A biocatalytic hydroxylation-enabled unified approach to C19-hydroxylated steroids

Steroidal C19-hydroxylation is pivotal to the synthesis of naturally occurring bioactive C19-OH steroids and 19-norsteroidal pharmaceuticals. However, realizing this transformation is proved to be challenging through either chemical or biological synthesis. Herein, we report a highly efficient method to synthesize 19-OH-cortexolone in 80% efficiency at the multi-gram scale. The obtained C19-OH-cortexolone can be readily transformed to various synthetically useful intermediates including the industrially valuable 19-OH-androstenedione, which can serve as a basis for synthesis of C19-functionalized steroids as well as 19-nor steroidal drugs. Using this biocatalytic C19-hydroxylation method, the unified synthesis of six C19-hydroxylated pregnanes is achieved in just 4 to 9 steps. In addition, the structure of sclerosteroid B is revised on the basis of our synthesis.

C 19-hydroxylated steroids belong to a unique family of natural products with diverse bioactivities (Fig. 1a). For example, stereonsteroid A and B 1 exhibit excellent cytotoxicity against P-388 and HT-29 cancer cell lines. Besides the general cardiotonic activity, strophanthidol 2 , 19hydroxysarmentogenin [3][4][5] and ouabagenin [6][7][8] also possess antitumor activities. Bufadienolides bryophyllin A and C 9 show strong insecticidal activity against third instar larvae of silkworm (Bombyx mori) (LD 50 3-5 μg g −1 ). However, isolating these natural products from their sources is very low-yielding. For instance, bryophyllin A-C were isolated in 0.00018%, 0.000043%, and 0.00003% yields (w/w), repectively 10,11 , which severely impedes further biological and pharmaceutical studies. Consequently, chemical synthesis is considered as an alternative approach for producing these steroidal compounds 4,5,7,8 . In recent years, selective C-H oxidation has emerged as one of the important reactions used in the chemical synthesis of steroidal compounds 12,13 . However, functionalization of the C19 methyl group in steroids through C-H oxidation still remains as a challenge. The synthetic strategies developed so far include the Norrish-type II photoreaction and oxidative fragmentation method 7,8 for steroids containing a C11 carbonyl group, and the oxygen radical-induced remote C-H hydroxylation under Barton's nitrite ester photocleavage 14,15 or Suάrez-type conditions ( Fig. 1b) [16][17][18] . Despite the effectiveness in C19-hydroxylation, all of these methods required prefunctionalization at specific positions of steroids, and necessitated strict reaction conditions and multi-step chemical transformations, often leading to low synthetic efficiency. Therefore, developing synthetic strategies for direct steroidal C19-hydroxylation is highly desirable.
Biocatalysis is emerging as a powerful and valuable tool for organic synthesis 19,20 , and has the potential to be more efficient than chemical synthesis, particularly for functionalization of unactivated C-H bonds 21,22 . For example, P450 or other radicalmediated enzymes have been used to directly introduce functional groups to inert carbon moieties with high regio-and enantioselectivity 20,23 . Combining biocatalytic C-H activation and chemical synthesis could enormously shorten the synthetic routes in natural product synthesis, as were demonstrated in a few of recently disclosed successful approaches towards some interesting natural products [24][25][26][27][28][29][30] .
Inspired by these reported work, we start our exploration on using a biocatalytic approach to directly introduce the C19hydroxyl group into steroids. It has been documented that a few filamentous fungi, belonging to the Pellicularia genus, able to hydroxylate the steroidal C 19 -H bond 31 . It interests us that the strain and substrate pair, Thanatephorus cucumeris NBRC 6298 and cortexolone (1), are able to produce the C 19 -OH cortexolone (2) (Fig. 1c) with a 20% conversion 31,32 . Albeit the low conversion, this method looks attractive to us owing to its easy operation and low cost, especially given that 2 could serve as a versatile intermediate to various important C19 functionalized steroids (Fig. 1c), and the currently available access of 2 has to go through a 13-step synthesis (Supplementary Fig. 1) 8 .
Herein, we report our work on developing a steroidal C19 hydroxylation strategy, and a unified efficient synthesis of six bioactive pregnanes from 2 which examples the power of this strategy.

Results
Identification of the C19-hydroxylase. We initially repeated the previously reported procedure of C19 bio-hydroxylation 31,32 . Consistent with the reports, the conversion of 1 to 2 in our hands was 20.3% (entry 1 in Table 1). Besides 2, 11β-hydroxycortexolone (3, 12.5% conversion) was also obtained as well as a number of minor side-products (trace III in Fig. 2). These side products were formed probably due to the non-specific activity of C19 hydroxylase or the existence of other enzymes. The responsible gene was cloned from the T. cucumeris, and Nine p450 and one p450 reductase genes that robustly transcribed were identified through transcriptome analysis (Supplementary  Table 1). These p450 genes were cloned under the strong promoter PamyB, and individually transformed into the model filament fungus Aspergillus oryzae NSAR1 33 for evaluation of the activity. To our delight, tcP450-1, one of the p450 genes, is confirmed to be able to convert 1 to 2. However, like T. cucumeris, this recombinant strain also produced 3 and other minor side products (Fig. 2, trace II), suggesting that TcP450-1 has a relaxed specificity and generates all observed cortexolone derivatives. Based on the ratio of each product, the maximum conversion of 2 by TcP450-1 is inferred to be 48% (calculated by dividing its peak area into the sum of all products' peak area), significantly higher than the current value of 20.3%, and indicating a potential for further improvement.
Optimization of the C19 hydroxylation activity. In spite of easy manipulation, basal activity in A. oryzae is also able to convert 1 into a reduced product, C 20 -OH cortexolone (4) (Fig. 2, trace I and II), which results in very low conversion of the C 19 -OH product. During this study, Lu et al also identified the C 19 -OH hydroxylase and overexpressed it in Pichia pastoris 34 . Although no reduced products were generated, the methylotrophic yeast system still results in very low conversion. As the usefulness of both heterologous systems are limited, we, therefore, go back to optimize the original T. cucumeris system. We then worked on the optimization of fermentation conditions. As 1 was found partially precipitated in the fermentation broth, we reduced the concentration 1 from 0.5 g L −1 to 0.25 g L −1 which, as a result, increased the conversion of 1 to 27.0% (entry 2 in Table 1 and Supplementary Fig. 2). Subsequent using liquid media to develop a seed culture could continue to improve the conversion to 29.0% (entry 3 in Table 1 and Supplementary Fig. 2).
As TcP450-1 is an ion-dependent enzyme, and metal ions could play a significant role in cell growth and enzyme activity, we evaluated the effect of Mg 2+ , Zn 2+ , Mn 2+ , and Fe 2+ (each 1 mM L −1 ) on the substrate conversion. It was found that Fe 2+ was the best among the ions evaluated: the conversion could be increased to 34.9% and 45.1% when the concentration of Fe 2+ added into the fermentation culture was 1.0 mM L −1 and 1.5 mM L −1 , respectively (entries 4 and 5 in Table 1 and Supplementary  Figs. 3 and 4).
As the obtained highest conversion is close to the theoretic value (48%), we shifted the focus onto optimization of the enzymatic reaction. Although protein engineering is generally an effective way to improve an enzymes' substrate specificity 20,22 , this direction could be very challenging for TcCYP450-1 due to its membrane-binding properties and lack of reliable homologous structural models. Therefore, instead of changing the enzyme we chose to modify the substrate. It was described that progesterone, a steroid structurally similar to cortexolone (1), was not able to be hydroxylated in C19 31 , which implied that the hydroxyl substitution at C17 or C21 in 1 could be crucial for the enzyme's specificity. Therefore, engineering these hydroxyl groups might be helpful. To test this assumption, 21-acetyl-cortexolone (5) and 17-acetyl-cortexolone (6) 35 are employed (Fig. 2). Feeding of 5 to the broth culture resulted in an identical product profile to that of 1 (trace V in Fig. 2). To our delight, when 6 was used as the substrate, the formation of 11β-OH and other minor side products was greatly reduced, and 2 was formd in a significantly high conversion (80%) (entry 6 in Table 1; trace VI in Fig. 2  Furthermore, the key intermediate C 19 -OH 17-acetyl-cortexolone (7) was identified in the early stage of biocatalysis (trace VI in Fig. 2 and Supplementary Fig. 5). Hence, the C17 acetyl group in 6 played a role in not only the C19 hydroxylation specificity, but also the C19 hydroxylation activity as 6 was able to be fully converted into 2 within 24 h ( Supplementary Fig. 5). Thus, we postulated that 6 is quickly hydroxylated into 7 which is then gradually hydrolyzed into 2 regioselectively, and the difference in hydroxylation and hydrolysis reactivities ensures the high C19specficity. The high efficiency of this biocatalytic approach is obvious considering that the precedent chemical approach requires 13 steps from cortisone acetate and suffers from a very low overall yield (2%) (Supplementary Fig. 1) 8 . Following this method, we scale up the biotransformation and obtain over five grams of 2.
Synthetic versatility of C 19 -OH cortexolone (2). Armed with this scalable and robust C19 hydroxylation method, we started to examine its versatility by transforming 2 into other synthetically useful intermediates (Fig. 3). The first exampled transformation was the treatment of 2 with basic hydrogen peroxide 36 . We were delighted to find that the selective β-face epoxidation of 2 took place to provide epoxide 8 as the sole diastereomer (58% yield) whose formation, we believe, was stereo-specifically directed by the C19 hydroxy group in 2. When 2 was subjected to Pd/C mediated hydrogenation conditions, the A/B cis-configurated product 9 was formed with moderate diastereoselectivity. DDQmediated desaturation of the A ring in the protected form of 2 with its 19,21-diols protected by acetyl group 37 , intermediate 10 was obtained with 81% yield over two steps from 2. The allylic oxidation of C6 in 2 was realized through a two-step sequence involving the formation of enol acetyl ester and PhIO-mediated oxidation 38 , resulting in a β-allylic alcohol 11 (overall yield of 52% from 2). Additionally, the TMSI-promoted 39 selective C17 dehydroxylation of 2 resulted in the highly desired 19hydroxydeoxycorticosterone (19-OH-DOC) 12 40 in 84% yield (β:α = 20:1). This transformation is intriguing in terms of its high chemo-and diastereo-selectivity since the C19,21 diols as well as C3,20 carbonyl groups remained untouched during the process. Literature precedents 41,42 showed that the α-hydroxyl ketone motif in the side chain of 2 and 12 could act as a functional handle for direct appendage of butenolide motif (13), which is ubiquitous in cardiotonic steroids. As another important exampled application, 2 was transformed to 19-hydroxyandrostenedione 14 in 84% yield through a facile oxidative cleavage of the side chain. Both 12 and 14 are key intermediates in pharmaceutical industry as were used in the preparation of more than ten 19-nor steroidal drugs such as 19-nor-DOC 40 , norethisterone, and ethinyloestradiol (Supplementary Figs. [6][7][8]. It is worth noting that the preparation of either 12 or 14 has to be through more than 8 steps using the current conventional chemical methods (see Supplementary Figs. 6, 9-11 for details).
Collective synthesis of C 19 -OH pregnanes. To further demonstrate the usefulness of this strategy, we have accomplished the unified synthesis of five bioactive pregnanes in stereonsteroid family, with intermediate 12 as the starting point (Fig. 4). Our embarking on these biologically important stereonsteroids started from constructing the trans A/B ring junction. Considering the previously incorporated C19 hydroxy group would take a deleterious directing effect, as demonstrated in the preparation of 9, the two hydroxy groups on C19 and C21 were protected as TBS ethers to provide 15. Then, a one-pot reaction involving Pd/Ccatalyzed hydrogenation of C=C double bond and subsequent NaBH 4 -mediated reduction of 3,20-diketones was conducted to deliver 16 which was a key intermediate since it already possessed the desired trans-fused A/B ring system. Removal of the TBS protecting group from 16, followed by NaIO 4 -mediated oxidative cleavage of vicinal diol in the side chain lead to aldehyde 17 in excellent yield (95% over two steps). After a Wittig olefination was performed in 93% yield, the synthesis stereonsteroid A (18) was achieved in just five steps from 12. Then, three natural pregnanes, sclerosteroid A (20), ceratosteroid C and D (21 and 19), were concurrently obtained through selective acetylation of 18 using acetic anhydride. Moreover, ceratosteroid C (21) was further transformed into stereonsteroid B (22) quantitatively through a facile Dess-Martin periodate oxidation. These five     Encouraged by the successful step-economic synthesis of the five stereonsteroid compounds, we shifted our efforts to the synthesis of sclerosteroid B which was reported to possess promising anti-inflammatory activity 43 . As depicted in Fig. 5, the synthesis also started from 12, which was transformed to the conjugated enol acetyl ester 24 through concurrent acetylation of the C19, C21 diol in an excellent yield. Then, one-pot procedure involving NaBH 4 -mediated reduction and subsequent NaIO 4mediated oxidative cleavage of the in situ formed vicinal diols in the side chain to provide the aldehyde 25 in 75% yield 45 . It is worth mentioning that, through this one-pot process, three transformations had been achieved, including diastereoselective reduction of the C3 carbonyl, reduction of the C20 carbonyl, as well as removal of the acetyl protection of the C19,21 diol. Finally, 25 was transformed to the reported structure of sclerosteroid B (26) through a one-pot Wittig olefination and bis-acetylation. However, we found that the spectra of the synthetic sample 26 do not match with what was reported 43 . After careful examination of the spectral data, we surmised that the internal double bond might be assigned incorrectly in the reported structure, and C4-C5 position might be the correct location for the double bond. To confirm our hypothesis, we started our synthesis of the proposed molecule in just two steps from 12 by following a similar strategy. After 12 was transformed to aldehyde 27 through a one-pot process of Luche reduction 46 and NaIO 4 -mediated oxidative cleavage of the side chain, another one-pot operation involving Wittig olefination and acetylation of C19,21 diol transformed 27 into the revised sclerosteroid B (28) in 78% yield. Gratifyingly, the NMR spectra of the synthetic sample 28 is consistent with those reported. Thus, we concluded that 28 is the correct structure of sclerosteroid B whose synthesis was achieved in just two steps from 12. In the biological assay, both 28 and 26 showed cytotoxicity activity against human colon cancer cell line HT29 by inducing cell apoptosis (IC 50 = 59.76, and 169.50 μM for 28 and 26, respectively) (see Supplementary Fig. 12). These results suggest that the location of the double bond is important for the anticancer activity.

Discussion
C19 hydroxylation is a unique feature of some natural-occurring bioactive steroids, but these 19-hydroxylated steroids are normally in low abundance in nature, which has severely impeded their biological and medicinal investigations. Worse still, this modification is very challenging from both the chemical and biological synthesis perspective. Traditional chemical approaches for steroidal C19 hydroxylation require prefunctionalization at specific positions and multi-step chemical transformations (see Fig. 1) 7,8,[14][15][16][17][18] ; necessitate harsh reaction conditions and suffer from low synthetic efficiency. To the best of our knowledge, only one example of biocatalytic direct C19 hydroxylation has been reported so far (from 1 to 2) 31 , but the synthesis suffered from very low efficiency (20% conversion).
To provide good solutions to this synthetic challenge, we turned our focus on developing an efficient and broadly applicable approach to access the 19-hydroxylated steroids by combining the power of biocatalysis and synthetic chemistry. Initially, we cloned the C19-hydroxylase TcP450-1. Through heterologous expression in the A. oryzae NSAR1, we conjectured that its low activity could be majorly stemmed from poor reaction specificity (generating other hydroxylated products). However, protein engineering to alter its substrate specificity could be a very difficult task since this protein has two transmembrane segments ( Supplementary Fig. 13), and no similar proteins could be used in homologous structural modeling. To address this problem, we employed an unusual strategy-manipulating the substrate instead of engineering the enzyme. Fortunately, we identified that acetylation of the C17-OH of cortexolone (1) could significantly improve the enzyme's specificity and efficiency (conversion from 20.3 to 80%), allowing C 19 -OH-cortexolone (2) be formed in a scalable way.
The versatile C 19 -OH-cortexolone (2) could be transformed into an array of synthetically useful intermediates (8)(9)(10)(11)(12)(13)(14), among which 19-hydroxydeoxycorticosterone (19-OH-DOC, 12) and 19-OH-androstenedione (14)  them were prepared from 2 in just one step in excellent yields. In contrast, these two compounds are currently made in 8 to 11 steps by conventional chemical methods with low efficiency, and nasty reagents are required as well 40,47,48 (Supplementary Figs. 6,[9][10][11]. 12 retains the 17-β configurated stereocenter which is ubiquitous in bioactive steroids. Moreover, the C17 α-hydroxy ketone side chain of 12 provides very useful handles for further structural manipulations. For instance, by reacting with Bestmann ylide, this hydroxymethyl ketone moiety can be efficiently converted into the γ-crotonolactone 41,42 , the core structure of the large cardenolide family. As this work afforded a very convenient approach to access 12, a general synthetic strategy was conceived for preparation of C19-OH 17-β steroids with using 12 as the key precursor. The effectiveness of this strategy was well-validated through our achievement of the unified synthesis of six C19hydroxylated pregnanes as well as the synthesis and structural revision of sclerosteroid B. The primary C19-OH precursor 2 has already been commercialized through our collaboration with Adamas, a chemical reagent company. Because the C19 hydroxymethyl group in 14 can be readily removed through the facile oxidative decarboxylation, 14 has become the highly valuable common intermediate to manufacture the 19-nor steroidal drugs in industry. Currently, more than ten 19-nor steroidal drugs are made employing 14 as the key intermediate ( Supplementary Figs. 7, 8). Nevertheless, current process for the preparation of 14 is very lengthy (8 steps) and suffers from low efficiency and toxic reagents ( Supplementary Figs 10-11 for details). In stark contrast, by using our method, 14 could be afforded in just one step from 19-OH-cortexolone (2) or 3 to 4 steps from cortexolone 21-acetate (RSA) ( Supplementary  Fig. 14). It is worthwhile mentioning that RSA is an inexpensive precursor for making hydrocortisone and produced in bulk in the steroid industry (the current price is less than 100 USD kg −1 ). Furthermore, considering the reagents and reaction conditions employed, it is evident that our method is much greener and more economical than the current chemical synthesis, so it has the potential to replace the manufacturing process of 14.
In addition to the above above-mentioned work, we predict the protection group in 17-OH could lay a solid foundation for further expanding the substrate scope of C19 bio-hydroxylation. Exploration of the relationship between the enzymes' specificity and C17-OH protection groups as well as other steroid variants with 17-acetyl are ongoing in our laboratory and will be reported in due course.

Methods
Transcriptome analysis of T. cucumeris. T. cucumeris was inoculated into the YD liquid medium and cultured at 30°C, 200 rpm for 2 d before transferred into a 250 mL Erlenmeyer flask containing 100 mL of YD liquid medium (5% (v/v)). Cortexolone (50.0 mg, final concentration 0.5 g L −1 ) is dissolved in DMF (0.5 mL) and filtered for sterilization, then added to the medium. The fermentation of the culture is then continued under the same conditions until a small amount of product is detected by LC-MS. The mycelium was centrifuged down and immediately stored in liquid nitrogen. The total RNA was extracted for transcriptome sequencing at Genewiz (Suzhou, China).
Heterologous Expression of TcP450-1 in Aspergillus oryzae. The gene of tcP450-1(P450 19OH ) obtained by PCR amplification of the cDNA of T. cucumeris using primer tcP450-1-F/R (Supplementary Table 2) and cloned into pTAex3 at the EcoRI site to yield the expression plasmid pWHU2490. A. oryzae NSAR1 was grown on PDA for 5 d. The seed culture was inoculated into a 250 mL Erlenmeyer flask containing 50 mL of DPY medium. After incubation at 30°C for 2 d, mycelia were collected and washed twice with an aqueous solution of NaCl (0.9% w/w). Protoplast was prepared by using Yatalase (Takara; 3.0 mg mL −1 ) and Lysing Enzymes (Sigma; 4.5 mg mL −1 ) in Osmotic Medium (0.6 M of MgSO 4 , 10 mM of Na 2 HPO 4 , pH 5.8) at 30°C for 3 h. The Osmotic Medium containing Protoplasts was filtered and centrifuged at 1157 × g for 5 min, then washed with trapping buffer (0.6 M D-Sorbitol, 0.1 M Tris-HCl, pH 7.0) and STC buffer (1.2 M of D-Sorbitol, 10 mM of CaCl 2 , 10 mM of Tris-HCl, pH 7.5). Then, the protoplasts were diluted by STC buffer till 10 8 -10 9 cells mL −1 , and used for plasmid transformation immediately 33,49,50 .
Plasmids pWHU2490 and pTAex3 were added into the protoplast in order to generate the corresponding TcP450-1 expression strain and control strain, respectively. After the protoplast was incubated on ice for 60 min, 1 mL PEG4000 solution (250 g L −1 PEG4000, 100 mM CaCl 2 , 600 mM KCl, 50 mM Tris-HCl pH 7.5) was added. After 30 min incubation at room temperature, STC buffer (4 mL) solution was added. The transformation system was then poured onto the Czapek-Dox agar plate supplemented with appropriate nutrients, and incubated at 30°C for 3-7 d. Transformants were picked out and streaked onto the corresponding auxotrophic Czapek-Dox plates for growth for a few days. The recombinant strains (mWHU2487 and mWHU2488) were then transferred to the corresponding auxotrophic Czapek-Dox liquid medium, and shaken at 30°C for 2 d to extract DNA. The genotype of the transformants mWHU2487 and mWHU2488 were verified by PCR using primers Pamy-gt-F and Tamy-gt-R (Supplementary Table 2, Supplementary Fig. 15).
Biotransformation with T. cucumeris. Using liquid media to develop seed culture: T. cucumeris was inoculated into the YD liquid medium and cultured at 30°C at 200 rpm for 2 d 31,32 . Then, 5 mL seed culture was transferred to a 250 mL Erlenmeyer flask containing 100 mL of YD liquid medium, and incubation was continued under the same culture conditions for 2 d. Using solid media to develop seed culture: T. cucumeris was grown on PDA for 7 d, crushed and inoculated into a 250 mL Erlenmeyer flask containing 100 mL of YD liquid medium, and then further cultured at 30°C at 200 rpm for 2 d. Cortexolone (50.0 mg, final concentration 0.25 g L −1 or 0.5 g L −1 ) was dissolved in DMF (0.5 mL) and sterilized by filtration, and added into the culture broth. After the culture was continued to be fermented under the same conditions for 3 d, it was subjected to EtOAc extraction (3 × 100 mL EtOAc for each 100 mL of the culture, sonicated for 15 min) The combined organic extracts were dried under reduced pressure to result in metabolites which were subsequently re-dissolved with MeOH (2 mL) and filtered through a 0.22 μm membrane filter to remove particles. HPLC analysis was per- Biotransformation with recombinant A. oryzae strain. Recombinant A. oryzae strains are inoculated into a 250 mL Erlenmeyer flask containing 100 mL MPY medium (maltose-polypeptone-yeast extract: 3% maltose, 1% polypeptone, and 0.5% yeast extract) 49 . The strains were cultured at 30°C and 200 rpm for 2 d. Cortexolone (50.0 mg) was dissolved in DMF (0.5 mL, final concentration 0.25 g L −1 ), were sterilized by filtration, then added to the medium. The culture was continually incubated under the same culturing conditions for 5-10 d, and then subjected to EtOAc extraction (3 × 100 mL EtOAc for each 100 mL of the culture, sonicated for 15 min) The combined organic extracts were dried under reduced pressure to result in metabolites which were subsequently re-dissolved with MeOH (2 mL) and filtered through a 0.22 μm membrane filter to remove particles before HPLC analysis.
Biotransformation with different metal ions. T. cucumeris was into the YD liquid medium and cultured at 30°C and 200 rpm for 2 d. Then, 5 mL seed culture was is transferred into a 250 mL Erlenmeyer flask containing 100 mL of YD liquid medium, and the incubation was continued under the same culture conditions for 2 d. A solution of cortexolone (25.0 mg in 0.5 mL of DMF, final concentration 0.25 g L −1 ) and a solution of MgSO 4 , ZnSO 4 , MnCl 2 or FeSO 4 (final concentration 1.0 mM L −1 ) were sterilized by filtration, and added into the medium. After the culture was fermented under the same conditions for 3 d, it was subjected to EtOAc extraction (3 × 100 mL EtOAc for each 100 mL of the culture, sonicated for 15 min). The combined organic extracts were dried under reduced pressure. Metabolites were subsequently re-dissolved with MeOH (2 mL) and filtered through a 0.22 μm membrane filter to remove particles before HPLC analysis.
Biotransformation using the optimized condition. T. cucumeris was inoculated into the YD liquid medium and cultured at 30°C and 200 rpm for 2 d. Then, 5 mL seed culture was transferred into a 250 mL Erlenmeyer flask containing 100 mL of YD liquid medium, and incubation was continued under the same culture conditions for 2 d. A solution of substrate 1, 5 or 6 (25.0 mg dissolved in 0.5 mL of DMF, final concentration 0.25 g L −1 ) and an aqueous solution of FeSO 4 (1.5 mmol L −1 ) are filtered for sterilization, were sterilized by filtration, and added into the medium, and fermented under the same conditions for 3 d, and then subjected to EtOAc extraction (3 × 100 mL EtOAc for each 100 mL of the culture, sonicated for 15 min). The combined organic extracts were dried under reduced pressure to result in metabolites which were subsequently re-dissolved with MeOH (2 mL) and filtered through a 0.22 μm membrane filter to remove particles before HPLC analysis.
Reporting summary. Further information on research design is available in the Nature Research Reporting Summary linked to this article.

Data availability
The sequence of the P450 genes reported in this paper has been deposited in GenBank under accession number MK309346-MK309355. The X-ray crystallographic coordinate for structure of 23 reported in this study has been deposited at the Cambridge Crystallographic Data Centre (CCDC) (number 1853184), and can be obtained free of charge from CCDC via www.ccdc.cam.ac.uk/data_request/cif. The raw data underlying Supplementary Figs