Preparative Production of Functionalized (N- and O-Heterocyclic) Polycyclic Aromatic Hydrocarbons by Human Cytochrome P450 3A4 in a Bioreactor

Polycyclic aromatic hydrocarbons (PAHs) and their N- and O-containing derivatives (N-/O-PAHs) are environmental pollutants and synthetically attractive building blocks in pharmaceuticals. Functionalization of PAHs can be achieved via C-H activation by cytochrome P450 enzymes (e.g., P450 CYP3A4) in an environmentally friendly manner. Despite its broad substrate scope, the contribution of CYP3A4 to metabolize common PAHs in humans was found to be small. We recently showcased the potential of CYP3A4 in whole-cell biocatalysis with recombinant yeast Komagataella phaffii (Pichia pastoris) catalysts for the preparative-scale synthesis of naturally occurring metabolites in humans. In this study, we aimed at exploring the substrate scope of CYP3A4 towards (N-/O)-PAHs and conducted a bioconversion experiment at 10 L scale to validate the synthetic potential of CYP3A4 for the preparative-scale production of functionalized PAH metabolites. Hydroxylated products were purified and characterized using HPLC and NMR analysis. In total, 237 mg of fluorenol and 48 mg of fluorenone were produced from 498 mg of fluorene, with peak productivities of 27.7 μmol/L/h for fluorenol and 5.9 μmol/L/h for fluorenone; the latter confirmed that CYP3A4 is an excellent whole-cell biocatalyst for producing authentic human metabolites.


Introduction
Polycyclic aromatic hydrocarbons (PAHs) are environmental contaminants that are frequently formed during the combustion of organic material. PAHs can easily cross cell membranes through diffusion due to their lipophilic nature. However, although the parent PAH molecules themselves are not considered carcinogenic, their metabolites can form DNA 'adducts' leading to mutagenesis and cancer formation [1].
Previous studies identified human liver P450 monooxygenases [23][24][25], such as CYP2A6, CYP2A13, and CYP1B1, as key enzymes responsible for the metabolic activation of PAHs in the human body [26][27][28]. Although the human CYP3A4 is renowned for its broad substrate scope and for metabolizing 50% of all FDA-approved drugs [25,29], its contribution towards PAH degradation was found to be comparatively small [27]. Hence, the metabolism of PAHs by P450 3A4 has not attracted much attention in the literature, despite the discovery that these compounds induced the CYP3A4 PXR promoter which enhances its expression [30].
Human P450s especially possess the potential to hydroxylate high-value lead compounds such as authentic human metabolites [31] or natural products [20,32,33]. However, low solubility due to the generally hydrophobic nature of PAHs was identified as a major limitation for efficient P450 biocatalysis when compared to chemical methods [34]. Synthesizing hundreds of milligrams of metabolites is often regarded as a synthetic challenge with human P450s due to their performance and requirement for co-factors [35][36][37][38]. Being able to scale-up production to achieve sufficient volumetric productivities and high yields with high oxygen transfer rates is therefore challenging too. The high redox requirements of monooxygenases were additionally reported as a major hindrance for their use at preparative scale [39]. Albeit not necessarily greener, the biocatalysis processes can be highly sustainable if renewable feedstocks are used in a resource-efficient manner [40].
Including stirred-tank cultures in the research process is important to facilitate later scale-up, as traditional shake-flask or microtiter plate cultures lack the same monitoring and control possibilities [41,42]. In addition, shaken cultures differ from stirred cultures in oxygen mass transfer, heat transfer, and agitation speed/type. Such scale-ups commonly cause problems, especially when using human CYPs due to their inherent instability [37,43], as well as strong dependency on the availability of the redox partner [35,38]. A key parameter that can be controlled in a bioreactor is dissolved oxygen that can be tweaked with temperature, aeration rate, headspace pressure, and stirring speed. This offers an advantage over shake-flask experiments for oxygen-transfer-rate-sensitive enzymes such as monooxygenases.

Materials and Methods
All solvents and chemicals were purchased from SigmaAldrich/Merck (Steinheim/ Darmstadt, Germany), VWR International (Fontenay-sous-Bois, France), Carl Roth GmbH (Karlsruhe, Germany) or InvivoGen (San Diego, CA, USA) in best available purity. HPLC gradient solvents were used for the HPLC measurements. HPLC tubes were bought from Macherey-Nagel (Düren, Germany) and the corresponding caps and inserts from Bruckner Analysentechnik (Linz, Austria) In experiments (2.1) an Agilent Technologies 1100 Series HPLC was used and in experiments (2.2) a Shimadzu Prominence LC-20AD HPLC was used. In both cases, separation was carried out via a Kinetex C18 (100 Å; 50 × 4.6 mm; 2.6 µm) reverse-phase column. The strirred-tank steel-jacket Bioengineering NLF22 10 L bioreactor from Bioengineering AG (Wald, Switzerland) was used for experiment (2.2). K. phaffii cells with expressed P450 3A4 were obtained from bisy GmbH (Hofstaetten, Austria) as lyophilized powder for experiments (2.1) and as biomass in carbon source exhausted basal salts medium [57] for experiment (2.2). OD measurements were executed with an Eppendorf BioPhotometer plus. NMR spectra were recorded on a Bruker Avance 400 spectrometer (operating at 400 MHz for 1 H NMR and 101 MHz for 13 C NMR) in CHCl 3 -d with tetramethylsilane (TMS) as internal standard purchased by Carl Roth.

Broad PAH Substrate Screening and HPLC
Whole-cell biotransformation for the substrate screening using commercial K. phaffii cells with co-expressed human P450 3A4 monooxygenase and human cytochrome P450 reductase (0.5 mM substrate in 400 µL, OD 600 = 100, 28 • C, 320 rpm, 17 h) and the corresponding HPLC analysis (254 nm) were performed in the same way as reported before [45].

Bioreactor Biotransformations
The bioreactor biotransformation was performed using a commercial K. phaffii human P450 3A4 catalyst from 6 L of cell biomass at an OD 600 of 200 with a dry cell weight that was measured as 90 g/L. Temperature was maintained at 28 • C and the pH at 7.4 with 20% ammonium solution and 20% phosphoric acid. Dissolved oxygen was maintained at 30% for the duration of the biotransformation, by changing stirring speed. Headspace pressure was maintained at 0.5 bar and aeration rate was 10 L/min. To initiate the biotransformation, 498 mg of fluorene was dissolved in 30 mL of hot (65 • C) methanol and added to 6 L of cell broth for a final concentration of 0.5 mM. Four 50 mL samples were taken at each time point to analyze the kinetics of the bioconversion. Time points were 0, 15, 30, 45, 60, 90, 120, 180, 300, 420, 540, 660, 1440, and 1560 min. Each sample was mixed with 50 mL of methanol to stop the biotransformation, vortexed and centrifuged (4000 g, 4 • C, 30 min), and 100 µL supernatant was taken for HPLC analysis. The percentage amounts of each compound in the samples were calculated with calibration curves from commercially bought reference compounds. The supernatants of samples at time points 0, 420 and, 1560 min were loaded in batches onto a C18 flash column (Chromabond ® Flash RS 40 C 18 ec, 15-40 µm) obtained from Macherey-Nagel (Düren, Germany) for chromatographic purification using a Cytiva ÄKTA Pure 25 M (Marlborough, MA, US) device using the following method: Water containing 0.1% acetic acid (A) and methanol (B) was used for elution at room temperature (2 mL/min, 280 nm) in the following ratios: 0 mL: A/B 80/20; 120 mL: A/B 80/20; 200 mL: A/B 0/100; 330 mL A/B 0/100. The chromatographic fractions were pooled together for each time point and evaporated using a rotary evaporator. The wide mouth volumetric flask was then washed with (3 × 10 mL) EtOAc, transferred into a smaller volumetric flask, and evaporated under a stream of nitrogen. The obtained masses were off-white in color with a yellow fluorescent tint. The samples were weighed using a precision scale, then dissolved in deuterated chloroform for analysis using NMR.

NMR
NMR experiments were performed for the characterization and quantification of the biotransformation products. NMR spectra were recorded at room temperature for each compound and product mixture (20 mg/mL) that was initially purified as mentioned above.

Broad PAH Substrate Screening
Firstly, the substrate scope of P450 3A4 towards PAHs, their Nand O-containing derivatives, as well as already hydroxylated PAH compounds were investigated ( Figure 1). Notably, half of the 14 tested PAHs were accepted by P450 3A4 encompassing especially 3-cyclic, medium molecular weight (MMW) PAHs such as 1 (49% conversion; Figure S1) and acenaphthene (4, 51%; Figure S2). Authentic reference compounds allowed the identification of 2 and 3 as well as 1-hydroxyacenaphthene (5) and 1-acenaphthenone (6) as the products, respectively. Hence, in the latter case, we could confirm the literature results [28]. Interestingly, several new peaks were observed by HPLC for the conversion of 9-methylanthracene (7, 9%; Figure S3), whereas anthracene (0%) was not tolerated and phenanthrene (6%) only just. Apparently, the 9-methyl group of 7 is targeted by P450 3A4 as the activated benzylic position, as indicated by the appearance of a sharp, new peak. However, further smaller peaks were also observed, indicating either follow-up reactions with the product 9-hydroxymethylanthracene or that the 9-methyl group is required for active site crowding to allow the reaction [58]. None of the 2-cyclic, low-molecular-weight (LMW) PAHs such as indene or naphthalene were accepted, except for azulene (8, 35%; Figure S4), resulting in the formation of 1-hydroxyazulene (9) as indicated by an authentic reference. Higher molecular weight species (HMW) consisting of four fused cycles, such as fluoranthene (12%) and pyrene (10, 35%; Figure S5), were converted, as was to be expected by the renowned large active site cavity of P450 3A4 [59][60][61]. The conversion of 10 was particularly fascinating because at least five new product peaks could be observed, indicating a strong diversification of 10 by this human P450, which could be harnessed to produce several hydroxylated metabolites simultaneously. Generally, CYP3A4 seemed to accept mostly MMW and HMW species of PAHs, with azulene as the only LMW that was accepted.
In case of derivatives already bearing oxygen-containing substituents, some LMW PAHs were accepted. However, it seems likely that their communal methoxy group was dealkylated by P450 3A4, which is a common reaction of human liver P450s [62].
The conversion of the N-PAH substrates indole (68%), phenanthridine (11, 35%; Figure S6), and acridine (13; 5%) was in direct contrast to the observed activity towards non-heterocyclic containing equivalents of indene, phenanthrene, and anthracene. Therefore, the presence of the heterocyclic nitrogen atom within the heterocyclic compounds must be a prerequisite for being accepted by P450 3A4 as a substrate. The appearance of blue color suggested the formation of indigo from the electron-rich indole substrate. In addition, the presence of only one dominant product peak for the conversion of 11 implied the specific oxidation of one electron-rich position, possibly at N. Hence, we hypothesize N-oxide formation, which is a commonly occurring reaction for P450 3A4 [32]. Although O-PAHs are more electron-rich and thus more easily oxidized, the enzyme showed no discernible activity towards the selected O-PAH compounds. The lack of conversion of O-PAHs with low molecular weight could again be explained by the large CYP3A4 active site. In the case of dibenzofuran, it was further evidence for 2 being the product of the conversion of 1 because the O-heteroatom was occupying position 9.

Bioreactor Biotransformation
With the (N-/O)-PAH being accepted as substrates of P450 3A4, we aimed to scale up the biotransformation catalyzed by this K. phaffi-based whole-cell biocatalyst, from microtiter plates to a 10 L pilot scale fed-batch stirred-tank reactor, in order to show its potential for preparative purposes and thus tackle this highly restricting factor for the integration of biocatalysis in synthetic chemistry.
Compound 1 was selected as an exemplary substrate because it is a structural motif of many natural products [3,4], was hydroxylated with a relatively good percentage to 2 (49%), and was further oxidized to the corresponding ketone (3), allowing the illustration of a potential cascade reaction [63,64], which are important in biocatalysis as they enable more efficient synthesis routes by reducing the need for additional steps [34]. For the preparation of the CYP3A4 catalyst for this specific biotransformation, the most commonly cited K. phaffii fermentation buffer and cultivation conditions from the Invitrogen corporation P. pastoris fermentation guide [57] were used.
The quantity of metabolites commonly used for in vitro and safety pharmacology testing are in the range of 0.1-100 mg [35], which we aimed to produce. We succeeded to isolate 237 mg of 2 and 48 mg of 3, and also recovered 212 mg of 1. To better characterize the kinetics and product diversification, we sampled a time series of 200 mL samples at  The conversion values for the points at 540 min, 660 min, and 1440 min were all approximately within the same error range, indicating that the conversion had slowed down considerably after the former time point. This is most likely due to the lack of glucose for intracellular NADPH regeneration. Other factors, such as high Km of the enzyme for this substrate, product inhibition or poor substrate partition coefficient due to its hydrophobic nature could also contribute to the slowdown. The high oxygen transfer rates achievable in a bioreactor could contribute to CYP3A4 degradation through H 2 O 2 formation, especially if the cell chassis lacks the energy for reactive oxygen species disposal. The HPLC chromatograms for timepoints at 0, 420, and 1560 min are shown in Figure 3. The most abundant product formed was 2, with a small amount of 3 present. The three molecules absorb differently at 254 nm, especially the ketone, likely due to the electron π-π * transition. Therefore, calibration curves were needed to assess the yield more accurately.
We calculated the yields as 27.9% for 2 at 420 min and 57.6% at 1560 min. Yields for 3 were 1.5% at 420 min and 3% at 1560 min ( Figure 3). To confirm our results, we chromatographically purified samples taken at 0, 420, and 1560 min and analyzed their mol% yields using NMR, which is a more accurate method of quantifying PAHs [65]. The conversion could be monitored via the chemical shift of the hydrogen at 9-position ( Figure 4). As 1 was converted (oxidized) into 2 and subsequently 3, H-9 at 3.91 ppm (H a -orange highlight) first shifted downfield to 5.58 ppm and then disappeared. Therefore, the biotransformation of sample 1 into 2 and 3 could also be monitored by the decrease and increase in peak integration of the singlets at 3.91 ppm (H a ) or 5.58 ppm (H b -red highlight), respectively. The formation of 3 could also be confirmed by the appearance of peaks at 7.48 ppm (blue highlight) that belong to aromatic protons of fluorenone in a new chemical environment. Maximal yield was 47.6 mol% of 2 and 9.7 mol% of 3 (Table 1).  It appears as if HPLC measurements overestimated the yield of 2 and underestimated the yield of 3. Similar variance between analytical methods was reported before in the literature [65].
The most time efficient point was reached around 9 h after initiation. The long activity time showed that the yeast cells expressing CYP3A4 were indeed a stable biocatalyst with enzyme activity present, albeit diminished, even after 26 h.
Volume productivity (space-time yield) is a key sustainable catalysis metric [66]. After 7 h the bioreactor reaches its most productive stage, with productivities of 27.7 µmol/L/h for 2 and 5.9 µmol/L/h for 3. With regard to sustainability, the reaction is inefficient after 9 h, when conversion leveled off, as shown in Figure 2. As a comparison, a very commonly cited biotransformation performed by CYP3A4 is the biotransformation of testosterone, where 6β-hydroxytestosterone is the major of multiple products [44]. Productivity numbers for the production of 6β-hydroxytestosterone were 6 µmol/L/h for CYP3A4 expressed in E. coli cells (100 g of wet cell mass, 1 L jacketed vessel, 48 h reaction time) [35,67], and 22 µmol/L/h for CYP3A4 expressed in K. phaffii cells (200 g of wet cell mass, 2 L jacketed vessel, 8 h reaction time) [44]. The productivity numbers we achieved in this study show that CYP3A4 is slightly worse at fluorene biotransformations compared to its model substrate, i.e., testosterone. In the second study, the total combined yield for all testosterone products was 70%; in the present study, we report a 57% combined conversion. A recently reported biotransformation of fluorene by CYP5035S7 from Polyporus arcularius, reported a combined yield for its two products, 2-hydroxyfluorene (35%) and 2,7-dihydroxyfluorene (18%), of 53% [45]. Although the two products formed are different, compared to CYP3A4, the total yield is similar.
During our broad PAH screening, we detected a 49% combined yield for 1, meaning that scaling up the reaction from a 96-microtiter plate scale to a 10 L preparative scale improved yields slightly. It is likely that any of our screening reactions could be scaledup in the same way and would improve volumetric productivity while maintaining the same yield. The frequently used media and bioreactor conditions we have chosen in this study matched the efficiency of microtiter plates, although could be further optimized. Oxygenation of the bioreactor could be increased drastically above 30% dissolved oxygen, by increasing aeration, headspace pressure, and agitation. Increased agitation would also assist in improving substrate mass transfer rates, as some CYP3A4 substrates, including PAHs, are quite hydrophobic and can form a biphasic layer. Supplementing the reaction buffer with a small amount of glucose or methanol could improve cofactor regeneration and extend the productivity of the reaction, which we investigated previously [44].
In total, we produced 237 mg of 2 and 48 mg of 3. Although this yield is acceptable for the production of highly valuable pharmaceutical compounds, it is a concern that would need to be addressed in order to be economical in other industries [35]. For this to be achieved, efforts of both protein engineering, strain development, and bioprocessing will need to be combined.

Conclusions
In the present study we achieved the conversion of PAHs and their N-, O-and hydroxylated derivatives by the human P450 CYP3A4 expressed in K. phaffi, which proves that this system is a suitable whole-cell biocatalyst to produce oxidized PAHs in sufficient amounts for metabolic or pharmaceutical development. Several PAHs, hydroxylated PAHs, and N-PAHs were accepted as substrates, whereas none of the O-PAHs we selected were converted. Fluorene (1) was selected as the model substrate for the scale-up study because of its high yield, its positions as a structural motif in natural products, and the production of fluorenone (3) from fluorenol (2), which could illustrate a potential cascade reaction. Scale-up was performed in a 10 L jacketed steel bioreaction with 6 L of reaction broth volume using the most cited K. phaffii fermentation conditions from the Invitrogen Pichia pastoris fermentation guide. The employed cost-effective buffer systems will likely ensure a successful scale up of the developed PAHs products. Multiple samples were taken during the biotransformation to illustrate the kinetics of the reaction. All samples were analyzed using HPLC equipped with a UV-Vis detector and three representative time interval samples were analyzed using NMR. The final yield of 2 and 3 were 47.6% and 9.7%, respectively. Peak reactor volume productivity was reached after 7 h and was calculated as 27.7 µmol/L/h for 2 and 5.9 µmol/L/h for 3. This productivity was sustained for up to 9 h of reaction time, after which it fell substantially. Total productivity after 26 h of reaction time was 9 µmol/L/h for 2 and 1.8 µmol/L/h for 3. In total, we produced 237 mg of 2 and 48 mg of 3. These results show that human CYP3A4 seems to participate in the metabolism of PAH compounds to a significant extent by accepting many of them and their derivatives as substrates. We demonstrated the possibility of scaling up the biotransformations of even such uncommon substrates for CYP3A4 to produce preparative scale product amounts for synthetic exploitation, the production of authentic human metabolites, or other valuable pharmaceutical compounds.