An efficient and regioselective biocatalytic synthesis of aromatic N‐oxides by using a soluble di‐iron monooxygenase PmlABCDEF produced in the Pseudomonas species

Summary Here, we present an improved whole‐cell biocatalysis system for the synthesis of heteroaromatic N‐oxides based on the production of a soluble di‐iron monooxygenase PmlABCDEF in Pseudomonas sp. MIL9 and Pseudomonas putida KT2440. The presented biocatalysis system performs under environmentally benign conditions, features a straightforward and inexpensive procedure and possesses a high substrate conversion and product yield. The capacity of gram‐scale production was reached in the simple shake‐flask cultivation. The template substrates (pyridine, pyrazine, 2‐aminopyrimidine) have been converted into pyridine‐1‐oxide, pyrazine‐1‐oxide and 2‐aminopyrimidine‐1‐oxide in product titres of 18.0, 19.1 and 18.3 g l‐1, respectively. To our knowledge, this is the highest reported productivity of aromatic N‐oxides using biocatalysis methods. Moreover, comparing to the chemical method of aromatic N‐oxides synthesis based on meta‐chloroperoxybenzoic acid, the developed approach is applicable for a regioselective oxidation that is an additional advantageous option in the preparation of the anticipated N‐oxides.


Summary
Here, we present an improved whole-cell biocatalysis system for the synthesis of heteroaromatic N-oxides based on the production of a soluble di-iron monooxygenase PmlABCDEF in Pseudomonas sp. MIL9 and Pseudomonas putida KT2440. The presented biocatalysis system performs under environmentally benign conditions, features a straightforward and inexpensive procedure and possesses a high substrate conversion and product yield. The capacity of gram-scale production was reached in the simple shake-flask cultivation. The template substrates (pyridine, pyrazine, 2aminopyrimidine) have been converted into pyridine-1-oxide, pyrazine-1-oxide and 2-aminopyrimidine-1oxide in product titres of 18.0, 19.1 and 18.3 g l -1 , respectively. To our knowledge, this is the highest reported productivity of aromatic N-oxides using biocatalysis methods. Moreover, comparing to the chemical method of aromatic N-oxides synthesis based on meta-chloroperoxybenzoic acid, the developed approach is applicable for a regioselective oxidation that is an additional advantageous option in the preparation of the anticipated N-oxides.

Introduction
Aromatic N-oxides (ArN?O) have been widely used as structural motifs in various fields due to their increased reactivity towards either electrophilic or nucleophilic agents compared to regular nitrogen heterocycles (Mfuh and Larionov, 2015). The appropriate N-oxides of pyridines, pyrazines, pyrimidines and quinolines are precursors for a huge variety of C2-functionalized bioactive compounds used in medicine today (Wang et al., 2016). No less important are N-oxides themselves as N ? O moiety is present in various pharmaceuticals, chemical catalysts, agriculture agents and pyrotechnic compounds, and they are promising units in future studies as well (Dyer et al., 2018). Quinoxaline 1,4-di-N-oxide derivatives show a broad range of biological properties, including anti-tubercular, antimicrobial, antitumoral, antitrypanosomal and antiinflammatory activities, thus possess potential applications in human and veterinary medicines (Srinivasarao et al., 2020). In silico molecular investigation showed that pyridine N-oxide-based antiviral compounds are potential inhibitors against human SARS infection, also they are more potent than common antiviral drugs chloroquine and hydroxychloroquine (Ghaleb et al., 2020). Chiral heteroaromatic N-oxides have been intensively studied over the last few years as excellent organocatalysts for stereoselective reactions, including allylation, propargylation, allenylation, ringopening of meso-epoxides and many more (Wrzeszcz and Siedlecka, 2020). Also, heterocyclic N-oxide-based compounds have been investigated as possible fluorogenic scaffolds that can be easily applied in the design and synthesis of small-molecule fluorescent probes (Ma et al., 2020). However, the most common methods for ArN?O synthesis such as oxidation with metachloroperoxybenzoic acid (mCPBA) (laboratory scale) or hydrogen peroxide (industrial scale) are hazardous and possess some serious limitations (V€ or€ os et al., 2014). Also, ArN?O can be prepared by various rearrangements and ring-closing reactions (Chucholowski and Uhlendorf, 1990;Nesi et al., 1992); nevertheless, the existing organic chemistry methods are not able to fully address existing demands and, usually, they do not match the increasing environmental requirements. Therefore, the establishment of biocatalytic methods for ArN?O production seems a promising alternative that can offer selectivity and sustainability. There are several reports on the transformation of N-heteroaromatic compounds into N-oxides both by whole-cell biocatalysts (Kutanovas et al., 2013;Mitsukura et al., 2013;Stankevi ci ut _ e et al., 2016) and application of purified enzymes (Ullrich et al., 2008). Most recently, metabolic engineering approaches that offer biosynthesis of specific pyrazine and phenazine N-oxides using cellular components as starting materials have been introduced (Guo et al., 2020;Morgan and Li, 2020). Despite the recent progress, the application of biocatalytic methods is limited by the complicated synthesis procedure, biocatalyst's inactivation and insufficient substrate scope. However, the main issues remain the low productivity, practicality and usability as none of the described biocatalytic approaches seriously challenges chemical oxidizers used in a laboratory-scale production of ArN?O.
In this study, we describe a new whole-cell biocatalysis system for ArN?O synthesis based on production of soluble di-iron monooxygenase PmlABCDEF. By substituting E. coli host cells with certain Pseudomonas strains, we simplified the biotransformation procedure and vastly increased the productivity. Various pyridine, pyrazine and pyrimidine derivatives could be converted into appropriate N-oxides on a preparative scale in the simple shake-flask cultivation. This improved synthesis method was used to yield specific pyrazine and pyrimidine oxidation products that were difficult to obtain by employing a standard mCPBA-based technique.

Selecting a host for PmlABCDEF monooxygenase
Non-heme iron-dependent monooxygenases are multicomponent enzymes, which utilize dioxygen to catalyse a variety of different reactions including hydroxylation, epoxidation, sulphoxidation and require NAD(P)H as an electron donor (Leahy et al., 2003). They are known to be immensely challenging for heterologous protein biosynthesis, which often leads to relatively low activity in comparison to some other oxygenases, thus diminishing the potential biotechnological application (Torres Pazmiño et al., 2010). Despite our recent success in producing enzymatically active soluble di-iron monooxygenase in E. coli and applying it for N-oxidation purposes (Petkevi cius et al., 2019), the issues concerning productivity have not been fully addressed. Although the formation of pyrazine N-oxide was increased by employing the bioreactor fermentation, we were not able to boost the specific N-oxidation activity in the simple shake-flask biotransformation by any measures. Thus, we sought a microbial host that would be more suitable for the synthtesis of PmlABCDEF enzyme and the oxygenase-based whole-cell biocatalysis in general. Pseudomonas species feature a flexible redox metabolism, they show very high robustness against extreme environmental conditions or the presence of toxic substrates/products and inhibiting solvents and as a result, they have been intensively studied for potential application in industrial biotechnology (Poblete-Castro et al., 2012). Our early attempts to produce PmlABCDEF monooxygenase in various Pseudomonas species identified two successful cases. One is Pseudomonas putida KT2440, a versatile microbe, well known for its use in whole-cell biocatalysis (Poblete-Castro et al., 2012). The other selected host, Pseudomonas sp. MIL9, came from our laboratory's collection of microorganisms. Based on 16S rRNA phylogenetic analysis (Fig. S1), the closest relative species are Pseudomonas umsongensis and Pseudomonas moorei both of which are members of P. jessenii subgroup in genus Pseudomonas (Gomila et al., 2015). A particular interest in this microbe as a host was the presence of nonheme di-iron monooxygenase-like genes (GenBank JAFEHE010000000, contig_00034). Apparently, due to a frameshift mutation, these genes do not translate into catalytically active enzyme (based on the inability of a wild type strain to perform any non-heme di-iron monooxygenase-related transformations and a lack of the enzymatic activity of the recombinant proteins (data not shown). However, we speculated that the presence of such genetic material indicated the capability of Pseudomonas sp. MIL9 to provide an excellent cellular machinery for the biosynthesis of the heterologous nonheme di-iron monooxygenase such as PmlABCEDF. Thus, we opted to investigate and compare E. coli and both Pseudomonas biocatalysts for their performance in the production of aromatic N-oxides.

Biotransformation efficiency in different bacterial species
The comparison of biotransformation effectiveness between strains was made under varying reaction conditions. The main differences occurred in plasmid vectors, which were used to express pmlABCDEF. Pseudomonas species harboured a hybrid, arabinose-inducible vector, which was made by fusing regulatory elements of pBAD24 into the chassis of pBBR1MCS plasmid.
Although this expression vector was tested in E. coli (E. coli bw25113), the performance of conversions did not match those of E. coli BL21 transformed with pET_pmlABCDEF plasmid; thus for further experiments, the later one was used for E. coli instead. Pyridine, pyrazine and 2-aminopyrimidine were chosen as model substrates for the reason that they are the key building blocks of substrate scope described in this study. We compared the reaction parameters between resting and growing cells. Incubation temperature of 30°C was used for resting cells assay while growing cells biocatalysis was performed at 20 and 30°C, respectively. All strains were cultivated in 100 ml flat-bottomed flasks with 20 ml of LB medium containing 1.0% of glucose. For a resting cell assay, a final concentration of 2 g DCW l -1 was used for all biocatalysts. Utilizing resting cells of E. coli as biocatalysts and an initial concentration of substrate of 75 mM, pyridine, pyrazine and 2-aminopyrimidine were converted into pyridine-1-oxide (PNO), pyrazine-1-oxide (PyrazNO) and 2-aminopyrimidine-1-oxide (2ANO) with an average conversion of 17%, 18% and 12%, respectively (Fig. 1). The increase in conversion was observed when Pseudomonas resting cells were used instead. Under these conditions, P. putida KT2440 strain was able to convert 34% of pyridine, 28% of pyrazine and 25% of 2-aminopyrimidine. Quite similar results were obtained, when resting cells of Pseudomonas sp. MIL9 were employed. This biocatalyst transformed pyridine, pyrazine and 2-aminopyrimidine into appropriate Noxides with conversion yields of 30%, 33% and 22%, respectively. Interestingly, growing cells at 20°C for both E. coli and Pseudomonas did not show any noteworthy increase or decrease in conversion effectiveness compared to analogous reactions performed with the resting cells at 30°C.
However, the critical point was reached performing biotransformations when growing cells were used at 30°C. The biosynthesis of PmlABCDEF in E. coli BL21 cells was shown to be susceptible to cultivation temperature as 20°C being optimal; induction at 30°C results in complete loss of N-oxidation capacity. Unlike whole cells of E. coli BL21, both Pseudomonas strains, maintained catalytic activity at 30°C if the growing cells were used for biocatalysis. Moreover, the later approach resulted in the boost of a conversion degree. Growing cells of P. putida KT2440 on average converted 86% of pyridine, 96% of pyrazine and 92% of 2-aminopyrimidine when an initial concentration of 75 mM of the substrate was used. Apparently, for this biocatalyst, the used concentration of substrates did not affect the cell growth (average cell concentration after 24 h of biotransformation was 2.8 g CDW l -1 , 2.9 g CDW l -1 and 2.6 g CDW l -1 in the reaction broth with PNO, PyrazNO and 2ANO, respectively) as it was comparable to that without any substrate (2.8 g CDW l -1 ). On the other hand, whole cells of Pseudomonas sp. MIL9 completely transformed 75 mM of pyrazine to a corresponding pyrazin-1-oxide with traceable amounts of pyrazin-1,4-dioxide suggesting that even higher concentrations of pyrazine can be used. However, this biocatalyst seemed to be affected by high concentrations (75 mM) of pyridine and 2aminopyrimidine as the cell growth was weakened (average cell concentration after 24 h of biotransformation was 1.1 g CDW l -1 in the reaction broth with PNO and 1.3 g CDW l -1 with 2ANO) compared to pyrazine conversion (1.9 g CDW l -1 ) and growth without substrate (2.0 g CDW l -1 ) ( Fig. 1). Nevertheless, 59% conversion of pyridine and 68% for 2-aminopyrimidine were substantially higher than those achieved by resting cells or growing cells at 20°C. We opted to elucidate the best performance conditions for N-oxidation using Pseudomonas growing cells producing PmlABCDEF.

Optimization of biotransformation in Pseudomonas strains
The comparison of conversions using different strains suggested two key parameters for improvement: (i) duration of biotransformation and (ii) a proper way of substrate addition. A relatively high (75 mM) starting concentration of a substrate, in some cases, inhibited biomass growth that most likely affected the overall conversion. Also, unlike the system with E. coli BL21 cells, the duration of the reaction was not limited to 24 h using Pseudomonas cells as in both cases bacteria still featured catalytic activity. Moreover, the supplementation of the reaction broth with glucose to regenerate redox equivalents such as NADH seemed to have only a positive effect since it was able to improve cell growth as well as the performance of biotransformation. After a rough optimization, glucose was introduced in portions, starting with an initial concentration of 1.0% and a constant load of an additional 0.5% before every overnight cultivation ( Fig. 2). Hence, ArN?O production was performed with growing Pseudomonas cells complying with their maintenance needs and substrate supplementation until the end of productive biotransformation. All reactions were performed in 100 ml flat-bottomed flasks containing 20 ml of LB and 1.0% of glucose. Extra portions of glucose (250 µl of 40% solution) were added before every phase of overnight cultivation. Gene expression was initiated by adding 200 µl of 20% L-arabinose solution at the beginning of the exponential growth (5 h from the start of the cultivation). This was followed by the supplementation of 0.6 mmol of an appropriate substrate, resulting in a final concentration of 30 mM. The progress of the bioconversion was monitored by TLC and HPLC-MS to keep the track of substrate consumption as well as to maintain constant substrate feed. For the most part, supplementation of 0.6 mmol of the substrate was executed every 5-6 h, followed by the addition of 0.8-1.0 mmol for each phase of overnight cultivation. Biocatalysis using Pseudomonas sp. MIL9 cells reached their peak after 43 h of biotransformation (48 h of total cultivation) as no significant production increase was detected beyond this point (Fig. 2). The total amount of substrates added into the reaction reached 3.6 mmol, which would correspond to a total concentration of 180 mM. Pyridine exhibited the average conversion of 91%, pyrazine -87% and 2-aminopyrimidine -72%. The determined final product concentrations were 157 mM for PNO, 141 mM for PyrazNO and 119 mM for 2ANO (Table 1). Apparently, in this assay, the presence of substrates or an accumulation of products did not influence the cell growth as average cell concentrations (2.0 g CDW l -1 , 2.2 g CDW l -1 and 1.8 g CDW l -1 for reactions with PNO, PyrazNO and 2ANO, respectively) were similar to ones reached during cultivation without any substrate (2.1 g CDW l -1 ). On the other hand, whole cells of P. putida KT2440 demonstrated the catalytic activity for a longer period. The biotransformation was called complete after 67 h from the first substrate addition, as no substantial product formation was detected afterwards. During this period, a total of 4.6 mmol of each substrate were added, corresponding to an end concentration of 230 mM. The average conversion of 87% was observed for pyridine, 98% for pyrazine and 79% for 2aminopyrimidine. After measuring the final product concentrations, we found out 189 mM of PNO, 199 mM of PyrazNO and 165 mM of 2ANO. During this type of biotransformation, there was no significant impact on the growth of P. putida KT2440 strain regarding the accumulation of reaction products. The cultivation of P. putida KT2440 under analogous conditions only without any addition of a substrate resulted in an average cell concentration of 4.8 g CDW l -1 , which was similar to cultivations when N-oxides were produced (4.7 g CDW l -1 for PNO, 4.4 g CDW l -1 for PyrazNO and 4.1 g CDW l -1 for 2ANO). Because of the longer biotransformation time, utilizing P. putida KT2440 whole cells resulted in greater product titres, though Pseudomonas sp. MIL9 biocatalyst featured slightly higher average productivity (Table 1). The clear differences between these biocatalysts regarding conversion performances appeared in the first phase (0-30 h) of the biotransformation. During this period, P. putida KT2440 cells exhibited N-oxide production rates of 0.28 g l -1 h -1 for PNO, 0.24 g l -1 h -1 for PyrazNO and 0.31 g l -1 h -1 for 2ANO, and they were similar to the average productivity of the complete conversion. However, Pseudomonas sp. MIL9 strain displayed productivities of 0.39 g l -1 h -1 for PNO, 0.39 g l -1 h -1 for PyrazNO and 0.37 g l -1 h -1 for 2ANO (Fig. 2). Interestingly, it seemed that the catalytic activity was maintained mainly during the exponential growth and begun to reach a plateau through the stationary phase. It seems that a biooxidation capacity closely correlates with the growth phase in which cells stay metabolically active and are capable of the regenerating of reducing equivalents such as NADH since the addition of glucose at the stationary phase did not enhance the conversion.

Biocatalysis versus chemical oxidation
Once we established the biocatalysis system for preparative scale ArN?O synthesis, we aimed to challenge mCPBA-based oxidation as this is the most typical chemical reagent used for the laboratory-scale synthesis of ArN?O (V€ or€ os et al., 2014). For some pyrazine derivatives, N-oxidation with mCPBA results in unwelcome side reactions including the formation of di-Noxides or isomeric mono-N-oxides as in the case of asymmetric pyrazines bearing various alkyl substituents (Sato, 1985;Butler and Cabrera, 2013). However, di-Noxide formation from compounds 2a and 6a was easily handled in the case of a biocatalytic method (Fig. 3). A simple monitoring of the biotransformation was enough as di-N-oxides started to form after the prolonged exposure when a full conversion of the initial substrate has already been achieved. It seemed that to some degree, PmlABCDEF-based oxidation was able to deal with regioselectivity issues as well. Asymmetric pyrazine derivatives 6a, 11a and 13a were converted by a biocatalytic N-oxidation into single products 6b, 11b and 13b, respectively, all of which bear an oxygen moiety at the less hindered nitrogen. For compounds 11a and 13a, this must be driven more by the chemical nature (deactivating lone-pair electrons of nitrogen by ortho electronwithdrawing substituents) rather than enzymatic preference. Nevertheless, the biocatalytic oxidation of 6a was an explicit example of regioselective synthesis as only mono-N-oxide 6b was produced. Together with the previously described production of 3,5-dimethylpyrazine 1-oxide and 2,3,5-trimethylpyrazine 1-oxide (Petkevi cius et al., 2019), this type of biocatalysis makes a strong case for the selective synthesis of asymmetric alkylpyrazine N-oxides.
Various monosubstituted pyrimidines and methylpyrimidines, in general, are very susceptible to side reactions (decomposition, annular carbon oxidation, ring-opening) accompanying N-oxidation in the presence of different peracids (Jovanovic, 1984). Moreover, unsymmetrical pyrimidines can potentially yield two isomeric compounds making all of these compounds very challenging to oxidize selectively. In this study, the substrates 7a and 12a were transformed to the appropriate single products employing both synthesis methods, biocatalysis and mCPBA oxidation. Upon close investigation of literature (Jovanovic, 1984) and NMR data, those products were identified as 7b and 12b, respectively; an attack to less hindered nitrogen atom was observed. This was also supported by our previous investigation where it was shown that particular features of undesirable substrates for PmlABCDEF included two orthosubstituents. Unlike chemical oxidation, a biocatalytic conversion of compounds 4a, 14a and 17a produced only a single product in each sample. As expected, compounds 4a and 14a were transformed into N-oxides 4b and 14b bearing oxygen moiety at the N-1 position. Interestingly, HPLC-MS data of the bio-oxidation product of 17a did not match those of 17b and 17c that were produced by chemical oxidation. The analysis of NMR data revealed the formation of compound 17d (4-amino-2-methoxypyrimidin-5-ol), a case where an expected Noxidation shifted to a ring hydroxylation at C-5 position instead. Although it is an exceptional case in PmlABCDEF-based biocatalysis, this is a noteworthy example for further studies as a selective ring hydroxylation in heteroaromatic compounds is a very difficult reaction to achieve using the known methods of chemical synthesis.

Discussion
The results of this study not only provide important data on the biocatalytic synthesis of ArN?O but also might share some insights on the biotechnological application of the non-heme di-iron monooxygenases in general. The ability of these biocatalysts to oxidize a variety of different compounds for biocatalytic applications has been widely studied, though little progress has been made to produce them in various recombinant hosts such as E. coli or Pseudomonas putida (Torres Pazmiño et al., 2010). Thus far, monooxygenases such as soluble methane monooxygenase (sMMO) of Methylosinus trichosporium OB3b (Sullivan et al., 1998) (Nichol et al., 2015). However, the aforementioned studies were focussed on investigating genetic or enzymatic properties rather than the application for the synthesis of targeted compounds due to the low productivity of biocatalytic systems. This has been an indication that there are additional host factors that are vital for optimal monooxygenase function, and it was demonstrated in the study of three-component alkane monooxygenase (AlkB) of Pseudomonas oleovorans GPo1. It was shown that despite considerable production of alkane monooxygenase in some E. coli hosts, the most active recombinants showed in vivo alkane-oxidation rates no higher than that of the native host strain (Staijen et al., 2000). Authors suggested that the intracellular environment (protein misfolding and incorrect processing, incomplete iron incorporation, the improper ratio of enzyme components) places significant and rather specific restrictions on the synthesis, stability and activity. The importance of host selection was also demonstrated in the recent study of di-iron monooxygenases, where a three-component monooxygenase from Rhodococcus wratislaviensis was produced as an active form in recombinant Rhodococcus erythropolis cells and applied for the synthesis of amethyl-D-serine (Hibi et al., 2021). In our case, the heterologous production of PmlABCDEF in E. coli resulted in the biosynthesis of active form only at 20°C or lower, while induction at 30°C produced insoluble proteins and exhibited no catalytic activity (Fig. S2). On the contrary, biosynthesis of PmlABCDEF as well as catalytic activity was favoured at 30°C in Pseudomonas hosts. We should keep in mind that PmlABCDEF shares high similarity to counterparts of Pseudomonas genus as it shows overall~90% amino acid sequence identity to YHS domain-containing proteins as well as putative monooxygenases of various Pseudomonas species implying pmlABCDEF originated from this genus. Thus, we can make an educated guess that re-introducing pmlABCDEF genes into specific Pseudomonas hosts provided a suitable intracellular environment and enabled us to achieve stable and efficient protein biosynthesis. Also, unlike the metabolism of glucose in E. coli, Pseudomonas strains are known to employ the Entner-Doudoroff pathway, the Embden-Meyerhof-Parnas pathway and the pentose phosphate pathway for the metabolism of glucose (Nikel et al., 2015). Ultimately, this biochemical network favours the processing of biochemical resources towards the production of NAD(P)H rather than the generation of ATP (Mart ınez-Garc ıa and de Lorenzo, 2019). The elevated levels of reductive equivalents make Pseudomonas species a phenomenal host for redox-intensive reactions such as PmlABCDEF-based N-oxidation. Interestingly, the whole-cell systems, including those exploiting non-heme di-iron monooxygenases, have been attributed to ArN?O production, which helps to put into perspective the capabilities of the presented synthesis method. For instance, oxidation of pyridine to pyridine N-oxide has been demonstrated employing phenoldegrading bacteria Diaphorobacter sp. J5-51, Acinetobacter sp. SJ-15, Acinetobacter sp. SJ-16, Acidovorax sp. J5-66 and Corynebacterium sp. JOR-20 as this activity was assigned to the presence of phenol monooxygenases (Sun et al., 2014). After 70 h of cultivation under optimum conditions, 100 mg l -1 (1.3 mM) of pyridine was completely transformed. Additionally, phenol hydroxylase gene with six components pheKLMNOP was heterologously expressed in the non-pyridine-degrading Pseudomonas sp. CO-44 that resulted in detection of pyridine N-oxide, though any noteworthy boost of productivity has not been stressed out. The capability of Burkholderia sp. MAK1 whole cells to oxyfunctionalize a variety of pyridine derivatives, including several N-heterocyclic ring systems to corresponding N-oxides, was linked to the activity of atypical non-heme di-iron monooxygenase HpdABCDE (Petkevi cius et al., 2018). The induced resting cells possessed a capacity to consume up to 0.5-1.0 mM concentration of a substrate in 100 ml reaction buffer over 24-48 h (Stankevi ci ut _ e et al., 2016). However, the expression of monooxygenase gene in hpdABCDE-minus mutant did not significantly improve catalytic capabilities. More advanced biocatalysis systems for ArN?O synthesis include Verticillium sp. GF39 fungi cells (Mitsukura et al., 2013), Rhodococcus jostii TMP1 whole cells (Kutanovas et al., 2013) and artificial metabolic pathway in Pseudomonas chlororaphis HT66 (Guo et al., 2020). The later approach produced 1hydroxyphenazine N 0 10-oxide in product titre of 143.4 mg l -1 , Verticillium sp. GF39 formed 5 mM 1methylisoquinoline N-oxide from 1-methylisoquinoline with a molar conversion yield of 100% after a 10 h incubation at 20°C and Rhodococcus jostii TMP1 apparently was able to catalyse N-oxidation of 2,3-dimethylpyrazine and 3-ethylpyridine utilizing 8-10 mM concentration of a substrate in 96 h. Keeping all that in mind, the application of PmlABCEDF monooxygenase stands out as an exceptional biocatalysis system for producing ArN?O. Not only it features a simplified procedure and broad substrate spectrum, the recent upgrade improved the productivity up to the level that it could rival chemical synthesis methods.

Conclusions and future perspectives
The results presented in this study show that production of PmlABCDEF monooxygenase in Pseudomonas putida KT2440 and Pseudomonas sp. MIL9 strains serves as a powerful bio-oxidation catalyst. The use of two different biocatalysis systems enabled diverse synthesis approaches for aromatic N-oxides. Pseudomonas sp. MIL9 was preferred for quick and productive biotransformation, while Pseudomonas putida KT2440 was used in case maximum production of the targeted compound was needed. The later strain is a well-described biocatalyst; thus, additional improvements for ArN?O production, including metabolic engineering and gene editing, are possible. It seems that Pseudomonas sp. MIL9 strain is also a promising host for all sorts of biotransformation. Most recently, the genome analysis of the closely relative strain Pseudomonas umsongensis GO16 has shown the metabolic versatility and the potential for biotransformations (Narancic et al., 2021). Since the genome sequencing data of Pseudomonas sp. MIL9 are also available (GenBank: JAFEHE010000000); further investigations, including the comparison and analysis of genetic data, should help to develop new biotransformation platforms.
The recombinant Pseudomonas strains were able to catalyse N-oxidation of various pyridine, pyrazine and pyrimidines with high conversion and product yield. Compared with known biocatalysis techniques for aromatic N-oxide production, the presented approach emerges as the most productive biotransformation method by far. The achieved level of productivity allows using of this method for practical application on the laboratory-scale level and in some cases to replace a more common mCPBA-based oxidation if a certain regioselective synthesis is required. The development of a suitable biocatalysis chassis allows key decisions on the design of this biotransformation method to be made more rapid and effective under conditions that are closer to industrial settings. Similar to other microbial-based industrial processes, a transition from a test tube reaction to a small bioreactor frequently encounters serious obstacles. In this case, a big challenge lays underneath the difficulty of downstream processing for recovering synthesized N-oxides of interest. A possible solution may be a proper two-phase liquid-liquid cultivation system with an aqueous and an organic phase. In recent years, such new reactor configurations have been developed making downstream processing easy as well as dealing with toxicity effects (Verhoef et al., 2009). Also, a new biotransformation matrix should implement changes in the reaction medium, shifting from undefined nutrient broth to defined mineral medium. There are a variety of different defined media used for Pseudomonas cultivation, including a recent approach where high cell density (CDW of 102 g l -1 ) cultivation of Pseudomonas putida KT2440 was reached even without a supply of oxygen-enriched air (Davis et al., 2015). Integration of such growth strategy into existing or currently developed pipelines of utilizing renewable feedstocks or industrial wastes has the potential for the development of a sustainable bioproduction of desired N-oxides. The presented biocatalysis system is an exceptional example of non-heme di-iron monooxygenases application, which are not limited to N-oxidation reactions and other synthesis routes including hydroxylation, sulphoxidation and epoxidation should follow. Together with a naturally high tolerance of Pseudomonas species to extreme and toxic conditions as well as flexible genetic modifications, these microbial cell factories have an excellent starting position for further development of them into the effective production platforms for various chemicals, which are not accessible so far.

Bacterial strains and growth conditions
The bacterial strains used in this study are listed in

Transformation of bacterial strains
The introduction of plasmid DNA into E. coli strains was carried out via heat shock. Standard procedures and techniques were used for the transformation and the preparation of chemically competent cells. The transformation of Pseudomonas strains was executed according to the modified protocol of Cho et al. (1995). The portion of overnight culture (0.5 ml) was transferred into 50 ml of fresh LB medium and was grown with shaking at 30°C until the early log phase (OD 600 0.3-0.5). The cells were harvested by centrifugation at 4000 g for 10 min. at 4°C, washed with 5 ml of ice-cold glycerol solution (10%) and re-centrifuged. This washing procedure was repeated two more times, and finally, cells were suspended in 0.5 ml of ice-cold glycerol solution (10%). The resulting mixture was divided into five aliquots of 100 ll that were immediately used for transformation, unlike E. coli, Pseudomonas cells (mostly for the MIL9 strain) could not be electroporated with high efficiency after being frozen. 100 ll of cell suspension was mixed with 5 ll of plasmid DNA (~100 ng ll -1 ) and transferred to chilled 0.2 cm gap cuvettes. Following the delivery of the pulse (12 kV cm -1 , a time constant 5 ms), the cells were mixed with 0.9 ml of SOB medium and shaken for 1 h at 30°C before transferring onto LB-agar plates containing appropriate antibiotic (cultivated at 30°C overnight).

Biotransformation conditions for whole-cell biocatalysis
The detailed description of conversion settings using E. coli BL21 (DE3) as a host was outlined in our previous study (Petkevi cius et al., 2019). The appropriate Pseudomonas strain (KT2440 or MIL9, respectively) transformed with Pml_pBAD2 plasmid was grown overnight in 10 ml of LB medium supplemented with kanamycin (final concentration 40 lg ml -1 ) and glucose (final concentration 1.0 g ml -1 ). The portion of overnight culture (0.5 ml) was transferred into 50 ml of fresh LB medium supplemented with kanamycin (final concentration 40 lg ml -1 ) and glucose (final concentration 1.0 g ml -1 ). The incubation was executed with shaking (200 rpm) at 30°C until the early log phase (OD 600 0.3-0.5), and then, 0.5 ml of 20% arabinose solution was added. This was followed by the addition of 0.5-1.5 mmol of the appropriate substrate, resulting in starting substrate concentration of 10-30 mM. Depending on the bio-convertibility of the used substrate, 1 to 4 additional portions of 0.5-1.5 mmol were introduced into the reaction mixture during the course of the complete biotransformation. Before every phase of overnight cultivation, 625 µl of 40% glucose solution was added to the cultivation broth. Pseudomonas strains were also tested with each substrate for the possible unspecific activity or intrinsic oxidation capabilities as samples with appropriate cells without Pml_pBAD2 plasmid were prepared (no such activity was found for substrates used in this study).

Isolation of the reaction products
The reaction mixture was separated from biomass by centrifugation (4000 g for 30 min). The volume of the supernatant was reduced under vacuum to a volume of 5-10 ml, and it was transferred to the separation funnel. Then, water phase was washed with 20-40 ml of chloroform at least three-four times. The organic phases were combined and dried with anhydrous Na 2 SO 4 . The resulting solution was evaporated under reduced pressure to give a crude product. Additionally, a purification procedure by silica gel flash chromatography with CHCl 3 -MeOH (5:1) could be applied to obtain a refined product. However, this purification step was mandatory on two occasions: (i) when a biotransformation product could not be extracted due to poor solubility in the organic phase, the supernatant of the conversion mixture was dried out under vacuum and gently resuspended in a mixture of chloroform and methanol and transferred onto chromatography column; (ii) the anticipated N-oxide was synthesized employing mCPBA. All isolation stages were monitored by TLC. The purity of the final product was verified by high-performance liquid chromatographymass spectrometry (HPLC-MS) and the chemical structure was confirmed by nuclear magnetic resonance (NMR) spectroscopy.

HPLC-MS analysis
The sample (0.5 ml) of the whole-cell biocatalysis reaction was transferred to a 1.5 ml tube and mixed with an equal part of acetonitrile. After the mixture was centrifuged (12 000 g for 5 min), 0.5 ml of supernatant was analysed using a high-performance liquid chromatography system. HPLC-MS analysis was performed using a high-performance liquid chromatography system (Shimadzu, Japan) equipped with a photo diode array (PDA) detector and a mass spectrometer (LCMS-2020; Shimadzu) equipped with an ESI source. The data were analysed using the LabSolutions LCMS software.

H NMR and 13 C NMR
NMR spectra were recorded in DMSO-d 6 or CDCl 3 on an Ascend 400: 1 H NMR -400 MHz, 13 C NMR -101 MHz (Bruker, MA, USA). Chemical shifts (d) are reported in ppm relative to the solvent resonance signal as an internal standard.

Analytics
Conversion is expressed as ((I a -F a )/I a ) 9 100%, where I athe initial amount of a substrate before the conversion, F athe final amount of a substrate after conversion. The substrate's amount was determined by integrating the absorbance area of a particular peak in the HPLC chromatogram. If no substrate was left, the conversion is defined as complete (> 99%). Volumetric productivity was defined as the amount of product (g) produced per volume (l) of the reaction per time (h) under optimum conditions. Product yield is expressed as P t (mol)/ P f (mol), where P ttheoretical amount of product formed based on the degree of conversion; P fthe amount of product found after the biotransformation. Cell dry weight (CDW) was determined after biotransformation was completed. A titre of the reaction was described as the total amount (g) of product formed in a conversion mixture (l) during the whole-cell biocatalysis.

Genome sequencing and analysis
This Whole Genome Shotgun project has been deposited at DDBJ/ENA/GenBank under the accession JAFEHE000000000. The version described in this paper is version JAFEHE010000000.