Cryptomeridiol, a typical eudesmane diol, is the active principle component of the antispasmodic Proximol. Although it has been used for many years, the biosynthesis pathway of cryptomeridiol has remained blur. Among terpenoid natural products, terpenoid cyclases are responsible for cyclization and generation of hydrocarbon backbones. The cyclization is mediated by carbocationic cascades and ultimately terminated via deprotonation or nucleophilic capture. Isoprene precursors are, respectively, converted into hydrocarbons or hydroxylated backbones. A sesquiterpene cyclase in Tripterygium wilfordii (TwCS) was determined to directly catalyze (E,E)-farnesyl pyrophosphate (FPP) to unexpected eudesmane diols, primarily cryptomeridiol. The function of TwCS was characterized by a modular pathway engineering system in Saccharomyces cerevisiae. The major product determined by NMR spectroscopy turned out to be cryptomeridiol. This unprecedented production was further investigated in vitro, which verified that TwCS can directly produce eudesmane diols from FPP. Some key residues for TwCS catalysis were screened depending on the molecular model of TwCS and mutagenesis studies. As cryptomeridiol showed a small amount of volatile and medicinal properties, the biosynthesis of cryptomeridiol was reconstructed in S. cerevisiae. Optimized assays including modular pathway engineering and the CRISPR–cas9 system were successfully used to improve the yield of cryptomeridiol in the S. cerevisiae. The best engineered strain TE9 (BY4741 erg9::Δ-200-176 rox1::mut/pYX212-IDI + TwCS/p424-tHMG1) ultimately produced 19.73 mg/l cryptomeridiol in a shake flask culture.
Cryptomeridiol has been found in Cryptomeria japonica and as a constituent of plant volatile oils with pharmacological properties including antioxidant , antihyperlipidemic and antibacterial  activities. A variety of sesquiterpenoids can be synthesized from cryptomeridiol, since it is a eudesmane-type sesquiterpene with unique stereochemical configurations and activated sites that can be replaced . Until now, the synthetic process of cryptomeridiol has been widely appreciated. For example, it can be chemically synthesized from (−)-elemol via oxymetallation conversion [4,5]. Moreover, the chemical-microbiological synthesis of cryptomeridiol has also been developed, though the process is derived from the semi-synthesis of 11-hydroxyeudesmanolides . A relatively short and efficient hemi-synthesis process for cryptomeridiol was conducted using ilicic acid isolated from Dittrichia viscosa (L.) as a starting material . Nevertheless, the most efficient hemi-synthesis described thus far obtained cryptomeridiol in three steps with overall yields of 52%  (Figure 1). Using internal enzymes for synthetic steps is likely to be an appropriate and economical way to produce cryptomeridiol.
Synthetic routes to produce cryptomeridiol and proposed intramolecular proton transfer mechanism (A–D) represent chemical procedures as follows.
The internal key enzymes for generating cryptomeridiol may include sesquiterpene cyclase and cytochrome P450 depending on the structure. Sesquiterpene cyclase plays a critical role in generating hydrocarbons with complex backbones. The conserved DDxxD domain of this enzyme catalyzes the elimination of the phosphate diester in FPP to initiate the multistep cyclization cascade of isoprene . At the termination of cyclization, a carbocation ultimately eliminates a proton to form a double bond or generate a hydroxyl with the participation of water . For double or multiple hydroxyl groups, the hydroxylated reactions commonly require cytochrome P450 . Recently, Liang et al. reported a novel sesquiterpenoid synthase from maize that can generate eudesmane-2, 11-diol and minor cryptomeridiol. The dihydroxylated products directly synthesized by a single reaction from FPP . In this study, a simpler mechanism than the above hypothesis was found, in which a sesquiterpene cyclase in T. wilfordii was unexpectedly found to directly produce double hydroxyl eudesmane diols, similarly to the ZmEDS from maize. But this enzyme generates predominantly cryptomeridiol. Using this single enzyme for the one-step synthesis of cryptomeridiol will greatly economize and accelerate its production process.
Microbial fermentation via microbial cell factories turned out to be highly competitive in its ability to produce large quantities of industrial compounds. Here, a synthetic of cryptomeridiol was well established in Saccharomyces cerevisiae. Some key genes in synthetic cryptomeridiol pathways, including tHMG1, IDI, and ERG20, were overexpressed to increase the production of cryptomeridiol, while the knockout of rox1 and the promoter of ERG9 led to decreasing squalene content and increasing cryptomeridiol yield in yeast. In a preliminary shake flask culture, the best engineered strain TE9 (BY4741 erg9::Δ-200–176 rox1::mut/pYX212-IDI + TwCS/p424-tHMG1) ultimately produced a yield of 19.73 mg/l cryptomeridiol after fermentation for 3 days.
Cloning of TwCS
Total RNA was extracted from a suspension of Tripterygium wilfordii cells using the CTAB method . Then, ∼ 1 µg of total RNA was utilized to synthesize the first strand of cDNA using a PrimeScript 1st Strand cDNA Synthesis Kit (Takara Bio Group, Japan). According to the original mRNA sequence in the transcriptome data, specific 5′ or 3′ race primers (Supplementary Table S1) were designed to amplify the lacking 3′ and 5′ ends of TwCS using a SMART™ RACE cDNA Amplification Kit (Clontech, CA, U.S.A.) according to the manufacturer's instructions. The full-length cDNA was then cloned using Phusion High-Fidelity DNA Polymerase (New England Biolabs, Ipswich, MA, U.S.A.). All of the PCR products were purified and cloned into the pEASY-T3 Cloning Vector (TransGen Biotech, Beijing, China), which was transformed into Escherichia coli Trans5α competent cells (TransGen Biotech, Beijing, China). The final full-length cDNA of TwCS was confirmed by sequencing.
Bioinformatics and transcription level analysis
The TwCS sequence was blasted against the NCBI database (https://blast.ncbi.nlm.nih.gov/Blast.cgi). Bioinformatics analyses, such as the isoelectric point (pI) and molecular mass (MW), of the deduced amino acid sequences were analyzed online (http://www.expasy.org/). A multiple alignment analysis of TwCS and other corresponding sequences downloaded from GenBank was performed using DNAMAN8.0 software (Lynnon Biosoft, Quebec, QC, Canada), and a phylogenetic tree was constructed in MEGA5.1 software (Arizona State University, Tempe, AZ, U.S.A.) using the Neighbor-Joining method with 1000 bootstrap repetitions .
The transcription level of TwCS was analyzed using quantitative real-time PCR (qRT-PCR). Seven organs from T. wilfordii including root bark, root phloem, root xylem, stem bark, stem without periderm, leave and flower were performed as described previously .
Plasmid and strain construction
All of the primers used for constructing plasmids are summarized in Supplementary Table S1. The plasmid and yeast strains used in this study are shown in Supplementary Table S2. To identify the function of TwCS and to improve the production of sesquiterpenes, the pYX212 and pET24a vectors were used as expression vectors and p424GPD was used for overexpressing tHMG1 . The TwCS was amplified by PCR with primers flanking NdeI and BamHI sites. The PCR products were purified and ligated into pET24a vector using the T4 DNA ligase (New England Biolabs, Ipswich, MA, U.S.A.).
A gene expression module containing a promoter, a functional gene and a terminator was used for homologous recombination (Supplementary Figure S1). Phusion High-Fidelity DNA Polymerase and the Gene JET Gel Extraction Kit (Thermo Fisher Scientific, Shanghai, China) were used to amplify and purify all of the fragments, respectively. The TPIp promoter and the pYX212t terminator were derived from the pYX212 vector, while the TEF1p promoter, the FBA1t terminator and the functional genes (ERG20 and IDI) were PCR-amplified from S. cerevisiae BY4741 genomic DNA. The fusion enzymes of ERG20 and IDI relied on a widely used GGGS linker encoding sequence ‘GGT GGT GGT TCT’. All of the modules were assembled with purified fragments of a promoter, a gene, a terminator and a promoter for the adjacent module, as described in a previous study . Subsequently, purified modules, including TPIp-IDI-FBA1t-TEF1p, TPIp-ERG20-FBA1t-TEF1p, TPIp-IDI/ERG20-FBA1t-TEF1p, TPIp-ERG20/IDI-FBA1t-TEF1p, TEF1p-TwCS-pYX212t and TPIp-TwCS-pYX212t, were inserted into the pEASY-T3 cloning vector. After verification by sequencing, the PCR-amplified modules were mixed with linearized pYX212 vector which was digested with BamHI (New England Biolabs, Ipswich, MA, U.S.A.) and transformed into yeast strain BY4741 by electroporation at 2.5 kV, 25 µF, and 200 Ω in a 4-mm gap electroporation cuvette using a Bio-Rad Gene Pulsers. Transformants were selected on synthetic drop-in medium (added 20 g/l glucose) without uracil and/or histidine where appropriate. The selected colonies were cultured at 30°C for 2–3 days, and plasmids were extracted using an E.Z.N.A. Yeast Plasmid Mini Kit (Omega Bio-tek, Inc., Doraville, GA, U.S.A.). Then, positive strains were verified by PCR using the plasmids as templates.
To further develop a highly efficient metabolically engineered strain, the CRISPR/Cas9 system was used to construct the strain BY4741. Specific gRNAs targeting erg9 and rox1 were designed as described in a previous study . A single gRNA expression vector for the erg9 promoter was constructed by amplifying linear p426S-NR52p-gRNA (Addgene reference number: 43803) and inserting a 20-bp target gRNA in the single gRNA expression cassette. Then, multiple gRNA expressing M plasmids (erg9 and rox1) were inserted into the p426-SNR52p-gRNA vector. A detailed description of the construction of double gRNA expression cassettes was described previously .
The primers used for site-directed mutagenesis were listed in Supplementary Table S1. The R268A, R270A, W277H, W277F, D529A, and P524A mutagenesis were generated by PCR using an overlap extension strategy with pET24a-TwCS as a template. DpnI (New England Biolabs, Ipswich, MA, U.S.A.) was used for digesting the template in the PCR product. The digested product was transformed into E. coli Trans5α competent cells. The mutagenesis was verified by sequencing.
Protein purification and enzymatic reaction
The recombinant vector pET24a-TwCS and the six mutants were transformed into E. coli BL21 (DE3). The culture was induced with 0.2 mM isopropyl-thiogalactopyranoside at 16°C for 18–24 h. Cells were harvested by centrifugation at 4000 g for 15 min at 4°C, and then subsequently, the pellet was resuspended in buffer containing 20 mM Tris (pH 8.0), 250 mM NaCl, and 20 mM imidazole. His-tagged TwCS protein and mutant enzymes were eluted using an imidazole gradient (20–300 mM) and concentrated using an Amicon Ultra-15 concentrator (Millipore).
The enzymatic reaction consisted of 50 mM HEPES (pH 7.5), 10 mM MgCl2, 5 mM dithiothreitol, and 5% (v/v) glycerol, purified protein (100 µg), and 10 µl FPP/GPP/GGPP (Sigma), which were overlaid with hexane and incubated overnight at 30°C.
To determine Michaelis–Menten of TwCS, 1 µg of protein was used in a 50 µl reaction mixture containing 40 mM Tris–HCl (pH 8.0), 100 mM NaCl, 10 mM MgCl2, 5 mM dithiothreitol, 5% (v/v) glycerol, and 10–400 µM FPP. The mixture was incubated at 30°C for 10 min, after which the mixture was added with 10 µl of stop solution (0.2 M EDTA and 0.4 M NaOH, pH 8.0) . The reactions were then extracted with 150 µl of n-heptane for three times, concentrated using nitrogen gas, and analyzed by GC–MS.
To build the three-dimensional structure of TwCS, its sequence was searched in PDB (https://www.rcsb.org/). The TwCS was similar to the 5-epi-aristolochene synthase and the identity between these two enzymes was 46%. A molecular model of TwCS was constructed using the SWISS Model (https://www.swissmodel.expasy.org/). The crystal structure of 5-epi-aristolochene synthase with 2 units of Mg2+ and 1 molecular trifluorofurnesyl diphosphate (PDBID: 5EAU) was selected as a suitable template in view of these two enzyme's high value of GMQE and identity. The substrate FPP was refined using elBow program in PHENIX suite. Then, the model of TwCS and FPP was performed molecular docking as recently described . The results of this docking run were shown in PyMOL.
Extraction, purification, and structure elucidation
The fermentation broth of the engineered S. cerevisiae strains was extracted with n-hexane four times. The hexane phase was collected, concentrated using a rotary evaporator and dried with nitrogen. Then, the essential oil extracts were separated and purified by mass-mediated preparative HPLC with fraction collection using a Waters 2767 sample manager. A Waters XSelect CSH Prep C18 OBD (5 µm; 19 mm × 150 mm) preparative column was employed for the isolation of impurities. Mobile phase A consisted of 0.1% (v/v) formic acid in water and mobile phase B consisted of acetonitrile. The flow rate was kept at 20 ml/min. The structures were elucidated using NMR spectroscopy. All of the NMR data were acquired with a BRUKER ACANCE III 600 MHz spectrometer using TMS as an internal standard.
Cultivation in S. cerevisiae and quantification of sesquiterpenes
Sesquiterpene-producing strains were evaluated through bioreactor cultivation. Synthetic dropout medium (FunGenome Company, Beijing, China) with 20 g/l glucose as the carbon source was used to pre-culture the corresponding auxotrophic strains. The medium for a 3 l stirred-tank bioreactor was composed of 8 g/l synthetic dropout medium without uracil and histidine, 10 g/l (NH4)2SO4, 10 g/l KH2PO4, and 1.0 g/l MgSO4·7H2O. Additionally, the culture medium was supplemented with 2 ml/l trace metals (EDTA 15 g/l, ZnSO4·7H2O 0.45 g/l, MnCl2 1 g/l, CoCl2·6H2O 0.3 g/l, CuSO4·5H2O 0.3 g/l, Na2MoO4·2H2O 0.4 g/l, CaCl2·2H2O 0.45 g/l, FeSO4·7H2O 0.3 g/l, H3PO3 0.1 g/l, KI 0.1 g/l) and 1 ml/l vitamin solution. Fifty percent NH3·H2O was used to buffer the pH of the medium. Strains were pre-cultured in shake flasks at 30°C for 48 h with shaking at 230 rpm and collected by centrifugation. Then, in the 3 l stirred-tank bioreactor (Eppendorf BioFlo/CelliGen 115), 1 l of fermentation medium was inoculated with the pre-culture cells. A glucose solution (500 g/l) was fed periodically to maintain growth of the strains. A concentrated medium with 40 g/l synthetic dropout medium lacking uracil and histidine and 100 g/l (NH4)2SO4 was fed for fermentation.
To test the sesquiterpene production of every strain, 50 ml cultures were started after inoculating 500 µl pre-cultures. Synthetic dropout medium (FunGenome Company, Beijing, China) with 20 g/l glucose as the carbon source was used to pre-culture the corresponding auxotrophic strains. The strains were grown at 30°C with shaking at 230 rpm in defined minimal medium with 20 g/l glucose. After 72 h, the OD600 of all the strains were measured. Each culture broth was added to an equal volume of n-hexane and was kept shaking at 200 rpm for 2 h after which it was ultrasonically extracted twice. The organic layer was pooled and evaporated. The samples were condensed to a final volume of 1.0 ml and were analyzed by GC–MS. The isoprenoids were quantified using a Thermo TRACE 1310/TSQ 8000 gas chromatograph (splitless; injector temperature 250°C) with a TG-5 MS (30 m × 0.25 mm × 0.25 µm) capillary column. The GC conditions were as follows. The oven temperature was first kept constant at 50°C for 2 min, it was then increased to 280°C at a speed of 8°C/min and held for 10 min at this final temperature. The injector temperature and the ion trap heating temperature were both set at 250°C. The analog β-eudesmol was applied in the preparation of a standard curve to relatively quantify the isoprenoid content of each strain.
Results and discussion
Cloning, sequence analysis, and gene expression analysis of TwCS
T. wilfordii is an important medicinal plant that produces diverse terpenoid natural products . Apart from some reported investigation of its terpene synthase, many of its natural products remain uncharacterized. A novel terpene synthase from a suspension of T. wilfordii cells was cloned; the responsible DNA fragment contained a 1662-bp open read frame that was designated TwCS. The predicted molecular mass and isoelectric point of the deduced protein were 63.67 kDa and 5.06, respectively. A multiple sequence alignment revealed that TwCS contains an arginine–arginine (proline)/tryptophan conserved (RR(P)x8W) domain at amino acids 12–22 in its N-terminus and a DDxxD active site located at amino acids 305–309 (Supplementary Figure S2). These characteristic motifs are found in sesquiterpene synthases from other plants, such as Xanthium strumarium (AMP42988) , Artemisia annua (AAF61439) , and Santalum spicatum (AEF32536) , which strongly suggested that TwCS could be a sesquiterpene synthase . In addition, over the course of verifying the uncharacterized TwCS products by screening the evolutionary relationships between TwCS and other terpene synthases (TPSs) from diverse subfamilies (including TPS-a, TPS-b, TPS-c, TPS-d, TPS-e/f, TPS-f, and TPS-g), a phylogenetic tree was constructed using a Clustal W alignment of the amino acid sequences of various TPSs. The proteins clustered into seven groups in the tree, and the TwCS deduced protein sequence was grouped into a cluster with the TPS-a eudicot subfamilies, while the monocot sesquiterpene synthases formed a separate clade (Supplementary Figure S3). TwCS was nearly identical with germacrene A synthase from A. annua . The gene expression analysis of seven different organs showed that the highest TwCS expression was observed in the root phloem. Transcript levels of TwCS were somewhat low in stem bark and flower and could not be detected in all of the tested organs (Supplementary Figure S4).
Functional characterization of TwCS in vitro and in vivo
To further explore the function of TwCS in vitro, the TwCS was ligated into a pET24a vector and expressed in E. coli BL21 (DE3). Purified protein (Supplementary Figure S5) was incubated with FPP and analysis of the reaction mixtures by GC–MS showed a new peak at 21.46 min that had specific mass fragments at m/z 189, 164, 149, 109, and 59 and was predicted to be β-eudesmol (Figure 2) . However, the GC–MS analysis of the β-eudesmol (CAS: 473-15-4) standards revealed a single peak at 21.16 min (Supplementary Figure S6), which was definitely different from the enzymative product. The retention time and mass fragments of epi-eudesmol (CAS: 15051-81-7) did not correspond to those of this product either.
GC–MS analysis of products catalyzed by purified TwCS proteins.
The activities of putative terpene synthases were examined relative to a negative control, an extract prepared from E. coli harboring the same expression vector without an inserted TPS gene. As shown in Figure 3, the activities were negligible when GGPP was used as a substrate by extracts prepared from E. coli expressing the TwCS, little monoterpene synthase activity was recorded with GPP; however, when incubated with FPP, TwCS exhibited very substantial sesquiterpene synthase activities (Figure 2). TwCS kinetic parameters were determined to have a Km of 95.27 µM for FPP, a Vmax of 2.766 pmol/µg/min and a kcat of 2.98 × 10−3 s−1 (Supplementary Figure S7).
Substrate specificity of TwCS using GPP, FPP, and GGPP as substrates.
The functional characterization was performed in vivo through S. cerevisiae BY4741. A recombinant plasmid for the overexpression of farnesyl pyrophosphate synthase (ERG20) was constructed, and the TwCS gene was then inserted. This plasmid was then transformed into BY4741, producing strain TE1. GC–MS analysis of fermentation extracted from TE1 revealed the presence of several eudesmane-type sesquiterpene product peaks at 18.94 min (compound 1), 19.27 min (compound 2), and 21.46 min (compound 3) on the mass chromatogram (Figure 4). Compounds 1 and 3 were the main components of the sesquiterpene products, while compound 2 was a minor product. The compound 1 peak at 18.94 min had dominant mass fragments at m/z 204, 189, 161, and 59, which matches the data reported for γ-eudesmol . The compound 2 peak at 19.27 min that gave mass fragments at m/z 204, 189, 161, 149, and 59 was predicted to be α-eudesmol through comparison to the literature MS data and recent study  (Supplementary Figure S8). Compound 3, which was the most abundant of the three compounds, was the same compound as the enzymatic product.
GC–MS analysis of products from cultures of yeast strain BY4741 expressing TwCS.
TwCS is a cryptomeridiol synthase
To more fully confirm the product of this novel TPS, a large-scale fermentation was performed to produce sufficient quantities for structural analysis by NMR. After performing extraction and separation procedures, ‘purified compound 3’ was obtained and analyzed further. The structure of compound 3 was elucidated through detailed analysis, including MS, 1H NMR, 13C NMR, and DEPT spectral data (Supplementary Table S3 and Figures S9–S11), and the relative configuration of compound 3 was confirmed through 1H-1H COSY, HSQC, HMBC, and NOE (Supplementary Figure S12–S15), all of which indicated that the compound was cryptomeridiol [26,27]. Cryptomeridiol is the active principle of Proximol®, an antispasmodic drug. Cryptomeridiol is composed of two hexatomic rings, which each has a hydroxyl group. Other cyclic products of sesquiterpene synthase, including germacrene A , santalene , and τ-cadinol , have no or just one hydroxyl group. Terminal isopentenyl of santalene can generate a hydroxyl via P450 enzymes . The TwCS enzyme shows biofunction in generating cryptomeridiol. A crystal study of diterpene synthase determined that its multistep cyclization cascade is ultimately terminated by deprotonation or nucleophilic attack [31,32]. The specific double hydroxyl group of cryptomeridiol suggested that the structure of TwCS may contain two active sites to capture water molecules to carbocation. TPSs have similar αβγ or αβ, α-domain architectures that create the structure of sesquiterpene and perform the cyclization reactions . The α domains host the class I active sites, which have conserved DDxxD and (N,D)xx(S,T)xxxE domains that bind metal to initiate the cyclization of FPP, while the class II active sites are presented in the βγ interface and contain a conserved DxDD motif that is also responsible for the protonation-dependent mechanism.
Key residues for TwCS catalysis
TwCS showed above 40% similarities to other sesquiterpene synthases, but they catalyze the formation of different terpenes. To screen the relationship between the function and the structure, providing insights into the catalytic mechanism of TwCS, the optimized natural substrate FPP was docked into the active site of TwCS (Figure 5). Overall, the modeling shows that FPP can be accommodated in the TwCS active site, consistent with previous studies [11,33] (Supplementary Figure S16).
Docking of FPP in the active site of TwCS (A) and relative activity of site-directed mutagenesis (B)
Two active site residues, R268 and R270, are in close proximity to the diphosphate group of FPP and could potentially stabilize the substrate oxygen ion through H-bonding or salt bridge. To explore these possibilities, R268 and R270 were mutated to alanine. The R270A mutant showed no sesquiterpene products produced, while the R268A showed significant effect, resulting in very low enzyme activity of 2.8% relative to wild-type TwCS, suggesting that R268 and R270 play a key role in catalysis, probably combining diphosphate group and involved in FPP binding (Figure 5). Sequence alignments with related sesquiterpene enzymes indicate that the W277, D529, Y534 residues are highly conserved. Furthermore, mutation of W277 to histidine, W277 to phenylalanine, and D529 to alanine also result in large decreases in the activity of TwCS, suggesting that the aromatic residues may provide potential hydrophobic interaction and π-stacking interactions. However, the P534A mutation did not have a significant impact, with the mutant retaining ∼73% activity relative to the wild type. Interestingly, this D529 was located in a disordered loop and far away from the ligand in the docking result (Figure 5). Starks et al. speculated the proposed catalytic mechanisms of the cyclization of eudesmane and suggested that D525 was a key residue in the process of carbocation rearrangement . This disordered loop (523V–537L) could be wobbly and D529 played a vital role in catalysis. The W277H mutant exhibited a more significant drop in activity than W277F, indicating that the hydrophobic residue may interact with FPP, not the imidazole ring.
Increase in cryptomeridiol content by overexpressing of tHMG1, IDI, and ERG20
Metabolic pathway engineering and fermentation production are important for commercial production of natural products. A full artemisinic acid biosynthetic pathway was developed for the production of the antimalarial drug artemisinin, yielding fermentation titers of 25 g/l of artemisinic acid . Additionally, microbial cell factories provide new synthetic routes for the production of terpenoid flavors and fragrances . Bisabolene, another sesquiterpene that serves as a biofuel precursor, can also be produced using synthetic biology tools . Thus, S. cerevisiae is a key cell factory platform for the production of pharmaceuticals, industrial chemicals, and biofuels .
Cyptomeridiol biosynthesis can also be well established in S. cerevisiae. On the basis of strain TE1, two approaches to optimize productivity were applied, including the overexpression of genes in the mevalonate pathway and the construction of protein fusions. Since condensation reactions of isopentenyl diphosphate and dimethylallyl diphosphate form the basic carbon chain of terpenes, isopentenyl diphosphate isomerase (IDI), which catalyzes the transition between isopentenyl pyrophosphate (IPP) and dimethylallyl pyrophosphate (DMAPP), play a vital role in the biosynthesis of sesquiterpene. In yeast, ERG20 gene encodes FPP synthase which catalyzes the condensation of IPP and DMAPP initially elongating to the result 15-carbon compound FPP . Therefore, improving ERG20 activity will likely alter the flux of FPP intermediates and thus play a key role in the up-regulation of final terpene products in yeast.
For the overexpression strategy, the IDI and ERG20 modules were individually cloned into the pYX212 vector to enhance the supply of precursors. The resulting strain that expressed the ERG20 module, TE4, produced 0.90 ± 0.053 mg/l compound 1 and 4.24 ± 0.005 mg/l compound 3 under shake flask culture conditions. The other strain, TE5, with up-regulated by IDI yielded 1.09 ± 0.008 mg/l compound 1 and 4.57 ± 0.132 mg/l compound 3 (Figure 6 and Supplementary Figure S17). The overexpression of IDI provided more IPP and DMAPP for the biosynthesis of sesquiterpene and appeared to be more important for increasing their production than ERG20. Because IPP is also used for the synthesis of isoprene and ergosterol, a fusion protein of IDI and ERG20 might enhance the efficiency of IPI conversion to FPP . Thus, the IDI–ERG20 and ERG20–IDI fusion proteins were constructed. The addition of the IDI–ERG20 module (strain TE2) decreased the production of compounds 1 and 3 compared with that of TE4 and TE5. The ERG20–IDI module led to improving production compared with strain TE3, but its production was still lower than that of strains TE4 and TE5. Weise et al.  measured the ratio of DMAPP/IPP and its influence on isoprene emission, and the ERG20 and IDI fusion proteins may influence this proportional balance.
Production of compounds 1 and 3 by recombinant yeasts.
tHMG1, a truncated hydroxy-3-methylglutaryl coenzyme A (HMG-CoA) reductase gene, shows higher catalytic activity and has been widely applied in the optimization of the mevalonate pathway to supply the metabolic flux for terpene biosynthesis . An early study showed that the overexpression of tHMG1 resulted in an 88% increase in the miltiradiene yield compared with that of HMG1. Thus, we directly transformed p424-tHMG1 into strains TE3 and TE5, and the additional expression of HMG1 in the TE3 background (TE7) led to a 1.7-fold increase in cryptomeridiol, resulting in 7.75 ± 0.122 mg/l, as well as a 64% increase to 5.68 ± 0.174 mg/l in strain TE6 (Figure 6).
Cryptomeridiol synthases engineering using CRISPR/Cas9
In addition to plasmid gene alteration, genome engineering using the CRISPR/Cas9 system was successfully applied to improve the yield of sesquiterpene. The influential steps in the pathway that were manipulated included squalene synthase, encoded by the ERG9 gene, and a DNA-binding protein, encoded by the rox1 gene. ERG9 catalyzes the reductive dimerization of FPP to generate squalene . Metabolic engineering projects aiming to promote the production of plant sesquiterpenes and other terpenoids in S. cerevisiae have reduced the squalene pathway by down-regulating ERG9 in particular . The deletion of the upstream activating sequence (UAS) in the strong native promoter of ERG9 combined with the overexpression of multiple other genes resulted in significantly increased cryptomeridiol productivity (strain TE8, Figure 7). Strain TE8 produced 13.47 ± 0.880 mg/l cryptomeridiol in a shake flask culture. Rox1 is an important regulator of anaerobic genes in S. cerevisiae. Anaerobically induced regulation is involved in the mevalonate pathway as well as the biosynthesis of sterols and unsaturated fatty acids . Rox1 can down-regulate the expression of specific ERG genes for environmental adaptation . Strain TE9 in which rox1 was deleted and ERG9 was down-regulated in the background of strain TE3 demonstrated 2.5-fold higher production of cryptomeridiol than strain TE3, reaching a final yield of 19.73 mg/l cryptomeridiol and 1.86 mg/l y-eudesmol (compound 1) under shake flask culture conditions (Supplementary Figure S18).
CRISPR/Cas9 editing of the erg9 promoter and rox1 improved sesquiterpene production.
Sesquiterpenoids, an enormous class of natural compounds, have diverse biological and ecological functions, such as resisting or attracting microbes, insects and herbivores . Numerous sesquiterpenoids are found in both higher and lower plants, and some have practical applications in agriculture, industry and medicine. Henquet et al.  found a drimenol synthase and drimenol oxidase involved in the biosynthesis of insect deterrent drimanes. Jin et al.  identified several sesquiterpene synthases which provide a tool for biosynthesizing volatile organic compounds in Cananga odorata. The biosynthesis of sesquiterpene derives only from simple C5 isoprene building blocks, IPP, and its isomer DMAPP. The specific matching principle and prenyltransferases participate in the cascade of C5-units, and consequently, FPP (15 carbon) is generated and converted into a cyclic sesquiterpene . The TwCS protein is a functional sesquiterpene synthase that generates three sesquiterpenes, mainly cryptomeridiol. The cryptomeridiol structure is similar to the common core structure of sesquiterpene alkaloids, and a potential cyclization reaction was presumed to exist between cryptomeridiol and the dihydro-agarofuran sesquiterpene (Supplementary Figure S19). Since the wild and commercial cultivated T. wilfordii are limited scale and have an almost trace ability to synthesize bioactive secondary metabolites, the profitable and sustainable bioengineering solution has become a potential and additional filling for the T. wilfordii ’s wide spread practical applications .
In a conclusion, we cloned and identified a novel sesquiterpene cyclase that mainly produces cryptomeridiol. The cyclized sesquiterpenes produced by this enzyme were confirmed to have the characteristic (CH3)2COH group. Cryptomeridiol is an important component of volatile oils that has anti-inflammatory, antioxidant and antimicrobial activities [51–53], and its structure shows double hydroxyl sesquiterpene alcohols, which is unlike other sesquiterpene products catalyzed by known sesquiterpene cyclases from dicotyledon. The terpenoid biosynthetic pathway is an essential foundation for metabolic engineering of cryptomeridiol. We successfully engineered a yeast strain and intentionally improved the production of cryptomeridiol by genetically manipulating the mevalonate pathway and squalene pathways as well as using a transcription regulator. Consequently, we obtained the final yeast strain TE9, which yielded 19.73 mg/l cryptomeridiol in a shake flask culture, an amount that is more than 38-fold higher than the original TE1 strain. Our findings describe the first known biosynthesis of double hydroxy sesquiterpene compounds in dicotyledon. Moreover, the engineered yeast strain provides insights into the role of metabolic engineering in supplementing high value compounds and it is a good reference for other novel sesquiterpene synthases.
W.G. and L.H designed the experiments. Y.T. designed the research, performed most of the experiments and drafted the manuscript. P.S. and X.Z. planned some preliminary experiment. P.S., H.G., T.H. and J.C. performed some experiments. Yi.Z., Yu.Z. and L.G. analyzed data. All authors approved the final manuscript.
The authors thank Dr Yan-long Yang for helpful analysis on sesquiterpene's conformation. The authors also thank Professor Reuben J. Peters for helpful introductions while preparing the manuscript. The authors acknowledge Yuhong Xiang in Capital Normal University for supporting SYBYL-based computing services.
This work was financially supported by the National Natural Science Foundation of China , National High Technology Research and Development Program of China [863 Program: 2015AA0200908], and the Support Project of High-level Teachers in Beijing Municipal Universities in the Period of 13th Five-year Plan [CIT&TCD20170324] to W.G., Beijing Municipal Education Commission [KM201710025012] to X.Z., and Key project at central government level: the ability establishment of sustainable use for valuable Chinese medicine resources  to L.H.
The Authors declare that there are no competing interests associated with the manuscript.