Preparative‐Scale Biocatalytic Oxygenation of N‐Heterocycles with a Lyophilized Peroxygenase Catalyst

Abstract A lyophilized preparation of an unspecific peroxygenase variant from Agrocybe aegerita (rAaeUPO‐PaDa‐I‐H) is a highly effective catalyst for the oxygenation of a diverse range of N‐heterocyclic compounds. Scalable biocatalytic oxygenations (27 preparative examples, ca. 100 mg scale) have been developed across a wide range of substrates, including alkyl pyridines, bicyclic N‐heterocycles and indoles. H2O2 is the only stoichiometric oxidant needed, without auxiliary electron transport proteins, which is key to the practicality of the method. Reaction outcomes can be altered depending on whether hydrogen peroxide was delivered by syringe pump or through in situ generation using an alcohol oxidase from Pichia pastoris (PpAOX) and methanol as a co‐substrate. Good synthetic yields (up to 84 %), regioselectivity and enantioselectivity (up to 99 % ee) were observed in some cases, highlighting the promise of UPOs as practical, versatile and scalable oxygenation biocatalysts.


Introduction
The oxygenation of non-activated carbon atoms remains one of the most important reactions in synthetic chemistry, for the synthesis of bulk chemical precursors and the preparation of late-stage oxygenated intermediates and drug metabolites. [1,2] As toxic reagents and/or harsh reaction conditions often needed when using abiotic oxidants, oxygenase enzymes are of significant interest, particularly for the functionalization of complex molecules with high regioand enantioselectivity. [3][4][5][6] Cytochromes P450 (P450s) are known to catalyze the selective oxygenation reactions of a range of substrates [7][8][9] including bulk chemicals such as alkanes [10] and aromatics, [11] as well more complex molecules like steroids, [12] terpenes [13] and pharmaceuticals. [14] Despite considerable research on recombinant P450s expressed in heterologous hosts, their use in preparative scale oxygenation is limited, with only a few industrial examples having been developed. [15][16][17] This is due to challenges associated with the heterologous expression of P450s (particularly eukaryotic P450s), poor turnover rates, low stability, and the mode of electron transport to the active site heme in P450s, which requires a nicotinamide cofactor (NAD(P)/H) and auxiliary redox proteins (e.g. cytochrome P450 reductases and ferredoxin reductases/ ferredoxins).
The development of alternative enzyme systems for hydroxylation that avoid these disadvantages is therefore important. In 2004, Hofrichter and co-workers reported the discovery of heme-dependent unspecific peroxygenases (UPOs) in fungal species, such as the enzyme from Agrocybe aegerita (AaeUPO). [18][19][20] This new class of oxygenase enzymes can catalyze the oxygenation of a range of unactivated carbon atoms in alkanes, [21] alkenes [22] aromatics [20] and drug molecules, [23,24] but crucially, at the expense only of the hydrogen peroxide as the stoichiometric oxidant, without the need for NADPH or auxiliary electron transport proteins. Recently developed heterologous systems for the expression of UPOs [25][26][27][28] have facilitated the application of both AaeUPO and other enzymes to the oxygenation of a number of substrates, including a 115 mmol-scale production of butanol from butane in a bioreactor. [29] These practically simpler hydroxylation biocatalysts can be considered much more suitable than P450s for scalable applications.
This study is focused on exploring the oxygenation of Nheterocyclic compounds (Scheme 1). N-heterocycles feature heavily in pharmaceutical compounds, and have been the target of biocatalytic oxygenations by various alternative systems. [30][31][32][33] The oxygenation of N-heterocycles, like pyridine derivatives 1 and 3, were identified as reactions of interest (Scheme 1A), and hence their oxygenation was explored using AaeUPO, in addition to a much wider array of N-heterocyclic systems including alkyl pyridines, bicyclic N-heterocycles and indoles (Scheme 1B). The results show that, in contrast to cell-free P450-based systems, AaeUPO can be readily applied to the hundreds-of-milligrams scale synthesis of oxygenated N-heterocyclic intermediates, with remarkable promiscuity, when applied as an easy-to-use cell free lyophilized powder. Significantly, the method was used successfully in 27 preparative scale examples (30-810 mg). As expected, reaction yields and selectivities varied depending on the substrate; nonetheless, with moderate-to-good isolated yields generally observed (up to 84 %), and high regioselectivity and enantioselectivity (up to 99 % ee) in several cases, AaeUPO clearly has much promise as a practical, versatile and scalable oxygenation biocatalyst.

Results and Discussion
Prior to this study, the expression of useful amounts of AaeUPO in E. coli had not been reported. Therefore, recombinant production of the enzyme was carried out using the yeast host Pichia pastoris. The gene encoding the 9-point mutant of AaeUPO, described by Alcalde and coworkers [25,26] was cloned into a modified pPICZα vector and equipped with a C-terminal histidine tag to give a variant rAaeUPO-Pada-I-H as described previously. [34] This mutant sequence was chosen as the wild-type AaeUPO was reported not to express well in P. pastoris. [25,26] 19 L of fermentation in Pichia resulted in culture supernatant that was concentrated and lyophilized to give 138 g of powdered enzyme.

Oxygenation of Alkylpyridines
rAaeUPO-Pada-I-H was then used in a first screen with a range of alkylpyridines. Thus, alkylpyridines 5 to 16 (10 mM in 5 mL of 50 mM KPi buffer pH = 7.0) were reacted with the enzyme (Scheme 2) with slow addition of two equivalents of hydrogen peroxide by syringe pump at a rate of 0.08 eq h À 1 (Supporting Information Section 4 Method 1). 4-Alkylpyridines 5-8 were not transformed by rAaeUPO-Pada-I-H, which we propose may be because of coordination of the unhindered N-atom in these substrates to the heme iron of the enzyme. Similarly, 3-picoline 9, 2-picoline 13, 2,4,6-trimethylpyridine 18 and 3-methoxy-2-picoline 19 were not transformed. Interestingly the behaviour with 9 contrasts with that reported for wild-type AaeUPO, in which a mixture of products with oxygenation at both the the methyl group and the N atom was observed 35] although the substrate concentration was in that case was much lower (500 μM). However, alkylpyridines with more shielded N-atoms gave more acceptable conversions. For example, 3-ethylpyridine 10, was converted into a 9 : 1 mixture of the (R)-enantiomer of alcohol 20 with > 98 % ee and the corresponding ketone 21, in 98 % conversion overall. 2-Ethylpyridine 14 was also transformed, albeit with only 3 % conversion into alcohol 22 under these conditions. 3-Isopropyl pyridine 12 and 2isopropyl pyridine 16 were transformed into tertiary alcohols 24 and 25 with 17 % and 3 % conversions respectively. 3-n-Propylpyridine 11 and the 2-n-propyl isomer 15 were converted to benzylic alcohols 26 (4 %) and 27 (15 %). Finally, 2-methoxy 5-methyl pyridine 17 was transformed into alcohol 28 and aldehyde 29 with 30 % and 70 % conversion respectively. The enantioselectivity of most of the transformations that afforded chiral products was not recorded at this point, although as is shown later in the manuscript, the enantioselectivity was very high in all cases measured.
AaeUPO is known to be deactivated by hydrogen peroxide [36] and in our hands rAaeUPO-Pada-I-H lost 50 % of activity even at 1 mM H 2 O 2 , measured using the veratryl alcohol peroxygenase assay (Supporting Information, Sections 5 and 6; Figure S1). [23] In an attempt therefore to reduce the peroxide stress on the UPO, a system for in situ generation of peroxide was evaluated, using alcohol oxidase from Pichia pastoris (PpAOX) and methanol, based on the method previously described by Hollmann and coworkers. [36] Three different amounts of PpAOX (0.3, 1.0 and 3.0 U mL À 1 ) were tested for each substrate (Supporting Information Section 4, Method 2), as it was observed that different substrates required different amounts of PpAOX for high conversions. PpAOX demonstrated no activity towards these or other substrates when applied without rAaeUPO-PaDa-I-H. The best results showed that superior conversions could be achieved in many cases using in situ peroxide generation with selectivity conserved. In this manner, 3-isopropyl pyridine 12 was now transformed into 24 with 80 % conversion and 2-ethylpyridine 14 was hydroxylated to form the (R)-benzyl alcohol 22 and ketone 23 with 36 % and 29 % conversion respectively. The low conversion with 4-alkylated substrates e.g. 4ethylpyridine 6 compared to their 2-and 3-ethyl isomers, such as 14 and 10, was investigated by comparing the inhibition of rAaeUPO-Pada-I-H with these compounds using a competition-based format of the 5-nitro-1,3-benzodioxole assay for UPOs described previously (Figure 1, see Supporting Information Section 7 for further details). [37] In this way, IC 50 values for each of substrates 14, 10 and 6 were obtained (Supporting Information Section 7, Table S1). The very low value of 0.016 mM recorded for 6 is reflective of the strong binding of the available nitrogen in this substrate to the heme iron, militating against productive binding for oxygenation.

Oxygenation of Bicyclic Substrates
The screening approach was extended to the rAaeUPO-Pada-I-H catalyzed biotransformation of bicyclic compounds such as 1, 3 and a range of bicyclic N-heterocycles 30-36 (Scheme 3) using either syringe pump addition of hydrogen peroxide or in situ generation using PpAOX (Supporting Information Section 8).
Products were identified by GC-MS analysis and compared with authentic standards. Scale-up of several of the reactions later (see below) also allowed additional verification of product identities by various characterisation methods. First, 2,3-cyclohexenopyridine 3 was transformed into 5-hydroxy product 4 with 75 % conversion using slow addition of H 2 O 2 . 5,6,7,8-tetrahydroisoquinoline 30 was transformed into a mixture of three regioisomeric alcohols (37 a-c, approximately 1 : 1 : 1 ratio) with a combined conversion of 64 %, also using slow addition of H 2 O 2 . Lower selectivity was revealed in the transformation of 1,2,3,4tetrahydro(iso)quinoline derivatives 31 and 32, with several unstable products (e.g. 38 and 39) formed with either mode of peroxide delivery, which were later found to lead to the formation of oligomers upon scale-up. Quinoline 33 was oxidized at various positions in its benzenoid ring, yielding epoxide 41, its ring-opened derivative 42 and dihydroxy epoxide 43 with 81 % combined conversion with slow addition of H 2 O 2 . The same conditions also saw isoquinoline 34 converted into N-oxide 44 with high selectivity. In contrast, quinaldine (2-methylquinoline) 35 was oxidized on the aromatic ring to form a phenolic product (unknown regioisomer, 32 % conversion) and at the benzylic position, to form 45 with 10 % conversion. Oxidation of 3-methylisoquinoline 36 yielded N-oxide 46 as the major product (82 %) and aldehyde 47 as a minor product (8 %), each again using syringe pump addition of peroxide. The formation of N-oxides from both 34 and 36 is likely indicative of the greater availability of the nitrogen atom for oxidation in isoquinolines.

Transformation of Indoles
Three indole derivatives 48, 49 and 50 were also tested, and each was converted smoothly into 2-oxindole products; 51, 52 and 53 were formed in 100 %, 85 % and 85 % conversions respectively, using slow addition of H 2 O 2 in the case of 48 and 50, and PpAOX in the case of 49 (Scheme 4).

Scaleable Oxygenations Using rAaeUPO-Pada-I-H
To optimize the rAaeUPO-Pada-I-H/PpAOX system for scaleable N-heterocycle oxygenation, a design of experiments (DoE) approach was applied to the transformation of pyrindan 1 into chiral alcohol 2 (Supporting Information Section 9). Notably, 2 is a key intermediate in the synthesis of patented Ghrelin O-acetyltransferase inhibitors, [38] which underlines the utility of AaeUPO in the production of useful chiral feedstock molecules. As input factors, the substrate concentration (5-50 mM), the temperature (20-40°C), the rAaeUPO-H loading (1-10 U mL À 1 ), and the rAaeUPO-H to PpAOX ratio (20 : 1-5 : 1) were chosen. Ten reactions were run using the factors described (Table S2) containing two centre-points for reproducibility and eight factorial points, with a combination of factors. As responses, the GC-based conversion to the desired hydroxylated product 6,7-dihydro-5H-cyclopenta[b]pyridin-5-ol 2 and the GC-based conversion to by-products (54-56) were chosen. The results for the ten runs are shown in Table S3. Analysis of the DoE experiments suggested that the best results for obtaining 2 would be obtained at a 40 mM concentration of 1 at 20°C using 10 U mL À 1 UPO at a 5 : 1 ratio to PpAOX (2 U mL À 1 ). Informed by this result, a scale-up reaction using 40 mM of 1 on a 300 mg scale was performed using 10 U mL À 1 rAaeU-PO-PaDa-I-H at a 5 : 1 ratio with PpAOx, at 20°C. After 115 h, GC analysis suggested that products 2, 54 and 55 were formed with 63 %, 4 % and 17 % conversion, as well as a small amount (2 %) of the product 56, with hydroxylation in the 6-position (Scheme 5). Following purification using flash chromatography (R)-2 was isolated in 34 % yield and with 88 % ee.
Following the success of this approach, the rAaeUPO-PaDa-I-H/PpAOX system was then applied to other preparative scale reactions, starting with the oxygenation of alkyl pyridines 10, 14, and 15 on scales ranging between 100-810 mg of substrate (Table 1, and Supporting Information Section 10). All yields quoted refer to isolated material following column chromatography. It was not always possible to fully separate the products in cases where mixtures of products were obtained, with full details of the isolated materials for all preparative scale reactions included in the Supporting Information.
Next, a range of disubstituted alkyl pyridines was tested, in order to explore substituent and electronic effects ( Table 2). 2-Methoxy-5-methylpyridine 17 was converted into primary alcohol 28 and aldehyde 29 with 31 % and 47 % isolated yields respectively. These results confirm that the biocatalyzed oxygenation method can be applied to methyl pyridines as well as ethyl pyridines. Pyridine 60, which contains both an ethyl and methyl group, was oxygenated primarily on the ethyl benzylic position to form 61 a in 73 % yield, with a low (3 %) yield of 61 b. Dimethyl pyridine 62 showed clear selectivity for oxygenation on the meta-rather than ortho-methyl group, affording oxidised products 63 a-c in 84 % combined yield. 3,5-Dimethyl pyridine 64 afforded a mixture of two oxidised products 65 a,b.
2,5-Disubstituted pyridines were the most effective substrates tested: for example, electron rich substrate 66 affording alcohol 67 a in high ee, alongside over-oxidised ketone product 67 b in good combined yield. Electron deficient 2,5-disubstituted pyridines also proved to be effective substrates, with ethyl pyridines 68-71 all being selectively converted into secondary alcohols 72-75 in good isolated yields (54-84 %) and with excellent ee (up to > 99 %) in all cases measured.
The preparative oxygenation of various bicyclic Nheterocycles was then examined on scales ranging from 30-530 mg. In these cases, a relatively high amount of PpAOX was generally required to reach synthetically useful yields when using this H 2 O 2 generation mode (H 2 O 2 method A). This prompted us, in most instances, to perform reactions with slow H 2 O 2 addition (H 2 O 2 method B), with superior products yields usually observed using this method; for example, substrate 3 was hydroxylated predominantly at the 5-position to give (R)-alcohol 4 (> 99 % ee) in both cases, but in either 28 % or 82 % isolated yield using PpAOX or slow H 2 O 2 addition respectively. Interestingly, the identity of the side products formed was different depending which peroxide source was used, as the 6-hydroxy product 76 was formed (although in only 2 % yield) when using PpAOX, whereas the 7-hydroxy compound 77 was formed (in 13 % isolated yield) when slow addition of H 2 O 2 was employed (Scheme 6).
A range of other bicyclic N-heterocycles was also examined in preparative scale reactions (Table 3). In the case of tetrahydroisoquinoline 30, the enzyme was selective for oxygenation to form alcohol products, albeit with relatively low regioselectivity; benzylic isomers 37 a and 37 c and the 7-hydroxy compound 37 b were recovered in similar yields of 12 %, 22 % and 15 % respectively. Oxidation of quinoline 33 gave the epoxide 41 and the dihydroxy epoxide 42 in 60 % and 21 % isolated yields respectively. The oxidation of isoquinoline 34 on the same scale gave the N-oxide 44 and dihydroxy compound 78 in 34 % and 20 % yields, respectively. These results contrast with the hydroxylation of quinoline 34 by P450 BM3 variants, which furnished only aromatic hydroxylation products. [33] Noxide formation was also the major reaction pathway when methyl-substituted isoquinoline 36 was tested, with N-oxide 46 and aldehyde 47 obtained in 59 % and 12 % yield respectively.
Attention then turned to electronically diverse tetrahydroquinoline substrates. Substrates 79, 81 and 83 (which all contain electron poor 2-substitutents) were selectively oxygenated (with no apparent over-oxidation to ketones), but with modest regioselectivity, forming alcohol products (80 ac, 82 a-d, 84 a-c). In contrast, the analogous methoxysubstituted substrate 85 was hydroxylated with complete regioselectivity, affording alcohol 86 as a single regioisomer. Comparable regioselectivity was not conserved when switching to 3-substituted substrates 87 and 89, with regioisomeric mixtures 88 a-c and 90 a-c formed, although the selectivity for oxygenation to form alcohol over ketone products was again high.
Finally, preparative scale indole oxygenation reactions were performed (Table 4). Indoles 48 and 49 were converted into 2-oxindole derivatives 51 and 52 in 60 % and 75 % yield respectively. In addition, 3-methyl-2-oxindole 50 was oxidized to 2-oxindole 53 in 50 % yield, but also formed 3hydroxy derivative 63 in 29 % yield. This chiral alcohol product was found to be racemic, which may indicate that oxidation proceeds through epoxidation of the enol tautomer of 53 (i.e. a 2-hydroxyindole), followed by epoxide opening to form a cation at the indole 3-position and trapping by water, thus eroding any enantioselectivity imparted during the epoxidation.

Conclusion
In summary, a lyophilized rAaeUPO-PaDa-I-H has been shown to be an effective and promiscuous biocatalyst for the practical oxygenation of a wide range on N-heterocyclic substrates, showcased by 27 preparative scale biotransformations (30-810 mg). Various reactivity/selectivity trends have been identified, isolated yields are moderate-to-good in most cases, and enantioselectivity is generally excellent.
These oxygenation reactions offers several advantages in terms of selectivity and sustainability over conventional chemical methods. When comparing rAaeUPO-PaDa-I-H with established P450 oxygenations, it is true that P450s can offer superior selectivity in some cases, especially where mutant libraries give scope for improvements. [39,40] However, this must be considered against the requirement for the nicotinamide cofactor and electron transport proteins when using P450s. The results in this report certainly suggest that UPOs offer major advantages with respect to scalability, and furthermore it is anticipated that selectivity may be improved through reaction engineering as we learn more about this enzyme class; e.g. by optimising the mode and rate of hydrogen peroxide delivery (as demonstrated herein), as well as through mutational engineering of the UPOs, which has been used to improve selectivity and performance  in related biotransformations. [29,41] The simplicity of the reaction systems overall is another advantage, as is our ability to generate large amounts of a robust, easy-to-use lyophilised biocatalyst. Taken together, these features all point towards an extremely promising future for UPO catalysis for use in the scalable, biocatalytic oxygenation of industrially relevant molecules.