Unspecific peroxygenase enabled formation of azoxy compounds

Abstract

Enzymes are making a significant impact on chemical synthesis. However, the range of chemical products achievable through biocatalysis is still limited compared to the vast array of products possible with organic synthesis. For instance, azoxy products have rarely been synthesized using enzyme catalysts. In this study, we discovered that fungal unspecific peroxygenases are promising catalysts for synthesizing azoxy products from simple aniline starting materials. The catalytic features (up to 48,450 turnovers and a turnover frequency of 6.7 s^–1) and substrate transformations (up to 99% conversion with 98% chemoselectivity) highlight the synthetic potential. We propose a mechanism where peroxygenase-derived hydroxylamine and nitroso compounds spontaneously (non-enzymatically) form the desired azoxy products. This work expands the reactivity repertoire of biocatalytic transformations in the underexplored field of azoxy compound formation reactions.

Azoxy products have rarely been synthesized using enzyme catalysts. Herein, the authors report that fungal unspecific peroxygenases are promising catalysts for synthesizing azoxy products from simple aniline starting materials.

Introduction

Functional groups containing a nitrogen-nitrogen bond (e.g., N-N or N = N) are commonly encountered in natural products^1,2. The reactivity and structural diversity of these molecules are of significant interest in synthetic chemistry. Despite the identification of various compounds with nitrogen-nitrogen bonds in a variety of natural sources, enzymes catalyzing nitrogen-nitrogen bond formation are still underexplored^3–7. Among these, compounds containing an azoxy group (N = N → O) have been isolated from different taxa, including plants, fungi, and sponges^6,8–12 exhibiting a wide range of biological activities such as antimicrobial and cytotoxic effects⁸. Therefore, it is not very astonishing that modern organic chemistry also focusses on providing synthetic access to a wide range of these important products^13,14. Azoxy compounds can be synthesized via redox reactions using nitro- or amine-substrates. However, the traditional synthetic methods, such as reductive dimerization of nitrosobenzenes or oxidation of anilines (Fig. 1a), suffer from drawbacks including reliance on hazardous catalysts, poor selectivity and high environmental impact, which are increasingly undesirable in today’s chemical industry¹⁵.

Fig. 1. Methods for azoxy compound formation reactions.

a Chemical methods and (b) the so far established biocatalytic methods for azoxy compound synthesis directly using nitro- or amine-substrates. c The envisioned enzymatic oxidative azoxy compound synthesis by fungal peroxygenases with simplified electron transport chains.

Today, only a few enzymes have been reported to catalyze azoxy compound formation reactions (Fig. 1b), limiting their use in organic synthesis. Interestingly, reductive activation of nitro precursors appears to be the preferred synthetic strategy. Nicotinamide cofactor-dependent nitroreductases for example reduce nitrobenzene substrates, often yielding incomplete reductions and a mixture of reaction intermediates¹⁶. A notable work by Poelarends and coworkers demonstrated a nitroreductase/H₂O₂ system to convert nitrobenzenes into the corresponding azoxy derivates¹⁷. Recently, an oxidative azoxy-group formation step was proposed in the biosynthesis of azoxymycin and O-akylazoxymycins^5,18. The key enzyme (AzoC) was identified as a nonheme diiron monooxygenase (Fig. 1b). A similar oxygenase (VlmB homologes) was reported for participation in the biosynthesis of valanimycin via a cascade reaction (Supplementary Fig. 1)^6,7. However, study on AzoC and VlmB homologes are primarily of an enzymological and biochemical nature, focusing on elucidating biosynthetic pathways and enzyme mechanisms.

Fascinated by these reactions, we wondered if peroxygenases would offer a simpler, more robust and efficient, and most importantly, a scalable access to azoxy compounds from simple amine precursors (Fig. 1c). Over the past decade, unspecific peroxygenases (UPOs, EC 1.11.2.1) have emerged as promising biocatalysts for a wide array of challenging reactions, including selective C-H bond oxyfunctionalization, epoxidation, and heteroatom oxidation (N- or S-oxygenation)^19–22. We hypothesize that UPOs may also function as N-oxygenases, producing generating hydroxylamine- and nitroso-intermediates for spontaneous N-Nbond formation. This strategy would not only expand the reaction repertoire of naturally occurring heme-dependent enzymes but also augment the available enzyme toolboxes for the persistent challenge of azoxy compound synthesis.

Results and discussion

Reaction design and characterization

We began with investigating the unspecific peroxygenase from Agrocybe aegerita (AaeUPO, PaDa-I variant)^23,24 catalyzed transformation of 4-(trifluoromethyl)aniline (1, Fig. 2a) and found straightforward conversion into the desired azoxy product. After 8 h, 89.5% of the substrate was converted into the main azoxy product (1a, 8.2 mM, 94.5%), along with minor amounts of nitro- (0.6 mM, 1.2%) and azo- (0.5 mM, 4.0%) by-products detected (Fig. 2a). For AaeUPO, a total turnover number (TON = moles_product × moles_AaeUPO^-1) of 16900 was determined, aligning well with the generally high catalytic activity of AaeUPO and surpassing the performance of most documented synthetic catalysts (Supplementary Table 1). Control experiments confirmed that the reaction proceeded solely in the copresence of AaeUPO and H₂O₂. Other heme-containing enzymes such as cytochrome P450 BM3 and horseradish peroxidase (HRP) failed to yield the desired azoxy product (Fig. 2a), highlighting its unique catalytic prowess of AaeUPO in facilitating azoxy compound synthesis.

Fig. 2. Investigation of enzymatic azoxy compound synthesis.

Reaction conditions: a For P450 and HRP: [1] = 2 mM, 30 °C, 10 v/v% acetone, 12 h, 1 mL, [P450 BM3] = 1.46 mg mL⁻¹, [NADPH] = 10 mM, NaPi buffer (100 mM, pH 8.0) or [HRP] = 0.5 mg mL⁻¹, citrate buffer (100 mM, pH 5.0), [H₂O₂] = 1 mM h⁻¹. Reaction conditions unless specifically stated: [1] = 20 mM, [AaeUPO]= 500 nM, [H₂O₂] = 3 mM h⁻¹, NaPi buffer (100 mM, pH 8.0), 30 °C, 10 v/v% acetone, 8 h, 1 mL. b The influence of pH on the product formation: pH = 4 − 9. c The influence of H₂O₂ rate: [H₂O₂] = 2 mM h⁻¹ (square), 4 mM h⁻¹ (circle), 5 mM h⁻¹ (triangle), 6 mM h⁻¹ (inverted triangle), 12 h. d The initial reaction rate at varied AaeUPO concentration: [AaeUPO] = 200 – 800 nM, 2 h. e The setup of a 6 L reaction and the isolated product. Selectivity = [product_azoxy] / [products_{in total}] × 100%. The reported value is based on the mean value of triplicate experiments (n = 3).

Next, we further characterized the AaeUPO-catalyzed conversion of 1 to 1a. The pH range of the reaction was astonishingly broad with >75% conversion between pH 4 and pH 9. The highest conversion (96.3%) and selectivity (97.3%) were observed at pH 8 (Fig. 2b). The initial reaction rate depended on both the H₂O₂-addition rate (Fig. 2c) and the AaeUPO concentration (Fig. 2d). As the catalytic performance of AaeUPO was excellent (turnover frequencies of up to 6.7 s⁻¹ and total turnover number of 48,450), the synthetic potential of this reaction was evaluated on 6 L-scale corresponding to a 6000-fold volume increase (Fig. 2e and Supplementary Figs. 2–5). 38.5 g of the desired 1a were obtained within 11 h reaction time in 95.6% isolated yield. It is interesting to note that during this reaction, no intermediate accumulation of the hydroxylamine product was observable (Supplementary Fig. 5). In view of more sustainable chemistry, we performed an E-factor analysis²⁵ for the AaeUPO-catalyzed azoxy compound synthesis. An E-factor of 170 kg_waste kg_product⁻¹ at first sight does not seem very promising (Supplementary Table 2). Solvents i.e. water (82%), acetone (7.2%) and ethyl acetate (6.9%) accounted for >96% of this E-factor paving the way for more efficient and less environmentally damaging syntheses. We are confident that future optimization such as increasing the reagent loading as well as solvent recycling will drastically reduce the wastes generated. For comparison, we reported a rather low E-factor for enzymatic azoxy-synthesis (Supplementary Table 3), suggesting the great potential of this strategy to access useful organic compounds by biocatalysis.

Substrate and enzyme scope investigations

Having established the AaeUPO-catalyzed azoxy compound synthesis reactions, we next explored the product scope of this transformation (Fig. 3 and Supplementary Figs. 6–74). All the 25 aniline derivates tested were transformed in good to excellent conversion with chemoselectivities >70% (generally >90%). Interestingly enough, Hofrichter and coworkers reported predominant ring hydroxylation followed by AaeUPO-initiated polymerization under comparable reaction conditions²⁶. Our experiments, however, showed a clear preference for azoxy compound formation probably due to the pattern of the ring substitution. The arene substitution pattern had little influence on the outcome of the azoxy compound formation and a broad range of electron-donating and -withdrawing substituents were effectively converted. It is worth noting that in case of alkyl-substituted anilines, trace C-H hydroxylation was observed (Supplementary Table 4), especially the benzylic hydroxylation is well-documented for AaeUPO²⁷. Apparently, N-oxidation is highly favored in our findings.

Fig. 3. Substrate scope of AaeUPO-catalyzed azoxy compound synthsis.

The substituted anilines using the following conditions: [substrate] = 20 mM, [AaeUPO] = 640 nM, [H₂O₂] = 4 mM h⁻¹, NaPi buffer (100 mM, pH 8.0), 30 °C, 10 v/v% acetone, 12 h, 1 mL. TON = [1a –31a] / [AaeUPO]. The conversion was based on the area estimation. The yield was calculated from the amount of isolated product in a 50 mL scale, except for 1a in a 6 L, with selectivity indicated in parentheses. ^(a) The reaction time was 20 h. Duplicate experiments were performed (n = 2).

Next to aniline derivates aminopyridine and aminoquinoline derivatives were also smoothly converted, with chemoselectivities ranging from 70% to 98%. Thermodynamically, trans-configured azoxy compounds should be the favored products. This was confirmed in case of 1a, 18a, 19a, 20a and 22a, which were synthesized in semi-preparative quantities, crystallized, and had their crystal structures solved (Fig. 3 and Supplementary Tables 5–9). Hetero-coupling reactions principally are possible as well. The absolute configuration of each asymmetric azoxy compound was assigned by either the corresponding chemical standard or crystal structure (Supplementary Tables 10). However, starting with two different aniline derivatives resulted in a statistical mixture of the four conceivable homo- and hetero-coupling products (Fig. 3, 26a–31a and Supplementary Figs. 75–91). The low selectivity is consistent with the reductive azoxy compound formation²⁸. Additionally, we also tested aliphatic amine substrates including cyclopentylamine, n-butylamine, nonylamine and benzylamine, which, however, did not show any desired azoxy products formation with the copresence of AaeUPO and H₂O₂.

Other unspecific peroxygenases were also investigated. It was shown that Grogu, CciUPO, PabUPO-II and CfuCPO enabled the conversion of anilines to the corresponding azoxy-aromatic products in satisfactory conversion and selectivity (Supplementary Fig. 92), with TONs up to 50,710. The high catalytic turnover numbers and the varied source of microorganisms of the investigated UPOs suggest the potential of incorporating peroxygenases into natural or engineered metabolic pathways for the biosynthesis of azoxy bond containing compounds²⁹.

Mechanistic investigation

To gain mechanistic insights into the discovered azoxy bond formation by AaeUPO, molecular dynamics (MD) simulations and quantum chemical (QC) calculations were conducted using 4-aminobenzotrifluoride (ArNH₂) as a representative substrate. The key steps in the azoxy formation reactions involve the formation of aromatic hydroxylamine (ArNHOH) and nitroso (ArNO) intermediates (Supplementary Fig. 93). In these processes, AaeUPO first undergoes activation by H₂O₂, leading to the conversion of iron porphyrin from compound 0 (Cpd 0) to compound I (Cpd I). The mechanism of this activation process has been well-established in previous studies^30–32. Therefore, MD simulations were performed on the systems where AaeUPO is complexed with ArNH₂ and ArNHOH, respectively, in the presence of Cpd I (Fig. 4a, b). These enzyme-substrate (ES) complex structures show that the aromatic ring of substrates (ArNH₂, ArNHOH) engages in π-π stacking interactions with surrounding residues such as F69, F188 and F191. The -NH₂ and -OH groups form hydrogen bonds with the Cpd I in the corresponding ES structure. Moreover, these groups and Cpd I are also engaged in a hydrogen bond network involving water molecules, R189 and E196.

Fig. 4. The interactions between the substrate and active site residues.

Representative conformations from MD simulations for the enzyme in complex with ArNH₂ (a) and ArNHOH (b) in the presence of Cpd I. For clarity, most of the non-polar hydrogen atoms are omitted. Key distances are shown in angstroms.

For the conversion of aniline, two possible mechanisms, the N-oxidation mechanism and the hydrogen atom transfer (HAT) mechanism were considered because of the previous proposals for similar reactions by cytochrome P450 monooxygenases^33–35. The performed computational analysis reveals its adherence to the N-oxidation mechanism (Fig. 5a), in which the oxidation is initiated by the rate-determining direct attack of Fe=O center at the amino group with a barrier of 16.1 kcal mol⁻¹. Subsequently, the hydroxylamine is formed by an intramolecular proton transfer via two water molecules (Supplementary Figs. 94–95). In comparison, the barrier of HAT mechanism is slightly higher than the N-oxidation mechanism (by 1.4 kcal mol⁻¹, Supplementary Figs. 96–97).

Fig. 5. Mechanistic study of the AaeUPO-catalyzed azoxy compound synthesis from aniline.

The energy profiles for the catalytic oxidation of ArNH₂ to ArNHOH (a), the subsequent oxidation of ArNHOH to ArNO (b), the formation of azoxybenzene from ArNHOH and ArNO in water solution (c), as well as the relative binding energies of anilines in AaeUPO and P450 BM3 enzymes (d). The optimized full structures and spin densities are provided in supplementary materials.

Through two HAT processes, the hydroxylamine is further oxidized to the nitroso product (ArNO) by AaeUPO (Fig. 5b). The first HAT is preferred to take place on the hydroxyl group of the hydroxylamine intermediate, leading to the barrier-less formation of a transient oxygen radical (Supplementary Figs. 98–99). The second HAT occurs with a barrier of 3.3 kcal mol⁻¹, where the hydrogen from the -NH group is transferred to Fe-OH, resulting in the formation of ArNO along with the generation of one water molecule. As such, the calculations predicted that the conversion of ArNHOH to ArNO is rapid. This can be corroborated by the experimental observation that ArNHOH was not accumulated in the reaction course (Supplementary Figs. 100–101). Other considered pathways with higher energies for this conversion are given in Supplementary Figs. 89–90. Once the ArNHOH and ArNO intermediates are formed, they can undergo spontaneously chemical dehydration condensation in solution to generate azoxybenzene with a calculated barrier of 15.7 kcal mol⁻¹ (Fig. 5c, see also Supplementary Fig. 102 for optimized structures).

For the formation of the azoxy compound, QC results indicate that the ArNHOH and ArNO intermediates can readily condense to generate the azoxy product in solution. First, the -N=O group of ArNO undergoes a nucleophilic attack by ArNHOH, concerted with a proton transfer from the -NH group of ArNHOH to the -N=O group of ArNO, mediated by a water molecule. This results in the formation of an N-N bond between ArNHOH and ArNO. Subsequently, the intermediate formed in the first step is dehydrated to form the azoxy compound. The calculated energy barrier for the entire process is 15.7 kcal mol⁻¹. Moreover, the reaction is significantly exothermic, with the energy of the azoxy product being lower than that of the reactants by 31.1 kcal mol⁻¹. Thus, the condensation between ArNHOH and ArNO in solution is both kinetically feasible and thermodynamically favorable.

Although the AaeUPO-catalyzed azoxy compound synthesis showed predominant chemoselectivity, the studies on the observed minor nitro- and azo-products are valuable to decipher the differences between AaeUPO and the P450 monooxygenases. In the case of azobenzene formation by AaeUPO, QC calculations indicate that it is a process with a high energy barrier (26.6 kcal mol⁻¹), consistent with the experimental observation (Supplementary Fig. 105). As for the nitro product, its formation is speculated through the direct oxidation of ArNO by Cpd I. However, MD simulations of AaeUPO complexed with ArNO revealed that ArNO is not preferred to bind in the active site (Supplementary Fig. 106). To assess the chemical feasibility of this oxidation, QC calculations using a simplified model comprising only Cpd I and ArNO were performed. It was shown that Cpd I can readily oxidize ArNO to ArNO₂ with a barrier of 10.5 kcal mol⁻¹, suggesting that the difficulties in forming ArNO₂ indeed stem from the unfavorable binding of ArNO to AaeUPO (Supplementary Fig. 107). This is further supported by the much higher binding free energy of ArNO compared to ArNH₂, by ca 16.6 kcal mol⁻¹ obtained by using thermodynamic integration (Ti) method (Fig. 5d)³⁶. Experimentally, substituting aniline using 4-COCH₃ or 3,4-F instead of 4-CF₃ resulted in increased ratios of the nitro products. Indeed, the binding of these two nitroso compounds are favorable than ArNO characterized by lower relative binding energies (Fig. 5d). This explanation can also be applied to the inability of P450 BM3 to promote the azoxy bond construction. The binding energy of ArNO, compared to ArNH₂, is 8.6 kcal mol⁻¹ lower in P450 BM3 than in AaeUPO (Fig. 5d). This shows that ArNO has a higher affinity for the protein pocket of P450 BM3, making it more readily further oxidized to ArNO₂ by P450 BM3 than by AaeUPO.

In summary, chemical transformations using enzymes represent a frontier in modern chemistry. While protein engineering has proven to be a powerful approach in creating catalysts to expand the boundaries of chemistry, there is still large space to explore the unknown functions of natural enzymes. The latter holds great promise in expanding the reactivity repertoire of the natural biocatalysts³⁷. The envisioned azoxy compound synthesis reaction demonstrates the potential of unspecific peroxygenases as robust catalysts to achieve synthetic reactivities that have not realized by other heme enzymes. Fungal unspecific peroxygenases, a family of heme-dependent enzymes, may be particularly useful to enable efficient and sustainable functionalization of simple anilines to produce valuable azoxy containing chemicals. Furthermore, apart from the enzymatic methodology, the ability to construct diverse functional azoxy bonds can expand their application in agrochemicals, pharmaceuticals, pigments, and liquid crystals. Finally, the obtained structural and energetic insights not only provide valuable understanding of AaeUPO-catalyzed reactions but also offer guidance for the rational design of heme-dependent UPO and P450 BM3 with tailored activity. Notably, the calculations revealed that the less favorable binding of the nitroso-intermediate in AaeUPO, compared to P450 BM3, promotes more efficient synthesis of azoxy compounds over nitro-products. This finding can inform the strategic design of P450 BM3 mutants aimed at modifying product preference.

Methods

General procedures for the enzymatic azoxy compound synthesis

In a 2 mL transparent glass vial, 20 mM of the amine substrate (1) and 640 nM of AaeUPO (crude extract, Supplementary Figs. 110–111) were added to NaPi buffer (100 mM, pH 8) with 10 v/v% acetone. The reaction volume was adjusted to 1 mL after adding all reaction components. Afterwards, the H₂O₂ from an 800 mM stock solution was supplied by a syringe pump at a rate of 4.98 µL h⁻¹, corresponding to a concentration of 4 mM h⁻¹. The reaction vial was agitated in a thermal shaker at 800 rpm and 30 °C. During the reaction course, 100 µL of the reaction mixture was withdrawn and extracted with 200 μL of ethyl acetate (extraction ratio: 1:2) containing the inner standard (dodecane). After mixing and separation by centrifugation (10,000 × g, 1 min), the organic layer of ethyl acetate was dried over anhydrous Na₂SO₄ before analysis by GC or GC-MS. To quantify the conversion of the aniline into the azoxy product, calibration curves were used for the model reaction (Supplementary Figs. 112–114), whereas the other substrates were quantified by the peak area. Triplicate experiments were performed and three independent data were used for analysis. To investigated the substrate scope, in a 100 mL glass bottle, 40 mM of aniline substrate and 640 nM of AaeUPO (crude extract) were added to NaPi buffer (100 mM, pH 8.0) with 10 v/v% acetone. The reaction volume was adjusted to 50 mL using the same buffer. Then, H₂O₂ from an 800 mM solution was dosed by a syringe pump at a rate of 250 µL h⁻¹, corresponding to a concentration of 4 mM h⁻¹. The reaction was stirred at 400 rpms and 30 °C. Upon the completion of the reaction monitored by thin-layer chromatography, the mixture was extracted with 50 mL of ethyl acetate three times and dried over Na₂SO₄. The solvent was evaporated at 40 °C under reduced pressure. The target compounds were purified by flash chromatography (Biotage rapid preparation system) using 0–35% ethyl acetate in petroleum ether. The targeting compounds were analyzed by ¹H NMR, ¹³C NMR and GC-MS. The isolated yields were calculated based on the actual amount of azoxy products compared to the theoretical yield.

Computational study

To gain mechanistic insights into the AaeUPO-catalyzed oxidation of anilines, molecular dynamics (MD) simulations and quantum chemical (QC) calculations were performed using the computational models designed on the basis of the previously solved crystal structure (PDB ID: 6EKZ)³⁸. Details of the computational methods and the technical aspects are provided in the Supplementary Information. The optimized geometries and the energy profiles calculated by QC are shown in Supplementary Figs. 94–99, 103–105 and 107, and the spin densities of the optimized intermediates and transition states are summarized in Supplementary Tables 11–12. The root mean square deviation (RMSD) and distance analysis of the MD simulations are shown in Supplementary Figs. 106 and 108–109.

Reporting summary

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

Author contributions

H.L. developed the enzymatic nitrogen-nitrogen formation system and performed most of the experiments. Y.H., Z.Z., X.Z., X.W. and J.S. assisted the synthetic experiments. F.C., H.S. and X.S. conducted theoretical modeling. P.D. coordinated the project. H.L., F.C., F.H., X.W. and W.Z. cowrote the manuscript, with input from all authors.

Peer review

Peer review information

Nature Communications thanks Pedro Alexandrino Fernandes, Yuan-Yang Guo and the other anonymous reviewer(s) for their contribution to the peer review of this work. A peer review file is available.

Data availability

All data are available in the main text or the supplementary information file. Crystallographic data for the structures reported in this article have been deposited at the Cambridge Crystallographic Data Centre, under deposition numbers 2303483 (1a), 2303495 (18a), 2303477 (19a), 2303494 (20a), 2303524 (22a) and 2376593 (31a). Copies of the data can be obtained free of charge via https://www.ccdc.cam.ac.uk/structures/. Source data are provided with this paper. All data are available from the corresponding author upon request. Source data are provided with this paper.

Competing interests

The authors declare no competing interests.

Supplementary information

The online version contains supplementary material available at 10.1038/s41467-024-52648-0.