.5 μM, citrate buffer (50 mM, pH 4 − 5), or NaPi buffer (50 mM, pH 6 − 8), or Tris-HCl buffer (50 mM, pH 9), [H2O2] = 1 mM h−1, 30 °C, 12 h. Reaction conditions for (c), [THF] = 10 mM, [rAaeUPO] = 1.5 μM, Tris-HCl buffer (50 mM, pH 9), [H2O2] = 1 mM h−1, 30 °C, 24 h. The reported value is based on the mean value of two distinct experiments (n = 2).

Full size image

After the successful control of the product distribution, we selected the substrates to show rAaeUPO-catalyzed C-H bond oxyfunctionalization for lactone synthesis via the formed hemiacetal as an intermediate (Fig. 5). Lactones 1b, 2b, 3b, and 10b were formed with good conversion (56-76%). It is also possible to synthesize lactams (e.g., 17b and 18b) via hemiaminal, whereas the poor enzyme activity requires further improvement.

Fig. 5: Substrate scope of the cyclic ether oxidation reaction for lactone and lactam synthesis.
figure 5

Conditions: [substrates] = 10 mM, Tris-HCl buffer (50 mM, pH 9), 5% dimethyl sulfoxide (DMSO) in 10, 17 and 18, [rAaeUPO] = 1.5 μM, [H2O2] = 1 mM h−1, 30 °C, 24 h. The conversion was determined by GC. Conversion = [1-18b]/[1-18] × 100%. The reported value is based on the mean value of two distinct experiments (n = 2).

Full size image

Quantum chemical study on the mechanism and regioselectivity

While we were delighted to discover that UPO can facilitate the synthesis of cyclic hemiacetals, a reaction that has not yet been convincingly demonstrated by enzymes, however, two questions remain unanswered. The first concerns the detailed mechanism of hemiacetal formation under neat conditions, which contradicts the established ether cleavage mechanism under aqueous conditions. The second involves determining the exact mechanism for controlling the selectivity of both hemiacetal and lactone synthesis. To address these questions, molecular dynamics (MD) simulations and quantum chemical (QC) calculations were performed in the present study (Fig. 6a, b).

Fig. 6: Optimized structures and calculated energy profiles of rAaeUPO-catalyzed reaction.
figure 6

Optimized structures of the enzyme-substrate complexes with tetrahydrofuran 1 (a) and tetrahydrofuran-2-ol 1a (b). The “*” represents the atoms fixed in the geometry optimization. Nonpolar hydrogens of the protein and heme are hidden for clarity. Calculated energy profiles for the hydroxylation of tetrahydrofuran 1 to tetrahydrofuran-2-ol 1a and tetrahydrofuran-3-ol 1a’ (c) and the oxidation of tetrahydrofuran-2-ol 1a to γ-butyrolactone 1b (d).

Full size image

The QC calculations showed that the hydroxylation of tetrahydrofuran (1) to form tetrahydrofuran-2-ol (1a) follows the typical mechanism proposed for heme-dependent enzymes and the doublet spin state is the ground state of the reaction44,45,46. After the activation of the iron porphyrin by hydrogen peroxide to form the compound I (Cpd I), the reaction proceeds with two steps consisting of the rate-limiting hydrogen atom transfer (HAT) process producing a free radical intermediate and the following low-barrier hydroxyl rebound process. Importantly, the experimentally observed regioselectivity is reproduced by QC calculations. The energy barriers for the HAT from the α-carbon and β-carbon of 1 are 10.6 kcal mol−1 (TS1-A, Fig. 6c) and 13.1 kcal mol−1 (TS1’-A, Fig. 6c), respectively. Thus, the barrier for the HAT of the H on the α-carbon to give tetrahydrofuran-2-ol is 2.5 kcal mol−1 lower than that of the H on the β-carbon to generate tetrahydrofuran-3-ol. Analysis on the electronic structures of the transition states showed that the spin density of the α-carbon is −0.18 at TS1-A, while the spin density of the β-carbon is −0.42 at TS1’-A. The free radical forming in the HAT process is thus more delocalized in the former case, resulting in a lower barrier. This explanation can be generalized to all the considered substrates shown in Fig. 3. In the case of substrate 10, it presents a unique scenario with two α-carbons. When the HAT process takes place on the α-carbon atom between the carbonyl group and the aromatic ring (C1), electrons of the radical can be more delocalized because of the proximity to the π-system of the aromatic ring, as compared to when the radical forms at the other α-carbon atom (C3). This causes hydroxylation at the C1 position more favorable over the C3 position, resulting in the formation of product 10a.

In the step of the oxidization of 1a to form γ-butyrolactone (1b), interestingly the free radical intermediate after the HAT process is unstable and a proton transfer from the hydroxyl group of 1a to Glu196 takes place spontaneously during the optimization (Fig. 6d). The latter process is accompanied by an electron transfer from the substrate to the iron center. The following proton transfer from Glu196 to Fe-OH is highly exothermic.

Overall, the QC calculations indicate that the conversion of 1 to 1a consists of the rate-limiting HAT process forming a free radical and the subsequent hydroxyl rebound. For the over oxidation of 1a to 1b, the HAT process is concerted by a proton transfer from 1a to Glu196 due to the instability of the free radical intermediate created by the HAT. Furthermore, the regioselectivity observed in the conversion of 1 to 1a was reproduced by the calculations as it was shown that the selectivity is resulted from the enhanced delocalization of the radical at the α-carbon of the substrate in the formation of the free radical intermediate. Additionally, the superposition of all the optimized structures of the intermediates and transition states involved in the reaction shows that the residues in the rAaeUPO active site change very little during the entire catalytic cycle (Supplementary Fig. 40).

Umbrella sampling simulations on the pH-dependence

As discussed, an interesting finding is that the product selectivity is altered at different pH values. At pH = 4, 1a is the only product of the rAaeUPO-catalyzed oxyfunctionalization of tetrahydrofuran, while at pH = 9 the accumulated 1a is overoxidized to 1b (Fig. 5). To rationalize the reasons for this product selectivity, umbrella sampling simulations were performed for the binding of 1 and 1a to the active site of the enzyme at varied pH conditions (pH = 4 and 9) by modeling the ionizable residues in different protonation states (Fig. 7). The obtained potential of mean force (PMF) profiles shows that under both conditions the binding of 1 to the active site is much more favored than 1a. Moreover, two minima were identified on the PMF profiles for all considered cases at the distances between the oxygen atom of Cpd I and the oxygen of the substrate five-membered ring with values of ca 5 Å and ca 15 Å, which can be denoted as the substrate binding inside and outside of the active site, respectively. At lower pH, 1 displays almost the same probability for binding inside and outside, but 1a is prefers to bind outside (Fig. 7a). This explains why the conversion of 1a to 1b was not experimentally detected at pH 4. When the pH increases to 9, the binding preference inside the active site is much more preferred than that outside for 1, and the binding inside and outside are now both possible for 1a (Fig. 7a). Therefore, at pH = 9, the conversion of 1 to 1a easily occurs, and the oxidation of 1a to 1b can take place when 1a accumulates to a certain amount. This is highly consistent with the experimental results. The impact of phosphate buffer molecules at pH = 9 on the binding affinities of the substrate was also explored (Fig. 7b), and the calculations show that the presence of buffer molecules significantly enhances the binding of 1a to the active site, resulting in 1a being much more preferentially located in the active site. Thus, in light of the current simulations the differences in the catalytic performance of the rAaeUPO reaction between pH 4 and pH 9 can be attributed to the distinct binding preferences of 1 and 1a to the active site of the enzyme under the two pH conditions.

Fig. 7: Potential of mean force of 1 and 1a entering the active site under different condition.
figure 7

a PMF of tetrahydrofuran 1 and tetrahydrofuran-2-ol 1a entering the active site at pH = 4 (left) and pH = 9 (right). The umbrella sampling simulations were performed using the model with different protonation states of the ionizable residues in the enzyme. b PMF of tetrahydrofuran-2-ol 1a entering the active site at pH = 9 in the presence of buffer molecules in the solution. The curve shows the average of values of three independent PMFs, while the error bar shows the error deviation.

Full size image

In summary, this work used the strategy of reaction engineering and demonstrated that unspecific peroxygenase from Agrocybe aegerita was capable of synthesizing optically pure cyclic hemiacetal compounds through the selective C-H bond oxyfunctionalization of cyclic ethers. Under neat reaction conditions, hydrolysis of the hemiacetal products was remarkably circumvented, leading to a unexplored application of peroxygenases compared to the established demethylation reactions. A diverse range of substrates was converted into the corresponding hemiacetal products with up to >99% ee. Moreover, by switching the reaction conditions to aqueous conditions the hemiacetal products were further oxidized into lactones, which represents an in situ application of the hemiacetals. The formation of lactam products was also possible, however, the low enzyme activity needs to be addressed. In virtue of molecular dynamics simulations and quantum chemical calculations, we proposed reaction mechanisms for the peroxygenase-catalyzed conversion of cyclic ether to hemiacetal, and subsequently to lactone. The differences in the distribution of hemiacetal and lactone products can be explained by the binding preferences of the cyclic ether and hemiacetal to the enzyme’s active site. This work demonstrates an enzymatic method for hemiacetal synthesis and highlights the potential of peroxygenase as a versatile biocatalyst in synthetic chemistry.

Methods

Materials

Unless otherwise stated, all reagents and solvents were purchased from commercial suppliers (J&K Chemical, Bide Pharmatech Ltd., Macklin, Energy Chemical, etc.) and used without purification. Anhydrous toluene and dichloromethane were obtained by distilling the solvent using calcium hydride and stored under nitrogen. The resin carriers were purchased from Mreda, Macklin and Sunresin New Materials Co. Ltd. and used as described in the immobilization of rAaeUPO part of the Supplementary Information.

Enzyme preparation

The procedures of the recombinant expression and purification of the evolved nonspecific peroxygenase mutant (PaDa-I) from A. aegerita in P. pastoris were based on reported procedures29. Typically, the single colony from the P. pastoris clones were picked and inoculated in BMGY medium (5 mL). After incubation for 36 h at 30 °C, 1 mL of the preculture was transferred into 100 mL of BMGY medium and incubated for 24 h till OD600 reached around 1.5. Methanol was then added and the incubation was continued for 72 h. After the fermentation process, the culture broth of P. pastoris cells containing rAaeUPO was separated by centrifugation (10956 × g, 4 °C, 1 h). The supernatant was filtered through a 0.22 µm filter and kept at −80 °C for later use.

Typical procedures for hemiacetal synthesis

To a 1 mL transparent glass vial, 100 mg (corresponding to 1.45 μM) of immobilized rAaeUPO and 0.5 mL of substrate were added. Then, H2O2 from a stock solution (300 mM) was injected by a syringe pump into the mixture at rate of 6 mM h−1 (10 μL h−1). The reaction vial was sealed and agitated in a thermal shaker at 30 °C and 800 rpm. At intervals, aliquots were withdrawn, 20 μL of the mixture was mixed with 180 μL of ethyl acetate (containing 5 mM of dodecane as an internal standard). Then, the sample was dried over Na2SO4 and analyzed by gas chromatography (GC) to determine the product concentration. To measure the optical purity, HPLC was used after the product was derivatized by 2-methylbenzoyl chloride (acetic anhydride for 12a). Details of the GC and HPLC methods are shown in Supplementary Tables 4, 5 and Supplementary Figs. 6–18, 28–36. For the detailed experimental procedures and analytics of the hemiacetal synthesis, see Supplementary Information.

Computational study

To shed light on the reaction mechanism and regioselectivity of rAaeUPO-catalyzed oxyfunctionalization of the C-H bond of tetrahydrofuran, molecular dynamics (MD) simulations and quantum chemical (QC) calculations were performed on the basis of the previously solved crystal structure47. Optimized geometries and spin densities of the transition states and intermediates calculated by QC are shown in Supplementary Figs. 37–39. The root mean square deviation (RMSD) and distribution histograms of the MD simulations are shown in Supplementary Figs. 40, 41. Details of the methods are included in the Supplementary Information.

Reporting summary

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