Bering in mind the homology between the hydroxylase from H. serapedicae SmR1 (fdeE27) and the already described by us enzyme from R. glutinis KCh 735 (RgF8H25) in the first step we compared the amino acid sequences of the two enzymes (step A, see Fig. 1 for a diagram of the steps). We carefully analyzed the structural information of related enzymes and explored important motifs and suggestions for their use (step B). Further, the protein-coding sequence that meets the prediction criteria was cloned from synthetic genes and expressed in E. coli cells (step C). The investigated protein was produced, purified, and biochemically analyzed towards hydroxylase activity and regioselectivity (step D). The final step of the study focused on the stabilization of the resulting readily oxidizable products (step E).Figure 1Flow scheme of the approach for the characterization of hydroxylase from H. serapedicae SmR1 and obtaining of hydroxy product, with steps A–E.A blast search to date described F8H, RgF8H25 and LjF8H33 revealed significant homology to fdeE, with coverage rates of 53% and 81%, respectively. The next step was to analyze FMO-specific binding motifs. RgF8H, LjF8H, and fdeE contain a Rossmann fold motifs (GxGxxG) and a GD motif (GDAxHxxxPxxxxG). Interestingly, despite sharing the same activity, all these enzymes lack the characteristic motifs, suggested to connect the FAD and NAD(P)H binding sites (FxGxxxHxxx)25,33,42 in their structures, despite that activity is detected. It is speculated that this motif is necessary for the regeneration of the FAD without a redox protein partner (Fig. 2a), which might suggest that the three enzymes belong to the same new FMO subclass or share different NADPH-binding domains.Figure 2(a) Multiple sequences alignment encoding the amino acids: LjF8H—line 1, fdeE—line 2, and RgF8H—line 3. Highly conserved FAD and GD motifs were highlighted in boxes using red and blue underscores, respectively. (b) Phylogenetic analysis based on FMOs, CYPs, Rieske-type, plant derived FMOs, and DOXs sequences. Protein sequences downloaded from the NCBI were aligned using Clustal Omega software (https://www.ebi.ac.uk/Tools/msa/clustalo). The cladogram was constructed using iTOL software (https://itol.embl.de/). Labels used: yellow colour—C-8 hydroxylases, blue colour – C-3 hydroxylases, orange colour—C-6 hydroxylases, green colour—C-3’ hydroxylases, pink colour—C-3′5’ hydroxylases, light grey colour—C-2’ hydroxylase, and dark grey colour—C-2 hydroxylase. Accession numbers of amino acid sequences used are available in the Supplementary material (Table S1). (c) I, III: Aligned predicted 3D protein structures of fdeE, LjF8H, and RgF8H with many poses of the docked FAD cofactor. II, IV: Simplified representation of part of the structure of the II-fdeE or IV-RgF8H with FAD,naringenin and quercetin docked (highest score hits), residues that are identical in both proteins are omitted in the graphic, surfaces were colored according to the electrostatic potential. The models and docking experiments were made in UCSF Chimera and figure created in ChimeraX 1.6.1.A cladogram was constructed among known flavonoid hydroxylases (Fig. 2b). The sequence of fdeE, was grouped within F8H, a member of the FMO family, and shows the highest homology with RgF8H25 (query cover—53%, percent identity—29.26%), and LjF8H (query cover—81%, percent identity—25.69%). Aligned structures of the 3D models of fdeE, LjF8H, and RgF8H indicate that all three enzymes share a similar overall architecture, with 10–15% of residues being identical in the RR distance map. All proteins dock FAD in the same cavity, where isoalloxazine indicates substrate binding pockets with a tunnel responsible for high regioselectivity of the reaction. However, in the case of RgF8H, an additional loop (residues 250–257) with W254 facing FAD packs it more tightly than in the other two enzymes (Fig. 2c). All enzymes differ significantly in the size of the substrate binding pocket (fdeE > RgF8H > LjF8H), as shown in Supplementary Fig. S1, whereas the hydrophobic residues mostly overlap, and some are even identical. The polar or aromatic residues differentiate significantly, which may be responsible for variation in substrate specificity. Both fdeE and LjF8H share the same electrostatic surfaces with similar residues distributed in space, while in RgF8H, basic residues R224 and K369 are oriented towards the substrate, changing the surface potential, which should be responsible for different binding affinities and hence, different substrate specificities. Docking of naringenin and quercetin to RgF8H and fdeE results in almost all binding positions overlapping near the FAD with the correct orientation for C8 hydroxylation, with a few low-score results where the B ring faces FAD. This is interesting to note, as no reaction towards C3’ or C4’ (chrysin/pinocembrin) was observed in the case of RgF8H, and also no activity towards quercetin or other flavonols. Additionally, only in the case of fdeE isoflavone genistein might be docked in the correct orientation for hydroxylation, which was further confirmed experimentally (Supplementary Table S4).Confirmation of predicted activityThe enzyme was produced using transformed E. coli cells. A codon-optimized sequence of fdeE was ordered in pSEVA182 vector with specific restriction sites for Golden Standard MoClo assembly. Then, the two sets of vectors were prepared—for inducible expression with RhaS/pRhaBAD promoter and with pJ23100 strong constitutive synthetic promoter (Supplementary Table S2). Overexpression of protein upon induction with rhamnose in E. coli BL21 (DE3) cells and subsequent purification by IMAC Ni2+ affinity chromatography enabled in vitro reactions and protein analysis via SDS-PAGE. Based on SDS-PAGE analysis (for qualitative purity of purified protein), Bradford analysis and further calculation, we have shown that the production and purification of the enzyme is efficient (Fig. 3b, Supplementary S2) and of approximate levels of pure protein produced are > 100 mg/L of culture. Characteristic of FMO, the yellow color was observed for the purified enzyme fraction solution, what was confirmed with the the UV–Vis spectrium of purified fdeE with absorption spectra characteristic for FAD absorbance and FMOs (Fig. 3c).Figure 3(a) Scheme of the hydroxylation reaction of naringenin by fdeE. (b) SDS-PAGE gel electrophoresis of recombinant hydroxylases: marker (line 1), purified (2 × HisTrap) fraction of fdeE (line 2, full SDS-Page electrophoresis gel picture can be found in Supplementary material), (c) UV-Vis spectrum of purified fdeE. (d) The UPLC analysis of reaction products catalyzed by the recombinant protein in vivo and in vitro reaction (pink line—standard of 8-hydroxynaringenin, blue line—in vitro reaction, black line—in vivo reaction). The UPLC profiles were monitored by a photodiode array detector. 1: 8-hydroxynaringenin, 2: naringenin. (e) UV–Vis spectrum of 8-hydroxynaringenin.A constitutive vector was used for in vivo activity, while a rhamnose-induced vector with an additional cofactor regeneration system was used for the in vitro reaction (Fig. 3a). Analysis of the products obtained by UPLC-DAD revealed the presence of a peak overlapping with the retention time and UV spectrum of the 8-hydroxynaringenin standard, both in case of in vivo and in vitro reactions (Fig. 3d,e).Enzyme characterizationInvestigated flavonoids hydroxylase are flavin-containing enzymes, therefore an important parameter to evaluate is the uncoupling rate of the peroxyflavin intermediate formed upon reaction with molecular oxygen43. The uncoupling rate can be determined by measurement of coenzyme (NADPH or NADH) oxidation, after the initial saturation of the reaction buffer with oxygen44. In the case of purified fdeE over 30 min of incubation, the NADPH concentration did not change (Fig. S3a) indicating almost no uncoupling. For NADH, we observed a slight decrease in concentration (Fig. S3b), which may be due to trace contamination of other enzymes that we were unable to remove as well as detect. More importantly, the concentration of each coenzyme decreased when naringenin was added to the reaction mixture (Fig. S3), although the results clearly show that NADPH is more preferred by the described enzyme than NADH, nevertheless, both coenzymes were accepted and we observed product formation for both. However, due to the preferential use of NADPH by fdeE in the subsequent assays we performed to characterize the properties of the enzyme, we already used only NADPH as the hydride donor. Interestingly, the incubated enzyme retained the yellow color in solution over the incubation time (no change also in UV–Vis absorbance profile of the enzyme solution was observed upon addition of NADPH), which might be linked to stable hydroperoxyflavin, thus the superoxidized form of enzyme, although no characteristic absorption spectra for this intermediate was detected (Fig. S4)45,46,47. Hydrogen peroxide produced as a result of decupling might damage the cell and lead to unwanted oxidation of both substrate and/or product in vitro and in vivo reactions.Little is known about these FMOs, so we analyzed other biochemical characteristics of the enzyme to find out, whether sequence identity to RgF8H would result in similar substrate scope and optimal biochemical parameters. The results of these studies are presented in Table 1a and Supplementary Fig. S5. fdeE shows best activity in the range 20–30 °C, above > 60% (Fig. S5a). At temperatures above 40–45 °C, only negligible activity was observed, and protein aggregates were easily visible. The enzyme was most active in an acidic environment, with the optimum pH for the reaction equal to 5.0 (Fig S5b). However, under these conditions, the enzyme is rapidly denatured. The ThermoFAD assay showed that apparent melting temperature (Tmapp) in the buffer of pH ≤ 5.5 in the temperature range of 40–45 °C (Fig. S5f.). The use of a more alkaline environment significantly increased the stability of the enzyme. In buffer pH 6–6.5, denaturation of the enzyme was observed at 54 °C. The enzyme was most stable at pH 7–7.5 (denaturation at 65–70 °C) and pH 8–8.5 (62–69 °C), while at pH 9–10 it melts at about 58 °C. Furthermore, it was observed that the type of buffer also affected the stability of the protein (pH = 8, phosphate buffer—66 °C, Tris–HCl—69 °C buffer). Therefore, a phosphate buffer of pH 7.5 was used for the enzyme in further studies, ensuring both high stability and activity. The most interesting finding of this test is the sensitivity to buffer concentration and ionic strength. The best results were obtained with a 10–15 mM buffer, a slightly worse result was observed in a 25 mM buffer, while the use of 50–100 mM buffers was associated with a decrease in enzymatic activity by half (Fig S5d). Concerning ionic strength, a decrease in the relative activity below 70% was observed at salt concentrations of 0.3 M (Fig. S5e). The addition of a small amount of the co-solvent DMSO had a beneficial effect on the conversion (> 5%), while the addition of a larger amount of either solvent was associated with a decrease in enzyme activity (Fig. S5c). At temperatures above 40 °C, the enzyme lost its activity in less than one hour (Fig. S5g), while holding it at temperatures up to 40°C for three hours did not affect it. Due to the extreme instability of the enzyme during storage, complete inactivation was observed after 1 day at − 20 °C. The addition of glycerol maintained enzyme activity longer but did not stop enzyme inactivation. Furthermore, replacing the buffer after purification and before the storage also had a positive effect (Supplementary Fig. S6). For the purified fraction, the best storage buffer is 25 mM Tris–HCl pH 8.0 with the addition of 10% v/v glycerol. However, storage at 4 °C resulted in a loss of activity in less than one day after purification, and the addition of BSA (0.1–1.0 mg range) and/or FAD (10–1000 µM) did not affect stability despite the addition of glycerol. For this reason, some subsequent steps for this enzyme were carried out using crude protein extracts that retained enzyme stability > 80% after 30 days of storage at − 20 °C.Determination of reaction product structuresThe in vitro reaction was performed with the addition of a reducing agent (DTT) and internal standard (dibenzofuran), to determine substrate acceptance towards flavonoids and phenolic compounds. The results were assessed by UPLC-DAD and LC–MS analysis. The substrate library containing 49 compounds is shown in Supplementary Table S3. Regarding flavonoids, the enzyme showed broader substrate specificity than the RgF8H described earlier25 (Table 1b). fdeE was active towards 17 of tested substrates and hydroxylated flavanones and flavones, similar to the RgF8H, and also show activity towards isoflavones and flavonols. The location of the hydroxyl groups in the A ring of the substrate was noteworthy. A correlation between the enzyme activity and the presence of hydroxyl groups at C-5 and C-7 carbon atoms of the substrate was found. Each reaction product was characterized by UV–Vis absorbance maxima and by electrospray ionization mass spectroscopy (ESI/MS) (Table 1b and Supplementary Fig. S7), and detailed descriptions of the LC/MS analysis can be found in the Supplementary Files, section—Products identification. The similarity of the investigated bacterial hydroxylase to the eukaryotic enzyme RgF8H ends with activity towards flavones and flavanones, as fdeE has broader substrate scope. These results are complemented by the sequence homology of the two enzymes (53%) shown by in silico amino acid sequence analysis and also in 3D models putting the two enzymes on an equal footing in terms of mechanism of action rather than substrate acceptance (Fig. 2c). Interestingly, for the tested substrates, we observed enzymatic activity of both of them only against flavonoids, confirming the assumption of a specific flavonoid degradation pathway already described in H. seropedicae SmR126 and probably occurring in R. glutinis KCh735.The formation rate of hydroxyflavonoidsDue to the low stability of the hydroxylation products, studies of product formation and degradation over time were carried out. Five substrates were chosen for the in vivo study: naringenin, luteolin, apigenin, quercetin, and chrysin. We decided to limit the range of substrates tested and select those whose stabilized production may be of interest to the industry due to their already proven properties, yet very low stability.Flavonoids that differ in their hydroxylation ratio and hence susceptibility for further over-oxidation in the reaction mixture. Figure 4 demonstrates that all hydroxylation products were degraded after 24 h of cultivation. In addition, we also used RgF8H in our study to compare the activity of all F8-FMOs already identified and described in the available literature. The highest activity was observed for fdeE, which performed the hydroxylation reaction rapidly, and also the product was degraded within 6–8 h. Degradation of the product or substrate was closely related to the color change of the medium. Control cultures with naringenin, apigenin, and chrysin did not change color (Supplementary Fig. S8), while visible color changes were observed in control cultures with quercetin and luteolin, indicating that these substrates were degraded even by untransformed E. coli DH5α cells during the 24-h cultivation which is also confirmed by the UPLC results (Supplementary Figs. S9, 10). In the case of quercetin, a dark brown color was observed as early as 6 h in fdeE cultures.Figure 4Degradation of C-8 hydroxylation products over time in vivo at an initial substrate concentration of 1 mM: (a) naringenin, (b) luteolin, (c) apigenin, (d) quercetin, (e) chrysin.Studies on C-8 hydroxy derivatives of flavonoids have shown that hydroxylated products exhibit enhanced antioxidant properties48. This activity also has implications for preparative reactions, as the resulting products are susceptible to oxygen-induced degradation. Therefore, the oxygen required for the reaction may also promote product degradation. To test, whether lowering the redox potential of the entire reaction mixture, could affect product degradation without significantly affecting the progress of the reaction, we experimented with the mild addition of a reducing agent to an in vitro reaction with naringenin and quercetin as substrates. The best results were obtained using 5–25 mM dithiothreitol (DTT), followed by 5 mM thiosulphate, but with already half the reaction yield (Supplementary Fig. S11), similar results were obtained for sodium dithionite. In contrast, the addition of cysteine did not affect the stability of the reaction products. On the basis of the tests performed with the addition of a reducing agent to the in vitro reaction, we decided to investigate the stability and yield of the products using 25 mM DTT in the reaction mixture. The two enzymes fdeE and RgF8H were used, and naringenin, luteolin and, quercetin were tested as substrates. The highest conversions were observed for the fdeE enzyme for each of the three tested substrates, while reaction efficiencies of an order of magnitude lower were observed for RgF8H. For comparison, trials without the addition of a reducing agent were carried out, for which almost all products were degraded within 30 min (Fig. 5 DTT-). In the case of naringenin, half of the product was oxidized after 15 min with each of the enzymes used. fdeE converted > 99% of the luteolin and quercetin within 15 min but products were also not detected if the reaction lasted longer than 30 min, indicating their degradation.Figure 5Stability of C-8 hydroxylation products in vitro reaction over time at an initial substrate concentration of 0.1 mM: (a) naringenin, (b) luteolin, (c) quercetin, without and with 25 mM DTT addition.For test with and without the addition of DTT, for which we obtained the best results in our preliminary test (Fig. S11), we chose three compounds from our previous test, naringenin, luteolin, and quercetin. These substrates belong to different subclasses of flavonoids (flavanones, flavones, and flavanols), differ in the hydroxylation ratio, and among previously tested were the most prone for undesired degradation, which allowed us to verify the potential stability of formed products. A protective effect for 8-hydroxynaringenin was evident for the first 30 min of the reaction, running the reaction longer allowed half of the products to be retained for each of the enzymes tested (Fig. 5a). Different results were obtained for luteolin, where lowering the redox potential of the reaction environment resulted in stable yields as well as product retention. For fdeE, we observed > 99% conversion already after 15 min and no change in product concentration for a further 45 min. RgF8H was also active towards luteolin, but with a lower yield, and also for this enzyme, we were able to observe the reaction product after 60 min (Fig. 5b). During the in vivo and in vitro reaction, the hydroxylation product of quercetin at the C-8 position was degraded almost immediately. Application of DTT allowed the product to be retained even after 60 min of reaction. The bacterial enzyme—fdeE, showed activity towards this substrate, converting > 99% of quercetin within 5 min, and due to the lowering of the redox potential of the reaction medium, the product remained stable for another 60 min (Fig. 5c). We also observed that the addition of DTT to the reaction mixture affected both product stability and reaction rate. Lowering the redox potential in the reaction with naringenin and quercetin decreased the rate of product degradation, especially in the first 5 min of the reaction (Fig. 5a,c). A completely opposite relationship was shown for luteolin, for which the addition of DTT stabilized the hydroxyproduct, but also caused it to form almost half as fast in the first 5 min (Fig. 5b). This clearly shows that, when choosing a reducing agent, we need to be careful and thoroughly check their effect on the reaction process with individual substrates as both effects, lowering the product degradation and reaction rate can be observed.Additionally, an attempt to stabilize the products by methylation is not possible due to the demethylating activity of the hydroxylases. This was confirmed in the case of wogonin, which contains a methyl group at the C-8 carbon atom that was oxidized—leading to demethylation and conversion of wogonin to 8-hydroxychrysin (Supplementary Fig. S12). UPLC-DAD analysis of the reaction—retention time and UV–Vis spectrum indicated clearly the hydroxylation product of chrysin and the demethylation of wogonin.Kinetic parametersPreliminary tests showed that the UV spectra of the individual flavonoid compounds and the coenzyme, whose consumption we initially planned to use to determine the kinetic parameters of the enzyme, overlapped. Making spectrophotometric measurements of the time course of the reaction impossible. Being aware that it would be impractical to determine the kinetic parameters from the NADPH concentration, we decided to use parameters derived directly from the product formation progress curve. This approach does not require a vast amount of samples to be made and gives the possibility to estimate Vmax and Km from a single reaction49. We obtained the highest values of the kinetic parameters for luteolin and chrysin, while quercetin was the slowest transformed substrate (Table 2, Supplementary Fig. S13), which is in agreement with our previously obtained results. Due to the low stability of the compounds obtained and despite the use of the reducing agent in the reaction medium, it should be noted that the results achieved may be subject to errors, caused by the inevitable partial decomposition of the product. However, a clear conclusion of evolutionary adaptation of fdeE to degrade naringenin, the most common flavonoid in plants is clearly observed.
Table 2 Calculated kinetic parameters for five substrates, their initial reaction concentrations 0.05mM.8-Hydroxyquercetin stabilization, purification and identificationIn our previous studies, we obtained hydroxy derivatives of flavones and flavanones25, however, the characterization of fdeE enabled also a route to hydroxylated flavonols and isoflavones (Table 1b).Because of the many challenges involved in the preparation of these extremely valuable compounds containing so many hydroxyl groups (like photo-lability, insolubility in water, or arduous purification), we set out to develop a stable and efficient way to produce and purify them50,51,52. Quercetin was used as a substrate for this study, firstly because of its antioxidant properties, but more importantly because of the range of problems to be solved for its rapid oxidation that should occur during extraction and purification. Being aware of the problem of extracting the quercetin product, we tested three different approaches. The preparative conversion of quercetin after 24 h of reaction with fdeE ended at conversion of 17%, which was equivalent to 17.1 mg of 8-hydroxyquercetin obtained (Supplementary Table S5). Assuming low losses during purification, the amount of product obtained after purification should be about 5 mg, due to the division of the reaction mixture into 3 parts. Such results were obtained only for the sample extracted with ethyl acetate, 5.16 mg, thus, the purification efficiency was more than 90%. The use of 1-butanol resulted in a loss of more than half of the product. Lyophilization of the entire mixture proved to be the worst method of product purification, as degradation of the compounds during the chromatography was evident after already a dissolution and filtration. This was confirmed by the negligible mass of the final product obtained.The structure of the product was confirmed by NMR spectrometry, which unambiguously showed the introduction of the hydroxyl group to the C-8 carbon and evident peak shifts as the quercetin was added in 20% (mass/mass) to the analysis (Fig. 6 and Supplementary File: Figs. S14–S17). Detailed descriptions of the NMR results can be found in the Supplementary Files, under—Products identification.Figure 6(a) 1H-NMR (600 MHz, DMSO-d6) spectrum part of 8-hydroxyquercetin (major) and quercetin (minor) mixture—signals from hydroxyl groups. (b) 1H-13C NMR (HMBC) (600 MHz, DMSO-d6) spectrum of 8-hydroxyquercetin (major) and quercetin (minor) mixture with characteristic signals indicating specific site of hydroxylation.
Novel enzymatic route to the synthesis of C-8 hydroxyflavonoids including flavonols and isoflavones
