Enzymatic synthesis of valerena-4,7(11)-diene by a unique sesquiterpene synthase from the valerian plant (Valeriana officinalis)

Authors


D.-K. Ro, Department of Biological Sciences, University of Calgary, 2500 University Drive NW, Calgary, AB T2N 1N4, Canada
Fax: +1 403 289 9311
Tel: +1 403 220 7099
E-mail: daekyun.ro@ucalgary.ca

Abstract

Valerian (Valeriana officinalis) is a popular medicinal plant in North America and Europe. Its root extract is commonly used as a mild sedative and anxiolytic. Among dozens of chemical constituents (e.g. alkaloids, iridoids, flavonoids, and terpenoids) found in valerian root, valerena-4,7(11)-diene and valerenic acid (C15 sesquiterpenoid) have been suggested as the active ingredients responsible for the sedative effect. However, the biosynthesis of the valerena-4,7(11)-diene hydrocarbon skeleton in valerian remains unknown to date. To identify the responsible terpene synthase, next-generation sequencing (Roche 454 pyrosequencing) was used to generate ∼ 1 million transcript reads from valerian root. From the assembled transcripts, two sesquiterpene synthases were identified (VoTPS1 and VoTPS2), both of which showed predominant expression patterns in root. Transgenic yeast expressing VoTPS1 and VoTPS2 produced germacrene C/germacrene D and valerena-4,7(11)-diene, respectively, as major terpene products. Purified VoTPS1 and VoTPS2 recombinant enzymes confirmed these activities in vitro, with competent kinetic properties (Km of ∼ 10 μm and kcat of 0.01 s−1 for both enzymes). The structure of the valerena-4,7(11)-diene produced from the yeast expressing VoTPS2 was further substantiated by 13C-NMR and GC-MS in comparison with the synthetic standard. This study demonstrates an integrative approach involving next-generation sequencing and metabolically engineered microbes to expand our knowledge of terpenoid diversity in medicinal plants.

Database
The sequences of cDNAs described in this work are available in the GenBank database under the following accession numbers: VoTPS1, JQ437839; VoTPS2, JQ437840

Abbreviations
EF1

elongation factor 1α

EI

electron impact

FPP

farnesyl diphosphate

RI

retention index

sesqui-TPS

sesquiterpene synthase

Introduction

The root extracts of several plant species in the genera Valeriana and Nardostachys (Valerianaceae family) were commonly used by ancient Greek, Roman and Asian societies as sedatives and anxiolytics [1,2]. Although geographically different cultures use their own unique herbal sedatives, it is notable that many mild sedatives are derived from Valeriana or Nardostachys. Examples from these genera include Valeriana wallichii from India, Valeriana edulis from Mexico, Nardostachys jatamansii from the Himalayas, and Valeriana angustifolia from China. Among them, Valeriana officinalis (commonly known as valerian) has become the dominant herbal medicine in Europe and North America as a sedative for the relief of insomnia and anxiety. In Germany, valerian root extract has been approved as a natural sedative. In the USA, valerian extract is also accepted as ‘generally recognized as safe’ by the Food and Drug Administration. In a national survey conducted in 2002, approximately 2 million adults in the USA were reported to have consumed valerian to alleviate insomnia [3]. Certainly, valerian is one of the most commonly used medicinal plants, alongside ginseng, echinacea, St John’s wort, and chamomile.

The active ingredient responsible for the sedative and anxiolytic effects is believed to be present in valerian root, and its chemical identity has been pursued for decades. The major chemicals unique to valerian roots found in valerian oil are mainly composed of monoterpenoids, sesquiterpenoids, iridoid derivatives known as valepotriates, and alkaloids such as actinidine [4–6]. The medicinal properties of these compounds have been evaluated, and multiple lines of evidence have suggested that the C15 sesquiterpenoid valerenic acid is the principal chemical responsible for the observed sedative and anxiolytic effects [7,8].

At the molecular level, in vitro assays have shown that valerenic acid can specifically bind to the β-subunit of 4-aminobutyrate (a neurotransmitter) receptor type A with high affinity (nanomolar level) [9]. Valerenic acid appears to allosterically modulate the affinity of 4-aminobutyrate for the type A receptor [9,10]. Furthermore, it was shown that a point mutation of the type A receptor abolished the binding affinity of valerenic acid for the receptor in vitro, and mice carrying the point-mutated type A receptor did not show the anxiolytic effect of valerenic acid [10]. Intriguingly, the amide derivative of valerenic acid displayed stronger sedative efficacy in both in vitro binding assays and in vivo mouse experiments, offering the potential to develop synthetic drugs mimicking the valerenic acid skeleton [11]. In spikenard (Nardostachys grandiflora), the hydrocarbon valerena-4,7(11)-diene, instead of valerenic acid, accumulates as the major volatile natural product (Fig. 1), and inhalation of this volatile terpene in mice also showed sedative effects [12]. Of particular interest is that the described mode of function and the physiological efficacy of valerenic acid and its derivatives are closely related to those of the common synthetic drugs benzodiazepine and barbiturate, which have been prescribed to cure insomnia and anxiety for decades [13,14]. These data collectively suggest that valerenic acid and its natural and synthetic derivatives can influence neurotransmission and potentially be developed as mood-controlling medicines.

Figure 1.

 Biosynthesis of valerenic acid from FPP in valerian. Carbon numbers are given for valerena-4,7(11)-diene. VLS, valerena-4,7(11)-diene synthase.

Valerenic acid belongs to the subclass of terpenoids known as the sesquiterpenoids. The first step of sesquiterpenoid biosynthesis is catalyzed by sesquiterpene synthases (sesqui-TPSs), which catalyze the conversion of farnesyl diphosphate (FPP) to structurally diverse C15 sesquiterpene hydrocarbons. Such a sesqui-TPS is presumed to catalyze the synthesis of valerena-4,7(11)-diene from FPP in the valerian plant (Fig. 1). Sesqui-TPSs catalyze the cleavage of a diphosphate moiety from FPP, resulting in an unstable terpene carbocation. Subsequently, the catalytic contour of the substrate-binding site in sesqui-TPSs initiates and precisely guides a cascade of carbocation reactions (e.g. double-bond migration, hydride and methyl shifts, or deprotonation) that lead to the synthesis of a single terpene or multiple terpenes. To date, a number of sesqui-TPS genes of medicinal and industrial importance have been identified and characterized [15–18]. With the availability of such sesqui-TPS genes, the production of some sesquiterpenes has been markedly scaled up by the use of engineered microbial platforms [19–21]. Valerena-4,7(11)-diene, valerenic acid and their derivatives could also be produced by similar microbial systems; however, the corresponding sesqui-TPS gene for valerena-4,7(11)-diene biosynthesis has yet to be identified. Here, we report the identification and characterization of valerena-4,7(11)-diene synthase from Valeriana officinalis by the use of next-generation sequencing and metabolic engineering tools. From the reaction products, a possible mechanism for the unique C5–C6-fused sesquiterpene is proposed. This is the first report of a sesqui-TPS that is capable of synthesizing valerena-4,7(11)-diene.

Results and Discussion

Metabolite profiling of valerian root

Valerenic acid is known to accumulate in the root of the valerian plant (V. officinalis) [6]. To ensure the presence of valerenic acid in the V. officinalis root prior to 454 pyrosequencing, the volatile metabolites from V. officinalis root were analyzed by GC-MS. The metabolites from the root were identified by spectral match to the mass spectra library and to the authentic valerenic acid standard. The root sample contained a complex mixture of volatile compounds, but the two most abundant volatiles were identified as bornyl acetate and valerenal (Fig. 2A,D). Although the metabolite composition of valerian varies with the ecotype, these two compounds have been reported as major constituents in valerian root [4,6]. In our initial analysis, valerenic acid was not detected, but the valerenic acid standard also could not be measured at concentrations lower than 100 μm, probably because of its low volatility. To increase the volatility of valerenic acid, the valerian root extract and valerenic acid standard were derivatized (i.e. methylated) and reanalyzed by GC-MS. After derivatization, several new peaks appeared, and the retention index (RI) and mass fragmentation of one later-eluting compound coincided with those of the methylated valerenic acid (Fig. 2C). The valerenic acid content of greenhouse-grown valerian was found to be 0.56 ± 0.02 mg (n = 4) per gram of fresh weight, but no valerenic acid was detected in aerial parts (stem and leaves) of the plant. Metabolite analysis therefore confirmed that valerenal and valerenic acid are major terpenoid constituents of V. officinalis root. This result also suggested that valerena-4,7(11)-diene sesqui-TPS transcripts are specific to root, and are likely to be abundant.

Figure 2.

 GC-MS profile of volatile metabolites from valerian root. (A) Valerenic acid from valerian root can be detected after methylation, and it shows an identical RI and mass fragmentation pattern to those of the authentic standard (B, C). (D) Mass fragmentation of valerenal.

Transcript sequencing and candidate gene isolation

From the same valerian root sample analyzed by GC-MS, cDNA was prepared and subjected to 454 pyrosequencing. This deep transcript sequencing yielded a total of 949 214 reads with an average read length of 347 bp. After removal of repetitive, AT-rich and low-quality sequences, 759 335 high-quality reads were collected and assembled via the magpie bioinformatics platform (Bioinformatics Center, University of Calgary), with the mira algorithm [22]. The mira assembly of 454 pyrosequencing reads generated 55 093 unigenes, covering 42.3 Mbp of the total transcripts. From this sequence dataset, transcripts homologous to the previously reported sesqui-TPSs (e.g. amorpha-4,11-diene, 5-epi-aristolochene and germacrene A synthases) were retrieved by blastx homology search [17,23,24]. Two full-length valerian sesqui-TPS cDNAs were distinctly identified, owing to their abundance in the database, and they are referred to as VoTPS1 and VoTPS2. The read numbers for VoTPS1 and VoTPS2 transcripts constituted 0.03% (ranked 200th) and 0.04% (ranked 259th) of all reads, respectively. The amino acid sequences deduced from the ORFs of VoTPS1 and VoTPS2 shared 75% identity, with 563 and 562 amino acids, respectively (Fig. S1). These two sesqui-TPS clones appear to be unique to valerian, because the blast analysis showed that the closest terpene synthase to VoTPS1 and VoTPS2 was germacrene D synthase from Vitis vinifera, with only 53% amino acid identity. The proteins encoded by VoTPS1 and VoTPS2 did not possess any motif for plastid targeting, implying that they are not diterpene or monoterpene synthases, which are known to be localized to the plastid. Semiquantitative RT-PCR analyses of these two transcripts in valerian root and aerial tissues (stem and leaves) showed predominant expression patterns of VoTPS1 and VoTPS2 in valerian root (Fig. S2). Although a marginal level of VoTPS1 expression was detected in aerial tissues, the level of VoTPS1 transcripts in aerial tissues was found to be 178 ± 6-fold (n = 4) lower than that in root by quantitative PCR. Accordingly, we decided to focus on these two cDNAs because of their transcript abundance and specific expression in root.

Functional screening of VoTPS cDNAs in engineered yeast

We have previously shown that the in vivo assessment of sesqui-TPS clones in yeast allows technically reliable and cost-effective means for the characterization of sesqui-TPS [25]. To evaluate the biochemical activities of the two cDNAs, their ORFs were individually cloned under the Gal10 promoter in the pESC-Leu2d plasmid [26]. VoTPS1 and VoTPS2 cDNAs were then expressed in the EPY300 yeast strain, which was previously engineered to synthesize abundant FPP, an immediate precursor of sesqui-TPS [20,26]. The volatile metabolites synthesized from the transgenic EPY300 strain were sequestered by dodecane overlain in the culture. Accordingly, the dodecane fraction from the culture medium was analyzed by GC-MS, and newly synthesized terpenoids were analyzed in comparison with the electron impact (EI) mass fragmentation pattern and RIs of standards or MS library data. As a result, six new terpenoids unique to VoTPS1-expressing or VoTPS2-expressing yeast were identified (Fig. 3A; Table S1). The yeast expressing VoTPS1 produced predominantly δ-elemene and germacrene D (Fig. 3A, peaks 1 and 2, respectively), and a minor amount of germacrene B (Fig. 3A, peak 3). It was previously reported that germacrene C is unstable and thermally converted to δ-elemene in GC-MS analysis [27], so the appearance of δ-elemene in GC was further assessed at different GC inlet temperatures. Upon injection of the sample at an inlet temperature of 300 °C, a dominant peak of δ-elemene appeared, but this peak completely disappeared when the sample was injected at 150 °C inlet temperature, with a noticeable increase in the baseline (Fig. 3B). Germacrene C could not be detected, as it is continuously converted to the fast-eluting δ-elemene during its migration on the GC column, and thus increased the baseline. The published tomato recombinant germacrene B/germacrene C synthase [27] and germacrene D standards were used to unambiguously determine the chemical identities of the VoTPS1 products (Fig. S3). Therefore, the VoTPS1 clone encodes a multiproduct sesqui-TPS synthesizing germacrene C and germacrene D as major terpenes. On the other hand, the yeast expressing VoTPS2 produced a major terpene (Fig. 3A, peak 4) whose EI fragmentation and RI matched with those of valerena-4,7(11)-diene in the MS library (massfinder 4). As minor products, bicyclogermacrene and alloaromadendrene (Fig. 3A, peaks 5 and 6, respectively) were also detected.

Figure 3.

 Unique terpene compounds synthesized from the yeast expressing VoTPS1 or VoTPS2. (A) Dashed lines indicate the metabolites present in the vector-transformed control. Numbers are unique metabolites identified from VoTPS1-expressing or VoTPS2-expressing yeast. In comparison with the mass fragmentation data from the MS library and RIs, these were identified as follows: peak 1, δ-elemene; peak 2, germacrene D; peak 3, germacrene B; peak 4, valerena-4,7(11)-diene; peak 5, bicyclogermacrene; peak 6, alloaromadendrene. (B) The terpenes from VoTPS1-expressing yeast were analyzed at different inlet temperatures by GC-MS.

Characterization of the VoTPS2 product

To ensure that the compound at peak 4 of Fig. 3A was valerena-4,7(11)-diene, we attempted to purify the compound from the dodecane layer of the culture; however, it was difficult to separate the compound at peak 4 from dodecane, owing to their similar chemical properties. Therefore, instead of dodecane, a hydrophobic resin (Amberlite) was added to the yeast culture, and ∼ 0.7 mg of peak 4 was purified from the resin. When the purified product was analyzed by NMR, the chemical shifts from the 13C-NMR perfectly matched those of the published NMR signals of valerena-4,7(11)-diene [28,29] (Table S2). However, 1H-NMR data overlapped with other contaminants, making accurate integration and signal assignments very difficult. As an alternative approach, a commercially available natural product, valerenic acid, was used to synthesize valerena-4,7(11)-diene by sequential reductions of the C12 carboxylic acid. The chemically synthesized valerena-4,7(11)-diene standard and the purified compound were then analyzed on three different GC columns, including one chiral-selective column (DB1, DB-wax, and cyclodex B). The retention time of the synthesized standard and the compound at peak 4 were identical in all three columns, and they showed identical EI fragmentation patterns (Fig. 4). When the valerena-4,7(11)-diene standard was spiked with the purified compound and the mixture was analyzed by GC, perfectly symmetrical single peaks were obtained from all three columns. Therefore, the 13C-NMR and GC-MS analyses (i.e. EI fragmentation and retention time) confirmed that the enzymatically synthesized compound at peak 4 is valerena-4,7(11)-diene.

Figure 4.

 Validation of the VoTPS2 enzyme product (peak 4) as valerena-4,7(11)-diene. The EI fragmentation of peak 4 is identical to that of the synthesized standard, valerenia-4,7(11)-diene. A DB1-UI-MS column was used under the conditions given in Experimental procedures, and C8–C40 alkane standard was used to calculate the RI.

In vitro characterization of VoTPS1 and VoTPS2

To examine the catalytic properties of the VoTPS1 and VoTPS2 recombinant enzymes, N-terminal His-tagged enzymes were produced in Escherichia coli and purified with Ni-affinity columns. From 1 L of E. coli culture, 0.7 mg of VoTPS1 and 0.3 mg of VoTPS2 were purified with > 95% purity (Fig. S4). When the purified VoTPS1 or VoTPS2 was incubated with FPP in vitro, the major terpenes synthesized were essentially the same as those from transgenic yeast (Fig. 5A,C, peaks 1–6). However, other minor unknown terpenes (Fig. 5A,C, peaks 7 and 8) were also identified from the in vitro assays. When the kinetic properties of VoTPS1 and VoTPS2 were measured by the use of [3H]FPP, hyperbolic FPP saturation kinetics were obtained for both VoTPS1 and VoTPS2 (Fig. 5B,D). The apparent Km and kcat values for VoTPS1 were determined to be 13.7 ± 2.5 μm and 1.0 (± 0.1) × 10−2 s−1 (n = 3), and those for VoTPS2 were 9.5 ± 1.6 μm and 1.3 (± 0.1) × 10−2 s−1 (n = 3). These values are similar to those reported for other sesqui-TPSs [30,31].

Figure 5.

In vitro enzyme assays of VoTPS1 and VoTPS2 recombinant enzymes. (A, C) Purified recombinant VoTPS1 and VoTPS2 were incubated with FPP to synthesize terpenes in vitro, and the respective GC-MS profiles are shown. An additional two compounds identified from the in vitro assays are shown in peaks 7 and 8. The identities of these minor compounds are unknown. (B, D) Plots of VoTPS1 (B) and VoTPS2 (D) reaction rates versus FPP concentration. FPP saturation curves were determined at pH 7.5 and fitted to a single catalytic site Michaelis–Menten model. Calculated Km and kcat values are shown.

Cyclization mechanism of valerenia-4,7(11)-diene

The carbon cores of most sesquiterpenes are formed by 10 carbons. However, valerena-4,7(11)-diene has a unique structural core composed of nine carbons. This nine-carbon core requires ring contraction by formation of a new C–C bond. This unusual reaction is catalyzed by VoTPS2 from Valeriana and possibly other closely related genera, such as Nardostachys. Interestingly, the closest homolog, VoTPS1, exhibiting 75% amino acid identity to VoTPS2, catalyzes the synthesis of germacrene B, germacrene C, and germacrene D, which are commonly found in multiple plant species outside of Valeriana and Nardostachys. We postulated that VoTPS2 recently diverged from VoTPS1 by gene duplication and neofunctionalization in one evolutionary lineage of the Valerianaceae family, and therefore a comparative mechanistic study of VoTPS1 and VoTPS2 may provide insights into the appearance of the unique valerena-4,7(11)-diene from more commonly found terpenes. In addition, structural observations of other minor terpenes (i.e. bicyclogermacrene and alloaromadendrene) produced by VoTPS2 may help to reveal the VoTPS2 mechanism for valerena-4,7(11)-diene synthesis. Taking the above into consideration, we propose one possible mechanism for the synthesis of valerena-4,7(11)-diene by VoTPS2 in Fig. 6. In this scheme, the germacrene bearing a C-6 carbocation is the central precursor for all four major terpenes produced from VoTPS1 and VoTPS2 [i.e. germacrene C, germacrene D, bicyclogermacrene, and valerena-4,7(11)-diene]. Other minor products (i.e. germacrene B and alloaromadendrene) can be coupled to the main reaction framework as shown in Fig. 6. The formation of germacrene B, germacrene C, germacrene D, bicyclogermacrene and alloaromadendrene can be explained by standard carbocation reaction mechanisms, such as hydride shift (indicated as an arrow in b in Fig. 6), deprotonation (a, c, e, g, and i), double-bond migration (d), and protonation (h). The simplest way to link valerena-4,7(11)-diene synthesis to this reaction scheme is to involve new C–C bond formation between C-6 and C-8. This reaction will evoke a unique ring contraction resulting in a nine-carbon core (reaction f). Subsequently, a cascade of deprotonation and reprotonation reactions (j, k, and l) will lead to the formation of valerena-4,7(11)-diene guided by VoTPS2. Further studies are necessary to understand how VoTPS2 promotes the new σ-bond formation between C-6 and C-8 from the germacrene C-6 carbocation while it suppresses all other apparently simpler reactions, such as deprotonation and allylic rearrangement. In summary, given the wide occurrence of germacrene B, germacrene C and germacrene D synthases in nature and the mechanistic link of valerena-4,7(11)-diene to the common carbocation precursor of germacrene B, germacrene C, and germacrene D, the unique VoTPS2 activity appears to have diverged from more common and simpler VoTPS1 activity.

Figure 6.

 A proposed mechanism for valerena-4,7(11)-diene formation catalyzed by VoTPS2 (valerenadiene synthase). The dashed rectangles indicate the VoTPS1 products, and the dashed circles indicate the VoTPS2 products. Letters (a–l) designate distinct carbocation reactions proposed to be catalyzed by VoTPS1 and VoTPS2. The rectangle with a solid line indicates the valerena-4,7(11)-diene biosynthetic mechanism that is proposed to have evolved in the genus Valeriana.

Conclusion

In the present study, the valerian transcriptome generated by 454 pyrosequencing and metabolically engineered yeast were used to identify two sesqui-TPSs (VoTPS1 and VoTPS2), that catalyze the synthesis of germacrene C/germacrene D and valerena-4,7(11)-diene, respectively. The chemical identity of the new terpene synthesized by VoTPS2 was confirmed by GC-MS and NMR analyses with the synthesized standard, and the purified recombinant VoTPS1 and VoTPS2 displayed competent kinetic properties. The formation of an unusual nine-carbon core implies that a distinctive ring-contraction mechanism is programmed in valerena-4,7(11)-diene synthase, and one possible mechanism is proposed in this work. Valerena-4,7(11)-diene and its derivatives (e.g. valerenic acid) are known sedatives, and therefore the new enzyme could be used to produce the terpene backbone for sedative and anxiolytic compounds.

Experimental procedures

Plant cultivation, and metabolite and RNA preparations

V. officinalis seeds were obtained from B&T world seeds (France). Seeds were germinated at 20 °C, and seedlings were grown in the University of Calgary greenhouse. Valerian roots were ground under liquid N2, and 100 mg of the ground tissue was extracted with 1 mL of ethyl acetate. The organic layer was partitioned by centrifugation at 20 000 g for 1 min, diluted 10 times, and analyzed by GC-MS with the conditions described below. Total RNA was isolated according to the published protocol [32], and 7 μg of double-stranded cDNA was prepared according to the supplier’s protocol (Invitrogen). A 454 GS FLX Titanium sequencer was used to sequence valerian cDNAs.

Plasmid construction for yeast expression

Full-length sesqui-TPS cDNAs (VoTPS1 and VoTPS2) were obtained by in silico analysis of the V. officinalis database from the PhytoMetaSyn project at the University of Calgary. VoTPS1 was amplified from valerian root cDNA with the forward primer 5′-AAGTGGATCCGCCATGGAGAGTTGCCTTAGTTTTTC-3′ and the reverse primer 5′-TCCAGCTAGCTTAATACGGAACACTTTCTACTAG-3′. The amplified PCR product was cloned with the TA-cloning kit (Promega, Madison, WI, USA), and subsequently digested with BamHI and NheI. The digested fragment was ligated to the respective multiple cloning site of the pESC-Leu2d plasmid [20,26]. VoTPS2 was amplified with the forward primer 5′-TAATGGATCCGCCATGGAGAGCTGCCTTAGTG ATC-3′ and the reverse primer 5′-AATTGCTAGCTTAACTCGGGATGCTCTCTACTAG-3′, and the resulting amplicon was digested with BamH1 and NheI, and cloned to the respective multiple cloning site of pESC-Leu2d. The two VoTPS1/VoTPS2 expression plasmids and an empty vector control were individually transformed into the EPY300 yeast strain according to the published method [33].

Quantitative transcript analysis

Semiquantitative RT-PCR analyses for VoTPS1 and VoTPS2 were performed with 250 ng of cDNA from V. officinalis root or aerial tissue for 30 cycles, with an annealing temperature of 55 °C. For VoTPS1, the primers used were as follows: forward primer, 5′-CTGTTTACGAACAAGACAAGTCATGCAAC-3′; and reverse primer, 5′-AAGTCACAAAGCGCACCAAATTCAGAACT-3′. For VoTPS2, the primers used were as follows: forward primer, 5′-TATCGTCGAACGATACATTATTAGCATCAG-3′; and reverse primer, 5′-CTTTGTAGAATACATTCATAAAGCATG-3′. Restriction enzyme maps of the resulting 921-bp (VoTPS1) and 1032-bp (VoTPS2) amplicons were obtained with EcoRV and HindIII separately, to confirm their sequence identities. Identical conditions and primers were used with 250 ng of RNA from V. officinalis root or aerial tissues as a negative control to rule out possible genomic DNA contamination. Elongation factor 1α (EF1) was used as an internal control, with the forward primer 5′-GACTGTCACACTTCTCACATTGCC-3′ and the reverse primer 5′-TCTCGACCACCATAGGTTTGGT-3′, and with 5 ng of cDNA from V. officinalis root or aerial tissues and the same PCR conditions as described above. Amplified VoTPS1/VoTPS2 and EF1 fragments were mixed and run in the same lane for visualization. Quantitative PCR of VoTPS1 was performed with the forward primer 5′- TGGTCAAAGCATCAACAATTATCGCT-3′ and the reverse primer 5′-CTTCTTCTTTTGTGGCACCATGTTGT-3′. Ten nanograms of cDNA from V. officinalis root or aerial tissues were used with an annealing temperature of 58 °C. The above-mentioned EF1 primers were also used as the reference gene.

Terpene standards

The natural product, valerenic acid, was purchased from Extrasynthese (France). The methyl ester of valerenic acid was prepared by the use of trimethylsilyl diazomethane in ether with 10% methanol, and was used as a standard for Fig. 2. Valerena-4,7(11)-diene was synthesized from valerenic acid according to the published protocol [28]. A detailed synthesis procedure is given in Doc. S1. (−)-Alloaromadendrene was purchased (Sigma-Aldrich, St. Louis, MO, USA), and germacrene D was a gift from the Keasling laboratory (University of California, Berkeley, CA, USA). The germacrene B/germacrene C synthase cDNA (AF035630) previously described [27] was chemically synthesized and cloned into the pGS-21a plasmid containing both an N-terminal His6 tag and a glutathione-S-transferase tag (Genscript, USA). The recombinant enzyme was purified following the procedure described below, and the His6 and glutathione-S-transferase tags were cleaved with EKMax enterokinase (Invitrogen) before enzyme assays were performed.

In vivo production of terpenoids in yeast

Transgenic yeast were inoculated in 2 mL of Synthetic Complete (SC) medium, without the amino acids His, Met, and Leu, and supplemented with 2% glucose, and the subcultures were cultured overnight at 30 °C and 200 r.p.m. The overnight culture was diluted 25-fold in 50 mL of SC medium lacking His and Leu, and supplemented with 2% galactose, 0.2% glucose, and 2 mm Met. The yeast cultures were overlaid with 5 mL of dodecane (10% of the culture volume) and grown at 30 °C at 200 r.p.m. for 3 days. Dodecane was used to trap any volatile terpenoids released during growth. The yeast cultures were then centrifuged at 3500 g for 5 min, and 1 mL of dodecane was extracted and diluted 100-fold in hexane for subsequent GC-MS analysis.

Heterologous expression in E. coli and protein purification

VoTPS1 and VoTPS2 genes were cloned into the expression vector pH9GW (provided by P. O’Maille, John-Innes Centre, UK), which contains an in-frame N-terminal His9 tag. E. coli (BL21AI) expressing either VoTPS1 or VoTPS2 was cultured at 37 °C until a D600 nm of 0.3 was reached, incubated for 30 min at 4 °C, and then induced with 0.2% arabinose for 24 h at 15 °C. The cultures were centrifuged (at 3500 g for 30 min at 4 °C), and pellets were resuspended in buffer A [25 mm Tris/HCl, pH 7.5, 100 mm NaCl, 10 mm imidazole, 10% (v/v) glycerol, 1 mm phenylmethanesulfonyl fluoride, and 1 mm dithiothreitol]. E. coli cells expressing VoTPS1/VoTPS12 were lysed by sonication. After centrifugation (30 min, 12 000 g, 4 °C), the protein supernatant was cleared with a 0.2-μm filter, and the recombinant enzymes were purified with a Bio-scale Mini-profinity IMAC cartridge (1-mL bed volume; Bio-Rad, Hercules, CA, USA) installed on the Bio-Rad FPLC system. Sample loading was performed at 1 mL·min−1, and the wash and elution steps were performed at 2 mL·min−1. A single wash step (10 mL of 50 mm Tris/HCl, 750 mm KCl, 10% glycerol, 40 mm imidazole) was performed following equilibration. VoTPS1 and VoTPS2 were eluted with a linear gradient (10 mL) of 0–500 mm imidazole in 50 mm Tris/HCl (pH 7.5), 100 mm KCl, 10% (v/v) glycerol, and 1 mm dithiothreitol. Fractions (1 mL) containing either VoTPS1 or VoTPS2 were pooled and concentrated with Amicon Ultra-15 concentrators (30-kDa cut-off). Protein was quantified with the Bradford method.

GC-MS analyses

Organic extracts of yeast strain EPY300 expressing VoTPS1 or VoTPS2 were analyzed with total ion scan and single ion mode (m/z 204) for product identification. Analysis was conducted on an Agilent 6890N GC system coupled to an Agilent 5975B mass spectrometer. Peaks pertaining to the expected parental mass of sesquiterpenes (m/z 204) were analyzed with an authentic standard, the NIST5/Wiley7 mass spectra library, massfinder 4, and/or by comparison with published literature. RIs were calculated by use of an alkane standard (C10–C40), and compared with the values in the literature and the massfinder 4 database. One-microliter samples were injected at an inlet temperature of 250 °C with a flow rate of 1 mL·min−1 helium on DB1-UI-MS and DB-Wax columns (30 m × 250 μm inner diameter and 0.25-μm film thickness). The initial temperature of the program was set to 40 °C, and this was followed by a linear increase of 10 °C·min−1 to a temperature of 220 °C. The Cyclodex B chiral column (30 m × 250 μm inner diameter and 0.25-μm film thickness) was also used to compare the retention behaviors of authentic valera-4,7(11)-diene and the terpene product from VoTPS2. The program used for chiral analysis was as follows: an initial temperature of 50 °C (held for 5 min) followed by a 5 °C·min−1 linear increase to 70 °C and a final ramp of 2.5 °C·min−1 to 200 °C.

Purification and NMR of valerena-4,7(11)-diene

One hundred milliliters of SC medium without His, Met, and Leu, and supplemented with 1.8% galactose and 0.2% glucose, was inoculated with 1.5 mL of overnight culture of the yeast strain EPY300 expressing VoTPS2. After 6 h, the culture was supplied with 1 mm Met and 5 mg of Amberlite XAD-4 (Sigma-Aldrich), which was washed with MeOH prior to use. After 3 days of growth at 30 °C and 180 r.p.m., the Amberlite resin was recovered by filtration, washed with distilled water, and submerged in MeOH. The suspension was extracted three times with 10 mL of hexane, and the combined supernatants were dried over Na2SO4 and evaporated with a gentle N2 stream to 200 μL. The concentrate was separated by silica column chromatography, and eluted with 10 mL of hexane. Fractions of 500 μL were collected and analyzed by GC-MS, and valerena-4,7(11)-diene-containing fractions were pooled and evaporated to dryness with a gentle N2 stream. For NMR analysis, the dried residue (0.7 mg) was dissolved in CDCl3, and spectra were recorded on an Ultra-shield Plus 600-MHz spectrometer (Bruker, Frankfurt, Germany) in CDCl3 at −20 °C. Chemical shifts were reported as parts per million relative to CDCl3.

Enzyme characterization

The assay incubation time and enzyme amount were determined to ensure that the initial velocity of the reaction was linear in the given conditions. [1-3H]FPP (Perkin Elmer; 23 Ci mmol−1) was spiked into 0.1 mm FPP substrates. Assays were carried out in 100-μL volumes with 1.5 μg of purified protein in each assay. The reactions were overlaid with 900 μL of hexane and incubated for 15 min at 30 °C. Reactions were terminated by addition of 100 μL of 0.5 m EDTA and 4 m NaOH, followed by 1 min of vortexing. Reaction mixtures were centrifuged (at 20 000 g for 1 min), and 500 μL of hexane was removed and mixed with 3.5 mL of scintillation cocktail. The total activity of the radioisotope-labeled product was analyzed by liquid scintillation counting (LS6500; Beckman Coulter, Brea, CA, USA). Kinetic properties were determined with substrate concentrations ranging from 0.25 to 50 μm, in triplicate for each concentration. Apparent Vmax and Km values were calculated with the Enzyme Kinetics Module in sigmaplot 12.

Acknowledgements

This work was supported by Genome Canada, Genome Alberta, the Government of Alberta, Genome Quebec, the McGill Innovation Centre, and MDEIE (Ministére du Développement économique, Innovation et Exportation) (to D. K. Ro), the Natural Sciences and Engineering Research Council of Canada, and a Canada Research Chair program (to D. K. Ro and J. C. Vederas). It was also supported by a grant from the Next-Generation BioGreen 21 Program (SSAC grant PJ008108), Rural Development Administration, Republic of Korea (to S. U. Kim and D. K. Ro).

Ancillary