Chemoenzymatic hybrid synthesis is an aspiring paradigm, which synergistically merges the advantages of biocatalysis and state-of-the-art synthesis to streamline synthesis of complex (natural) molecules.1,2 In their endeavor to generate meroterpenes, the Renata group demonstrated the power of chemoenzymatic synthesis by successfully implementing the regio- and stereoselective remote oxyfunctionalization ability of oxygenases into their synthesis routes.3 This development opened up entirely new retrosynthetic considerations in terpene synthesis by expanding the available scaffolds from the chiral pool.4 However, this ‘chiral pool’5 or ‘scaffold remodelling’ (SR)6 strategy is still limited to few precursors, which renders access to alternative cyclic terpene scaffolds complicated. It often entails low step- and atom-economy caused by strategic functional group manipulation and sequential chemistry (Fig. 1a). It is therefore not surprising, that despite their huge potential as bioactive compounds, terpenes are still highly underrepresented e.g., in Pharma (1.3% of APIs)7, as they are mainly produced via metabolic engineering of organisms or are extracted from overexploited plants with low biomass efficiency (usually 1-5% of plant material).8,9
To overcome this dichotomy, a de novo synthesis including a target-oriented and stereocontrolled cationic cyclization of the natural achiral linear precursor would dramatically shorten synthetic routes to cyclic terpenes (Fig. 1b, cf. Fig. S1). Nature uses this modular ‘divergent synthesis’ strategy for eons10 and despite the tremendous progress in the field of biomimetic stereocontrolled head-to-tail cyclization employing diverse catalysts e.g., Brønsted-acids11, transition metals12 or supramolecular cages13, their application in terpene synthesis is still highly underrepresented (<1%, Table S1). This is due to frequently required cryogenic conditions, alternative initiation motifs, limited stereocontrol, and inevitable strong nucleophiles as terminating groups of the cyclization, which collectively also excludes the use of broadly abundant natural linear precursors (Fig. 1b).14
Herein, we report on embracing the exquisite biocatalysis of the squalene-hopene cyclase (SHC) for streamlined and target-oriented hybrid synthesis of terpenes (Fig. 1c).
At the outset of this work, the promiscuity and application strategies of SHCs for specific targets have been demonstrated, which usually required tedious isolation protocols or huge host cell amounts (~400 g/L).15,16 It should be noted that during the preparation of this manuscript a study by Xiao et al. appeared that also applies the herein presented strategy for terpene synthesis.17 However, a comprehensive study on showcasing the stereoselectivity and enabling the synthetic potential of biocatalytic head-to-tail cyclizations in chemoenzymatic terpene synthesis on a broad substrate scope has yet to be addressed.
Generation of cyclic terpene precursors by engineered SHCs
To begin, we selected terpene targets from diverse areas of application and identified their inherent cyclic moiety (Fig. 2a, blue structures). Actinidiolide derivatives 1 are prominent bioflavors and insect pheromones, and are built from a C1 elongated monocyclic cyclogeraniol skeleton.18 Biopolymers based on 1,3‑dioxanone 2 include the drimane diol skeleton and are used e.g., in tissue engineering.19 Meroterpenes as metachromins 3 are applied in e.g., cancer treatment20 studies and incorporate the C13 megastigmane skeleton. Another representative of this compound family is the sponge-derived meroterpene 4 that bears potential in inflammation disease treatment21 and is made of the tricyclic ent‑isocopolane skeleton. The last and most complex targets, the α‑pyrone meroterpenes 5 offer a promising broad range of biological activities22 and can be derived from a sclareoloxide-like β-keto ester structure. We are aware that the targets presented herein lack functionalizations, such as the OH-group at position C-3 of 5, which emphasizes the important contributions of the Baran23 and the Renata group in the field of terpene oxidations.3 However, the focus of this work was to showcase the synthetic potential of biocatalytic cyclizations en route to terpenes, which is why prior, or late-stage oxidations were not considered.
With these retrosynthetic analyses in mind, we commenced screening our in-house SHC library with the appropriate linear terpenes starting with the smallest one E-geranyl acetate 6 (Fig. 2b). Biotransformations were carried out in vivo as this setup proved its efficiency for the membrane-bound enzyme.26 Fig. 2c shows the AacSHC wildtype compared with the two most interesting hits for each substrate. Moreover, to learn from the substrate-mutant relation, all substrates were evaluated in silico by docking them into the confined active site of the AacSHC wildtype (Fig. 2d and Fig. S2). The stereoselective cyclohydration of 6 would directly provide the precursor cyclogeranyl acetate hydrate 7 for the synthesis of 1 and thus overcome a three-step SR.18 To our delight variant G600R generated cyclohydration product with 95% selectivity and boosted the WT activity 7-fold.
Cyclohydration of E,E-farnesol 8 would result in the desired drimane skeleton for dioxanone 3 formation and shorten concurrent SR routes by two steps.27 While this substrate has already been the target of AacSHC-mediated cyclization by Hoshino and Hauer28,29, those studies employed purified enzyme and identified variants lacked full selectivity and conversion of 8, respectively. Among multiple hit positions (cf. Fig. S4), variant A306W achieved high conversion (89%) and excellent drimane diol 9 selectivitiy (94%). Notably, we found variant L36V/Y420F/G600L that enabled the stereocontrol of the final hydration and provided the opposite diastereoisomer S1-9 with a diastereoselectivity of 88% (Fig. S4 and NMR data).
In view of the sponge-derived meroterpene 4, we next focused on the cyclization of linear diterpene E,E,E-geranyl geraniol 10. Hoshino and co-workers reported the promiscuous tricyclization of 10 to 11 using purified AacSHC WT, albeit with a low yield of 12%.29 We identified multiple variants with increased acitvity, (Fig. S5), however, variant F605W produced the α-regioisomer ent-isocopolol 11 with 5‑fold improved conversion and ~90% selectivity. Additionally, mutations around the third transient carbocation (Fig. 2d, red circle) were identified that can direct a selective bicyclization of 10. Congruent to the tricyclization, most variants yielded product mixtures which emphasizes the challenge of selective cyclizations even with an enzyme (Fig. S5). Merely, the asparagine at position 600 resulted in 90% of bicyclic labdanol 12 with a conversion comparable to the WT. Interestingly, the in silico docked 12 in modeled G600N gives rise to a dual function of the asparagine that is anchoring the functional group30 and acting as a Brønsted-base (Fig. S6).31 Labrotary efforts towards 11 and 12 encompass five and eight steps, respectively.32,33
Finally, for α-pyrone meroterpene 5, we first evaluated the biocatalytic cyclization of E,E‑farnesyl acetone 13 to sclareoloxide 14. We chose this strategy due to the almost identically pre-folded natural 13 and non-natural terpene 15 in the SHC’s active site (cf. Fig. S3e and f). Promiscuous cyclization of 13 towards sclareoloxide 14 with purified AacSHC has been reported, albeit with very low conversions <1%.34 Our survey yielded variant F601D/F605L, among multiple other hits (Fig. S7), which surpassed the WT 10-fold, while ensuring selectivity of 99%. The introduced aspartate may anchor the keto-group and therefore lock the substrate in the right pre-folding faster or act as Brønsted-base to activate the final nucleophile. Subsequently, we used non-natural terpene 15 with the distinguished SHC variants. As 15 is prone to decarboxylation to 13 (Fig. S8), the Brønsted-acid catalyst has to overcome this side reaction and select one out of three potential intramolecular final nucleophiles. Intriguingly, variant F601D/F605L exhibited excellent conversion (99%, 5-fold higher compared to WT) and selectivity (98%) towards the desired cyclic product 16, which emphasizes the precise conformational substrate control of the SHC. Cyclic 14 can be prepared by SR in 3 steps35 whereas 16 has only been racemically cyclized yet.36
In summary, we were able to produce all desired (and more in the SI) terpene scaffolds with high conversions and excellent stereoselectivities, empowered by the tunable shape-complementary pre-folding of substrates in the confined active site of the biocatalyst. Our data demonstrate that mechanism-guided enzyme engineering in the sphere around the desired transient carbocation, as showcased for substrate 10 in Fig. 2d, enables the direction of cationic head-to-tail cyclizations in terms of regioselective α-/g-deprotonation, stereoselective hydration, and cascade progress. Leveraging this knowledge, which builds on our prior findings,30,37,38 paves the way to the cyclization of dozens of carbon skeletons with divergence potential to tens of thousands of natural products.39 To the best of our knowledge there is no chemical catalyst that is able to control a stereoselective one-step cyclization of these unbiased terpenes to such an extent (cf. Table S2 and supporting chromatograms), especially with an alkene as the terminal nucleophile (substrates 6, 8, 10). A limitation of the presented strategy is that, surprisingly, no single β‑deprotonation product could be detected.
Technical and mechanistic investigations on the biocatalyst preparation (727)
Having ascertained the stereocontrolled cyclization, we intended to provide a concise setup to translate SHC biocatalysis from lab to liter scale. As a model reaction for initial investigations, we chose the broadly studied promiscuous cyclization of E-geranyl acetone 17 using AacSHC variant G600M30 (Fig. 3a). First, we proved the ability to use lyophilized E. coli whole-cells, which would drastically simplify application of SHCs as a storable powder. To our delight, biotransformations showed no difference in conversion, and long-term stability (Fig. 3b). To improve substrate availability, four cyclodextrins were tested, which disclosed 2‑hydroxypropyl-β-cyclodextrin (2HPβCD) as the best candidate in terms of substrate conversion (Fig. S9). Next, we proved scalability with a previously30 elaborated protocol using 5 g/L (25 mM) substrate, 10 gCDW/L cells, 14 g/L (10 mM) 2HPβCD and buffer in a 5 L reactor stirring with 100 rpm at 30 °C what demonstrated a slow but operationally stable catalyst for 84 days and yielded 22.4 g (90%) of cyclic 18 (Fig. 3c). We ascribed the operational stability to the fact that SHC does not require a viable cell, but parts of the cell membrane, where the enzyme is embedded (see Thermolysis protocol in the supporting information), are enough to drive the biocatalysis. This hypothesis was strengthened by comparing the growth of freshly expressed with the lyophilized cells on an Agar-plate, which disclosed that few cells survive the lyophilization process. Further evidence was provided by fluorescence microscopy which showed that mainly cell debris remains (cf. Fig 3d and e, Fig. S10).
Intrigued by this data, we next examined parameters that may influence the space-time-yield of the biocatalysis. Therefore, buffer, cyclodextrin and the combination thereof were evaluated as additives. Moreover, terpene type (as defined by their partition coefficient P = [octanol/water], Fig. 3f, blue numbers), terpene concentration, cell concentration, and temperature were evaluated to get a coherent picture (Fig. 3f). To keep catalyst loading low we used 10 gCDW/L cells in all setups.
While the more hydrophilic terpenes (+)‑β‑pinene 19 and E-geranyl acetone 17 were better converted to (+)‑a‑pinene 20 and trans-hexahydrochromene 18 in the absence of 2HPβCD, the more hydrophobic substrates 13 and 10 were generally better cyclized in its presence. Vice versa the chelating citrate buffer improved the cyclization of the more hydrophilic substrates 19 and 17 (cf. Fig. S11 for non-chelating buffer), however, impeded the cyclization of more hydrophobic ones. In the worst-case using buffer completely disrupted the reaction (see 10 at 50 °C, 50 mM). The combination of both or increasing the cell density did not show clear trends in either direction (Fig. S12). Intuitively, the beneficial effect of 2HPβCD rises with increasing P of the substrates (Fig. S13) and, finally, it is generally advisable to employ a thermostable enzyme variant at higher substrate concentrations, which is known correlation40 (Fig. S14).
Tempted by the substrate-dependent effects of 2HPβCD in our setup, we calculated the transfer free energies of the substrates geranyl acetate 6, geranyl acetone 17, homofarnesol 21, geranyl geraniol 10 from water into the core of 2HPβCD (ΔGH2Oà2HPβCD) versus the membrane (ΔGH2OàCH), where the enzyme naturally sources its substrates, using the double decoupling method (for more information and selected calculated products, such as ambroxide 22 see Table S8). This data disclosed that (a) membrane partition, and encapsulation of terpenes by 2HPβCD from the water phase is generally beneficial for the system (cf. ΔGsolvH2O and transfer free energies), (b) both processes are competing, and (c) 2HPβCD encapsulation of more hydrophilic substrates 6, 17, 21 is stronger than their membrane diffusion. It should be noted that both transfers are reversible processes, as otherwise biocatalysis would be inhibited.
To sum up, our data set disclosed that the SHC is independent of a viable cell, which ensures long-term storage as well as operational stability. However, the setup is largely dependent on the sum of abiotic stressors, such as temperature, buffer, and terpenes, acting on the biological system of membrane and membrane-bound enzyme (Fig. S15). The pivot herein is to recognize the substrate’s level of hydrophobicity and based on that augment biocatalysis by setting the other parameters with special focus on temperature and internal organic reservoirs such as cyclodextrin. Presumably, the encapsulating host not only improves the solubility of the more hydrophobic molecules but also weakens their hydrophobic effect on the biological system by reversible encapsulation. These findings emphasize the practicability and flexibility of the biocatalyst setup for synthetic purposes.
Implementing stereocontrolled head-to-tail cyclizations into hybrid synthetic routes
The elaborate substrate-focused protocol thus paved the way for the hybrid syntheses of (mero-)terpenes devised in Fig. 2a. Tetrahydroactinidiolides, such as 1, are usually prepared via C-1 homologation of the appropriate terpene such as 7.18 Conversely, we envisioned a catalytic strategy that readily de-homologizes cyclic enolethers, such as 18, (Fig. 4a + b) inspired by the work of Moulines et al.35 First, we examined this transformation in two partial reactions: an iodine-mediated and a peroxidative rearrangement using enolether 18 (Fig. S16). Evaluation of the iodine reaction conditions disclosed that catalytic (5 mol-%) amounts of sodium iodide in the presence of excess H2O2 (successively added) and acidic buffer are enough to yield semi-pinacol rearrangement product P-1 (60:40 dr) (Fig. 4c) almost quantitatively (97%). For the peroxidative rearrangement, we employed the candida antarctica lipase (CALB) that can generate peracetic acid from H2O2 and ethyl acetate (EtOAc),41 which served as the solvent simultaneously. As stated by the authors of ref. 35, the reaction had to be carried out at 50 °C and yielded lactone 1 with 70% in the best scenario (Fig. S16A). Conveniently, the oxidative rearrangement consumes both semi-pinacol diastereoisomers (Fig. S17). Finally, we combined both reactions in one pot using sclareoloxide 14 as the substrate EtOAc as the solvent and substituting sodium iodide with lugol’s iodine (KI/I2) to overcome solubility issues of the iodide catalyst. Careful evaluation of the reaction conditions led to the optimal setup consisting of 50 mM substrate shaken in EtOAc, 1 mol-% Lugol’s iodine, 20 wt-% CALB, and 4 v/v-% of 30% H2O2 (10 eq., successively added) at 50 °C for 5 h. Our proposed mechanism is depicted in Fig. 4c: Initially (i), CALB generates acetic and peracetic acid (AcOH and AcOOH) from water and EtOAc, which in combination with H2O2 is used for the acidic-oxidative generation of “I+” (ii) (not further specified as iodine species are in a complex equilibrium).42 The subsequently formed iodonium intermediate is then nucleophilically attacked by water to form a semiacetal, which immediately generates ketone P-1 after deprotonation and semi-pinacol rearrangement (iii). P-1 then forms a Bayer-Villiger product in the presence of AcOOH, generated by CALB, EtOAc, and H2O2. After protonation, rearrangement, and cleavage of AcOH, the transient oxonium-ion reacts with AcOO-, which finally rearranges to form lactone 1 (iv).
Employing SHC biocatalysis, we generated 17, Z-17 and 13 to chiral enolethers 18, 23, 14 with high yields (82-92%) and subsequently edited their skeletal constitution using the iodine/lipase protocol to form trans-tetrahydroactinidiolide 1, cis-tetrahydroactinidiolide 25 and (+)-sclareolide 26 with moderate yields of 33-64% in one pot. Next, (−)-g-dihydroionone 24, which was prepared via directed cationic cyclization of Z-17 with 89% yield, was coupled with 2-acetyl resorcinol in a pyrrolidine-catalyzed tandem aldol/ intramolecular Michael addition, i.e. the Kabbe reaction,43 to generate chromanone 3 with 90% yield and 50:50 dr. Notably, L‑proline and 2-methoxy-pyrrolidine were tested as alternative catalysts with potential stereocontrol over the Michael addition which, however, did not result in any conversion (data not shown). Solid drimen diol 9 was cyclized with a yield of 88% and transformed into carbonate 2 via potassium carbonate (K2CO3) mediated carbonylation with a moderate yield of 33%. Labdane 12 and ent-isocopolane 11 were also generated with high yields of 78% and 90%, respectively, from linear 10 and could potentially be transformed within three steps to sponge-derived meroterpenes 4 and 27 e.g., by transition metal-catalyzed cross-electrophile coupling of iodinated 12 and 11 with iodinated methoxy-benzoic esters as demonstrated in ref. 3. Finally, α-pyrone meroterpenes 28 and 29 that constitute the carbon skeleton of pyripyropenes and phenylpyropenes, were generated from chiral 16 and nicotinoic as well as benzoic ester in a base-mediated tandem g‑acylation/ intramolecular annulation reaction with yields of 64 and 28%, respectively (for a summary see Table S3).
Conclusively, we could prove that the exquisite catalysis of cyclases can be harnessed for target-oriented synthesis of terpenes. Scaffold remodeling approaches to cyclic terpenes, while being testimonials of chemical creativity, can be shortened by up to 90% to essentially one diastereo- and enantiopure cyclization (cf. Fig. 2b), which is easily applicable as well as scalable, and provides yields up 92% at the same time. Combined with strategic interdisciplinary synthesis and catalysis we thus provided access to high molecular complexity in only two or three steps. Thus, the stage is set for novel or drastically shortened retrosynthetic logic in de novo terpene synthesis.