Seeds of S. miltiorrhiza were collected from Anhui Province and the plants were grown in nursery of Shanghai Chenshan Botanical Garden. For experimental analysis, the plants were grown in greenhouse at 25°C under a light intensity of 150 μmol of photons⋅m^-2⋅s^-1 with 14-h-light/10-h-dark cycle. Chemicals were purchased from Sigma-Aldrich, TCI and Fluka.

For hormone treatments, plants of 3-month-old, when biosynthesis of secondary metabolites was active and the transcription levels of the corresponding enzymes were high, were used. The plants were sprayed with the following solutions: 5 mM salicylic acid (SA), 100 μM ethylene (ET), 100 μM abscisic acid (ABA), 50 μM methyl jasmonate (MeJA), and 50 μM gibberellin A₃ (GA₃), in addition of a dimethyl sulfoxide (DMSO) solution (5%) as control. Leaf were then collected for isolation of total RNAs 2 h post-spray. Each experiment was performed with at least three biological replicates.

The following materials from the 6-month-old plants in bloom were harvested for extraction of chemicals: primary root (6 cm length from the apical); stem (the second to the fourth internodes from the top); the third pair of true leaves; and the inflorescence. The 6-month-old plants contained a high level of terpenoids and suitable for extraction.

Fresh materials (0.5 g) were ground in liquid nitrogen, followed by adding 2.5 mL hexane to extract the terpenoids in a shaker at 30°C for 1 h. A short column of aluminum oxide overlaid with anhydrous MgSO₄ was used to purify the extracts, and the obtained hexane phase was analyzed by Gas chromatography–mass spectrometry (GC-MS) by the total ion chromatogram mode.

Total RNAs from the plant tissues were extracted using CTAB solution (2% CTAB, 2% PVP, 100 mM Tris-HCl, 25 mM EDTA, 2M NaCl, 2% β-mercaptoethanol) and treated with DNase I (Takara) according to the manufacturer’s protocol. Total RNA of 1 μg was reverse-transcribed using the ReverTra Ace qPCR RT Kit (TOYOBO, FSQ-101). Rapid amplification of cDNA end (RACE) was performed by the 5′- and the 3′-Full RACE Core Set Ver.2.0 (Takara) according to manufacturer’s instructions. PCR products were cloned into the pMD18-T Vector (Takara) for sequencing. Full-length cDNAs were obtained by assembling fragments obtained by RACE in combination with annotated unigene sequences.

For gene expression analysis, real-time PCR was performed with gene-specific primer pairs on the reverse transcripts derived from total RNAs, using the SYBR^® Premix Ex Taq^TM II (Perfect Real Time) kit (Takara) on a Mastercycler system (Eppendorf, Germany). ACTIN (HM231319) was used as internal reference (Yang et al., 2010). The relative expression value was calculated via the 2^-ΔΔCt method. All PCR primers used in this investigation are listed in Supplementary Table S1.

Terpene synthase sequences from other plant species were obtained through searching the GenBank database at NCBI. Alignment of protein sequences was performed with CLUSTAL W program of MEGA 6 software with default parameters. Phylogenetic tree was built with the neighbor-joining method with MEGA 6 program. Chloroplast signal peptide prediction was performed by SignalP¹ software.

The open reading frames of cDNAs were amplified with fast-Pfu DNA polymerase, the PCR products were then digested by EcoRI and HindIII, followed by ligation into pET-32a expression vector. After confirmation by DNA sequencing, the plasmids were transferred into Escherichia coli Rosetta (DE3) for protein production. Recombinant proteins were purified with Ni-NTA resin (Thermo) according to manufacturer’s manual. Protein concentrations were determined with the Bradford method using BSA as standard.

The TPS activity was assayed as described (Ruan et al., 2016). The reaction was conducted in a volume of 500 μL buffer containing 20 μg purified recombinant protein, 25 mM hydroxyethyl piperazineethanesulfonic acid (HEPES, pH 7.0), 5 mM MgCl₂, 5 mM dithiothreitol (DTT) and 40 μM FPP, and incubated at 30°C for 1 h unless otherwise indicated. The reaction mixture was extracted with 800 μL hexane and subjected to analysis by GC-MS. To determine the quantities of the products, nonyl acetate (10 ng/μL) was added to the hexane as internal standard.

Kinetic parameters of the enzymes were determined as described (Ruan et al., 2016). Briefly, the purified recombinant enzyme (5 μg) was added to each assay mixture containing FPP range from 2.3 to 184.6 μM. The reaction was allowed to proceed for 5 min at 30°C, stopped by the addition of 0.5 M EDTA (pH 8.0) and the reaction products were extracted with 800 μL hexane containing 10 ng/μL nonyl acetate as internal standard, followed by GC-MS analysis. Kinetic parameters were obtained with GraphPad Prism 5 software fitting the Michaelis–Menten curve.

To obtain enough amount of enzymatic product, a 500 mL reaction buffer contained 25 mM Hepes (pH 7.0), 5 mM MgCl₂, 5 mM DTT, 80 μM FPP and 20 mg purified protein was used. The reaction mixture was overlaid with 20 mL HPLC-grade pentane and incubated at 30°C overnight. The mixture was extracted with 3 × 100 mL pentane, the combined extracts was evaporated under reduced pressure to afford 4.5 mg sesquiterpene product.

¹H, ¹³C and 2D nuclear magnetic resonance (NMR) spectra were recorded on a Bruker AVANCEIII^TM 500 spectrometer. Chemical shifts were reported using tetramethylsilane as the internal standard. The optical rotation was determined by using a autopol I spectrometer (rudolph research analytical).

Nicotiana benthamiana plants were grown in greenhouse, and 6-week old plants were used for the transient expression. The full-length coding region of SmSTPS1 was amplified by PCR with the primers listed in Supplementary Table S1, the PCR products were firstly cloned into the entry vector pDonr207 and then into plasmid pEAQ-HT-DEST1 using the Gateway BP and LR Clonase II enzyme mix (Thermo Fisher), respectively. The expression vector pEAQ-SmSTPS1 was introduced into Agrobacterium (strain GV3101). For transient transformation, Agrobacterium cells harboring pEAQ-SmSTPS1 was infiltrated into the abaxial side of healthy leaves of N. benthamiana using a syringe. After culture in the greenhouse for 4 days, the infected areas were extracted and the chemicals were detected by GC-MS.

Gas chromatography–mass spectrometry analysis was carried out on Agilent 6890 Series GC System coupled to an Agilent 5973 Network Mass Selective Detector, with the carrier gas helium at 1 mL/min, splitless injection, a agilent HP-5MS column (5% phenyl methyl siloxane, length 30.0 m, diameter: 250.00 μm, film thickness: 0.25 μm) for non-chiral analysis. For volatile terpenes detection of S. miltiorrhiza, a temperature program from 60°C (5 min hold) at 5°C/min to 260°C was used; For enzymatical assay, the following temperature program was used: initial temperature of 60°C (5 min hold), increase to 250°C by 10°C/min, and ramp to 300°C by 50°C/min (5 min hold). Compounds were identified by comparison with NIST data base (National Institute of Standards and Technology) library and retention indices (Garms et al., 2010).

To analyze volatile terpenes in root, stem, leaf and inflorescence of S. miltiorrhiza, fresh tissues were extracted with hexane and subjected to GC-MS. In total 20 terpenes were detected, including 8 monoterpenes, 11 sesquiterenes, and 1 diterpenes (Figure 1). Of the four organs sampled, leaf contained the most diversified terpenes, with at least 18 compounds (8 monoterpenes, 10 sesquiterpenes) being identified by searching NIST database and retention indices (Garms et al., 2010). By contrast, root had the lowest diversity, as GC-MS identified only 1 diterpene (ferruginol). Notably, ferruginol was not present at a detectable level in the aerial tissues examined.

We searched S. miltiorrhiza transcriptomes (Yang et al., 2013) for TPSs, and amplified cDNAs by 5′- and 3′-RACE. Three cDNAs, namely SmSTPS1, SmSTPS2, and SmSTPS3 (GenBank accession numbers are KY432512, KY432513, and KY432514), were isolated. They all encode proteins of 546 amino acids, which are highly identical to each other with the sequence identity of 97∼99%. Based on S. miltiorrhiza genome sequence database (Xu et al., 2016), SmSTPS1, SmSTPS2, and SmSTPS3 are duplicates arranged as a cluster in the genome. Searching of National Center for Biotechnology Information (NCBI) non-redundant protein database showed that the three proteins share a high (62%) sequence identity to germacrene A synthase PatTpsCF2 (AY508728) from Pogostemon cablin (Deguerry et al., 2006), another Lamiaceae species which is particularly rich in sesquiterpenes (Swamy and Sinniah, 2015). The three proteins of S. miltiorrhiza contain the conserved domain DDxxD involved in the coordination of divalent ions and water molecules, and a modified NSE/DTE domain [ND]D[LIV]x[ST]x₃E at the C-terminal involved in metal cofactor binding (Whittington et al., 2002). In addition, a modified RRx₈W motif exists at the N-terminal (Figure 2), which is related to cyclic products formation (Martin et al., 2010; Liu et al., 2014).

Terpene synthases of plants are divided into eight clades or subfamilies designated TPS-a through TPS-h, with a minimum identity of 40% in each clade, and most angiosperm sesquiterpene synthases are clustered into TPS-a subfamily (Bohlmann et al., 1997; Pazouki and Niinemets, 2016). SmSTPS1, SmSTPS2, and SmSTPS3 appear to be typical sesquiterpene synthases (Figure 3).

The three putative terpene synthases were then expressed in E. coli cells (Supplementary Figure S1). The purified fusion proteins were incubated with farnesyl diphosphate (FPP) as the substrate, and Mg²⁺ as a metal cofactor. SmSTPS1, SmSTPS2 and SmSTPS3 all converted FPP into a single and same 204 Da sesquiterpene product, corresponding to the chemical formula C₁₅H₂₄ (Figure 4). Its MS spectrum is closely resemble those of valencene by searching NIST database, however, the retention time of this enzymatic product in GC-MS analysis is different to that of authentic (+)-valencene (Supplementary Figure S2), suggesting that it may share the same skeleton with valencene.

Kinetic analysis with FPP in the presence of Mg²⁺ showed that the K_m values of SmSTPS1, SmSTPS2 and SmSTPS3 are 25.35, 24.20, and 10.44 μM, the estimated k_cat are 2.09 s^-1, 2.07 s^-1 and 1.53 s^-1, and the k_cat/K_m values are 8.27 × 10⁴ s^-1⋅M^-1, 8.56 × 10⁴ s^-1⋅M^-1, and 1.48 × 10⁵ s^-1⋅M^-1, respectively (Table 1). Thus, these three synthases have similar catalytic activities, though SmSTPS1 and SmSTPS2 are slightly more efficient. These parameters are in the normal range of plant TPSs.

To elucidate the structure of the enzymatic product of three putative terpene synthases, we collected 4.5 mg this compound through large scale incubation experiment. The ¹H and ¹³C NMR spectroscopic data of this sesquiterpene product (Table 2) were analogous to those of (+)-eremophilene (Schifrin et al., 2015), a stereoisomer of (+)-valencene, but their MS data are different; furthermore, in GC-MS separation of (+)-eremophilene and (+)-valencene mix, peak of (+)-valencene eluted before (+)-eremophilene (Schifrin et al., 2015), whilst in our GC-MS analysis of (+)-valencene and this esequiterpene product mix, (+)-valencene eluted after this compound (Supplementary Figure S2). All of these suggest that these compounds are structurally similar. The detailed 2D NMR analysis allowed the construction of the structure of the sesquiterpene (Supplementary Figures S3–S8). The ¹H-¹H chemical shift correlation spectroscopy (COSY) correlation of an olefinic proton (δ_H = 5.36) to H-2, and its heteronuclear multiple bond correlation (HMBC) correlations to C-3, and C-9 indicate the presence of C1-C10 double bound. The other olefinic protons (δ_H = 4.74, J = 15 Hz) have HMBC cross peaks with C-7 and C-13 methyl group, locating the exo-isopropylidene group at C-7. The HMBC correlations of C-15 methyl group to C-3, C-4, and C-5; And C-14 methyl group to C-4, C-5, C-6, and C-10 attached these two methyls at C-4 and C-5, respectively. Thus, this compound has the same planar structure as valencene and eremophilene, which harbors three chiral centers of C-4, C-5, and C-7, indicating the existence of eight stereoisomers.

The GC-MS and NMR data of this compound are different to those of (+)-valencene, (-)-4-epieremophilene (Zhao et al., 2004) and (+)-eremophilene (Schifrin et al., 2015), and thus their enantiomers (-)-valencene, (+)-4-epieremophilene, and (-)-eremophilene, only leaving (+)-5-eremophilene and (-)-5-eremophilene as two possible candidates. The rotating frame overhauser effect spectroscopy (ROESY) data are in agreement with this assignment. The optical rotation analysis gave a [α]²⁴ value of -29.4 (ethanol). Taken together, this compound was unambiguously determined as (-)-5-epieremophilene, which is neither isolated nor synthesized previously. And these three sesquiterpene synthases are S. miltiorrhiza (-)-5-epieremophilene synthases.

Because (-)-5-epieremophilene was not detected from the extracts of S. miltiorrhiza plants, we adopted a transient expression system to produce recombinant proteins in leaves of N. benthamiana (Peyret and Lomonossoff, 2013; Sainsbury and Lomonossoff, 2014), in order to further assay the activity of S. miltiorrhiza (-)-5-eiperemophilene synthase in planta. After 4 d of the infiltration of Agrobacterium harboring pEAQ-SmSTPS1, the leaves were gathered and extracted. We found that, compared to the extracts from control N. benthamiana leaves, an extra peak with the retention time of approximately 16 min appeared when SmSTPS1 was expressed, which was determined to be (-)-5-epieremophilene by GC-MS (Figure 5).

Relative expression levels SmSTPS1, SmSTPS2, and SmSTPS3 in root and aerial organs were analyzed by quantitative real-time PCR (qRT-PCR). The three (-)-5-epieremophilene synthase genes were mainly expressed in leaf and inflorescence, and their transcript levels were low or undetectable in stem and root. In general, SmSTPS1 was expressed at a higher level than the other two, and its transcripts were most abundant in leaf. SmSTPS2 was weakly expressed and exhibited a clear expression only in leaf, whereas SmSTPS3 was expressed predominantly in inflorescence (Figures 6A–C). Thus, the three STPS genes were differentially regulated in S. miltiorrhiza plants.

Next, we examined responses of the three sesquiterpene synthase genes to phytohormone treatments. Since SmSTPS1, SmSTPS2, and SmSTPS3 were all expressed in S. miltiorrhiza leaves, changes in their expression levels in leaves were analyzed after hormone treatments. We found that, although the three sesquiterpene synthases share high similarities in sequences and catalytic activities, they responded differentially to different phytohormones at transcription level. SmSTPS1 was induced strongly by ethylene (ET) and moderately by MeJA, abscicic acid (ABA) and gibberellin A₃ (GA₃) (Figure 6D), SmSTPS2 responded not only to ET but also to ABA (Figure 6E), and SmSTPS3 was up-regulated slightly only by ABA (Figure 6F).