Stereoselective chemo-enzymatic oxidation routes for (1R,3E,7E,11S,12S)-3,7,18-dolabellatriene

The diterpene (1R,3E,7E,11S,12S)-3,7,18-dolabellatriene from the marine brown alga Dilophus spiralis belongs to the dolabellanes natural product family and has antimicrobial activity against multi-drug resistant Staphylococcus aureus. Recently, we generated a CotB2 diterpene synthase mutant (W288G), which instead of its native product cyclooctat-9-en-7-ol, generates (1R,3E,7E,11S,12S)-3,7,18-dolabellatriene. In vivo CotB2 W288G reconstitution in an Escherichia coli based terpene production system, allowed efficient production of this olefinic macrocycle. To diversify the 3,7,18-dolabellatriene bioactivity we evaluated chemical and enzymatic methods for selective oxidation. Epoxidation by acetic peracid, which was formed in situ by a lipase catalyzed reaction of acetic acid with H2O2, provided efficient access to two monooxidized dolabellanes and to a novel di-epoxidated dolabellane species. These compounds could act as synthons en-route to new dolabellanes with diversified bioactivities. Furthermore, we demonstrate the almost quantitative 3,7,18-dolabellatriene conversion into the new, non-natural compound (1R,3E,7E,11S,12S,18R)-dolabella-3,7-diene-20-ol by hydroboration–oxidation with an enantiomeric excess of 94%, for the first time.

The efficient total chemical synthesis or biotechnological production process for structurally complex terpenoids such as dolabellanes, arteminisin, or taxol remains challenging (Kusama et al., 2000;Lozama and Prisinzano, 2009;Malik et al., 2011). Total chemical synthesis of functionalized natural products often involves multiple steps, resulting in racemic intermediates, and low product yields. In this respect, the synthetic assembly of olefinic macrocycles containing numerous stereocenters and often display elaborate functionalization patterns, such as glycosylation or addition of aromatic residues, are particularly demanding. Biotechnological routes for stereoselective production of functionalized terpenes are an increasingly attractive alternative to synthetic chemistry but at present remain rather inefficient . While biosynthetic routes for generation of olefinic terpene macrocycles in heterologous microbial production systems gain significant momentum due to advances in synthetic biotechnology , the identification and reconstitution of enzymes that functionalize terpene macrocycles in the desired heterologous production host, remains a bottleneck . In particular, efficient application of specific P450-type enzyme activities that can hydroxylate aliphatic diterpene macrocycles, en-route to highly functionalized terpenoids, often fail in target-oriented biosynthesis approaches (Zhang et al., 2011).
A strategy to circumvent these obstacles, is the efficient cell based production of terpene macrocycles in suitable Escherichia coli or Saccharomyces cerevisiae production strains (Engels et al., 2008;Ajikumar et al., 2010;Brück et al., 2014), followed by in vitro functionalization utilizing various chemical or enzymatic processes. These semisynthetic routes can provide efficient access to functionalized, bioactive terpenoids, but depend on the availability of the respective terpene synthase and its functional reconstitution in an efficient in vivo production system as well as the implementation of selective chemical synthesis steps.
Several chemical strategies have been used in semisynthetic approaches to convert terpenoids to diverse compounds. (Toyooka et al., 1993;Baloglu and Kingston, 1999;Morrison and Hergenrother, 2014;Turconi et al., 2014). However, these chemical strategies conventionally rely on available functionalized precursors, which are derived from isolated natural products or biotechnological production routes that provide specific biosynthetic intermediates. In certain instances the initial hydroxylation of the macrocycle has been accomplished by P450 type oxidoreductases (Le-Huu et al., 2015). By contrast, selective chemical hydroxylation of terpene macrocycles is extremely challenging due to the presence of chemically almost equivalent C-H bonds. Nevertheless, allylic hydroxylation by SeO2-mediated oxidation represent one successful approach. However, this particular reaction suffers from poor selectivity (Katsuyama et al., 2002).
The biosynthesis of dolabellane macrocycles starts, like all terpenes, with the stepwise condensation of the universal terpene precursors isopentenyl-diphosphate (IPP) and dimethylallyl diphosphate (DMAPP) yielding the aliphatic diterpene precursor geranylgeranyldiphosphate (C-20). In the subsequent cyclisation reaction by a diterpene synthase, a highly reactive carbocation is created by cleaving of the pyrophosphate group, which propagates the cyclisation through stepwise stereoselective connection of the carbon skeleton. The CotB2 W288G mutation attenuates the cyclisation cascade for the production of cyclooctat-9-en-7-ol by premature proton elimination (Meguro et al., 2015) to yield dolabellatriene (Figure 2). In analogy, it can be assumed that a native dolabellane synthase may have a similar reaction mechanism. In this study, we applied both enzymatic and chemical strategies to biotechnological production of dolabellane (2) in order to provide access to a broad spectrum of natural and non-natural dolabellanes.

General Experimental Procedure
All chemicals were obtained from standard sources at the highest purity grade. NMR spectra were recorded in CDCl 3 with an Avance-III 500 MHz (Bruker) at 300 K. 1 H NMR chemical shifts are given in ppm relative to CHCl 3 (δ = 7.26 ppm). 13 C NMR chemical shifts are given in ppm relative to CDCl 3 at δ = 77.16. The 2D experiments (HSQC, COSY, NOESY) were performed using standard Bruker pulse sequences and parameters.

Bacterial Strains, Genes, and Vectors
The E. coli strains XL-1 Blue and BL21 (DE3) were used for cloning and diterpene production. All strains and plasmids were obtained from Novagen/Merck Millipore. Genes were synthesized by Life Technologies GmbH featuring the appropriate restriction sites and adjusting codon usage for E. coli.
Extraction and Isolation of (1R,3E,7E,11S,12S)-3,7,18dolabellatriene The hydrophobic beads were recovered by filtration, washed with water and subsequently dried. Subsequently, these beads were extracted three times with 200 mL ethyl acetate. The organic phases were combined, dried with MgSO 4 . The solid formed after vacuum evaporation was used in analytical procedures. GC-MS and -FID analysis was carried out with the crude extract, which was dissolved in 1 mL ethyl acetate prior to injection. Purification of (1R,3E,7E,11S,12S)-3,7,18dolabellatriene was carried out by flash chromatography using hexane on silica (Silica gel 40, Sigma-Aldrich) as solid phase.

In vitro Assay
For the in vitro assay pET28a(P450 BM3 _F87A) and, respectively, coli BL21(DE3) by standard transformation methods. From an overnight culture (8 h cultivation at 37 • C, LB media supplemented with 30 μg/mL kanamycin) the cells were grown to an OD 600 0.5 in 100 mL LB media. Subsequently, 1 mM isopropyl IPTG, 1 mM δ-aminolevulinic acid (ALA), 0.5 mM FeSO 4 × 7H 2 O and 5 μg/L riboflavin was added to the cultures. Protein expression was done overnight at 25 • C. Next, the cells were harvested and centrifuged (10.000g for 20 min at 4 • C). The cell pellet was resuspended in 5 mL 50 mM kaliumphosphate buffer (pH 7.5). The cells were lysed by sonification using a Sonoplus HD2070 (Bandelin Electronic) performing five repeats on ice (1 min on and 3 min off at 50% power) and subsequently centrifuged (20.000g for 20 min at 4 • C). For the assay, 100 μL supernatant were added to 400 μL 50 mM kaliumphosphate buffer (pH 7.5) supplemented with 200 μM (1R,3E,7E,11S,12S)-3,7,18-dolabellatriene, which was added in absolute ethanol to a final assay concentration of 0.5% (v/v). The reaction was started by adding 100 μM NADPH/H + at 25 • C. After 3 h the reaction was stopped by addition of 500 μL ethyl acetate. The reaction mixture was vigorously mixed and the organic phase analyzed by GC-MS.

Optimizing Biotechnological Dolabellatriene Production
Recovery of (2) from dried algae biomass is limited in yield and supply. Ioannou et al. (2011) were able to isolate 0.15 mg/g from the brown algea D. spiralis. Due to the high diversity of dolabellane in this brown algae the isolation of (2) also involves extensive purification steps. In our previous research we were able to produce small amouts of (2) (1.7 mg/L) by an E. coli strain overexpressing the recombinant diterpene biosynthesis pathway with the DXP pathway enzymes isopentenyl diphosphate isomerase (idi), 1-deoxyxylulose-5-phosphate synthase (dxs), 1-deoxy-D-xylulose-5-phosphate reductoisomerase (dxr), geranylgeranyl diphosphate synthase (crtE) and the CotB2 W288G diterpene synthase mutant (Janke et al., 2014). More recently, we have achieved the microbial biosynthesis of taxadiene and several other diterpenoids with yields up to 80 mg/L using an E. coli strain overexpressing, additional 2-C-methyl-D-erythritol 4-phosphate cytidylyltransferase (ispD) and 2-C-methyl-D-erythritol 2,4-cyclodiphosphate synthase (ispF) (Schrepfer et al., submitted). In this study we have used this improved diterpene production system to improve our previous reported production yields (Janke et al., 2014). By integration of the CotB2 W288G synthase mutant into this enhanced production strain we could increase the yield for dolabetatriene to 8 mg/L in conventional shake flasks cultures. The dolabellatriene was recovered from eight independent 1 l batches of organic extracts. These extracts contained on average 8 mg/L of the target product. The variations between each batch were minor. This scalable biotechnological production route enables independent access of (2), which is a prerequisite for subsequent chemical functionalization studies reported here.
In this study, we devised selective chemical oxidation strategies for (2) to generate new dolabellanes with enhanced or diversified bioactivities. Conventionally, complex biocatalytic functionalization of diterpene olefins requires initial oxidative activation of the olefinic macrocycle, which in vivo is carried out by highly substrate specific P450-type oxidoreductases. Since there are no biocatalysts known in nature that can functionalize the dolabellatriene macrocycle, we evaluated chemical and enzymatic methodologies for the regio-and stereoselective oxidation of (2). The specific challenge in any terpene centered semisynthetic approach is the regio-and stereoselective functionalization of chemically very similar C-H bonds. In nature selectivity is provided by steric demand of amino acid residues in the binding pocket of P450 type enzymes .

Epoxidation of Dolabellatriene by Lipase Mediated in situ Generated Peracetic Acid
In our primary enzymatic based functionalization approach we focused on production of epoxidized variants of (2). The structure of (2) is in possession of three olefinic bonds. Two are located within the 11 membered ring and show an equal substitution pattern, while the third bond is a terminal alkene with a low degree of substitution. The selective epoxidation of the alkene bonds within the bicyclic carbon skeleton provides the opportunity for the synthesis of (5) and (6), respectively (Figure 3; Ioannou et al., 2011). Since epoxidation by chemical methods is strongly in favor for higher substituted and therefore more reactive dienes, the terminal olefinic bond is discriminated. Previously, the semisynthetic transformation of (2) to the natural, bioactive compounds (5), and (6) has been demonstrated using m-chloroperbenzoic acid, a highly reactive peracid (Ioannou et al., 2011). We evaluated the selectivity of the time resolved epoxidation of 15 mg (2) using the less reactive acetic peracid. The reaction was performed in ethyl acetate, containing acetic peracid that was formed in situ by reaction of immobilized lipase B (C. antarctica) with H 2 O 2 and catalytic amounts of acetic acid (Da Silva and Nascimento, 2012). The progress of the reaction was monitored by GC-MS analysis ( Figure 3B). After 4.5 h the highest concentration of a mixture containing equal amounts of (5) and (6) was detected [58% (5.1 mg, 56% (5); 4.1 mg, 44% (6)]. At 23 h the yield of (5) and (6) decreased (53% [4.8 mg, 57% (5); 3.7 mg, 43% (6))], while the new, diexpoxidated diterpene species (7) was formed. The yields of (5) and (6) observed in this study are similar to those reported for the equivalent reaction in the presence of m-chloroperbenzoic acid, which provided 56% [48% (6); 52% (5)] after 30 min (Ioannou et al., 2011).
The conversion of (5) and (6) to (7) was a rather slow process with a 43% (7.2 mg) yield obtained after 50 h, when the reaction was terminated. In this timeframe (5) and (6) decreased to 37% [55%, 3.2 mg (5); 45%, 2.7 mg (6)]. Interestingly, toward the end of the reaction a tri-epoxidated diterpene (8) was generated with a limited yield of 1.8% (0.3 mg). Examination of the stereochemistry indicated that the epoxidation of (2) was very selective. Since the structure of the non-natural dolabellane (7) has not been reported so far, we carried out a comprehensive structural analysis (Figure 3C, Supplementary Material). GC-MS and NMR analysis revealed that about 90% of the di-epoxidated (7) had the same configuration. Key NOE correlations from NOSEY NMR spectroscopy were observed between H-3/H-7, H-3/H 3 -15, H-2β/H-3, H-7/H-9β, H-7/H 3 -15 and H-2α/H-12 ( Figure 3C) indicating that the C15-methyl group is at the same side of the molecule as the hydrogens H-3 and H-7 positioned at the epoxy groups. Consequently, the epoxides in (7) are likely to display a (S, S) configuration (Supplementary Material). Therefore (7) can be described as (1R,3S,4S,7S,8S,11S,12S)-3,4,7,8-diepoxy-dolabella-18-ene. The preference for a certain epoxide conformation can be explained by the higher thermodynamic stability of the (S, S) epoxide compared to the (R, R) counterpart. Nevertheless, during epoxidation the formation of several side products were observed by GC-MS analysis, which may include formation of the (R, R) epoxides. A time resolved monitoring of the lipase mediated dolabellatriene epoxidation allows selective enrichment of either (5), (6), or (7), respectively. Specifically, increased concentrations of (5) and (6) can be obtained by premature termination of the reaction after 4.5-23 h. Separation of the resulting products (5) and (6) can be achieved according to previously reported protocols (Ioannou et al., 2011). In contrast to procedures leading to the predominant formation of either (5) or (6), enrichment of the di-epoxy diterpene (7) requires extended reaction times beyond 30 h. The long reaction times for this side reaction can be reasoned as epoxidation of the terminal olefin group is particularly slow. The resulting products (5), (6), or (7) may be used in a variety of consecutive reactions due to the diversity of epoxide chemistry.

Stereoselective Hydroxylation of Dolabellatriene by P450 BM3
In our next enzymatic based functialization approach we focused on functionalization of the less reactive terminal olefinic bond (C-20 C-18) of dolabellatriene. We aimed to transform the olefinic bond into a hydroxyl group, which potentially leads to the production of the more potent dolabellane (4) (Ioannou et al., 2011) or related structures.
Prior to our chemical synthesis approach we elucidated the application of the well-studied hydroxylase P450 BM3 derived from B. megaterium. The enzyme is known to catalyze the hydroxylation of long chain and saturated fatty acids. P450 BM3 is a fusion protein of a P450 catalytic unit and its corresponding redox system, which displays the highest turnover rate of any currently reported P450 enzyme system (Whitehouse et al., 2012). Consequently, it has been heavily engineered to hydroxylate natural and non-natural substrates, including macrocyclic diterpenoid skeletons (Urlacher and Girhard, 2012). In the case of the sesquiterpenoid macrocycle, amorpha-4-11-diene (9) an engineered P450 BM3 (F87A/A328L) was able to oxidize the terminal olefinic group to artemisinic-11S,12epoxide (9a) (Dietrich et al., 2009). More recently, it has been reported that P450 BM3 F87A is also capable of oxidizing the bihydroxylated diterpenoid β-cembrenediol (Le-Huu et al., 2015). Since epoxides can be used in a variety of consecutive reactions, application of an equivalent approach to generate (10) (Figure 4) would be very useful. As the epoxidation of (2) is more sterically demanding than (9), we primarily evaluated the feasibility of this strategy by in silico docking experiments (Supplementary Material). The docking experiments did provide an indication that (2) may be a substrate for P450 BM3 (Supplementary Figure S12). Consequently, we generated mutants P450 BM3 F87A and/or A328L, which have been reported to achieve epoxidation of amorpha-4-11-diene. The P450 BM3 mutants were recombinantly expressed and tested in an in vitro assay supplemented with (2). At the end of the assay (3 h) the terpenoids were extracted, concentrated and analyzed by GC/MS. However, none of the assays showed the formation of an oxidized diterpene. The conversion of (2) by the P450 BM3 mutants might be very slow, suffer from inhomogeneous distribution of the substrate and/or result in yields below the limit of detection. Subsequently, we introduced the P450 BM3 mutants in our E. coli based dolabellatriene production system to increase the reaction time and therefore the quantity of potential products. The E. coli strain in this in vivo assay was cultivated for 5 days followed by a detailed GC/MS analysis. Again we were not able to detect any hydroxylated products, indicating that (2) was not a substrate for any of the P450 BM3 mutants under our investigation. Based on these experimental results, we conclude that our in silico docking experiments may have underestimated the steric demands of (2) in the enzyme active site. Alternatively it may be argued that an initial introduction of at least one hydroxyl function in the diterpene macrocycle may assist substrate binding and orientation in close proximity to the catalytically active heme iron center. This molecular templating interaction in the active site of P450 BM3 may explain the successful hydroxylation of β-cembrenediol (Le-Huu et al., 2015). Conversely, for the smaller amorpha-4-11-diene no additional functionalization may be required to mediate binding in close proximity to the catalytically active center. More generally, it can be stated that at present in silico calculations do provide any guarantees that particular reactions will take place. However, feedback with experimental data combined with rapid improvements in bioinformatic algorithms may in future facilitate the precise prediction of enzymatic reactions.
Alternative enzyme activities, such as chloroperoxidase, which have been reported to oxidatively activate a variety of terpenes may eventually be capable of utilizing (2) as a substrate (Águila et al., 2008;Piantini et al., 2011). We are currently following up with a different strategy, involving high throughput mutagenesis of the class one hydroxylase CotB3, which is naturally involved in formation of cyclooctatin. The rationale behind this approach is the structural similarity of (2) with the native CotB3 substrate cyclooctat-9-en-7-ol.

Stereoselective Hydroxylation of Dolabellatriene via a Hydroboration Strategy
As current enzymatic strategies for hydroxylation of (2) were not successful, we evaluated alternative chemical routes for regioand stereoselective hydroxylation. In order to regio-selectively access the terminal alkene, we evaluated monohydroboration for target orientated hydroxylation. Bulky dialkylboranes are reported to be exceptionally regioselective for terminal alkenes on various substituted dienes (Graham et al., 2011). The resulting trialkylboranes can be used in numerous subsequent reactions to produce alcohols, amines, alkylation, and C-C coupling. Mechanistically, hydroboration is generally an antimarkovnikov reaction, resulting in formation of a primary alcohol after oxidation of the dialkylboranes. Subsequently, we utilized 5 mg of (2) as a substrate, which was primarily reacted with 9-borabicyclo[3.3.1]nonane (9-BBN-H) followed by an oxidative step with H 2 O 2 /NaOH. The resulting reaction mixture was analyzed by GC-MS/FID, resulted in detection of two hydroxylation products of (2). One of these alcohols was formed in an enantiomeric excess (ee) of 94%. The data indicated that the hydroboration reagent preferentially added to one specific side of the terminal olefinic group. The yield of the prominent alcohol (11) was determined. After purification the yield of the prominent alcohol (11) was 84% (4.5 mg) with respect to the substrate and was subsequently elucidated by 1D and 2D NMR spectroscopy (Supplementary Material). This is the first instance that compound (11) is reported.
Structure analysis confirmed the stereoselective formation of a primary alcohol (11) at the terminal olefinic bond (C-20) (Supplementary Figure S11). The outcome of the hydroboration reaction primarily depends on the position of the terminal alkene group. While in the educt (2), the bond between C-12 and C-18 allows free rotation, NOE correlations between H-20α/H 3 -19 and H-20β/H2-13 in (2) clearly assign the position of the terminal alkene group to point away from the 11-membered ring of (2) ( Figure 5B). Consequently, mechanistic considerations indicate that the terminal alkene group in (2) (see Figure 5A) is only accessible to si face in the syn addition of 9-BBN, while the re side is sterically hindered due to the CH 2 groups at C9/10 and the C15 methyl group, respectively ( Figure 5A). The stereochemistry of the newly formed stereo center at C18 was elucidated by NOESY NMR spectroscopy. Key NOE correlations of (11) were observed between H 2 -20/H-13, H 2 -20/ H 3 -19, H 2 -20/H-18, H 3 -19/H-9β, H 3 -19/H-18 and are illustrated in Figure 5B. In agreement with the observed NOE correlations we suggest the formation of the alcohol (1R,3E,7E,11S,12S,18R)-dolabella-3,7diene-20-ol ( Figure 5C). This is the first report of a stereoselective hydroboration of a triene-diterpenoid macrocycle.

CONCLUSION
We demonstrated that a sustainable, biotechnological production of the bioactive (1R,3E,7E,11S,12S)-3,7,18-dolabellatriene (2) by the mutant diterpene cyclase CotB2 W288G is feasible. This independent route provides access to sufficient material used in subsequent enzymatic and chemical approaches to generate an oxidatively activated macrocycle. Our strategies allowed access of a variety of natural as well as new, nonnatural dolabellanes with high regio-and stereo-selectivity. In particular, we could demonstrate that monohydroboration of the differentially substituted terminal triene 3,7,18-dolabellatriene is highly stereoselective. Subsequent oxidation yields the new, non-natural compound (1R,3E,7E,11S,12S,18R)-dolabella-3,7diene-20-ol (11). Consequently, hydroborations may be used in dolabellanes macrocycles to selectively functionalize the terminal alkene group. In a complementary approach we could demonstrate the efficient epoxidation of (2) by peracetic acid, which was formed in situ using a lipase catalyzed conversion of acetic acid and H 2 O 2 , resulting in products (5), (6), (7), (8). By temporal monitoring of the resulting reaction mixture it is possible to selectively enrich these products. In this study we could demonstrate that combining biotechnological, chemo-enzymatic, and chemical reaction cascades for the efficient and scalable production of oxidatively activated diterpene macrocycles. The resulting reaction products may inherently display improved bioactivity compared to their parent compound. Alternatively, these chemically activated compounds may serve as synthons for more elaborate synthesis toward highly functionalized natural and non-natural dolabellanes.