Asymmetric Cation‐Olefin Monocyclization by Engineered Squalene–Hopene Cyclases

Abstract Squalene–hopene cyclases (SHCs) have great potential for the industrial synthesis of enantiopure cyclic terpenoids. A limitation of SHC catalysis has been the enzymes’ strict (S)‐enantioselectivity at the stereocenter formed after the first cyclization step. To gain enantio‐complementary access to valuable monocyclic terpenoids, an SHC‐wild‐type library including 18 novel homologs was set up. A previously not described SHC (AciSHC) was found to synthesize small amounts of monocyclic (R)‐γ‐dihydroionone from (E/Z)‐geranylacetone. Using enzyme and process optimization, the conversion to the desired product was increased to 79 %. Notably, analyzed AciSHC variants could finely differentiate between the geometric geranylacetone isomers: While the (Z)‐isomer yielded the desired monocyclic (R)‐γ‐dihydroionone (>99 % ee), the (E)‐isomer was converted to the (S,S)‐bicyclic ether (>95 % ee). Applying the knowledge gained from the observed stereodivergent and enantioselective transformations to an additional SHC‐substrate pair, access to the complementary (S)‐γ‐dihydroionone (>99.9 % ee) could be obtained.


Introduction
Ionones are significant contributors to the appealing scents of many flowers and fruits,i ncluding violets,r oses,o r raspberries. [1] They belong to af amily of natural products known as apocarotenoids,which are derived from carotenoids by oxidative cleavage catalyzed by carotenoid oxygenases. [2] An efficient synthetic access to racemic ionones by cationolefin cyclization of pseudoionone (1)was discovered already in the late 19 th century by Tiemann and Krüger (Scheme 1). [3] Accordingly,i onones were among the first commercially utilized synthetic fragrance ingredients,featured for example in the iconic fragrance Vera Violetta (Roger &G allet, 1893).
However,t he most obvious synthesis route toward these compounds is currently missing, namely the asymmetric cation-olefin cyclization of pseudoionone (1)o rasuitable derivative thereof.I ta ppears that carbocation formation at the unpolar isoprene end of the linear chain in combination with enantiospecific folding of the linear C 13 precursor to form am onocycle is difficult to achieve with classical asymmetric catalysis.
In contrast, squalene-hopene cyclases( SHCs), which belong to the class II terpene cyclases,are capable of locking linear terpenoid substrates in defined chiral conformations, which allows to achieve polyene cyclizations with perfect stereocontrol. Consequently,S HCs have great potential as industrial biocatalysts for the production of enantiopure cyclic terpenoids.Awidely spread model reaction is the cyclization of the linear C 30 triterpene squalene (7)i nto the pentacyclic products hopene (8)and hopanol (9), through the generation of five new CÀCbonds and nine new stereocenters (Scheme 3). [6] Ther eaction is initiated by the protonation of the unactivated terminal isoprene unit with the unusually acidic middle aspartate of the DXDD active site motif.T he excellent chemo-, regio-, and stereocontrol over the polycyclization cascade is achieved through pre-folding of the substrate in ap roduct-like conformation, stabilization and shielding of the highly reactive carbocation intermediates from side reactions,and aselective termination through base assisted proton elimination or addition of water. [7,8] Te rpene cyclases from the SHC family are promiscuous enzymes and accept molecules ranging from C 10 monoterpenoids [9] to C 35 squalene analogues, [10] and the cyclization reaction can be initiated through protonation of unactivated olefins,c arbonyls,and epoxides. [11] This is in contrast to other main families of class II terpene cyclases:o xidosqualene cyclases are limited to substrates containing an epoxide functional group for initial protonation, [12] while class II diterpene cyclases such as ent-copalyl diphosphate synthases are generally only active towards the diphosphate containing substrate geranylgeranylpyrophosphate. [13] Importantly,S HCs have proven to be highly evolvable: Engineered SHC variants with not more than three mutations enabled av iable industrial-scale process to obtain Ambrofix TM , [14,15] as well as dramatically increased activity and altered chemo-and stereoselectivity of cyclization reactions with mono-and sesquiterpenoids,s uch as geraniol, [11] farnesol, [16] or citronellal. [17] Alimitation of SHCs,however,istheir strict (S)-enantioselectivity at the stereocenter formed after the first cyclization of all polyisoprenoids tested so far (an overview of products is given in reviews [18,19] ).
Here,w er eport our efforts to gain enantio-complementary access to valuable monocyclicterpenoids such as (R)-and (S)-g-dihydroionone (5)v ia SHC catalysis.E ven though the natural diversity of SHC sequences is vast, [20,21] most of the work on non-native substrates has thus far focused on two enzyme variants from Alicyclobacillus acidocaldarius (AacSHC) and Zymomonas mobilis (ZmoSHC1) [19] and only one study reported as creening panel consisting of 12 wildtype enzymes. [22] Thus,t oi dentify enzymes capable of synthesizing (R)-and (S)-g-dihydroionone (5), we opted for as creening approach based on an SHC wild-type library, which included 18 novel SHC homologs.B uilding on the ability of an ewly identified SHC from Acidothermus cellulolyticus to generate the monocyclic (R)-g-dihydroionone ((R)-5), we optimized the enzyme by directed evolution and could improve the conversion of nerylacetone ((Z)-2)to(R)g-dihydroionone ((R)-5)bytwo orders of magnitude to 79 % in 48 h. It should be noted, that during the preparation of this manuscript, astudy by the Hauer group was published, which similarly reports the biocatalytic production of (R)-g-dihydroionone ((R)-5)b ya ne ngineered SHC from Alicyclobacillus acidocaldarius. After five rounds of directed evolution, the authors identified an AacSHC variant with four mutations,w hich exhibited excellent selectivity (99.5 % ee)a nd conversion (89 %) in seven days. [23] In our report, we thus confirm the exciting observation that it is possible to obtain (R)-selective monocyclizations via SHC biocatalysis (> 99 % ee)y et using the distinct AciSHC Scheme 2. GC odour thresholds (GC-OTH) of g-dihydroionone enantiomers (5)a nd conversion of (S)-5 to (À)-a-ambrinol (6). enzyme (51.6 %s equence identity to AacSHC). In addition, we observed that all of our AciSHC variants exhibited exquisite selectivity in the transformation of the geometric geranylacetone (2)isomers:While the (Z)-isomer yielded the desired monocyclic (R)-5 product, the (E)-isomer led to the formation of the bicyclice nolether (S,S)-4.B iochemical and docking studies helped us to understand the mechanistic basis of the observed sterodivergent and enantioselective cyclization reactions.H arnessing this knowledge,w eu ltimately succeeded to additionally obtain the enantio-complementary (S)-5 (> 99.9 % ee)t hrough the application of an appropriately chosen SHC-substrate pair.

Results and Discussion
In our quest to create an efficient biocatalyst for the enantioselective production of (dihydro-)ionones,w ea imed to identify an SHC enzyme with the capability to generate monocyclic products from either (E/Z)-geranylacetone (2)or (E/Z)-pseudoionone (1). AacSHC, [24] ZmoSHC1, [24] and engineered variants of these enzymes [25] were previously reported to be inactive towards 1 and were found to convert 2 exclusively into the bicyclic product 4. Consequently,w e chose to explore the SHC diversity beyond these heavily studied variants by setting up ac omprehensive screening panel of 31 wild-type enzymes,s elected to span all major clades of the phylogenetic tree ( Figure S1). Thes creening library consisted of 13 previously characterized class II terpene cyclases from the SHC-family and 18 novel SHC homologs,which were identified through the presence of two defining PFAM domains for type II triterpene cyclases (PF13249, PF13243) and the SHC-family specific DXDD active site motif (Table S2). As thermostable enzyme scaffolds can be superior starting points for protein engineering and directed evolution approaches, [26] ten of the novel sequences were explicitly chosen to originate from thermophilic bacteria.
To characterize our SHC library and evaluate the biocatalysts potential for (dihydro)ionone production, we overexpressed the enzymes in E. coli BL21(DE3) and carried out whole-cell biotransformations with 10 mM squalene (7), 10 mM (E/Z)-geranylacetone (2), and 10 mM (E/Z)-pseudoionone (1). Product formation was analyzed using gas chromatography coupled to mass spectrometry equipped with af lame ionization detector (GC-MS-FID) (Figure 1). Nineteen of the investigated SHCs showed activity towards at least one substrate.Notably,ten of the active enzymes correspond to novel SHC homologs,w ith sequence identities to experimentally characterized variants between 52.6 %and 82.9 %. These results validate our bioinformatic search strategy,a nd the new enzymes further expand the toolbox of available SHCs for biocatalysis.
While (E/Z)-geranylacetone (2)w as converted by 15 members of our SHC panel (Figure 1), our screen did not identify any SHC homologs with activity towards pseudoionone (1), possibly due to steric and/or electronic effects of the conjugated g,d-double bond of 1,w hich is the distinguishing feature from 2 ( Figure S2). Analyzing the (E/Z)-geranylace-tone (2)c onversion data in more detail, we identified AciSHC,anovel SHC homolog from the thermophilic bacterium Acidothermus cellulolyticus,a sap ossible candidate for further development. While conversion of the C 13 substrate 2 into the bicyclic enol ether (4)w as widespread among the SHC panel, AciSHC was the only enzyme included in the panel that generated two additional minor products Figure 1. Characterizationo fthe wild-type SHC library with respect to the enzymes' activity towards squalene (7)and (E/Z)-geranylacetone (2). Whole-cell biotransformations were carried out by supplementing cell lysate with 10 mM substrate in 50 mM citrate buffer at pH 6 containing 0.8 %(7)or0.2 %(2)ofT riton-X-100. The SHCs are ordered based on phylogenetic relationship. Highlighted in yellow is AciSHC, the only wild-type SHC converting 2 into monocyclic products 5 and 10.P roducts 11 and 12 could not be structurally assigned.
Intrigued by these results,w ec reated as equence alignment of the active pocket [20] of the 14 SHCs,w hich mainly convert (E/Z)-geranylacetone (2)i nto the bicyclic product 4 and, in three cases,the structurally unassigned product 12 and compared it to the amino acid distribution of AciSHC ( Figure S3). Surprisingly,t he sequence alignment revealed that the active site of AciSHC appears to be similar in construction as those of the remaining enzyme panel:Ofthe 36 residues lining the substrate-binding pocket, only I41, located more than 18 away from the catalytic acid D380, was found to be unique in AciSHC ( Figure S3). Thus,w e proceeded to investigate the unusual product selectivity of AciSHC by constructing its homology model based on the crystal structure of AacSHC (PDB ID:1 SQC;i dentity: 51.62 %; similarity:0 .45) using SWISS-MODEL [27] followed by docking studies of (E)-2 and (Z)-2 using the software tool AutoDock Vina. [28] Both substrate stereoisomers afforded adocking state with aproductive pre-chair conformation for monocyclization, however, no "all" pre-chair state as required for the formation of the bicyclic product was found (Figure S4).
Thus,e ven though the identified substrate poses did not fully explain the experimentally observed product distribution, our docking results led us to speculate that already slight changes in the active pocket geometry might result in alternative pre-folding states of 2.I nt his way,t he enzyme could channel the substrate either into ac ationic cascade necessary for the formation of the bicyclic enol ether (4)o r allow termination of the reaction after as ingle ring-forming event to yield 5.I nt he latter case,d eprotonation of the exocyclicm ethylene group could occur through D378, which in our model of AciSHC is situated at adistance of 2.6 from the hydrogen of the relevant carbon C-11. Thep resence of D378, acting as acatalytic base,could explain the unexpected selectivity for the formation of the energetically unfavorable exocyclicd eprotonation product 5 over 10 ( Figure S5).
As g-dihydroionone (5)i sacompound of particular interest for the flavour and fragrance industry,w ea imed to improve the activity and selectivity of the AciSHC catalyzed conversion of 2 into 5 using structure-guided directed evolution. Based on the above-mentioned docking studies of (E)-2 and (Z)-2 into ah omology model of AciSHC,w e chose 14 sites for NNK single-site saturation libraries.I nt he first evolution round, we focussed on residues around 2 with the aim to improve pre-folding. In addition, we targeted the large unoccupied space in the active pocket to limit potentially unproductive binding modes known to occur for small substrates in other SHCs (Figure 2a). [11] Overall, we screened 90 clones for each of the fourteen libraries in deep-well plates amounting to the analysis of > 1200 enzyme variants.T he screening revealed variants with 2.9 to 5.4-fold increased conversion of 2 into 5 in the libraries A169X, P263X, A310X, G606X, and I613X (Figure 2b). With the exception of variant G606T,h ydrophobic residues were favoured substitutions, and while increased bulk seemed beneficial at sites A169, P263 and A310, smaller amino acids were preferred at I613.
Because all beneficial sites were located in the same area of the active pocket of AciSHC,i ts eemed plausible that epistatic interactions between the amino acid residues might occur. Going forward, we therefore opted to combine all beneficial mutations and the respective wild-type amino acid in afive-site combinatorial library,resulting in alibrary size of 288 variants.F ollowing library construction by overlap extension PCR, we screened 720 clones for an estimated coverage of 92 % [29] (Figure S6). Theb est variant for the conversion of (E/Z)-2 identified in the second evolution round was dubbed AciSHC_R2.1 (A169P,A 310M, G606C, I613V) and achieved aconversion of 2 into 5 of 21.4 %, a30fold increase over the wild-type enzyme.I no ur quest to understand the basis of the increased activity in the engineered AciSHC variants,wesequenced the top ten variants of the second evolution round, all exhibiting aconversion of 2 to 5 of more than 14.4 %. In this analysis,wefound nine unique protein sequences with an average of 3.8 mutations.A stonishingly,n os ingle mutation was present in all variants,w ith the best single site variant, A310F,only occurring in one of the optimized enzymes (Table S3). These findings could indicate that AciSHC can harbor multiple active site geometries, which can induce ap roductive pre-folding of 2 for efficient cyclization into 5.
Both products were obtained in excellent optical purity: Using variant AciSHC_R2.3, g-dihydroionone (5)(> 99 % ee) was formed in the non-natural laevorotatory form, which could be assigned to the absolute (R)-configuration based on the work of Brenna et al., [5] whereas the laevorotatory bicyclice nol ether (4)( > 95 % ee)c orresponded to the (S,S)-configuration as evidenced by comparison to Serra et al. [30] Thus,the SHC enzymes produced the two products 4 and 5 in opposite enantiomeric forms,aprocess which can be described as as tereodivergent and enantioselective conversion of the (E)-and (Z)-isomers of 2.Even when amixture of (E/Z)-2 was used as substrate, 5 was produced as the (R)enantiomer and 4 as the (S,S)-enantiomer with near to perfect enantioselectivity with all tested variants (Scheme 4, Table S4).
Going forward, we targeted to evaluate the broader synthetic implications of the observed stereodivergent transformation of geometric isomers by the AciSHC variants.Thus, we set out to transfer our insights to an additional enzyme with the goal to synthesize the (S)-enantiomer of g-dihydroionone ((S)-5), akey intermediate in the synthesis of (À)a-ambrinol (6). [31] Building on our previous results,w e hypothesized that for the synthesis of the natural (S)enantiomer of g-dihydroionone ((S)-5), we would require as uitable geranylacetone ((E)-2)s ubstrate with am asked carbonyl group to prevent the formation of the bicyclic enolether 4.T ot hat end, the industrially-proven AacSHC variant AacSHC_215G2 was employed for substrate screening in whole-cell biotransformations.Whereas no conversion was observed with dioxolane (E)-13,w ed etected the formation of am onocyclicp roduct with intact acetate group from (E)-14 (Scheme 5). To our surprise,the product was not the expected exo-methylene derivative 16,b ut its hydrated derivative 15,f ormed with perfect enantio-and diastereocontrol. Intrigued by this observation, we repeated the biotransformation with AacSHC_215G2 and (Z)-14,yielding, as expected, the g-dihydroionone derivative 16,a gain with opposite absolute configuration compared to 15.T hese observations prove that the sense of asymmetric induction is   determined solely by the geometry of the double bond in the substrate and is not influenced by the presence of the racemic acetate-bearing chiral center. It is also worth mentioning that no deacetylated product was observed despite the use of aw hole cell biocatalyst, where hydrolase-mediated ester hydrolysis could have been expected.
To further explore the scope of different SHC/substrate combinations,weperformed whole-cell biotransformations of pure (E)-and (Z)-2 with AacSHC_215G2. To our surprise and complementary to the earlier described SHC variants, AacSHC_215G2 converted (Z)-2 to (R,S)-4 with perfect enantioselectivity,d emonstrating that SHCs can fold a( Z)substrate in such am anner as to form a cis-fused bicycle,i n line with the Stork-Eschenmoser hypothesis. [32] Finally,( E)-2 was converted to (S,S)-4 with perfect enantioselectivity and high yield on gram scale by AacSHC_215G2. Thec hemical transformation of 15 to (S)-5 (Scheme 6) proved the absolute configuration of 15 and provided the first access to the natural (+ +)-enantiomer of g-dihydroionone (S)-5 via asymmetric cation-olefin cyclization. Thesame optical purity of (S)-5 was obtained when tangerinol (14; E/Z 3:2) of commercial quality was used. Similarly, 16 was transformed in two steps to (R)-5. [33] To better understand the mechanistic basis of these stereodivergent and enantioselective reactions,wegenerated homology models of the engineered AciSHC_R2.3 and AacSHC_215G2 variants using SWISS-MODEL [27] followed by molecular docking of (Z)-2 and (E)-2 as well as (Z)-14 and (E)-14,respectively,using Autodock Vina. [28] In the homology model of AciSHC_R2.3, (E)-2 showed areactive all pre-chair conformer for generation of abicyclic product, while for (Z)-2 the second chair was unfolded and the carbonyl-group too distant for an intramolecular nucleophilic attack (Figure 4a). Accordingly,t he polycyclization cascade is expected to be interrupted by deprotonation, leading to monocyclic products 5 or 10 (Figure 4b). Thepre-folding of the initial chair for (Z)-2 and (E)-2 was nearly identical. Accordingly,t he absolute configuration of the newly generated stereocenter resulting from the first cyclization is expected to be defined by the configuration of the double bond ( Figure 4). Reflecting our findings for AciSHC,t he docking study on AacSHC_215G2 revealed anearly identical prefolding of the initial pre-chairs for (Z)-14 and (E)-14,suggesting that the enantioselectivity of the cyclization reaction is again determined by the configuration of the double bond of the substrate ( Figure S8).

Conclusion
By screening acomprehensive SHC enzyme library,which expands the current SHC toolbox by ten active enzymes,w e identified the novel AciSHC capable to cyclize nerylacetone ((Z)-2)i nto the monocyclic( R)-g-dihydroionone ((R)-5). To the best of our knowledge,t he recent study by the Hauer group [23] and this work are the first examples of SHCs accepting oxygenated isoprenoids with a( Z)-configurated internal double bond, as well as affording a(4aR)-stereocenter after the first cyclization.I nterestingly,b oth studies identified similar hotspots in the enzyme active site influencing the monocyclization reaction, albeit in two different enzyme scaffolds with only 51.6 %i dentity ( Figure S10 and S11). Notably,through our combinatorial enzyme engineering approach, we identified several highly active AciSHC variants comprising divergent active site geometries,which can afford the necessary pre-folding of 2 to obtain monocyclization products.O ur findings therefore indicate that, depending on the enzyme starting scaffold, cyclization cascades cannot only be controlled through the introduction of anchoring hydrogen bonds as shown by Hauer et al. [23] but also through the appropriate choice of the geometric substrate isomer.T ransferring this knowledge to the industrially applied AacSHC_215G2 variant, we could highlight that stereodiver-  gent and enantioselective transformations of geometric isomers could indeed prove to be ag eneral principle in SHC catalysis.T hrough appropriate substrate engineering and downstream processing,w ec an obtain access to both enantiomers of atarget product via SHC biocatalysis,including the industrially highly relevant chiral building block (S)-gdihydroionone ((S)-5). Overall, this work provides an exciting opportunity of tuning the absolute configuration of the cyclized products using substrates with defined double bond stereochemistry and highlights the possibility to control the polycyclization cascade through substrate engineering.