Functional Divergence of Diterpene Syntheses in the Medicinal Plant Salvia miltiorrhiza^,

Positive selection and the divergent evolution by exon/intron patterns driving the fast divergence of copalyl diphosphate synthases underlie the biosynthesis of specialized diterpenes.

Abstract

The medicinal plant Salvia miltiorrhiza produces various tanshinone diterpenoids that have pharmacological activities such as vasorelaxation against ischemia reperfusion injury and antiarrhythmic effects. Their biosynthesis is initiated from the general diterpenoid precursor (E,E,E)-geranylgeranyl diphosphate by sequential reactions catalyzed by copalyl diphosphate synthase (CPS) and kaurene synthase-like cyclases. Here, we report characterization of these enzymatic families from S. miltiorrhiza, which has led to the identification of unique pathways, including roles for separate CPSs in tanshinone production in roots versus aerial tissues (SmCPS1 and SmCPS2, respectively) as well as the unique production of ent-13-epi-manoyl oxide by SmCPS4 and S. miltiorrhiza kaurene synthase-like2 in floral sepals. The conserved SmCPS5 is involved in gibberellin plant hormone biosynthesis. Down-regulation of SmCPS1 by RNA interference resulted in substantial reduction of tanshinones, and metabolomics analysis revealed 21 potential intermediates, indicating a complex network for tanshinone metabolism defined by certain key biosynthetic steps. Notably, the correlation between conservation pattern and stereochemical product outcome of the CPSs observed here suggests a degree of correlation that, especially when combined with the identity of certain key residues, may be predictive. Accordingly, this study provides molecular insights into the evolutionary diversification of functional diterpenoids in plants.

Salvia miltiorrhiza, a Lamiaceae species known as red sage or tanshen, is a traditional Chinese medicinal herb that is described in Shen Nong Ben Cao Jing, the oldest classical Chinese herbal book, which dates from between 25 and 220 C.E. The lipophilic pigments from the reddish root and rhizome consist of abietane quinone diterpenoids (Nakao and Fukushima, 1934), largely tanshinone IIA, cryptotanshinone, and tanshinone I (Zhong et al., 2009). These are highly bioactive. For example, tanshinone IIA exerts vasorelaxative activity, has antiarrhythmic effects, provides protection against ischemia reperfusion injury (Zhou et al., 2005; Gao et al., 2008; Sun et al., 2008), and exhibits anticancer activities (Efferth et al., 2008; Lee et al., 2008; Wang et al., 2008; Gong et al., 2011). In addition, tanshinones have been reported to have a broad spectrum of antimicrobial activities against various plant pathogens, including rice (Oryza sativa) blast fungus Magnaporthe oryzae (Zhao et al., 2011). Although tanshinones are mainly accumulated in the roots, trace amounts of tanshinones have been detected in aerial organs as well (Hang et al., 2008).

Diterpenoid biosynthesis is initiated by diterpene synthases (diTPSs), which catalyze cyclization and/or rearrangement of the general acyclic precursor (E,E,E)-geranylgeranyl diphosphate (GGPP) to form various hydrocarbon backbone structures that are precursors to more specific families of diterpenoids (Zi et al., 2014). Previous work has indicated that tanshinone biosynthesis is initiated by cyclization of GGPP to copalyl diphosphate (CPP) by a CPP synthase (SmCPS1) and subsequent further cyclization to the abietane miltiradiene by a kaurene synthase-like cyclase (SmKSL1), so named for its homology to the ent-kaurene synthases (KSs) required for GA plant hormone biosynthesis (Gao et al., 2009). Miltiradiene is a precursor to at least cryptotanshinone (Guo et al., 2013), and RNA interference (RNAi) knockdown of SmCPS1 expression reduces tanshinone production, at least in hairy root cultures (Cheng et al., 2014). The identification of SmCPS1 and SmKSL1 has been followed by that of many related diTPSs from other Lamiaceae plant species (Caniard et al., 2012; Sallaud et al., 2012; Schalk et al., 2012; Brückner et al., 2014; Pateraki et al., 2014). These largely exhibit analogous activity, particularly the CPSs, which produce CPP or the stereochemically related 8α-hydroxy-labd-13E-en-15-yl diphosphate (LDPP) rather than the enantiomeric (ent) CPP relevant to GA biosynthesis.

To further investigate diterpenoid biosynthesis in S. miltiorrhiza, we report here a more thorough characterization of its diTPS family. A previously reported whole-genome shotgun sequencing survey (Ma et al., 2012) has indicated that there are at least five CPSs, although only two KSL genes in S. miltiorrhiza (Supplemental Table S1). Intriguingly, based on a combination of biochemical and genetic (RNAi gene silencing) evidence, we find that these diTPSs nevertheless account for at least four different diterpenoid biosynthetic pathways, each dependent on a unique CPS, with the KS presumably involved in GA biosynthesis seeming to be responsible for alternative diterpenoid metabolism as well. In addition, our studies clarify the evolutionary basis for the observed functional diversity, with investigation of gene structure, positive selection, molecular docking, and mutational analysis used to explore the driving force for the functional divergence of these diTPSs. Moreover, we report metabolomic analysis, also carried out with SmCPS1 RNAi lines, which enables prediction of the downstream steps in tanshinone biosynthesis.

RESULTS

Molecular Analysis of the S. miltiorrhiza diTPSs

Using the working-draft genome sequences of S. miltiorrhiza (Ma et al., 2012), we identified seven complementary DNA (cDNA) sequences encoding diTPSs from an inbred line (bh2-7). The deduced amino acid sequences showed that five contain the (D,E)XDD motif required for the protonation-initiated cyclization reactions catalyzed by CPSs (Prisic et al., 2004) and are defined here as SmCPS1 to SmCPS5, while two have the DDXXD motif involved in binding the divalent magnesium ions required for heterolytic cleavage/ionization of the allylic diphosphate ester bond catalyzed by terpene synthases such as KSs (Christianson, 2006) and are defined here as SmKSL1 and SmKSL2, respectively (Supplemental Table S1).

Phylogenetic analysis revealed distinct conservation of these diTPSs. SmCPS1 to SmCPS3 cluster with other previously characterized CPSs from the Lamiaceae, all of which (including SmCPS1) produce CPP or the stereochemically analogous LDPP. On the other hand, SmCPS4 and SmCPS5 cluster with CPSs that have been previously shown to produce ent-CPP, generally for GA biosynthesis. As previously reported (Hillwig et al., 2011), SmKSL1 has undergone loss of the N-terminal γ-domain usually found in KSLs (although not most other terpene synthases), which has been a characteristic of the KSLs involved in more specialized diterpenoid metabolism from the Lamiaceae. By contrast, SmKSL2 retains the γ-domain and clusters with other dicot KSLs, many of which have been shown to act as KSs involved in GA biosynthesis (Fig. 1).

Figure 1.

Phylogeny of diTPS genes in S. miltiorrhiza. The phylogenetic relationship was reconstructed using the JTT models by PhyML 3.0 with 68 representative characterized diTPS (Supplemental Table S2). Numbers on branches indicate the bootstrap percentage values calculated from 100 bootstrap replicates. Physcomitrella patens CPS/kaurene (PpCPSKS) was used as outgroup. Blue lines show diTPS genes involved in more specialized diterpenoid metabolism from Lamiaceae, and red lines show the loss of N-terminal γ-domain found in KSLs. Red-marked enzymes show diTPS from S. miltiorrhiza.

Biochemical Characterization of the S. miltiorrhiza diTPSs

To determine the biochemical activity of these diTPSs, in vitro enzyme assays were carried out using crude extracts of recombinantly expressed proteins, separately combining each SmCPS with SmKSL1 or SmKSL2 and feeding GGPP as substrate. As expected, combining SmCPS1 and SmKSL1 led to the previously reported production of miltiradiene. Notably, combining SmCPS2 and SmKSL1 also led to production of miltiradiene (Fig. 2A; Supplemental Fig. S1A). Kinetic analysis indicates that SmCPS1 has both higher affinity (K_m = 0.54 μm versus 0.95 μm) and activity (>18-fold higher catalytic constant) with GGPP than SmCPS2 (Fig. 2A). Nevertheless, this observation raises the potential for redundancy between SmCPS1 and SmCPS2. No product was observed when either SmCPS1 or SmCPS2 was combined with SmKSL2. No product was observed when SmCPS3 was combined with either SmKSL1 or SmKSL2. Intriguingly, combining SmCPS4 and SmKSL2 led to production of an unknown compound, while combining SmCPS5 and SmKSL2 led to the production of ent-kaurene, the diterpene precursor to GAs (Fig. 2A; Supplemental Fig. S1A).

Figure 2.

Functional identification of diTPS in S. miltiorrhiza. A, Total ion chromatography of diterpene products from in vitro assays with kinetic constants of SmCPS1 and SmCPS2. k_cat, Catalytic constant. B, Total ion chromatography of the assay with purified SmCPS4 and CfTPS2. The assay product was extracted directly by hexane (no treatment) or after the treatment of alkaline phosphatase (treatment). C, Total ion chromatography of the assay combines SmCPS4 with SmKSL2, together with the identified enzyme CfTPS2 (CPS) and CfTPS3 (KSL) from C. forskohlii. Assays all with 30 μm GGPP as substrate. D, Proposed pathway to ent-13-epi-manoyl oxide in S. miltiorrhiza. OH, Hydroxyl; OPP, diphosphate. E, Total ion chromatography of hexane extracts from different organs in S. miltiorrhiza. The compounds are labdenediol (1), manoyl oxide (2), and ent-13-epi-manoyl oxide (3).

To identify the compound produced by SmCPS4 and SmKSL2, the observed mass spectra was used to search a number of publically available databases, revealing a close match to that of the known diterpenoid 13-epi-manoyl oxide (Supplemental Fig. S1B; Demetzos et al., 2002). This presumably arises from cyclization of LDPP, suggesting that SmCPS4 produces this intermediate. However, SmCPS4 is more closely related to ent-CPP-producing CPSs rather than the previously characterized LDPP synthases from Lamiaceae, whose products are enantiomeric. Previous mutational analysis has shown that the CPS from Arabidopsis (Arabidopsis thaliana) involved in GA metabolism (AtCPS), which then produces ent-CPP, can be easily diverted to the production of ent-LDPP (Potter et al., 2014). Thus, the stereochemistry of the SmCPS4 product was further investigated. Among the previously identified LDPP synthases is one from Coleus forskohlii, CfTPS2 (a CPS homolog), and this Lamiaceae species also encodes a subsequently acting KSL, CfTPS3, that reacts with LDPP to produce manoyl oxide (Pateraki et al., 2014). Both SmCPS4 and CfTPS2 reacted with GGPP to produce LDPP, detected as labdenediol by gas chromatography (GC)-mass spectrometry (MS) following dephosphorylation (Fig. 2B; Supplemental Fig. S1B). To determine if these were enantiomeric, CfTPS2 and SmCPS4 were separately incubated with either SmKSL2 or CfTPS3, respectively, and GGPP as substrate. Notably, combining SmCPS4 with CfTPS3 did not lead to the production of manoyl oxide (Fig. 2C). In addition, combining CfTPS2 with SmKSL2 led to predominant production of manoyl oxide with relatively less 13-epi-manoyl oxide (Fig. 2C). These results indicate that SmCPS4 produces ent-LDPP (8β-hydroxy-ent-CPP). Therefore, the product of SmCPS4 and SmKSL2 appears to be ent-13-epi-manoyl oxide (Fig. 2D).

Physiological Roles of S. miltiorrhiza diTPSs

The tissue-specific expression pattern of the S. miltiorrhiza diTPSs was investigated by quantitative reverse-transcription (qRT)-PCR analysis of various organs, including leaves and roots of 3-d-old seedlings and the root periderm, cortex, and xylem, as well as stem, leaf, sepal, petal, stamen, pistil, and immature seeds of adult plants at the flowering stage (Fig. 3). SmCPS1 and SmKSL1 exhibit closely coordinated expression, and their transcripts were found at extremely high levels in the root periderm, consistent with their role in biosynthesis of the tanshinone pigments, which accumulate in this tissue. SmCPS2 and SmCPS3 were most highly expressed in seedling leaves, although SmCPS2 also is highly expressed in the petals of adult plants. SmCPS4 was most highly expressed in the sepal, and SmCPS5 was most highly expressed in the stem. SmKSL2 was most highly expressed in the root xylem, although this was expressed throughout all tissues of adult plants to some extent.

Figure 3.

qRT-PCR analysis of transcript levels of seven diTPS genes in 12 organs including leaves and roots of 3-d-old seedlings, and the root periderm, cortex, and xylem, as well as stem, leaf, sepal, petal, stamen, pistil, and immature seeds of adult plants at the flowering stage. The expression level was normalized to that of Actin. Data are means from three technical replicates of at least three biological replicates.

To investigate the diterpenoid natural product arsenal of S. miltiorrhiza, both untargeted and targeted metabolomics analyses were carried out using liquid chromatography (LC)-quadrupole time-of-flight (qTOF)-MS and GC-triple quadrupole (QqQ)-MS, respectively. Notably, although S. miltiorrhiza has not been previously reported to produce (ent-)13-epi-manoyl oxide, this diterpenoid was found here, mainly in sepals, with lower amounts found in petals, leaves, and stems, but not in roots (Fig. 2E; Supplemental Fig. S1B). This tissue-specific accumulation pattern matches SmCPS4 mRNA expression with a correlation coefficient of 0.97, indicating that the production of ent-13-epi-manoyl oxide in S. miltiorrhiza depends on expression of SmCPS4 (as well as the more ubiquitous expression of SmKSL2).

As previously reported (Hang et al., 2008), tanshinones were found not only in the root and rhizome, but also in aerial tissues of S. miltiorrhiza. Given the distinct tissue-specific expression patterns of SmCPS1 and SmCPS2, both of which can produce the relevant CPP intermediate, it seemed possible that these are not redundant but, instead, contribute separately to tanshinone biosynthesis in the roots and aerial tissues, respectively. Moreover, as the only identified ent-CPP synthase, it also seemed likely that SmCPS5 is involved in GA metabolism.

Genetic Evidence for Distinct Diterpenoid Pathways in S. miltiorrhiza

To further investigate the distinct roles of the various SmCPSs in S. miltiorrhiza diterpenoid biosynthesis, RNAi gene silencing was carried out targeting either SmCPS1 or SmCPS5 separately. RNAi knockdown of SmCPS5 (Supplemental Fig. S2A) resulted in dwarf transgenic plants with significantly shorter pinnately compound leaves, shorter and narrower top leaves, and smaller flowers compared with wild-type plants and could be rescued (i.e. normal growth restored) by applying GA₃ to the T0 generation plants (Supplemental Fig. S2, B–D). Further, T1 generation plants showed a 3:1 segregation ratio for the dwarf phenotype, with severe stunting of shoots as well as dark-green leaves (Supplemental Fig. S2E). These phenotypes are associated with GA deficiency (Margis-Pinheiro et al., 2005), indicating that SmCPS5, presumably together with SmKSL2, function in the GA biosynthetic pathway in S. miltiorrhiza.

To investigate the potential redundancy between SmCPS1 and SmCPS2, RNAi targeting SmCPS1 was carried out, generating five lines exhibiting silencing of approximately 90% (Fig. 4A). These SmCPS1-RNAi plants exhibited an obvious white-color root phenotype compared with wild-type roots, which had the characteristic reddish color associated with tanshinones (Fig. 4B). Root-directed metabolite analysis demonstrated that cryptotanshinone, tanshinone IIA, and tanshinone I were dramatically reduced, from 14,465 ± 995, 2,311 ± 13, and 2,010 ± 117 μg g^–1 in wild-type plants to 0.10 ± 0.01, 1.2 ± 0.2, and 14 ± 1 µg g^–1 in the SmCPS1-RNAi plants, respectively (Fig. 4C). As a control, the water-soluble polyphenolic acid content of the plants, including lithospermic acid B and rosmarinic acid, also was investigated. The lack of any statistically significant differences between the wild-type and SmCPS1-RNAi samples (Fig. 4C) indicates that down-regulation of SmCPS1 expression specifically affects tanshinone biosynthesis (i.e. this did not interfere with the production of at least these polyphenolic acid metabolites).

Figure 4.

Phenotype and metabolic profiles caused by down-regulation of SmCPS1 in T0 generation plants. A, qRT-PCR analysis of transcript levels of five CPS genes in root of SmCPS1-RNAi and the wild type (WT). Expression was normalized to that of Actin. The error bars show the sds from mean value (n = 3 experiments). SmCPS2 is not expressed in these samples. B, The phenotype of down-regulation of SmCPS1. C, Quantitative analysis of five major compounds including cryptotanshinone (CT), tanshinone IIA (T-IIA), tanshinone I (T-I), lithospermic acid B (LAB), and rosmarinic acid (RA) in root of SmCPS1-RNAi lines and the wild type. The error bars show the sds from mean value (n = 3 experiments). Asterisks indicate significant difference at P < 0.01 compared with the wild type by Student’s t test. D, Cross section of the root. E, Summary of metabolite flux caused by down-regulation of SmCPS1. The red arrow shows the accumulated metabolites, and the blue arrow shows the reduced metabolites in root of SmCPS1-RNAi lines compared with the wild type. Underlined metabolites are identified by standard reference. FW, Fresh weight; OH, hydroxyl; OPP, diphosphate; R1, R2, and R3, substituent group. Bar = 100 μm.

Notably, tanshinone levels in the aerial tissues were not affected in the SmCPS1-RNAi plants (Table I). Given that SmCPS2 is predominantly expressed in these tissues (Fig. 3), these results suggest that SmCPS2 mediates tanshinone biosynthesis in aerial organs of S. miltiorrhiza independently of the SmCPS1-dependent pathway in the root periderm, such that SmCPS1 and SmCPS2 are not redundant.

Table I. Main compounds detected in aerial part and seedling of the wild type and SmCPS1-RNAi.

Organ	Concentrationa				Relative Concentrationb
	Cryptotanshinone		Tanshinone IIA		Miltirone		Trijuganone B		Tanshinone I
	Wild Type	RNAi	Wild Type	RNAi	Wild Type	RNAi	Wild Type	RNAi	Wild Type	RNAi
	ng g–1 fresh wt
Petal	0.2	0.2	2.2	2.2	12.1	12.0	19.9	20.2	10.6	11.0
Sepal	1.2	1.2	22.1	23.0	72.2	73.0	143.0	144.4	14.4	15.0
Young leaf	0.5	0.5	7.3	7.6	25.9	26.1	67.1	68.0	13.5	13.0
Young root	0.8	0.8	8.9	9.0	61.5	62.3	59.3	60.2	20.3	21.0

^aThe concentration obtained by calibration of standards.

^bThe relative concentration obtained by comparison with internal standard umbelliferone.

Metabolomic Analysis Clarifies Tanshinone Biosynthesis

Metabolomics analysis was carried out with not only the wild-type plant roots, but also SmCPS1-RNAi plant roots, with LC-electrospray ionization-qTOF-MS revealing 39 and GC-electron impact (EI)-QqQ-MS 19 metabolites with significantly reduced accumulation (i.e. a >2-fold change with P < 0.05; Supplemental Tables S3 and S4). On the other hand, four metabolites (by using GC-EI-QqQ-MS) with elevated accumulation also were identified (Supplemental Table S4). By comparison to known compounds, 21 of the metabolites with reduced accumulation in the roots of SmCPS1-RNAi plants were identified as diterpenoids, all of which were predominantly accumulated (>94.5%) in the periderm (Fig. 4D; Supplemental Tables S3 and S4). Of the metabolites exhibiting increased accumulation, three were identified as diterpenoids (geranylgeraniol, α-springene, and β-springene), and these were not detected in the periderm, cortex, and xylem of wild-type plant roots (Supplemental Table S4). These compounds presumably appear due to the down-regulation of SmCPS1, with accumulation of GGPP leading to hydrolysis to geranylgeraniol, with subsequent dehydration leading to α-springene and β-springene (Fig. 4E).

Of the 21 identified SmCPS1-dependent metabolites, 19 are tanshinones or plausible biosynthetic intermediates, while two are rearranged abietane diterpenoids (i.e. przewalskin and salvisyrianone; Fig. 4E). The 19 tanshinones and biosynthetically relevant metabolites could be further divided into five main groups according to the progressive modification of their carbon skeletons (Fig. 4E). Group I is simply composed of miltiradiene, with its planar cyclohexan-1,4-diene C ring. Group II compounds are dehydro-abietanes, with a characteristic aromatic C ring (i.e. abietatriene, ferruginol, and sugiol). Group III compounds are nor-abietatetraen-11,12-diones, with conversion of the C ring to an ortho-quinone (i.e. keto groups at C-11 and C-12), as well as aromatization of the B ring, along with loss of C-20 (e.g. miltirone and 4-methylene-miltirone). The 11 metabolites comprising group IV all are variously named tanshinones, with addition of the (dihydro)furan ring D. Group V contains three metabolites that have been additionally modified by the loss of one of the geminal methyl groups from the C-4 position, along with the presence of at least one double bond in the A ring.

Positive Selection for Divergent CPS Activity

Given the ease with which diTPSs can be diverted to alternative activity (Wilderman and Peters, 2007; Xu et al., 2007; Keeling et al., 2008; Morrone et al., 2008; Criswell et al., 2012; Potter et al., 2014), it is not clear what underlies the expanded nature and divergent activity observed in the S. miltiorrhiza diTPS family (i.e. selective pressure or genetic drift). Gene structure, specifically the number and placement of introns, has been associated with evolutionary descent in the terpene synthase gene family (Trapp and Croteau, 2001). Accordingly, the CPS and KSL genes of S. miltiorrhiza, Arabidopsis, and rice can be divided into three groups. Group I genes associated with GA metabolism each have the typical 15 exons and 14 introns for CPS genes and 14 exons and 13 introns for KS genes (Supplemental Fig. S3), which corresponds to the ancestral plant diTPS gene structure (Trapp and Croteau, 2001). Group II CPS/KSL genes, including SmCPS1, SmCPS3, SmCPS4, SmKSL1, OsCPS2, OsCPS4, OsKSL5, OsKSL6, OsKSL8, and OsKSL10, show diverged sequences and genomic architecture. In particular, intron loss (Zhang et al., 2014) relative to the conserved group I genes are often observed. For example, SmCPS1 has lost the 10th and 12th introns, SmCPS4 the 5th intron, and SmCPS3 the first four introns, while OsCPS2 has lost the 2nd and 3rd introns, and OsKSL5, OsKSL6, OsKSL8, and OsKSL10 all have lost the last intron (Supplemental Fig. S3). These architecturally divergent group II genes are involved in the biosynthesis of specialized metabolites in S. miltiorrhiza and rice. On the other hand, the group III genes SmCPS2, OsKSL4, and OsKSL7 exhibit conserved genomic architecture but divergent sequences and functions relative to the group I genes associated with GA biosynthesis (Supplemental Fig. S3).

To investigate whether the observed functional divergence of the CPS genes involved in biosynthesis of the tanshinones and other more specialized labdane-related diterpenoids is a function of positive selection, phylogenetic analysis of the protein-coding DNA sequences of CPS members from an array of angiosperms was carried out. The resulting phylogenetic tree (Fig. 5A) provided the basis for branch site model analysis conducted with the PAML package (Zhang, 2004; Zhang et al., 2005; Yang and dos Reis, 2011). This revealed a statistically significant signature for positive selection associated with functional divergence from the production of ent-CPP required for GA biosynthesis to the enantiomeric (i.e. normal) CPP involved in production of the tanshinones, as well as stereochemically related LDPP for other more specialized diterpenoid biosynthesis (branch c, nonsynonymous to synonymous substitution ratio = 5.64, P < 0.01; Fig. 5, A and B; Supplemental Table S5).

Figure 5.

Molecular evolution of CPS genes. A, Bold lines illustrate branches or genes evolved under positive selection with significant statistical support at P < 0.05 both in branch site model tests 1 and 2. Branch c shows the divergence of normal-CPP and normal-LDPP synthase from the ent-CPP synthase. d_N/d_S, Nonsynonymous to synonymous substitution ratio. B, Different stereoisomers of CPP and LDPP. C, Alignment of conserved residues. The asterisks indicate positive selection sites with posterior probability greater than 95% by Bayes Empirical Bayes analysis.

To more closely examine the basis for the observed functional diversification, the analysis was extended to individual codons in SmCPS1 versus SmCPS5. The positively selected sites identified by Bayes Empirical Bayes analysis correspond to Ser-362, Asp-389, Ser-415, and Pro-416 in SmCPS1, which are Trp-364, Ser-391, Thr-416, and Gly-417 in SmCPS5 (Fig. 5C). Models for both SmCPS1 and SmCPS5 were generated and revealed that three of these residues, Ser-362:Trp-364 (S-W) and Ser-415/Pro-416:Thr-416/Gly-417 (SP-TG) are part of the active site cavity, whereas Asp-389:Ser-391 (D-S) is more than 6 Å away from the active site cavity (Supplemental Fig. S4, A and B). Site-directed mutagenesis was performed to swap these residues between SmCPS1 and SmCPS5, except Thr-416 in SmCPS5, as this corresponds to Thr-421 in AtCPS, which has already been reported to be involved in catalysis (Köksal et al., 2014). Assays using the mutant enzymes reacting with GGPP alone or in combination with SmKSL1 or SmKSL2 showed that the mutant SmCPS1:S362W, in which Ser-362 (S) of SmCPS1 was replaced by the Trp (W) found at the same position in SmCPS5, was more than 100-fold less efficient in the production of CPP, and miltiradiene when assayed with SmKSL1, relative to the wild-type SmCPS1. Moreover, this mutant also did not yield any detectable product when incubated with SmKSL2 and GGPP (Supplemental Fig. S5), indicating that no ent-CPP is produced. The mutant SmCPS5:W364S also showed about 80-fold lower efficiency in production of ent-CPP, and ent-kaurene when assayed with SmKSL2 compared with wild-type SmCPS5. Similarly, this mutant also did not yield any product when incubated with SmKSL1. The other four mutant enzymes SmCPS1:D389S, SmCPS1:P416G, SmCPS5:S391D, and SmCPS5:G417P did not cause significant change in enzymatic activity relative to the parental/wild-type CPSs (Supplemental Fig. S5). Thus, while no change in product stereochemistry was observed, it does seem that the residue corresponding to the S-W position is important for catalysis in both SmCPS1 and SmCPS5 (Supplemental Fig. S4, C and D), despite their difference in product outcome, suggesting that this substitution has a role in enabling the production of normal versus ent-CPP and, hence, biosynthesis of the derived tanshinones.

DISCUSSION

The combined biochemical and genetic work reported here has defined the roles of the diTPS family in S. miltiorrhiza. Of the five SmCPSs, while SmCPS3 appears to be inactive, each of the other SmCPSs defines separate diterpenoid pathways. Rather than reflecting redundancy, the similar biochemical activity of SmCPS1 and SmCPS2 is coupled to their distinct roles in tanshinone biosynthesis in the roots versus aerial tissues, respectively. This discovery provides the possibility of using metabolic engineering strategies to enhance the production of tanshinones in aerial organs. Cultivation of the resulting plants could then provide annually renewable source materials for extraction of tanshinones by harvesting aerial tissues without destroying the entire plant.

Both SmCPS1 and SmCPS2 react with GGPP to form normal CPP in S. miltiorrhiza. Our phylogenic analysis indicates that SmCPS1 and SmCPS2 are in the same clade that also contains other CPSs from the Lamiaceae and closely related Solanaceae (Fig. 1) that similarly produce CPP and LDPP with analogous stereochemistry. These CPSs were likely derived from an ancestral CPS that produced ent-CPP for GA biosynthesis via early gene duplication and neofunctionalization that occurred at least before the divergence of the Lamiaceae and Solanaceae (Figs. 1 and 5), which has been estimated to have been 64 million years ago (48–75 million years; Zhang et al., 2012). During the evolution of Lamiaceae, two more gene duplication events likely happened, producing SmCPS3 as well as SmCPS1 and SmCPS2, which individually are representative of two more widespread clades (e.g. homologs to both are found in C. forskohlii), respectively. SmCPS2 retained the ancestral gene architecture (Supplemental Fig. S3) and exhibits a similar gene expression pattern as the SmCPS5 involved in GA biosynthesis (Fig. 3), whereas SmCPS1 has undergone more divergence, including intron loss (Supplemental Fig. S3) and altered transcriptional regulation (Fig. 3), along with exhibiting significantly higher catalytic activity than SmCPS2. These data suggest that tanshinone biosynthesis may have evolved early in the Lamiaceae, and the gene duplication leading to SmCPS1 and SmCPS2 enabled further localization and up-regulation of tanshinone biosynthesis in root periderm cells in some Salvia spp., including S. miltiorrhiza. The presence of distinct SmCPS1- and SmCPS2-dependent tanshinone pathways in the root periderm and aerial tissues, respectively, indicates that these labdane-related diterpenoids may play important roles in plant development and in adaptation to different stress conditions, which is a topic worth future investigation.

The ability of SmCPS4 to produce the enantiomeric form of LDPP was suggested by the ability of SmKSL2 to react with both ent-CPP (to produce ent-kaurene) and this ent-LDPP, but not the CPP product of SmCPS1 (or SmCPS2). The observed production of ent-13-epi-manoyl oxide in S. miltiorrhiza then suggests dual function for SmKSL2 (i.e. in GA and this more specialized diterpenoid biosynthesis). Regardless, identification of the ability of SmCPS4 and SmKLS2 provides access to this unique diterpenoid. In addition, the ability of SmCPS4 to produce ent-LDPP was presaged by its phylogenetic relationship to ent-CPP-producing CPSs. Based on the similar clustering of a previously identified LDPP synthase from Grindelia robusta (Zerbe et al., 2013), as well as the ability of a KS to selectively react with its product, it is suggested here that this GrTPS1 may produce ent-LDPP (Fig. 2D).

It is interesting to note that certain previously identified residues are consistent with the CPS enzymatic activity observed here. SmCPS5, shown here to play a role in GA biosynthesis, contains the His (His-326) associated with susceptibility to inhibition by Mg²⁺ (Mann et al., 2010), which has been suggested to serve a regulatory role in such CPSs (Prisic and Peters, 2007). In addition, SmCPS5 contains the His-Asn dyad (His-258/Asn-317; Fig. 5C) that has been suggested to act as the catalytic base and is conserved in CPSs that produce ent-CPP (Potter et al., 2014). By contrast, the ent-LDPP-producing SmCPS4 contains a Ser (Ser-299) in place of the corresponding Asn (Fig. 5C), consistent with the ability of such substitution of smaller residues to enable production of hydroxylated CPP (Potter et al., 2014). In addition, the investigation of residues showing signs of positive selection here indicates other positions important for altering stereochemical product outcome, although further experiments are required to determine how many and which residues are required. Nevertheless, altogether, the results reported here suggest that combining phylogenetic relationships with the identity of residues at positions of known catalytic relevance may be predictive for CPS catalytic activity.

MATERIALS AND METHODS

Plant Materials

The Salvia miltiorrhiza species has two different flower colors, purple and white. Varieties with purple flowers are distributed throughout China. The variety with white flowers (S. miltiorrhiza f. alba.) is only found in Shandong Province. S. miltiorrhiza f. alba is a rare and thus more valuable variety of Danshen. The white flower line bh2-7, inbred for 5 generations, was used in our study.

Plant Growth and Culture Conditions

bh2-7 seeds were surface sterilized with 5% (v/v) sodium hypochlorite and cultured on solid hormone-free Murashige and Skoog basal medium containing 30 g L^–1 Suc and 8 g L^–1 agar. Cultures were maintained at 25°C under a 16-h-light/8-h-dark photoperiod. Seedlings were transferred to pots filled with soil:vermiculite (3:1) mix and grown under the same temperature and light regime in a plant growth room. Dwarf plants of SmCPS5-RNAi were sprayed with 150 μm GA₃ solution to complement the phenotype in the T0 generation.

Genomic Sequence of diTPS Gene Family Members

To show the structural divergence that occurred between SmCPS1, SmCPS3, and SmCPS4, we cloned the full genomic sequence of each gene from bh2-7. Cetyl-trimethyl-ammonium bromide method was used to extract the genomic DNA and amplified with specific primers (Supplemental Table S6). Intron/exon structures were predicted using the Gene Structure Display Server (Guo et al., 2007).

Gene Expression Analysis

Plant samples were harvested and immediately frozen in liquid nitrogen. Total RNA was extracted using a modified cetyl-trimethyl-ammonium bromide protocol and treated with RNase-Free DNase I (Takara) to remove residual genomic DNA. RNA integrity and quality were checked by denaturing gel electrophoresis, and the absence of genomic DNA was confirmed by PCR using primers for Smactin, prior to reverse transcription. One to five micrograms of total RNA was reverse transcribed into cDNA using the SuperScript III reverse transcriptase and oligo(dT)_12–18 primer (Invitrogen), according to the manufacturer’s instructions. The synthesized cDNA was then diluted 10-fold. One microliter of this diluted template was used for subsequent qRT-PCR analysis with a total PCR reaction volume of 20 μL. qRT-PCR was performed with a SuperReal PreMix for SYBR Green Kit (TIANGEN) on a Corvett Rotor-Gene 3000 real-time PCR detection system. All reactions were performed using the following PCR conditions: initial denaturation step of 95°C for 10 min, followed by 40 cycles each of 95°C for 5 s, 60°C for 15 s, and 72°C for 20 s, with a final melting stage from 55°C to 95°C. A final dissociation step was performed to assess the quality of the amplified product. cDNA from a series of 5-fold dilutions were used for calibration, and the efficiency of the PCR amplifications was found to be in the range of 90% to 110%, which is considered desirable for quantitative PCR (Taylor et al., 2010). Relative expression levels were calculated as the ratio of the target gene transcript level to the transcript level of the housekeeping gene Actin (Smactin; Yang et al., 2010). Primer specificity was confirmed by direct cloning and sequencing of individual PCR amplification products. qRT-PCR was performed with three technical replicates of at least three biological replicates for each tissue or transformed plant line. The hot map of gene expression data was generated in the R software package.

Phylogenetic Analysis

Sixty-eight diTPSs with characterized functions were included in the phylogenetic analysis (Supplemental Table S2). To carry out the phylogenic reconstruction, multiple protein sequence alignments were performed with MAFFT version 7.012 employing the E-INS-I method (Katoh et al., 2005). Maximum likelihood trees were built using PhyML version 3.0 (Guindon et al., 2010). Specifically, PhyML analyses were conducted with the JTT substitution model, four rate substitution categories, and 100 or 1,000 bootstrap replicate analyses. The phylogeny was displayed using FigTree software (http://tree.bio.ed.ac.uk/software/figtree/).

Hairy Root and Plant Transformation for Knockdown of SmCPS1 and SmCPS5 by RNAi

A 289-bp gene-specific sequence including the 3′ untranslated region of SmCPS1 and a 399-bp gene-specific sequence in the 3′ end of SmCPS5 were amplified by PCR using cDNA as a template and then cloned using Gateway technology into the pK7GWIWG (II) binary vector (Limpenset al., 2005). Positive plasmids of pK7GWIWG-CPS1 and pK7GWIWG-CPS5 were identified with sequencing and restriction enzyme analysis and then introduced into Agrobacterium tumefaciens strain EHA105 and Agrobacterium rhizogenes strain ACCC 10060 by electroporation. Transformation of leaf explants from S. miltiorrhiza bh2-7 plants was carried out following previously described methods (Yan and Wang, 2007) with minor modifications. Single colonies of A. tumefaciens strain EHA105 cells harboring the various RNAi vectors were inoculated into 10 mL of liquid Luria-Bertani medium with 50 mg L^–1 spectinomycin and 100 mg L^–1 rifampicin and then grown on a shaker (180 rpm) at 28°C for 16 to 18 h. Cells were collected by centrifugation when the optical density at 600 nm reached 0.6 and were resuspended in 20 mL of liquid Murashige and Skoog medium. Leaves or petioles were cut into 0.5- × 0.5-cm pieces discs and precultured for 2 d on Murashige and Skoog basal medium supplemented with 2.0 mg L^–1 6-benzyladenine. The discs were then submerged with shaking in a bacterial suspension for 15 min and cocultured on the Murashige and Skoog basal medium for 2 d. The leaf discs were then transferred to selection Murashige and Skoog basal medium supplemented with 2.0 mg L^–1 6-benzyladenine, 50 mg L^–1 kanamycin, and 225 mg L^–1 timentin. After two to three rounds of selection (10 d each), the regenerated buds with expression of GFP were transferred to Murashige and Skoog basal medium supplemented with 25 mg L^–1 kanamycin for root formation and elongation. Rooted plantlets were further cultured on Murashige and Skoog basal medium for about 1 month. The plantlets (7–8 cm tall, with roots 5–6 cm long) were transplanted to soil and vermiculite (3:1) and covered by beakers to maintain humidity for 1 week and then gradually hardened off in pots in a greenhouse for further growth.

Hairy root cultures can be successfully obtained with S. miltiorrhiza, and we used both hairy root cultures and full plants to analyze the silencing effect of SmCPS5. Though hairy root cultures with silencing of SmCPS5 showed no significant phenotype, we found that it was an easy and fast system to analysis the effect of different silencing vectors. The A. rhizogenes-based transformation had a similar procedure as the A. tumefaciens-based transformation described above. When the hairy roots were 2 to 3 cm in length, expression of GFP was observed under a fluorescence microscope to identify positive lines. The positive hairy roots were excised and cultured on solid, hormone-free Murashige and Skoog basal medium containing 50 mg L^–1 kanamycin and reduced timentin from 225 mg L^–1 to zero in two or three selection cycles (15 d each). The rapidly growing kanamycin-resistant and GFP-visible lines with no bacterial contamination were then maintained at 25°C in the dark with Murashige and Skoog medium without ammonium nitrate and routinely subcultured every 25 to 30 d.

Metabolomics Profiling Using LC-qTOF-MS and GC-QqQ-MS

The metabolomics profiling data were acquired using a combination of two independent analytical platforms. LC-qTOF-MS analysis of methanol extracts was used for global unbiased metabolite detection. GC-QqQ-MS analysis of hexane extracts, optimized for detection of targeted intermediates, was used for the analysis of metabolites that were not readily detectable with the LC-qTOF-MS platform.

For LC-qTOF-MS, fresh plant root samples were frozen in liquid nitrogen and ground to a fine powder under continuous cooling. One hundred milligrams fresh weight of the powder was extracted in 2 mL of methanol, containing an internal standard (umbelliferone, 20 μg mL^–1). The extracts were sonicated twice for 15 min, centrifuged (1,500g) for 10 min, and then filtered through a 0.2-μm polytetrafluoroethylene syringe filter (Agilent). An aliquot of each filtrate (5 μL) was separated using an Agilent 1290 Infinity ultra high performance liquid chromatography system consisting of a binary pump, an autosampler, a column temperature controller, and a variable wavelength detector at 285 nm. The chromatography was performed using a ZORBAX RRHD SB-C18 column from Agilent Technologies (2.1 × 100 mm, 1.8 μm). The mobile phase consisted of 0.01% (v/v) formic acid in acetonitrile (A) and water containing 0.01% (v/v) formic acid (B). A gradient program was used as follows: linear gradient from 10% to 20% A (0–5 min), linear gradient from 20% to 40% A (5–7 min), linear gradient from 40% to 100% A (7–10 min), isocratic at 100% A (10–14 min), and linear gradient from 100% to 10% A (14–15 min). The mobile phase flow rate was 0.25 mL min^–1, and the column temperature was set at 30°C. Five reference standard compounds, namely salvianolic B, rosmarinic acid, crypotanshinone, tanshinone IIA, and tanshinone I were dissolved in methanol to create standard curves for use in absolute quantification calculations. To ensure that analytes were in the linear range and to exclude some artifactual peaks, samples were also diluted by 10- and 50-fold and reanalyzed.

MS was performed using an Agilent 6540 qTOF equipped with an electrospray ionization source operating in positive ion mode. The nebulization gas was set to 40 pounds per square inch. The drying gas was set to 10 L min^–1 at a temperature of 350°C, and the sheath gas was set to 11 L min^–1 at a temperature of 350°C. The capillary voltage was set to 4,000 V. The qTOF acquisition rate was set to 0.5 s. For full-scan MS analysis, the spectra were recorded in the range of mass-to-charge ratio (m/z) 100 to 1,000. Chromatographic separation, followed by full-scan mass spectra, was performed to record retention time and m/z values of all detectable ions present in the samples.

For GC-QqQ-MS, plant material was lyophilized for 48 h. One hundred milligrams of the lyophilized powder was extracted in 2 mL of hexane that contained two internal standards (tricosane, 6.5 μg mL^–1; and tetracosane, 0.8 μg mL^–1). The extracts were sonicated twice for 15 min and centrifuged (3,000g) for 10 min. The supernatant was evaporated under nitrogen, resuspended in 50 μL of hexanes or derivatized with 80 μL of N-methyl-N-(trimethylsilyl) trifluoroacetamide and 8 μL of pyridine at 80°C for 40 min, and then analyzed by GC-MS.

GC-MS analyses were performed on an Agilent 7890A GC system connected to an Agilent 7000B triple quadrupole mass spectrometer with EI ionization. A 1-μL portion of the extract was injected in splitless mode onto the column. The column used was a DB-5ms (30-m × 0.25-mm i.d., 0.25-μm film thickness; Agilent J&W Scientific) fused silica capillary column. Helium was used as the carrier gas for GC at a flow rate of 1.0 mL min^–1. The injector temperature was 280°C. The oven program was as follows: 50°C for 2 min, linear ramp at a rate of 20°C min^–1 to 200°C, and then followed with a linear ramp at a rate of 5°C min^–1 to 300°C, held at 300°C for 10 min. The transfer line temperature was 280°C.

Raw data were processed with Mass Hunter Qualitative Analysis software (Agilent). Mass Profiler Professional (Agilent) software was used to identify significantly different ion features. A series of filtration steps was performed to further filter the initial results. First, only features with abundances above 1,000 ion counts were selected. Second, features were passed through a quality control tolerance window of 0.1% plus 0.15 min and 5 μg mL⁻¹ plus 2.0 mD chosen for alignment of retention time and m/z values, respectively. Third, features that were not present in all biological replicates of any single sample group were removed. The data were normalized to the detected values for the internal standard peak; GC-QqQ-MS data were normalized to tetracosane. After alignment and normalization of the peaks of each sample, a single data set stored as a matrix was prepared. Fold change values were calculated as the ratio of mean SmCPS1-RNAi line feature values compared with the mean values for these features in the wild-type lines. Student’s t tests were then used to determine whether each feature was increased or decreased significantly. The aligned data were exported to SIMCA-P+ 12.0 for multivariate analyses (partial least-squares discriminant analysis).

Differentially accumulated features detected by LC-qTOF-MS were identified by automatic comparison to a personalized metabolite database established with METLIN software (Sana et al., 2008). Seven hundred sixty-two diterpenoids were collected from different Saliva spp., among them, 86 tanshinones and biosynthetically related metabolites came from S. miltiorrhiza. The putative metabolite peaks were tentatively identified by comparison with MS/MS spectra (Yang et al., 2006; Zhou et al., 2009) and reference standard compounds including sugiol, cryptotanshinone, tanshinone IIA, tanshinone IIB, tanshinone I, dihydrotanshinone I, przewalskin, salvisyrianone (BioBiopha), and miltirone (Faces Biochemical). Differentially accumulated features detected by GC-QqQ-MS were compared with the NIST 05 standard mass spectral databases and four reference standard compounds, including geranylgeraniol (Sigma), miltiradiene (Luqi Huang’s lab), and ferruginol and sugiol (BioBiopha).

In Vitro Assays

For in vitro functional assays, the full coding sequence of S. miltiorrhiza diTPS genes with specific restriction enzyme sites (Supplemental Table S6) was cloned into the pGEM-T vector (Promega), digested with corresponding restriction enzymes, and subcloned into the expression plasmid pET32a (Merck) to create pET32-CPSs and pET32-KSLs. AtCPS (AAA53632) and AtKS (AAC39443) were cloned from Arabidopsis (Arabidopsis thaliana). CfTPS2 (KF444507) and CfTPS3 (KF444508) were full synthesized. The expression, purification, and kinetic analysis of the recombinant proteins were performed as described previously (Hillwig et al., 2011). The constructs were transformed into Tuner (DE3) or Origami B (DE3) competent cells (Merck). Three to five positive colonies were cultured in Luria-Bertani medium with 50 mg L^–1 carbenicillin, and 0.1 to 0.4 mm isopropyl β-d-thiogalactopyranoside was added to induce the expression of the protein. Subsequently, cell pellets were collected and resuspended in assay buffer (50 mm phosphate, pH 7.4, 10% (v/v) glycerol, 2 mm dithiothreitol, and 10 mm MgCl₂) and sonicated for 10 s six times on ice. Lysate from the samples was centrifuged at 12,000g, and the resulting supernatant was used for the assays. The conversion of GGPP to CPP or LDPP was carried out by incubating 500 μg of pET32-CPS sample protein extract with 20 to 50 μm GGPP (Sigma) in a final volume of 250 μL of assay buffer for 2 to 4 h at 30°C. Assay mixtures were hydrolyzed (dephosphorylated) with 75 units of bacterial alkaline phosphatase at pH 8 for 16 h at 37°C to produce hexane-soluble products. GGPP was converted to kaurene, miltiradiene, or manoyl oxide by mixing 250 μg of pET32-CPS protein extract and 250 μg of pET32-KSL protein extraction with 20 to 50 μm of GGPP and incubated for 2 to 4 h at 30°C. Assay mixtures were extracted three times with an equal volume of hexane. The hexane fractions were pooled, evaporated under nitrogen, resuspended in 50 μL of hexanes or derivatized with 80 μL of N-methyl-N-(trimethylsilyl) trifluoroacetamide and 8 μL of pyridine at 80°C for 40 min, and then analyzed by GC-MS. To identify possible products, we obtained and analyzed a series of standard reference compounds including geranylgeraniol (Sigma), geranyllinalool (Sigma), (13E)-labda-8α,15-diol (BioBiopha), 13-epi-manool (BioBiopha), and sclareol (Sigma).

Purification of Recombinant CPS and KSL

Cell pellets were resuspended in 16 mL of protein lysis buffer (50 mm phosphate, pH 7.4, 300 mm NaCl, 10% (v/v) glycerol, 10 mm MgCl₂, and 20 mm imidazole) and sonicated for 10 s, six times, on ice. Lysate from the samples was centrifuged at 12,000g for 20 min at 4°C. The cleared lysate was transferred to prewashed nickel-nitrilotriacetic acid agarose beads and incubated for 30 min at 4°C. Thereafter, the nickel-nitrilotriacetic acid agarose beads were rinsed three times with 15 mL of washing buffer (20 mm phosphate, pH 7.4, 300 mm NaCl, and 100 mm imidazole). The tagged protein was then eluted by the addition of 6 mL of elution buffer (20 mm phosphate, pH 7.4, 300 mm NaCl, and 500 mm imidazole) to the bead bed. The buffer of the eluted proteins was then exchanged using a PD-10 column equilibrated with assay buffer (50 mm phosphate, pH 7.4, 10% [v/v] glycerol, 2 mm dithiothreitol, and 10 mm MgCl₂). The purified proteins were then identified by SDS-PAGE gel and quantified by Bradford assays (Genstar).

Enzyme Kinetic Analysis of SmCPS1 and SmCPS2

For kinetic assays, 20 nm purified SmCPS1 and SmCPS2 was used, with a 2-min reaction time at 25°C. Single-vial assays were used as described above. Assays were completed in triplicate with 0.5 to 20 μm GGPP for SmCPS1 and 0.5 to 8 μm GGPP for SmCPS2. Enzymes were deactivated at the end of the 2-min reactions by incubating the reaction vial at 80°C for 3 min, followed by quenching on ice. Three hundred molar purified SmKSL1 enzyme was then added to the vial in the single-vial assay for another reaction at 30°C for 2 h. Assays were analyzed via GC-QqQ using selected ion monitoring of m/z 57 (for internal standard tetracosane) and m/z 134 for the miltiradiene product. Miltiradiene concentrations were determined relative to the internal standard by using excess SmCPS1 and SmKSL1 in single-vial assays and allowing the reaction to proceed to completion (2 h). Kinetic parameters were determined by nonlinear regression using a Michaelis-Menten model implemented in GraphPad Prism 6.03.

Positive Selection Analysis

For molecular evolution analysis, 29 protein-coding DNA sequences for CPSs were used to construct phylogenetic trees, using PhyML3.0 (Guindon et al., 2010) under the general time-reversible nucleotide substitution model with four rate substitution categories. The branches with bootstrap values higher than 700 were used for branch site model tests 1 and 2 in the PAML package to detect whether positive selection had acted on particular amino acid sites within specific lineages (Zhang, 2004; Zhang et al., 2005; Yang and dos Reis, 2011). Branch site model uses a maximum-likelihood approach to calculate nonsynonymous to synonymous rate ratios. Likelihood ratio tests were performed, and the values of twice the difference between the log likelihood of different models were posteriorly transformed into exact P values using PAML 4.6 (Yang, 2007). The χ² distributions with the degree of freedom = 2 and degree of freedom = 1, which have been shown to be conservative under conditions of positive selection (Zhang, 2004), were used to perform tests 1 and 2, respectively. Probabilities of sites under positive selection were obtained using Bayesian approaches (Yang et al., 2005) implemented in PAML.

Homology Modeling, Molecular Docking, and Mutagenesis

Homology models were constructed within the SWISS-MODEL Workspace using the automatic alignment algorithm (Arnold et al., 2006). The crystal structures of CPS from Arabidopsis (Protein Data Bank nos. 3pya and 3pyb) were used as the templates (Köksal et al., 2011, 2014). The ligand docking modeling was performed with AutoDock Vina (Trott and Olson, 2010). Model visualization and binding site analysis were performed using PyMOL (http://www.pymol.org). Mutants were generated by whole-plasmid PCR amplification with overlapping mutagenic primers of the pGEM-T vector (Promega) clones and verified by complete gene sequencing prior to subcloning into the expression vector pET32a (Merck). The resulting constructs were heterologous expressed and analyzed as described above.

Sequence data from this article can be found in the GenBank/EMBL data libraries with the following accession numbers: SmCPS1 (KC814639), SmCPS2 (KC814640), SmCPS3 (KC814641), SmCPS4 (KP063138), SmCPS5 (KC814642), and SmKSL2 (KC814643).

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. EI mass spectrum of diterpenes in this study.

• Supplemental Figure S2. The dwarf phenotype caused by down-regulation of SmCPS5.

• Supplemental Figure S3. The intron/exon structure divergence of diTPS genes involved in labdane-related diterpenoid biosynthesis in Arabidopsis, rice, and S. miltiorrhiza.

• Supplemental Figure S4. Homology modeling and molecular docking results show the positive selection sites.

• Supplemental Figure S5. GC-MS analysis of the six mutant enzymes activity of SmCPS1 and SmCPS5.

• Supplemental Table S1. Information of diTPS gene family in S. miltiorrhiza.

• Supplemental Table S2. List of plant diTPS used in this paper.

• Supplemental Table S3. Different changed metabolites obtained from metabolomics analysis of wild-type and SmCPS1-RNAi lines by LC-qTOF-MS.

• Supplemental Table S4. Different changed metabolites obtained from metabolomics analysis of wild-type and SmCPS1-RNAi lines by GC-QqQ-MS.

• Supplemental Table S5. Summary of statistics for detection of positive selection for the CPS group.

• Supplemental Table S6. Information about primers used in this study.

Glossary

CPS

copalyl diphosphate synthase

RNAi

RNA interference

diTPS

diterpene synthase

GGPP

(E,E,E)-geranylgeranyl diphosphate

CPP

copalyl diphosphate

KS

ent-kaurene synthase

LDPP

8α-hydroxy-labd-13E-en-15-yl diphosphate

cDNA

complementary DNA

KSL

kaurene synthase-like

GC

gas chromatography

MS

mass spectrometry

qRT

quantitative reverse-transcription

LC

liquid chromatography

m/z

mass-to-charge ratio

EI

electron impact

qTOF

quadrupole time-of-flight

qQq

triple quadrupole