Introduction

Dahlia (Dahlia variabilis hort.) is a popular ornamental plant exhibiting a broad spectrum of flower colour from all hues of red, magenta and orange, to yellow and white1,2,3. The great range of colours, as well as flower shapes, is presumably a result of the octoploidy of garden dahlia (2n = 8x = 64), allowing a multitude of gene combinations of the two tetraploid parent species4. In contrast to many other members of the Asteraceae species, in which carotenoids contribute significantly to yellow hues, Dahlia flower colour is exclusively based on the accumulation of flavonoids and biochemically related compounds3 and in the specific case of yellow flowered dahlia the glucosides of the 6′-deoxychalcones, isoliquiritigenin (2′,4′,4-trihydroxychalcone) and butein (2′,4′,3,4-tetrahydroxychalcone) are responsible5,6. All flavonoids have a C6–C3–C6 skeleton, which is synthesized from phenylpropanoid and polyketide precursors. Chalcone synthase (CHS) catalyses the committed step in flavonoid biosynthesis forming naringenin chalcone from (hydroxyl-)cinnamoyl-CoA and malonyl-CoA. In most cases Naringenin chalcone is then converted to (2S)-flavanones, catalysed by the chalcone isomerase (CHI). 6′-Deoxychalcones, in contrast, are synthesized by the co-action of the chalcone reductase (CHR; chalcone polyketide reductase, PKR) with CHS7,8. Flavonoid 3′-hydroxylation is performed by flavonoid 3′-hydroxlase (F3′H), which shows a broad substrate specificity and is able to hydroxylate flavanones (Fig. 1a), flavones (Fig. 1b), flavonols (Fig. 1c) and flavanols (Fig. 1d).

Figure 1
figure 1

Reactions catalysed by chalcone 3-hydroxylase and flavonoid 3′-hydroxylase.

Full size image

In contrast to the well understood pathway to red anthocyanin and ivory flavone pigments in dahlia, the biosynthesis of yellow anthochlors remains somewhat of an enigma with regard to the involved enzymes and genes1,9,10. For a long time, it was assumed that F3′H is responsible for the formation of the 3,4-dihydroxy pattern of chalcones, as it shows high similarity to the flavonoid 3′-hydroxylation reaction catalysed by F3′H. However, we have shown that chalcone 3-hydroxylation is not catalysed by F3′Hs11 (Fig. 1e). Specific structural prerequisites can increase the chalcone acceptance of single F3′Hs to some extent12, but a cDNA sequence encoding an enzyme with particularly high chalcone 3-hydroxylase (CH3H) activity was detected in yellow flower petals of Cosmos sulphureus13. It can be assumed that CH3H probably evolved in the Asteraceae family from F3′Hs to specifically attract insects that favour yellow and blue over bright red flower colour14. Studies with the CH3H enzyme of Cosmos sulphureus (CsCH3H) showed that, besides chalcones, (2S)-naringenin, dihydrokaempferol, apigenin and kaempferol are also accepted as substrates13. Therefore, one can conclude that CH3H is a specialized F3′H as it exhibits an additional functionality compared to F3′H. Furthermore, it was shown that apple plants overexpressing CsCH3H exhibited an in increased level of 3-hydroxyphloretin15, while the in vitro assays of apple F3′Hs did not show dihydrochalcone hydroxylation activity16,17.

CH3H and F3′H belong to the large superfamily of cytochrome P450 dependent monooxygenases (CYP). Cytochrome P450 genes are subdivided and classified on the basis of amino-acid identity, phylogenetic criteria and gene organization18. Plant CYPs are heme containing proteins and cluster in two clades, where one clade frequently has conserved metabolic housekeeping functions and is related to CYPs of other phyla, whereas the other one is plant specific and involved in secondary metabolism19. CH3H and F3′H cluster in the CYP75B subfamily. The CYP75 gene family is divided into two subfamilies and belongs to clan7120. It comprises the flavonoid biosynthetic enzymes, which are involved in the biosynthesis of the majority of plant secondary metabolites21,22. The subfamily CYP75A contains flavonoid 3′ 5′-hydroxylases (F3′5′Hs) and the subfamily CYP75B contains F3′H as well as CH3H enzymes21,23. Most F3′5′Hs belong to the CYP75A subfamily, with some exceptions that cluster into the CYP75B subfamily and have also gained 5′-hydroxylase acitivity24,25,26,27. Whereas prokaryotic CYPs are usually soluble proteins, plant CYPs are typically bound by their N-terminal single transmembrane domain18 either at the cytosolic face of the endoplasmic reticulum, or the inner mitochondrial membranes, or the plastids19. Further, both mammalian and plant CYPs need a redox partner, the NADPH dependent cytochrome P450 reductase (CPR). CPRs are FAD- and FMN-containing enzymes and are also membrane anchored at the cytosolic face of the endoplasmic reticulum28. In the flavonoid pathway, several steps are catalysed by CYP enzymes, with F3′H being the most prominent29. The association with the endoplasmic reticulum membrane makes these enzymes more difficult to handle in every aspect, from enzyme preparation and characterization, production and purification of recombinant enzymes, to structural studies30. In this study we report for the first time the isolation of a cDNA sequence encoding CH3H from dahlia flowers and the production, purification and characterization of the recombinant enzyme possessing its transmembrane helix. Further, we can report for the first time that the hydroxylation of (2R)-naringenin is catalysed by CH3H with a high substrate specificity, even though (2R)-naringenin is not naturally formed by CHI.

Results and discussion

Cloning and sequence analysis of the DvCH3H

The cDNA from D. variabilis (Dahlia pinnata) encoding CH3H (deposited at the GenBank data library under accession number GQ479804) possesses an open reading frame of 1524 nucleotides, which was isolated with a degenerated primer designed in the 5′-noncoding region of the Cosmos sulphureus CH3H. The amino acids 4–23 of the deduced amino acid sequence are predicted by the program TMHMM31 to constitute the membrane anchor, which is attached to the membrane of the endoplasmic reticulum (Fig. 3). In order to obtain further information on the relationship of DvCH3H to other CYP75B sequences within the Asteraceae family, a BLASTP search using the NCBI database was performed and 50 sequences were obtained after filtering. DvCH3H shows a sequence identity of 70–84% to most of the obtained sequences (see Supplementary Fig. S1), while it shows 90% identity with CsCH3H and 84–87% sequence identity with F3′Hs from Tagetes, Dahlia, Rudbeckia, Zinnia, Bidens, Cosmos and Helianthus (see Supplementary Table S1). This is also reflected by the corresponding phylogenetic tree (Fig. 2). The enzymes exhibiting chalcone hydroxylase activity, namely the two CH3Hs from Dahlia pinnata (GQ479804 and BDE26439) cluster together with CsCH3H (ACO35755.1). The sequence BDE26439.1 has been released during the review process of this manuscript and is almost identical to GQ479804, with just two mutations within the membrane anchor and missing the last two amino acids of GQ479804. Notably, no other sequence clusters within this branch. However, it has been reported that F3′H from Tagetes erecta (ACO35756.1) showed weak chalcone hydroxylase activity in in vitro activity assays11, suggesting that the enzymes of this main branch might possess at least some of requirements necessary for chalcone hydroxylation. The substrate specificity of a CYP is determined by the amino acids within 6 substrate recognition sites (SRS)32,33,34 (Fig. 3) which are located in the proximity of the heme centre. In DvCH3H the typical LSXXG pattern is present in SRS1, which is presumably required for the hydroxylation of chalcones, as well as the loop region found in CsCH3H (FJ216429)13.

Figure 2
figure 2

Phylogenetic tree of available Asteraceae CYP75B amino acid sequences. The enzymes exhibiting chalcone hydroxylase activity are coloured red and the F3′H from Tagetes erecta showing weak chalcone hydroxylase activity is coloured orange. Note: The sequence of F3′H from Petunia hybrida (Q9SBQ9) has been included for comparison with F3′H from non-Asteraceae species.

Full size image
Figure 3
figure 3

Structural features of DvCH3H. (a) Sequence of DvCH3H. Secondary structure features were assigned using DSSP (Dictionary of Secondary Structures of Proteins)35 using the homology model of DvCH3H. Grey-shaded boxes highlight II, proline-rich hinge13,36; III, oxygen-binding motif; IV, EXXR triad13,37; V, the P450-pattern13,38. The blue framed box highlights the hydrophobic membrane anchor13 and the red framed boxes highlight the substrate recognition sites 1–6 (SRS1-6)32,33,34. (b) Homology model of DvCH3H. The SRS regions is colored dark blue and the heme is coloured red. The putative orientation in and on the membrane of the endoplasmic reticulum has been roughly estimated according to the discussion in the literature39.

Full size image

Purification

A purification procedure for the membrane-bound DvCH3H, recombinantly cultivated in controlled bioreactor runs in P. pastoris KM71H (for more details see40), was established by systematic investigation of each purification step. An overview is shown in Fig. 4.

Figure 4
figure 4

Unit operations necessary for the purification of recombinant DvCH3H. Investigated factors and ranges thereof are shown in the right column.

Full size image

Homogenization and centrifugation

For yeast cell disruption, it has been reported that mechanical cell disruption techniques such as high pressure homogenization, bead milling or sonication lead to higher recovery yields, when it comes to breakage of the fungal cell wall, in comparison to electrical, enzymatic, physical or chemical disruption methods41. We chose high pressure homogenization, as the technique is efficient, scalable and results are reproducible. Previous studies on cell disruption of yeast by high pressure homogenization showed that the number of passages and the pressure had the most significant impact42. Therefore, we kept those parameters high (1800 bar and 10 passages) and investigated the impact of the dry cell weight concentration on the homogenization process. The dry cell weight is taken for the calculation of the used biomass as it is much more accurate than the wet cell weight and necessary for the optimization of the homogenization process. For the wet cell weight, the remaining water in the cell pellet depends a lot on the centrifugation process and strongly biased the calculation of the used cell mass. Samples were homogenized at dry cell weight concentrations between 6 and 60 g/L and centrifuged at 20,000× g for 30 min afterwards. As shown in Table 1, there is a slight trend towards less product in the supernatant at higher biomass concentrations. However, as the recovery of target protein at 60 g DCW/L was only 6% lower than at 6 g DCW/L, we decided to persist with the high biomass concentration, as this allowed working with a more concentrated protein solution, reducing the volumes during subsequent purification steps. Nevertheless, in general, around 60% of the target protein was found in the cell pellet, which is discarded after the disruption procedure. We hypothesized that this could result from (I) poorly disrupted cells or (II) a too harsh centrifugation afterwards, which caused partial sedimentation of the target protein. To invesigate the disruption efficiency, colony forming units (CFUs) of frozen and thawed cells were compared to frozen, thawed as well as homogenized cells and we found that the homogenization process only led to a reduction of 55% of CFUs. However, due to cooling limitations, neither the pressure nor the number of passages could be further increased.

Table 1 Results of western blots with percentages of recombinant DvCH3H in supernatants (S) and pellets (P).
Full size table

Next, it was investigated whether a less harsh centrifugation procedure could lead to higher amounts of target protein in the supernatant. In literature, the seperation of cell debris is usually carried out by centrifugation between 1000 and 15,000× g for 15 min up to 30 min, which is a rather broad range43,44,45,46,47. To shed more light on the optimal conditions, g-forces between 1000 and 10,000, and centrifugation times between 5 and 20 min were investigated. The results in Table 1 show that a reduction in the g-force clearly leads to more protein in the supernatant with a recovery of up to 73%. At the same time, it was beneficial to keep the centrifugation time rather high. This can be explained as with shorter centrifugation times, the pellet is more voluminous, causing a lower recovery of target protein in the supernatant. Further reduction of the centrifugation force led to an incomplete separation of unbroken cells.

Ultracentrifugation

After cell disruption, the membrane protein fraction was collected by ultracentrifugation. It is generally recommended to pellet membrane proteins at 100,000–200,000× g for 1- 2 h43,44,45,48. We investigated the time needed to pellet recombinant DvCH3H at 200,000× g by analysing samples after 30 min and 60 min. As shown in Table 1, already after 30 min, no target protein was detectable in the supernatant anymore. Longer centrifugation times produced a more compact pellet which was more difficult to resuspend and solubilize in the subsequent step. Therefore, 30 min at 200,000× g was chosen for subsequent experiments.

Solubilization and ultracentrifugation

After pelleting the membrane protein fraction, the target protein needs to be solubilized out of the membrane by detergents. Therefore, we tested solubilisation in six different detergents at 4 °C over night45, always at 10 times their critical micellar concentration. The obtained results are shown in Table 1. LDAO solubilized the highest amount of target protein (36%), however, initial purification results showed that the protein was very unstable in this detergent. LDAO is considered a rather harsh detergent due to its zwitterionic nature as its charges can interact with non-hydrophobic parts of the protein potentially causing instability issues. OTG seemed to have a negative influence on protein stability, as a band of very low molecular weight was visible on the blot, which might originate from DvCH3H (see Supplementary Fig. S2b). Possible causes of this potential cleavage would have to be investigated in detail, however, exceed the scope of the manuscript. The detergents CHAPS and FC-12 did not solubilize a detectable amount of protein. DDM solubilized the second highest amount of target protein (27%) which is why this detergent was chosen for further experiments. Adjustment of the solubilisation time to 4 h, the detergent concentration to 20 mM and the protein concentration to 2 mg/mL allowed the extraction of 61% of the target protein (Table 1).

Chromatography

To further purify the recombinant transmembrane possessing his6-tagged DvCH3H, the solubilized membrane protein solution was loaded on immobilized metal ion affinity chromatography (IMAC) columns. Affinity chromatography can be quite challenging for tagged membrane proteins, as they sometimes bind very poorly to the resin due to the huge detergent micelles, which impair accessibility of the tag45. In order to reduce the detergent concentration in the load, the solubilized protein was diluted to a final DDM concentration to 2 mM with detergent free buffer. We were interested in whether a classic particle-based column or a monolithic column was superior for the purification. The results of two purification runs, one on a particle-based column and one on a monolithic column, with identical column volumes and similar metal ion capacities (15 µmol Ni2+/mL resin (particle based) and 23 ± 10 µmol Cu2+/mL resin (monolithic)) are compared in Table 2.

Table 2 Comparison of the IMAC purification on a monolithic and particle-based column.
Full size table

After 1:10 dilution of the load, rather high recoveries of DvCH3H were found, especially on the particle-based column, where 93% could be captured and eluted. On the monolithic column in contrast only 71% of DvCH3H were recovered in the eluate and binding of proteins in general seemed less specific as the purity of the eluate was only 32%. However, the eluate of the particle-based column was also only 56% pure, which is why an additional size exclusion chromatography was carried out. This allowed purification to more than 84%, as determined by analytical size exclusion chromatography (Fig. 5b), SDS-PAGE and Western blot (Fig. 5a).

Figure 5
figure 5

(a) SDS-PAGE and western blot and (b) SEC-chromatogram of the purified recombinant DvCH3H. Solid line indicates absorbance at 280 nm and the dashed line the absorbance at 425 nm, where the heme of the Tris-(2-carboxyethyl)-phosphin reduced recombinant DvCH3H has an absorbance maximum. The original SDS-PAGE and western blot are shown in Supplementary Fig. S3.

Full size image

Overall purification process

The entire purification process is summarized in Table 3. Values marked with an asterisk were theoretically calculated based on indirect measurements, when direct determination was not possible (for details see SI).

Table 3 Overview of the purification process.
Full size table

The overall recombinant DvCH3H recovery was 30%, with over 60% in each of the five purification steps. The most critical step was the solubilisation of the membrane protein. Advanced optimization with respect to the ionic strength or the pH of the solubilisation buffer might result in higher yields45.

The overall purification factor was 170, leading to at least 84% pure recombinant DvCH3H, as judged by analytical HPLC-SEC. This protein fraction was used for biochemical characterization of DvCH3H.

Biochemical characterization

Kinetic properties of recombinant DvCH3H

DvCH3H needs a cytochrome P450 reductase (CPR) as redox partner in order to be catalytically active. Therefore, microsomal preparations of yeast cells expressing Catharanthus roseus NADPH-cytochrome P450 reductase (CrCPR) were included in all activity assays. In the presence of microsomal preparations of recombinantly expressed CrCPR and NADPH as the electron donor, recombinant DvCH3H converted isoliquiritigenin to butein, naringenin to eriodictyol, apigenin to luteolin, kaempferol to quercetin, and dihydrokaempferol to dihydroquercetin. However, isoliquiritigenin and dihydrokaempferol were not only metabolised by DvCH3H to butein and dihydroquercetin, but also by yeast originating enzymes present in the microsome preparation of recombinant CrCPR to unidentified reaction products. Due to the high extent of these unknown products, the kinetic parameters of these substrates would be strongly biased and, therefore, could not be included in the kinetic studies.

For the other substrates, the pH optimum of the recombinant DvCH3H was investigated from pH 5.5 to pH 8.5 (see Supplementary Fig. S4). The pH optimum of DvCH3H for all tested substrates is between 7.0 and 8.0. DvCH3H shows the highest activity for naringenin between pH 7.0 and pH 7.5, for kaempferol between pH 7.5 and pH 8.0 and for apigenin at pH 7.5. For better comparability, the kinetic measurements were performed at pH 7.5. A NADPH concentration of 1.55 mM showed a sufficient excess, without limiting the activity of DvCH3H.

Kinetic analysis of recombinant DvCH3H was performed with the flavonol kaempferol, the flavone apigenin, and the flavanone naringenin. Usually, just a racemic mixture of naringenin is used as a substrate for investigations of enzymes involved in the flavonoid pathway. In the flavonoid pathway, however, only the (2S)-enantiomer is formed by CHI. A racemic mixture can be formed in planta, if the chalcones are isomerized chemically and in the absence of CHI. In this study, we therefore tested the racemic mixture and the pure enantiomers separately. The substrate concentration was varied in a range of 0.25–100 µM (Fig. 6) at a fixed concentration of 1.55 mM of NADPH, and the product formation was quantified by HPLC. The respective values for KM, vmax and kcat were calculated from the curves in Fig. 6 and are summarized in Table 4.

Figure 6
figure 6

Specific activity (v) of the recombinant membrane bound DvCH3H in dependence of the substrate concentration (S). The graphs show steady state kinetic measurements of DvCH3H using (a) (2S)-naringenin, (b) (2R)-naringenin, (c) racemic naringenin, (d) kaempferol and (e) apigenin as substrates.

Full size image
Table 4 Kinetic parameters of purified recombinant membrane bound DvCH3H.
Full size table

The kinetic data for apigenin, kaempferol, (2S)-naringenin, (2R)-naringenin and the racemic naringenin show KM values in a range below 1 µM, reflecting a high affinity of recombinant DvCH3H to all substrates (Table 4). The lowest KM value and the highest catalytic efficiency was observed with apigenin as substrate. In contrast, the highest KM value was obtained with (2S)-naringenin as substrate, however, the catalytic efficiency was slightly higher than the catalytic efficiency for apigenin. Unexpectedly, recombinant DvCH3H displayed a catalytic efficiency for (2R)-naringenin that is comparable to that of kaempferol. Furthermore, (2R)-naringenin showed a near two-fold higher affinity and a slightly higher turnover number than (2S)-naringenin, although it is not the naturally formed enantiomer in the flavonoid pathway.

The KM- and vmax value of the racemic naringenin basically reflects the values obtained with (2R)-naringenin, due to the significantly higher affinity and turnover rates of (2R)-naringenin in comparison to (2S)-naringenin. This indicates that recombinant DvCH3H would preferentially transform (2R)-naringenin to (2R)-eriodictyol, if racemic mixtures would be naturally available as substrate. A stereospecific interaction of naringenin enantiomers with various CYPs has been previously reported and is of pharmacological interest as naringenin glycosides are abundantly present in grapefruit juice, which is known to inhibit metabolization of drugs49. Interestingly, (2R)-naringenin was previously shown to be converted by the recombinant 2-oxoglutarate dependent flavonol synthase from Citrus unshiu to (−)-trans-dihydrokaempferol, but the subsequent conversion into the corresponding flavonol was not possible50. As other enzymes, such as DFR and flavone synthase I, are also stereospecific51, (2R)-naringenin conversion does not seem to be of physiological relevance in the latter steps of the flavonoid pathway.

From the natural substrates, the flavone apigenin seems to be the preferred substrate. Compared with the kinetic data reported for CH3H of Cosmos sulphureus12, recombinant DvCH3H shows a distinct difference in the flavonoid preferences. Recombinant DvCH3H has the highest catalytic efficiency for apigenin, whereas recombinant CsCH3H has the highest hydroxylation efficiency with kaempferol, apart from the preferred isoliquiritigenin. Unfortunately, chalcones could not be tested with the current test system. However, from the three flavonoid substrates tested, isoliquiritigenin, shows the closest structural similarity to apigenin due to the double bond in the C3-bond connecting rings A and B, and in contrast to the flavonol kaempferol, no hydroxyl group is present at the double bond. Based on these similarities, a high affinity for isoliquiritigenin and a high efficiency in conversion of this substrate can be hypothesized. Activity assays with purified CPR, as well as crystal structure analysis of substrate-enzyme complexes, will provide further insights into the differing binding modes of the different substrates.

Conclusion

This is the first report of the isolation of a CH3H cDNA sequence of dahlia flowers, as well as the recombinant production, purification and biochemical characterization of the enzyme. We developed a purification procedure for the transmembrane helix possessing DvCH3H by step-wise optimization of each downstream processing step. The final procedure allowed an overall recovery of 30%, with more than 60% in each purification step. The obtained DvCH3H was 84% pure and biochemically characterized regarding its pH optimum, as well as kinetic parameters on various substrates, including apigenin, kaempferol and the enantiomers of naringenin. In particular, we were able to detect the hydroxylation of (2R)-naringenin, which is not the enantiomer that is naturally formed by CHI. The results potentially lay the groundwork for future crystallization of DvCH3H, and thus understanding its ability to hydroxylate chalcones in position 3, a feature that is not inherent to the majority of F3′Hs.

Materials and methods

Plant material

The investigations were performed with petals of D. variabilis cv. Aurora’s kiss (Plant nursery Wirth, Vienna, Austria). The plant material was collected during the summers of 2007 and 2008, frozen in liquid nitrogen and stored at − 80 °C.

Cloning and recombinant expression of DvCH3H in Pichia pastoris

Extraction of mRNA and cDNA synthesis were performed with the µMACS mRNA isolation kit (Miltenyi Biotec, Bergisch Gladbach, Germany) and are described in detail in the SI and Supplementary Table S2. The cloning into the appropriate vectors, the transformation into Pichia pastoris as well as the recombinant expression were all performed as described previously40. The cultivated cells were harvested by centrifugation at 7000× g for 20 min at 4 °C, the supernatant was discarded and the cell biomass was frozen at − 20 °C until further use.

Sequence and phylogenetic analysis

The amino acid sequence of DvCH3H was used as a query for a BLASTP search in the NCBI database. The search was limited to Asteraceae species, the expect value was set to 1e-190 and the maximal target sequences was set to 250 sequences. The obtained sequences were aligned and partial sequences were omitted before the remaining 51 sequences were subjected to phylogenetic analysis by MEGA11 software52. The amino acid sequences were aligned using the muscle algorithm with default parameters followed by construction of a phylogenetic tree based on the maximum likelihood (Jones-Taylor-Thornton (JTT) model) with default parameters and 1000 bootstrap replicates.

Amino acid sequence diagram with secondary structure features

A homology model of DvCH3H was prepared with MODELLER v10.253 using the structure of CYP76AH1 (pdb: 5YM3) as well as the homology model of F3′H from Petunia hybrida (Q9SBQ9) obtained from the structure prediction website alphafold.ebi.ac.uk54. Molecular graphic images were generated with PyMOL (www.pymol.org) and secondary structure features were assigned using DSSP (Dictionary of Secondary Structures of Proteins)35.

Cloning and heterologous expression of CrCPR

The gene encoding a NADPH-dependent cytochrome P450 reductase CrCPR (GenBank: X69791.1) was codon-optimized for expression by a commercial supplier (Genscript, Piscataway, NJ, USA). For details of cloning and heterologous expression in Saccharomyces cerevisiae see SI.

Purification of transmembrane helix possessing DvCH3H

Homogenization and microsomal preparation

Unless stated otherwise, homogenization was carried out as follows: 60 g dry cell mass were dissolved per 1 L buffer A (50 mM Tris, pH 8, 100 mM NaCl, 10% Glycerol, 0.5 mM TCEP) by mixing with an immersion blender type 4179 (Braun, Neu Isenburg, Germany). For the calculation of the dry cell mass an aliquot of the cell suspension was dried at 105 °C. The resuspended cells were then homogenized at 1800 bar for 10 passages with a PandaPLUS2000 homogenizer (GEA, Düsseldorf, Germany). The homogenized biomass was centrifuged at 1000 × g for 20 min at 4 °C in a type Z 36 HK centrifuge (Hermle, Wehningen, Germany) unless stated otherwise.

Ultracentrifugation was carried out with a Sorvall WX Ultra Series WX 80 centrifuge (Thermo Fisher Scientific, Waltham, MA, USA) using a fixed angle Type 50.2 Ti Rotor (Beckman Coulter, Brea, CA, USA). The supernatant of the homogenized and centrifuged sample was filled into Optiseal Polyallomer Tubes (Beckman Coulter, Brea, CA, USA) and ultracentrifuged at 4 °C for 30 min at 200,000 × g unless stated otherwise.

The analyses of the different steps were carried out as follows: For the homogenization process, frozen and thawed cells and frozen, thawed and homogenized cells (at the same dry cell weight concentration) were diluted 1:2 to 1: 1,000,000 with sterile homogenization buffer and then streaked on YPD-zeocin plates. The plates were incubated for 3 days at 30 °C before counting the CFUs. For the homogenization and centrifugation efficiency, the cell pellet and supernatant were diluted to the original volume with buffer A and then analysed by use of western blots.

Optimization of the membrane solubilisation

The list of the tested detergents is shown in Table 5. For membrane solubilisation the pellets were redissolved in 1.5 times the original volume with buffer A by mixing with an Ultra-Turrax IKA T10 basic Instrument (IKA, Staufen, Germany). Concentrated detergent stock solution was added to a final concentration of 10 × critical micelle concentration (CMC) for each of the detergents and the solution was swayed on a PMR-30 Compact Fixed-Angle Platform Rocker (Grant Instruments, Royston, UK) at 4 °C for 4 h unless stated otherwise. Then, the mixtures were subjected to ultracentrifugation at 200,000 × g for 30 min. The solubilization efficiency was determined by diluting pellet and supernatant to the original volume with buffer A and subjecting the samples to western blots.

Table 5 Detergents used for membrane solubilisation, including their abbreviations as used in the following, as well as suppliers.
Full size table

Immobilized metal ion affinity chromatography

For preparative immobilized metal ion affinity chromatography (IMAC) an ÄKTA pure system (GE, Boston, MA, USA) was used. The supernatant of the detergent treated microsomes was diluted 1:10 with buffer A to reduce the detergent concentration. Additionally, a 500 mM imidazole solution in buffer A was added to reach a final concentration of 20 mM imidazole in the sample. The solution was loaded onto a pre-equilibrated column (either HisTrap FF column (GE, Boston, MA, USA) or CIM IDA-1 (BIA Separations, Ljubljana, Slovenia), both loaded with nickel ions) with a flow-rate of 127.5 cm/h. Elution was performed in a step gradient to 100% buffer B (50 mM Tris, pH 8, 100 mM NaCl, 10% glycerol, 0.5 mM TCEP, 500 mM imidazole, 2 mM DDM).

Preparative and analytical size exclusion chromatography

Preparative Size Exclusion Chromatography was carried out with a pre-equilibrated HiLoad 16/600 Superdex 200 pg (GE, Boston, MA, USA) for further purification. Analytical size exclusion chromatography was realized on an Ultimate HPLC 5000 (Thermo Scientific, Waltham, MA, USA) with a Superdex 200 Increase 5/150 GL (GE, Boston, MA, USA) (for details see SI).

Protein concentration determination SDS-PAGE and western blot

The protein concentration was measured by Bradford assays in crude samples (for details see SI). SDS-PAGE and western blot analysis are described in detail in the SI.

Enzymatic assay

The enzymatic assays of purified recombinant membrane bound DvCH3H were performed in triplicates with, apigenin, kaempferol, and naringenin ((2R)-, (2S)-, and racemic) as substrates. (2S)-naringenin and (2R)-naringenin were purchased from PlantMetaChem (Marburg, Germany). Apigenin, and kaempferol were obtained from Extrasynthese (Genay, France). (+ /−)-Naringenin and NADPH were purchased from Sigma-Aldrich (Vienna, Austria). The reaction mixtures contained 5 µl purified recombinant DvCH3H (50 mM Tris, 100 mM NaCl, 10% glycerol, 2 mM DDM, 0.5 mM TCEP, pH 8.0), 40 µL Saccharomyces cerevisiae INVSc1 microsomal preparation of recombinantly produced CrCPR, 1.55 mM NADPH, 10 µM substrate in 100 mM HEPES (4- (2- hydroxyethyl)-1-piperazineethanesulfonic acid), 200 mM NaCl, pH 7.5 in a final volume of 100 µL. After 30 min reaction time at 30 °C the reaction was stopped with 20 µL of 20% acetic acid in acetonitrile. After centrifugation at 16,000 × g for 5 min the reaction solution was filtered through a 0.22 µm PTFE membrane. For analysis, a Dionex UltiMate 3000 RSLC System (Thermo Scientific) equipped with a diode array detector (DAD) was used. The product identification was carried out by comparing the retention time as well as the characteristic spectral absorption of the corresponding standards (for more details see SI and Supplementary Fig. S5).

Product identification by LC–MS–MS

Enzymatic assays were performed as described before, except the reactions were stopped by mixing with 10 µL acetic acid and extracted with 70 µL ethyl acetate. After drying, the reaction mixture was dissolved in 40 µL methanol.

The reaction products luteolin, eriodictyol and quercetin were identified by high-performance liquid chromatography coupled to mass spectrometry (for more details see SI, Supplementary Table S3 and Supplementary Fig. S6).

Kinetic analysis

Experiments for determination of kinetic parameters of recombinant DvCH3H were performed by varying the substrate concentration. The amount of enzyme used was 1.25 ng for apigenin and kaempferol and 0.83 ng for racemic naringenin, as well as for (2S)- and (2R)- naringenin. Data analysis was carried out by nonlinear regression, mean values and standard deviations were calculated based on three replicates. Calculations and the graphs were prepared by employing the program OriginPro 2018 (OriginLab).

Ethical approval

The authors confirm that the cDNA clones were obtained from a commercially available Dahlia cultivar. As such, no permissions or licences were required under institutional, national or international guidelines or regulations. The study is in full compliance with the IUCN Policy Statement on Research Involving Species at Risk of Extinction and the Convention on the Trade in Endangered Species of Wild Fauna and Flora, since the commercial varieties used are neither endangered nor at risk of extinction. The cultivar used is listed in the manuscript, including its commercial availability.