Comparative Transcriptomics Unravel Biochemical ... · Comparative Transcriptomics Unravel Biochemical Specialization of Leaf Tissues of Stevia for Diterpenoid Production1 Mi Jung

Comparative Transcriptomics Unravel BiochemicalSpecialization of Leaf Tissues of Stevia forDiterpenoid Production1

Mi Jung Kim2, Jingjing Jin2, Junshi Zheng2, Limsoon Wong, Nam-Hai Chua, and In-Cheol Jang*

Temasek Life Sciences Laboratory, National University of Singapore, Singapore 117604 (M.J.K., J.J., J.Z., I.-C.J.);School of Computing, National University of Singapore, Singapore 117417 (J.J., L.W.); Department of BiologicalSciences, National University of Singapore, Singapore 117543 (J.Z., I.-C.J.); and Laboratory of Plant MolecularBiology, Rockefeller University, New York, NY 10065 (N.-H.C.)

ORCID IDs: 0000-0003-1241-5441 (L.W.); 0000-0002-8991-0355 (N.-H.C.); 0000-0001-9408-4273 (I.-C.J.).

Stevia (Stevia rebaudiana) produces not only a group of diterpenoid glycosides known as steviol glycosides (SGs), but also otherlabdane-type diterpenoids that may be spatially separated from SGs. However, their biosynthetic routes and spatial distributionin leaf tissues have not yet been elucidated. Here, we integrate metabolome and transcriptome analyses of Stevia to explore thebiosynthetic capacity of leaf tissues for diterpenoid metabolism. Tissue-specific chemical analyses confirmed that SGs wereaccumulated in leaf cells but not in trichomes. On the other hand, Stevia leaf trichomes stored other labdane-type diterpenoidssuch as oxomanoyl oxide and agatholic acid. RNA sequencing analyses from two different tissues of Stevia provided acomprehensive overview of dynamic metabolic activities in trichomes and leaf without trichomes. These metabolite-guidedtranscriptomics and phylogenetic and gene expression analyses clearly identified specific gene members encoding enzymesinvolved in the 2-C-methyl-D-erythritol 4-phosphate pathway and the biosynthesis of steviol or other labdane-type diterpenoids.Additionally, our RNA sequencing analysis uncovered copalyl diphosphate synthase (SrCPS) and kaurene synthase1 (SrKS1)homologs, SrCPS2 and KS-like (SrKSL), which were specifically expressed in trichomes. In vitro and in planta assays showedthat unlike SrCPS and SrKS1, SrCPS2 synthesized labda-13-en-8-ol diphosphate and successively catalyzed the formation ofmanoyl oxide and epi-manoyl oxide in combination with SrKSL. Our findings suggest that Stevia may have evolved touse distinct metabolic pathways to avoid metabolic interferences in leaf tissues for efficient production of diverse secondarymetabolites.

Stevia (Stevia rebaudiana), a perennial shrub belongingto the Asteraceae family, has been used for centuries inSouth America as a sweetener for herbal teas and foods(De et al., 2013). In addition to being a sweetening agent,Stevia has been used as a cardiotonic for hypertensionand heartburn and also to lower uric acid levels. Forreasons of food and medicine, there has been wide-spread cultivation of this plant throughout Europe,Asia, and North America (Philippe et al., 2014).

The sweetness of Stevia leaves is attributed to a groupof diterpenoid derivatives known as steviol glycosides(SGs), which are 150 to 300 times sweeter than canesugar (Saccharum officinarum). In contrast to artificialsweeteners, SGs are natural and can be used primarilyas a noncalorie sweetener and/or flavor enhancer.Stevia accumulates SGs in its leaves to as high as 20% ofthe dry weight, which are comprised of approximately10 major components in varying levels (Geuns, 2003).All of these compounds share a common backbone, asteviol aglycone, but with a different number of sugarmoieties (Glc, rhamnose, and Xyl; Obtani and Yamasaki,2002; Starratt et al., 2002). The quality and intensity of thetaste are determined by the C-position (C-13 or C-19) ofsteviol being modified and the number of the sugarmoieties (Obtani and Yamasaki, 2002).

As diterpenoid glycosides, SGs are synthesized froma 2-C-methyl-D-erythritol 4-phosphate (MEP) pathwayin plastids, which produces two isoprene units, iso-pentenyl diphosphate (IPP) and dimethylallyl diphos-phate (DMAPP) from pyruvate and glyceraldehyde-3-Pthrough a series of seven enzyme-catalyzed steps (Tottéet al., 2000; Kumar et al., 2012). The MEP pathway isalso important for the synthesis of many other terpe-noids, in particular mono-, di-, and tetraterpenes, be-cause IPP and DMAPP are their common precursors

1 This work was supported by the Temasek Life Sciences Labora-tory (grants to I.-C.J.) and the Singapore National Research Founda-tion (Competitive Research Programme Award no. NRF–CRP8–2011–02 to I.-C.J.).

2 These authors contributed equally to the article.* Address correspondence to [email protected] author responsible for distribution of materials integral to the

findings presented in this article in accordance with the policy de-scribed in the Instructions for Authors (www.plantphysiol.org) is: In-Cheol Jang ([email protected]).

M.J.K. and J.Z. performed the molecular and biochemical experi-ments; J.J. and L.W. performed the RNA-seq and transcriptome anal-yses; N.-H.C. coordinated the RNA-seq and transcriptome analysesand revised the article; I.-C.J. conceived the research, designed theexperiments, and wrote the article.

www.plantphysiol.org/cgi/doi/10.1104/pp.15.01353

2462 Plant Physiology�, December 2015, Vol. 169, pp. 2462–2480, www.plantphysiol.org � 2015 American Society of Plant Biologists. All Rights Reserved. www.plantphysiol.orgon July 13, 2018 - Published by Downloaded from

Copyright © 2015 American Society of Plant Biologists. All rights reserved.

http://orcid.org/0000-0003-1241-5441

http://orcid.org/0000-0002-8991-0355

http://orcid.org/0000-0001-9408-4273

mailto:[email protected]

http://www.plantphysiol.org

mailto:[email protected]

http://www.plantphysiol.org/cgi/doi/10.1104/pp.15.01353


(McGarvey and Croteau, 1995). Once the 5-carbon unitIPP is synthesized in plastids by the MEP pathway, IPPisomerase (IDI) catalyzes the reversible isomerizationof IPP to form DMAPP, and this enzyme might play aregulatory role in determining cellular DMAPP levels(Adam et al., 2002; Hunter, 2007).Among terpenoids, diterpenoids are classically de-

fined by 20-carbon isoprenoids derived from the com-mon precursor geranylgeranyldiphosphate (GGPP),which is catalyzed by GGPP synthase (GGPPS) insideplastids (Zi et al., 2014). Being required for the pro-duction of photosynthetic pigments and phytohormones(e.g., GA3 and abscisic acid), diterpenoid biosynthesisis essential for plant growth and development. Inaddition, plants also produce over 10,000 differentditerpenoids as secondary metabolites, which mayfunction as direct defense compounds against herbi-vores and microorganisms (Zerbe et al., 2013; Zi et al.,2014). The most commonly found version of diterpenoidsis labdane-type compounds such as GA3s, phytoalexins,sclareol, forskolin, taxol, and steviol, many of which are ofsignificant value to human life as biobased pharma-ceuticals, flavors, and fragrances (Zerbe and Bohlmann,2015).The structural diversity of different diterpenoids arises

from divergent evolution of specialized diterpenoid bio-synthetic pathways. These pathways share a commonprecursor GGPP and use a few types of enzymes, in-cluding different classes of diterpene synthases (diTPSs),cytochrome P450-dependent monooxygenases, andvarious modifying transferases (e.g., acyl-, methyl-, orglycosyltranferases), which may be unique for a givenplant species or family (Zerbe et al., 2013; Zi et al., 2014;Zerbe and Bohlmann, 2015). In angiosperms, special-ized diterpenoids are characterized by their commonbiosynthetic origins in the initial dual cyclization and/or rearrangement reactions by a sequential pair of classII and class I diTPSs (Peters, 2010). Class II diTPSscontain a DxDD motif, which is required for theprotonation-initiated cationic cyclization and rear-rangement of GGPP to distinct bicyclic diphosphateintermediates (Peters, 2010; Chen et al., 2011). Sub-sequently, class I diTPSs possessing DDxxD andNSE/DTE motifs for the binding of the substratediphosphate catalyze cleavage of the phosphategroup of intermediates and further cyclize or rear-range the resulting carbocations (Peters, 2010; Chenet al., 2011).In Stevia, SG biosynthesis shares three steps in com-

mon with GA3 biosynthesis pathway for the formationof kaurenoic acid, which requires two diTPSs, copalyldiphosphate synthase (CPS) and kaurene synthase(KS), and one member of the cytochrome P450 family,kaurene oxidase (KO; Richman et al., 1999; Humphreyet al., 2006). Because GA3 is an important phytohor-mone for plant growth and elongation of cells, it is notsurprising that these genes are conserved among higherplants (Hedden and Phillips, 2000). Moreover, an ad-ditional P450 monooxygenase, kaurenoic acid hydrox-ylase (KAH), is involved in the biosynthesis of steviol,

which provides the backbone for all SGs (Brandle andTelmer, 2007). The last stage of SG biosynthesis is theglycosylation of the diterpenoid steviol by UDP-glycosyltransferases, which transfer sugar residuesfrom activated sugars to an aglycone acceptor (Richmanet al., 2005; Brandle and Telmer, 2007).

The chemical complexity of lipophilic componentsproduced in Stevia leaves has been previously examinedby gas chromatography (GC)-mass spectrometry (MS)showing the presence of diverse mono- and sesquiter-penes and fatty acids (Markovi�c et al., 2008). Interest-ingly, apart from kaurene as a diterpenoid hydrocarbonprecursor for GA3s or steviol, other labdane-typediterpenoids such as manoyl oxide, sclareol, austroinulin,and sterebins are also detected in Stevia leaf extracts,which may confer biological and pharmacologicalproperties (Ibrahim et al., 2007; Markovi�c et al., 2008;Cho et al., 2013). However, the biosynthetic pathwaysor terpene synthases for these diterpenoids in Steviahave not yet been reported.

The biosynthesis and accumulation of many plantditerpenoids are restricted to specialized tissues orspecific cell types, which enables plants to effectivelysynthesize specific natural chemicals and to avoidmetabolic interference. For example, a diterpenoidforskolin accumulates in specialized root cork cells ofColeus forskohlii (Pateraki et al., 2014), diterpene resinacids in resin ducts of conifers (Zulak and Bohlmann,2010), and Z-abienol in tobacco (Nicotiana tabacum) leaftrichomes (Sallaud et al., 2012). This cell/tissue-specificaccumulation of diterpenoids may be associated withdifferential gene expression rather than their translo-cation and accumulation in specific cell/tissues. Al-though Stevia genes encoding CPS, KS, and KO relatedto GA3 or steviol biosynthesis have been shown to ex-press in leaf parenchyma (Humphrey et al., 2006), it isunclear where in the leaf tissues SGs are accumulated.In plants, specialized metabolites are stored in the placeof their synthesis, or in some cases, they can be trans-ported to other tissues (Alvarez, 2014).

The experimental approach of combining bioin-formatics and metabolite analysis has facilitated thecharacterization of biosynthetic pathways and relatedgenes for plant secondary metabolism (Bleeker et al.,2011; Zerbe et al., 2013; Jin et al., 2015; Zerbe andBohlmann, 2015). As the Stevia genome has not yetbeen sequenced, only limited sources of gene informa-tion are available from randomly selected ESTs (Brandleet al., 2002). Recently, Chen et al. (2014) reported a RNAsequencing (RNA-seq) data set of three Stevia geno-types comparing expression patterns of previouslyreported genes involved in SG biosynthesis in thesegenotypes, but there were no additional new genesidentified related to theMEP pathway, SG biosynthesis,or the synthesis of other terpenoids. Detailed orenriched transcriptome data sets from different tissuesof terpenoid-producing nonmodel plants not onlyprovide opportunities for the discovery of new genesand enzymes, but also explain the biosynthetic abil-ity of specialized terpene metabolism that may be

Plant Physiol. Vol. 169, 2015 2463

Diterpenoids Biosynthesis in Stevia Leaf Tissues

www.plantphysiol.orgon July 13, 2018 - Published by Downloaded from Copyright © 2015 American Society of Plant Biologists. All rights reserved.


spatially restricted to organs, tissues, or cell types(Zerbe et al., 2013; Jin et al., 2014; Zerbe andBohlmann, 2015).

Here, we performed comparative analysis of RNA-seqtranscriptomes of two different tissues, namely, tri-chomes and leaves without trichomes (leaf-trichomes).Comparison of transcripts between these two tissuesshowed that more than 20% of the assembled unigeneswere differentially expressed in trichomes, whereasgenes specifically expressed in the leaf-trichomes wereless than 12%. Most genes involved in SG biosynthesiswere predominantly expressed in leaf-trichomes whereSGs accumulated. Moreover, we found that SrCPS2 andKS-like (SrKSL), homologs of SrCPS and SrKS1, werespecifically expressed in trichomes. Taken together, ourresults show that comparative transcriptomic analysisalong with metabolite profiling provided useful infor-mation on the distinct biosynthetic pathways for spe-cialized diterpenoids that preferentially accumulate indifferent tissues of the Stevia leaves.

RESULTS

Metabolite Profiles in Stevia Leaves

SGs are known to accumulate mainly in Stevia leaves(Brandle and Rosa, 1992; Kumar et al., 2012); however,it is unclear whether SGs are synthesized and stored inspecialized tissue such as trichomes or oil cells in Stevialeaf tissues. We first investigated trichomes on thesurface of Stevia leaves using scanning electron mi-croscopy (SEM). Figure 1 shows that there were threedifferent types of trichomes: glandular, long non-glandular and short nonglandular. These three types oftrichomes could be distinguished by their differences inmorphology, structure, and size (Fig. 1). Among these,glandular trichomes in other plants have been shown toproduce, store, and secrete large amounts of diversesecondary metabolites such as terpenoids, phenyl-propanoids, and acyl sugars (Glas et al., 2012). To ex-amine if there is any differential distribution of terpenesin different cell/tissues of Stevia leaves, we first purifiedtrichomes (T) from young leaves using a standard glassbead abrasion method (Lange et al., 2000). We also pre-pared leaf tissues stripped of trichomes (leaves minus

trichomes [L-T]) by cold brushing (Wang et al., 2001).Both SEM and light microscopy confirmed the purity ofthe trichome preparation. Note that the preparation stillcontained amix of the three types of trichomes,which aredifficult to separate (Supplemental Fig. S1).

Next, we used HPLC and GC-MS to examine SGcontent and other diterpenoids in T and L-T samples.Figure 2A shows that most SGs were exclusivelyfound in L-T but not in T. In this HPLC analysis, eachSG was identified by comparison with the retentiontime of known standards. Whereas SGs were almostabsent in T, GC-MS analysis of T extract revealed awider range of terpenes, including mono, sesqui-, andditerpenes, compared with the extract derived fromL-T (Fig. 2B). Metabolites identified in T are listed inTable I.

RNA-seq Data of Stevia, de Novo Assembly, andAnnotation of Transcriptome

To compare transcriptomes of different parts ofStevia leaves, we sequenced RNA-seq libraries de-rived from T and L-T samples using an Illumina HiSeq2000. Illumina sequencing runs generated more than160 million high-quality strand-specific single-endreads of 101 bp from each of the two tissues. Thequality of Illumina sequencing outputs was high,as evaluated by FastQC (Supplemental Fig. S2). OurRNA-seq reads were de novo assembled into 53,793nonredundant unigenes, spanning a total of 95 Mbpof sequence with a GC content of 37.3%. The sequencelength of the assembled unigenes was at least longerthan 200 bp. The resulting assembly had a N50 value of1,482 bp. As a result, the assemblies produced 32,336contigs for L-T and 45,918 contigs for T, giving a total of53,793 unigenes (Table II).

To predict accurate gene annotations and to maxi-mize gene annotation percentages, the assembledunigenes were blasted against the National Centre forBiotechnology Information (NCBI) nonredundant pro-tein database and protein databases from Arabidopsis(Arabidopsis thaliana), Vitis vinifera, Glycine max, andOryza sativa. Among the 53,793 unigenes, 40,243(74.8%) unigenes were annotated through BLASTXsearchwith E-value# 1e-3.WemappedRNA-seq reads

Figure 1. Glandular and filamentous tri-chomes of Stevia. A, SEM images of Stevialeaves showing three types of trichomes. a,Glandular trichome. b, Long nonglandular fil-amentous trichome. c, Short nonglandular fil-amentous trichome. B, A detailed view of thethree types of trichomes of Stevia. Scale barsare included on all images.

2464 Plant Physiol. Vol. 169, 2015

Kim et al.


http://www.plantphysiol.org/cgi/content/full/pp.15.01353/DC1



Figure 2. Distribution of diterpenoids in Stevia leaf tissues. Tissue-specific abundance of metabolites was evaluated by ultra high per-formance liquid chromatography (A) and GC-MS (B). A, Comparison of SG content between leaf-trichomes and trichomes. Standardscontain amixture of nine knownSGs. Peak a, Rebaudioside (Reb)D. Peak b, RebA. Peak c, stevioside. Peak d, Reb F. Peak e, RebC. Peak f,dulcoside A. Peak g, rubusoside. Peak h, Reb B. Peak i, steviobioside. mAU, Milliabsorbance units. B, GC traces showing the differencebetween leaf-trichomes and trichomes. The arrow indicates the peak of camphor (10mgmL–1), the internal standard used in the assay. Thepeaks numbered in GC traces were identical to those listed in Table I.

Plant Physiol. Vol. 169, 2015 2465 www.plantphysiol.orgon July 13, 2018 - Published by Downloaded from

Copyright © 2015 American Society of Plant Biologists. All rights reserved.


onto the assembled transcripts to calculate their expres-sion levels using Bowtie (Langmead and Salzberg, 2012).RNA-seq by Expectation-Maximization was used to es-timate the abundance of assembled transcripts and todetermine the expression levels (Li and Dewey, 2011).

Differential Gene Expression between L-T and T

From the RNA-seq data, 32,336 and 45,918 unigeneswere observed to be expressed in L-T and T, respec-tively. The two expression patterns were clearly dis-tinct, as shown by the heat map (Supplemental Fig. S3).

Table I. Chemical composition of trichomes from Stevia

No.a Compounds RTb RIc Formula RAd

min %1 a-Pinene 9.07 922 C10H16 0.132 (+)-Sabinene 9.74 972 C10H16 0.153 b-Pinene 9.84 979 C10H16 1.104 b-Linalool 11.82 1,126 C10H18O 0.175 L-a-Terpineol 13.49 1,186 C10H18O 0.066 b-Elemene 16.82 1,389 C15H24 1.847 Caryophyllene 17.37 1,424 C15H24 2.158 a-Bergamotene 17.45 1,432 C15H24 0.949 a-Humulene 17.89 1,452 C15H24 1.1810 Germacrene D 18.29 1,484 C15H24 1.1011 b-Selinene 18.40 1,489 C15H24 0.2912 Elixene 18.52 1,492 C15H24 1.2713 g-Cadinene 18.74 1,513 C15H24 0.2414 d-Cadinene 18.82 1,524 C15H24 0.1715 (6)-trans-Nerolidol 19.27 1,561 C15H26O 0.9316 Farnesol 19.47 1,620 C15H26O 0.1717 Germacrene D-4-ol 19.69 1,660 C15H26O 0.5618 a-epi-Cadinol 20.58 1,638 C15H26O 0.2719 a-Cadinol 20.79 1,652 C15H26O 0.2820 Isoaromadendrene epoxide 21.08 1,579 C15H24O 0.4021 Ledene oxide-(II) 22.18 1,683 C15H24O 0.0922 [2-(4-Methoxy-phenyl)-[1,3]dioxolan-2-yl]-acetic acid 23.46 1,930 C12H14O5 0.1523 Manoyl oxide 25.32 1,988 C20H34O 1.0724 n-Octadecanol 25.95 2,053 C18H38O 2.3925 Ethanol, 2-[2-[4-(1,1,3,3-tetramethylbutyl)phenoxy]ethoxy]- 26.93 2,126 C18H30O3 0.2826 Stearyl acetate 27.30 2,160 C20H40O2 0.3327 3-Oxomanoyl oxide 27.90 2,219 C20H32O2 2.7328 (+)-Agathadiol 28.80 2,357 C19H30O2 4.8029 (+)-Copaiferic acid 28.96 2,369 C20H32O2 4.6230 4,8,13-Duvatriene-1,3-Diol 29.28 2,400 C21H28O2 0.7231 4-(3-Hydroxy-3-methylpentyl)-3,4a,8,8-tetramethyldecahydro-1-naphthalenol 29.57 2,415 C20H34O2 1.3032 Androstan-17-one, 3-methoxy-16,16-dimethyl-, (3b,5a)- 29.86 2,418 C22H36O2 1.9033 Labda-8(20),13-dien-15-oic acid, 19-hydroxy-, methyl ester, (E)- 30.45 2,429 C21H32O3 29.6434 Pregnane-18,20-diol, (5a)- 31.09 2,449 C21H36O2 5.6335 Allopregnane-3a,20a-diol 31.47 2,475 C21H36O2 12.2036 17-a-Hydroxypregnenolone 31.57 2,480 C20H32O2 9.9737 2-[5-(2,2-Dimethyl-6-methylene-cyclohexyl)-3-methyl-pent-2-enyl]-[1,4]benzoquinone 31.84 2,489 C21H28O2 2.9138 Isophthalic acid, 3,7-dimethyloct-6-enyl propyl ester 31.96 2,492 C21H30O4 3.3039 3-Ethyl-5-(2’-ethylbutyl)octadecane 32.13 2,599 C26H54 0.7640 Allopregnane-7a,11a-diol-3,20-dione 32.36 2,613 C21H32O4 0.3941 Ethanol, 2-[2-[2-[2-[p-(1,1,3,3-tetramethylbutyl)phenoxy]ethoxy]ethoxy]ethoxy]- 32.64 2,676 C22H38O 0.9742 n-Heptacosane 33.06 2,700 C27H56 0.45

aCompound listed in order of elution in an HP-5MS Ultra Inert (UI) column. b Retention time. c Retention indices calculated against C7-C30

n-alkanes on the HP-5MS UI column. d Relative amount. Ratio expressed against the sum of all peaks.

Table II. Overview of the assembly results of RNA-seq

Tissue Total Reads No. of Unigenes No. of Annotations % of Annotations

Leaf-trichomes 75,097,166 32,336 24,041 74.3Trichomes 85,902,623 45,918 32,692 71.2


Kim et al.




Among the unigenes, 3,807 unigenes were preferen-tially expressed in L-T, whereas 10,529 unigenes were Tspecific; the minimum differential expression level wasat least 4 times between the two samples (SupplementalFile S1). Note that more than 20% of the assembledunigenes were specifically expressed in T. Among thedifferential expressed genes, about 40% encoded eitherhypothetical proteins or were unannotated. Thesemight be unique genes with sequences divergent fromthose of other plants but nevertheless essential for tri-chome development in Stevia and/or partly be a resultof misassembled transcripts.To further analyze possible functions of the differ-

entially expressed unigenes from each sample, weassessed their Gene Ontology (GO) classifications byTrinotate (Quevillon et al., 2005). Figure 3 shows theGO terms for the more abundant unigenes in L-T or T.GO terms related to defense responses to fungus andvirus and responses to salicylic acid stimulus werehighly represented in T, reflecting the overall defensefunction of trichomes. However, unigenes from the GOterms of photosynthesis and chlorophyll biosynthesiswere enriched in L-T but not detectable in T. Moreover,other terms for cellular components related to chloro-plast envelope, thylakoid membrane, and stroma wereonly found in L-T. These results were not unexpectedbecause the L-T sample contained mesophyll cellswhere photosynthesis takes place. At the same time,our results showed that the T sample preparation wasnot contaminated by leaf tissues because Stevia tri-chomes are not implicated in photosynthesis by GOannotation. Although genes encoding enzymes andproteins related to photosynthesis are significantlyexpressed in glandular trichomes of tomato (Solanumlycopersicum) and tobacco (Harada et al., 2010; Cuiet al., 2011), they are not expressed in peppermint(Mentha3 piperita) glandular trichomes (Lange et al.,2000), indicating species differences between trichometypes.Because SGs were found in the L-T sample (Fig. 2A),

we hypothesized that the 3,807 unigenes that werepreferentially expressed in this sample may provide animportant resource for the identification of uniquegenes involved in SG biosynthesis and its regulation.

Expression and Molecular Analysis of the MEP Pathwayand IDI Genes

Fourteen genes encoding seven enzymes in the MEPpathway and IDI were identified from our SteviaRNA-seq data sets, including genes for four 1-deoxy-D-xylulose 5-phosphate synthases (DXSs), two each of1-deoxy-D-xylulose 5-phosphate reductoisomerase (DXR),1-hydroxy-2-methyl-2-(E)-butenyl 4-diphosphate reductase(HDR), and IDI, and one each of 2-C-methyl-D-erythritol4-phosphate cytidylyltransferase (MCT), 4-(cytidine59-diphospho)-2-C-methyl-D-erythritol kinase (CMK),2-C-methyl-D-erythritol 2,4-cyclodiphosphate synthase(MDS), and 4-hydroxy-3-methylbut-2-enyl diphosphate

synthase (HDS; Supplemental Fig. S4). We found fourDXS genes (designated as SrDXS1–SrDXS4) containingthe complete open reading frame (ORF). BecauseSrDXS that is designated here as SrDXS4 has beenpreviously reported (accession no. AJ429232; Tottéet al., 2003), three additional DXS genes, SrDXS1,SrDXS2, and SrDXS3 are newly identified in this study.Among four SrDXS proteins, SrDXS1 is the onlymember of the DXS1 clade, which is probably involvedin primary metabolism, whereas SrDXS2 and SrDXS4,which belong to the DXS2 clade, may be related tosecondary metabolism (Fig. 4A; Supplemental Fig. S5;Walter et al., 2002; Phillips et al., 2007). Some plantspecies have a third DXS protein cluster referred to asthe DXS3 clade, which has been proposed to participatein the synthesis of some products essential for plantsurvival (Cordoba et al., 2009, 2011). We found thatSrDXS3 was a member of the DXS3 clade (Fig. 4A). Weconfirmed differential expression of SrDXS genes byquantitative real-time (qRT)-PCR using the L-T andT samples. Figure 4B shows that SrDXS1 transcriptswere 4 to 5 times more abundant in L-T than T. On theother hand, SrDXS4, which was previously reported(Totté et al., 2003), was predominantly expressed inT. Interestingly, SrDXS2 transcripts were rarely de-tectable in L-T but were highly expressedmore than 200times in T.

From our RNA-seq data set, we were able to identifytranscripts for all previously reported DXR, MCT,CMK, MDS, HDS, and HDR (Totté et al., 2003; Kumaret al., 2012). Moreover, we found transcripts for anadditional copy of DXR and HDR, designated here asDXR2 and HDR2, respectively. Note that SrDXR1(accession no. AJ429233) and SrHDR1 (accession no.DQ269451) are previously reported genes (Totté et al.,2003; Kumar et al., 2012). Although SrDXR1 andSrDXR2 share 89.4% high amino acid identity(Supplemental Fig. S6), the expression pattern ofSrDXR1 and SrDXR2was clearly distinct in leaf tissues.Figure 4C showed that SrDXR1 expression was slightlyhigher in T compared with L-T by about 2.5-fold,whereas SrDXR2 transcripts were 20-fold more abun-dant in T. The expression of other intermediate genessuch as SrMCT, SrCMK, SrMDS, and SrHDS wasdetected in both L-T and T samples (Fig. 4C). All genesexcept SrMDS were slightly more expressed in T byabout 2- to 4-fold, but SrMDS expression was compa-rable in both samples.

SrHDR2 showed 86.7% amino acid identity withSrHDR1 (Supplemental Fig. S7). However, genes forthe two related proteins showed opposite expressionpatterns in L-T and T samples, but the differential ex-pression level of these genes was not significant (Fig. 4C).

Transcripts for two IDI genes, SrIDI1 and SrIDI2,were found in our Stevia RNA-seq data set. AlthoughSrIDI1 appeared to be the same as the previouslyreported SrIDI (accession no. DQ989585; Kumar et al.,2012), the protein encoded by SrIDI1 is 64 amino acidslonger than SrIDI at the N terminus, indicating thatSrIDImight be a partial sequence (Supplemental Fig. S8).












Figure 3. GO analysis of more abundantly expressed unigenes in leaf-trichomes (A) or trichomes (B). x Axis, log(1/P value).P value is the hypergeometric test result for each GO term.


Kim et al.



On the other hand, SrIDI2 is, to our knowledge, a newsequence not reported before. Our gene expressionanalysis showed that SrIDI1 was predominantlyexpressed more than 6 times in trichomes (Fig. 4C). By

contrast, SrIDI2 was expressed both in L-T and T.SrIDI1 and SrIDI2 share 68.4% identity at the aminoacid level, and both possess a transit peptide sequencefor plastidic targeting (Supplemental Fig. S8).

Figure 4. Analysis of genes involved in the MEPpathway from Stevia. A, Phylogenetic analysis ofDXSs from Stevia. The maximum likelihood treewas constructed by the MEGA 6 program from analignment of full-length SrDXSs with other plantDXSs. Abbreviations and accession numbers ofproteins are listed in Supplemental Table S2. B,Relative expression levels of SrDXS genes in leaf-trichomes (L-T) and trichomes (T). C, Relative ex-pression levels of SrDXR genes, SrMCT, SrCMK,SrMDS, SrHDS, SrHDR genes, and SrIDI genes inL-T and T. Transcript levels of genes expressed inL-T and T were measured by qRT-PCR. Amplifica-tion of Actin mRNAwas used for normalization.







Characterization of the GGPPS Gene Family

Analysis of our Stevia RNA-seq uncovered threeGGPPS genes designated as SrGGPPS1, SrGGPPS2, andSrGGPPS3. All three SrGGPPSs are clearly distinct fromgeranyl diphosphate synthase or farnesyl diphosphatesynthase family members and have two conservedAsp-rich domains, first aspartate rich motif (DDx9RR)and second aspartate rich motif (DDxxD), which arerequired for IPP and DMAPP substrate binding andcatalysis (Fig. 5A; Supplemental Fig. S9; Liang et al.,2002). SrGGPPS1 transcripts were almost undetectablein the L-T sample (Fig. 5B); rather, this gene was pre-dominantly expressed more than 400 times in T. Bycontrast, transcripts for SrGGPPS2 and SrGGPPS3weremore abundantly expressed in the L-T sample. All threeSrGGPPSs were predicted by the ChloroP program(http://www.cbs.dtu.dk/services/ChloroP) to beplastid localized. SrGGPPS1 and SrGGPPS3 had a pu-tative N-terminal plastid transit peptide sequence of 33and 45 amino acids, respectively, whereas SrGGPPS2was predicted to have only an extra six amino acids atthe N terminus, which may be too short to function as aplastidic transit peptide sequence (Supplemental Fig.S9). This was confirmed by subcellular localization ofeach SrGGPPS- yellow fluorescent protein (YFP) fusionprotein in Nicotiana benthamiana leaves using Agro-bacterium tumefaciens-mediated infiltration. Figure 5Cshows that SrGGPPS1 and SrGGPPS3 were clearly lo-calized in chloroplasts, whereas SrGGPPS2 was dis-tributed throughout the cytosol.

Expression Pattern of Genes for Diterpenoid SteviolBiosynthesis and Their Homologs

In addition to Stevia diTPSs, SrCPS, and SrKS(Richman et al., 1999), analysis of our Stevia RNA-sequncovered a homolog each for both SrCPS and SrKS,which were designated as SrCPS2 and SrKSL, respec-tively. Because both SrCPS2 and SrKSL were initiallyidentified as partial sequences from the RNA-seq dataset, we performed 59-RACE to clone the full-lengthcomplementary DNAs (cDNAs). Phylogenetic com-parison of the sequences of the full-length cDNAs withother related sequences assigned SrCPS2 and SrKSL tothe terpene synthase (TPS)-c and TPS-e/f subfamilies,respectively (Fig. 6A). These two families contain gen-eral or specialized diTPSs (Chen et al., 2011; Caniardet al., 2012).

Figure 5. Analysis of GGPPSs from Stevia. A, Phylogenetic analysis ofGGPPSs fromStevia. Themaximum likelihood treewas constructed by theMEGA 6 program from an alignment of full-length SrGGPPSs with otherplant GGPPSs, geranyl diphosphate synthases (GPPSs), small subunit I(GPPS.SSUIs), small subunit II (GPPS.SSUIIs), large subunit (GPPS.LSU),and farnesyl diphosphate synthases (FPPSs). Abbreviations and accessionnumbers of proteins are listed in Supplemental Table S2. B, Relative ex-pression levels of SrGGPPS genes in leaf-trichomes (L-T) and trichomes (T).

Transcript levels of genes expressed in L-T and T were measured byqRT-PCR. Amplification of Actin mRNA was used for normalization. C,Subcellular localization of SrGGPPs. YFP-fused SrGGPPSs (SrGGPPS1-YFP, SrGGPPS2-YFP, and SrGGPPS3-YFP) were transiently expressed inN. benthamiana leaves by A. tumefaciens-mediated infiltration and visu-alized 3 d postinfiltration using YFP channel of a confocal microscope. Auto,Chlorophyll autofluorescence; YFP, YFP channel image; Light, light micro-scope image;Merged,merged image betweenAuto andYFP. Bars =10mm.


Kim et al.



http://www.cbs.dtu.dk/services/ChloroP





SrCPS2 contains 777 amino acids, which is 10 aminoacids shorter than SrCPS, and shares 56.1% aminoacid identity with the latter (Supplemental Fig. S10).Like SrCPS, SrCPS2 possesses a DxDD motif for

protonation-initiated cyclization of GGPP; this motif isa characteristic of class II TPSs (Supplemental Fig. S10).SrCPS transcripts were approximately 5 times moreabundant in L-T than in T. By contrast, SrCPS2was only

Figure 6. Analysis of diTPSs fromStevia. A, Phylogenetic analysis ofSrCPS2 and SrKSL from Stevia.The maximum likelihood tree wasconstructed by the MEGA 6 pro-gram from an alignment of full-length SrCPS2 and SrKSL withother plant TPSs. Abbreviationsand accession numbers of pro-teins are listed in SupplementalTable S2. B, Relative expressionlevels of Stevia diTPS (SrCPS,SrCPS2, SrKS1, and SrKSL) andP450 (SrKO1 and SrKAH) genes inleaf-trichomes (L-T) and trichomes(T). Transcript levels of genes ex-pressed in L-T and T were mea-sured by qRT-PCR. Amplificationof Actin mRNA was used for nor-malization. C, Subcellular locali-zation of Stevia diTPSs. YFP-fusedSrCPS, SrCPS2, and SrKSL (SrCPS-YFP, SrCPS2-YFP, and SrKSL-YFP) were transiently expressedin N. benthamiana leaves byA. tumefaciens-mediated infiltrationand visualized 3 d postinfiltrationusing the YFP channel of a confo-calmicroscope. Auto, Chlorophyllautofluorescence; YFP, YFP chan-nel image; Light, light microscopeimage; Merged, merged imagebetweenAutoandYFP.Bars=10mm(SrCPS-YFP) and 20 mm (SrCPS2-YFP and SrKSL-YFP).









detectable in T, suggesting that it may function differ-ently in Stevia (Fig. 6B).

SrKSL encoded a 782 amino acid-long protein thatshares 61.2% identity with SrKS1-1 (hereafter referredto as SrKS1). It contains the conserved DDxxD andNSE/DTE motifs of class I TPS for metal-dependentionization of the prenyl diphosphate substrate(Supplemental Fig. S11). Similar to the relation betweenSrCPS and SrCPS2, the expression patterns of SrKS1andKSLwere clearly distinct from each other. Figure 6Bshows that SrKS1 was predominantly expressed morethan 5 times in L-T than T, but SrKSL expression levelwas more than 8 times higher in T compared with L-T.These results suggest that SrCPS and SrKS1 are in-volved in SG biosynthesis in L-T as reported (Richmanet al., 1999), whereas SrCPS2 and/or SrKSL are likely tofunction in the biosynthesis of other diterpenoids intrichomes.

Plant diTPSs, CPS, and KS have a transit peptidesequence at their N termini for plastidic targeting(Toyomasu and Sassa, 2010). Similarly, a putativeN-terminal plastidic transit peptide sequence of 52amino acids for SrCPS2 and 62 amino acids for SrKSLwas also predicted (Supplemental Figs. S10 and S11). Inconfirmation of this prediction, both SrCPS2-YFP andSrKSL-YFPwere found to be localized in chloroplasts ofN. benthamiana leaves (Fig. 6C).

Functional Characterization of SrCPS2 and SrKSL

The different spatial expression patterns betweenSrCPS and SrCPS2 suggest that SrCPS2 plays little rolein the synthesis of copalyl diphosphate (CPP) fromGGPP. To determine the catalytic activity of this diTPSin vitro, N-terminally truncated recombinant proteinsthat lack the putative transit peptide was expressed inEscherichia coli BL21(DE3). 6His-tagged purified re-combinant proteins were used for in vitro assays usingGGPP as the common substrate for diTPS activity. Re-action products were then dephosphorylated by alka-line phosphatase for analysis by GC-MS. Using SrCPSas a control, a single peak for copal-15-ol, which is adephosphorylated form of CPP (Supplemental Fig.S12A; Richman et al., 1999), was detected in the reactionmix. Moreover, SrCPS produced ent-kaurene in a se-quential reaction with SrKS1 (Supplemental Fig. S12, Band C; Richman et al., 1999). However, assay withSrCPS2 produced labda-13-en-8,15-diol (labdenediol)as a major component instead, suggesting that SrCPS2is a labda-13-en-8-ol (or copal-8-ol) diphosphate syn-thase (LPPS) that catalyzes the formation of labda-13-en-8-ol diphosphate (LPP) from GGPP (Fig. 7A). Wealso detected additional minor components, a racemicmixture of manoyl oxide and epi-manoyl oxide fromthe SrCPS2 product profile, but these were previouslysuggested to be the byproducts of a nonenzymaticreaction (Caniard et al., 2012; Zerbe et al., 2013;Pateraki et al., 2014). SrCPS2-catalyzed reaction pro-ducts were identified by comparison to authentic

standards of labdane-type diterpenes based on bothretention time and mass spectra (Fig. 7, A and B;Falara et al., 2010). Moreover, mass spectra of thethree compounds were nearly identical to the pub-lished fragmentation patterns of labdenediol, manoyloxide, and epi-manoyl oxide (Falara et al., 2010;Caniard et al., 2012; Zerbe et al., 2013; Pateraki et al.,2014).

With the newly identified SrKSL being highlyexpressed in trichomes, it seemed possible that SrKSLfunctions downstream of SrCPS2 to cyclize LPP.Therefore, we carried out further in vitro enzymeassays containing both SrCPS2 and SrKSL as enzymesources and GGPP as substrate. Figure 7C andSupplemental Figure S13A show that the coupled en-zyme assay with SrCPS2 and SrKSL produced bothmanoyl oxide and epi-manoyl oxide, but labdenediolwas no longer detected. Similarly, transient expres-sion of SrCPS2 and SrKSL in N. benthamiana usingA. tumefaciens infiltration also yielded manoyl oxideand epi-manoyl oxide in the infiltrated leaves (Fig. 8,A and B). These results suggest that SrKSL catalyzesthe conversion of LPP to manoyl oxide and epi-manoyl oxide. On the other hand, the expression ofSrCPS and SrKS1 in N. benthamiana produced ent-kaurene in the infiltrated leaves, as expected (Fig. 8,C and D).

However, in our in vitro enzyme assay combiningSrCPS2 and SrKS1, the peaks for manoyl oxide andepi-manoyl oxide were also detected (Fig. 7D;Supplemental Fig. S13B). This result was unexpected,because SrKS1 has been identified as a typical kaurenesynthase (Richman et al., 1999). Furthermore, in vivotransient assays in N. benthamiana where SrCPS2 wasinfiltrated alone or in combination with SrKS1 alsoresulted in the detection of these two peaks (Fig. 8A).These results suggest that SrKS1 and other endogenousdiTPSs of N. benthamiana may carry out this catalyticreaction as well. Nevertheless, because SrCPS2 andSrKSL have similar expression patterns in Stevia leaves,it is perhaps likely that they catalyze consecutive stepsin the conversion of GGPP to manoyl oxide and epi-manoyl oxide specifically in trichomes. Then, we alsotested the SrCPS and SrKSL combination. Interestingly,a unique peak similar to (-)-copalol (or ent-copalol)according to the National Institute of Standardsand Technology (NIST) library was detected in thiscombination (Fig. 7E). This compound was also ob-served by coexpression of SrCPS and SrKSL inN. benthamiana (Fig. 8C), confirming a specializedfunction of these two enzymes. However, it was nei-ther ent-copalol nor ent-kaurene, because its retentiontime did not correspond to the peaks for ent-copalolobtained from in vitro assay with SrCPS hydrolyzedby alkaline phosphatase or for ent-kaurene standard(Fig. 7E). Thus, the identification of this unique com-pound remains to be further investigated. Note thatwe only detected this peak when full-length SrKSLincluding the putative transit peptide was used in thereaction mix.


Kim et al.











Figure 7. In vitro characterization of diTPSs from Stevia. A, GC trace of reaction product from in vitro assay of recombinantSrCPS2 with GGPP as a substrate. Reaction product was dephosphorylated by alkaline phosphatase (AP) for GC-MS analysis.Peaks were compared with authentic standards. B, Mass spectra of peaks a, b, and c obtained from authentic standards andreaction product in A. C to E, GC traces of reaction products from coupled in vitro assays of recombinant SrCPS2 plus SrKSL (C),SrCPS2 plus SrKS1 (D), or SrCPS plus SrKSL (E) with GGPPas substrate. Peak a, Labdenediol; peak b, manoyl oxide; peak c, epi-manoyl oxide; peak d, unidentified compound; peak e, ent-copalol; peak f, ent-kaurene;m/z, mass-to-charge ratio; TIC, total ionchromatogram; EIC, extracted-ion chromatogram.





Figure 8. In vivo characterization of diTPSs from Stevia. SrCPS, SrCPS2, SrKS1, and SrKSL were transiently expressed inN. benthamiana leaves by A. tumefaciens-mediated infiltration. The compounds were analyzed 3 d postinfiltration by GC-MS.The resulting peaks were identified through comparison with reference mass spectra of the NIST 2014 library. A and C, GC tracesof hexane extracts fromN. benthamiana leaves expressing SrCPS2, SrCPS2 plus SrKSL, and SrCPS2 plus SrKS1 (A) or SrCPS, SrCPS


Kim et al.



DISCUSSION

Differential Distribution of Diterpenoids in Different CellTypes of Stevia Leaves

Stevia accumulates SGs largely in the leaves ratherthan other tissues such as stems and roots (Brandle andRosa, 1992; Kumar et al., 2012). Here, we found thatmost SGs are detected in L-T, where genes encodingCPS, KS, and KO related to steviol biosynthesis areexpressed (Figs. 2A and 6B; Humphrey et al., 2006).This implies that SGs are stored in the place of theirsynthesis and are not transported to other tissues orcells. On the other hand, Stevia leaf trichomes stored avariety of monoterpenes, sesquiterpenes, and otherlabdane-type diterpenes (Fig. 2B; Table I). In general,compounds that can be detected by GC-MS are rela-tively volatile, indicating that volatile organic com-pounds (VOCs) of Stevia mostly accumulate in thetrichomes. Our result is consistent with other reportsthat VOCs such as terpenoids or phenylpropanoids arestored in glandular trichomes with a storage cavity,e.g., peltate trichomes of mint and capitate trichomes ofbasil (Ocimum basilicum; Turner et al., 2000; Gang et al.,2002). Tobacco glandular trichomes are capable ofsynthesizing diterpenoids such as cis-abienol andduvatrienediol, and the latter compound could befound in Stevia trichomes as well (Fig. 2B; Table I;Keene and Wagner, 1985; Kandra and Wagner, 1988;Guo et al., 1994). These observations suggest that mostVOCs detected by GC-MS might be accumulated inglandular trichomes rather than the other two non-glandular trichome types. Our results demonstratedthat there is differential accumulation of diterpenoids indifferent cell types of the Stevia leaf tissues through abiochemical specialization.

RNA-seq Data of Stevia Leaf Tissues Distinguished GeneMembers Involved in SG or Other Labdane-TypeDiterpenoid Biosynthesis

The diterpene steviol, the aglycone of SGs, is knownto be synthesized through the MEP pathway from twoisoprene units, IPP and DMAPP (Totté et al., 2000, 2003;Supplemental Fig. S4). In addition to eight genes iden-tified previously (Totté et al., 2003; Kumar et al., 2012),our RNA-seq found additionally six genes (SrDXS1,SrDXS2, SrDXS3, SrDXR2, SrHDR2, and SrIDI2) en-coding enzymes involved in theMEP pathway and IDI.Among four SrDXS genes, only SrDXS1 transcriptsexpressed abundantly in the L-T sample (Fig. 4B). Itimplies that SrDXS1 rather than SrDXS4, which waspreviously implicated in SG biosynthesis (Totté et al.,2003), may be involved in SG biosynthesis in the L-T

sample. On the other hand, two genes, SrDXS2 andSrDXS4, encoding proteins of the DXS2 clade werepredominantly expressed in T (Fig. 4B). These resultssuggest that SrDXS2 and SrDXS4might be related to thebiosynthesis of other volatile terpenoids includinglabdane-type diterpenoids in trichome cells.

As mediators of the MEP pathway and an isomeraseof IPP or DMAPP, genes encoding DXR, MCT, CMK,MDS, HDS, HDR, and IDI of Stevia have been previ-ously reported (Totté et al., 2003; Kumar et al., 2012;Supplemental Fig. S4). Most genes except DXR, HDR,and IDI have been found in our RNA-seq data set as asingle copy, and the difference of their expression levelswas less than 2- to 4-fold in both L-T and T samples (Fig.4C; Supplemental Fig. S4), implying these genes func-tion both in primary metabolism and terpenoid bio-synthesis. SrDXR1, SrHDR1, and SrIDI1 have beensuggested to be involved in SG biosynthesis (Kumaret al., 2012). However, our RNA-seq and qRT-PCRanalyses showed that SrHDR1 and SrIDI1 transcriptswere more abundant in the T sample, suggesting theimportant role in the biosynthesis of other volatileterpenoids rather than SGs. Based on the expressionlevels shown in Figure 4C, it is more likely thatSrDXR1, SrHDR2, and SrIDI2 are associated with SGproduction.

In plant genomes, GGPPS is usually encoded by aGGPPS gene family comprising of two to 12 members(Beck et al., 2013). Among the three SrGGPPS genesidentified here, SrGGPPS1 has been previously impli-cated in SG biosynthesis (Kumar et al., 2012). However,SrGGPPS1 was in fact predominantly expressed in tri-chomes. Moreover, it was almost undetectable in theL-T sample where SGs are synthesized (Fig. 5B),which suggests that it is likely to function in the bio-synthesis of other diterpenoids rather than SGs. BothSrGGPPS2 and SrGGPPS3 identified here showedsimilar expression patterns expressing more abun-dantly in the L-T sample, but SrGGPPS2 was locatedin cytosol (Fig. 5, B and C). These results propose thatthe newly identified SrGGPPS3 is solely involved inSG biosynthesis.

It has been reported that SrCPS and SrKS1 are highlyexpressed in mature leaf tissue rather than stems androots (Richman et al., 1999; Kumar et al., 2012). Weconfirmed that both transcripts were more than 5 timesabundant in L-T where SGs are stored than in T (Fig.6B). However, their homologs, SrCPS2 and SrKSL,showed opposite expression patterns in L-T and T (Fig.6B). For this reason, we speculated that they mayfunction differently for the biosynthesis of other diter-penes in trichomes. Phylogenetic analysis further sup-ported a possible functional difference, as it placedSrCPS2 on the same branch as GrTPS1 (Fig. 6A), which

Figure 8. (Continued.)plus SrKS1, and SrCPS plus SrKSL (C). N. benthamiana leaves expressing p19 were used as a control. B and D, Mass spectraobtained from peak a, b, c, and d and reference spectra from the NIST 2014 library. Peak a, Manoyl oxide; peak b, epi-manoyloxide; peak c, ent-kaurene; peak d, unidentified compound; m/z, mass-to-charge ratio; EIC, extracted-ion chromatogram.








was identified as an LPPS from Grindelia robusta (Zerbeet al., 2013). Although SrKSL is closest to SrKS1, it alsofalls into same subgroup containing GrTPS6 (Fig. 6A),which was characterized as a manoyl oxide synthaseworking in conjunction with GrTPS1 (Zerbe et al.,2013). Our in vitro and in vivo enzymatic assays dem-onstrated this speculation, confirming that SrCPS2 is anLPPS (Fig. 7A). Genes encoding LPPS that catalyze theconversion of GGPP to LPP have also been reported in afew other plants. An SsLPPS sequence was detected bydeep 454 sequencing of cDNAs from clary sage (Salviasclarea) calices, where sclareol, a labdane-type diterpenealcohol, is mainly accumulated (Caniard et al., 2012).On the other hand, CfTPS1was highly expressed in theroot cork of C. forskohlii, where another labdane-typediterpenoid forskolin is synthesized (Pateraki et al.,2014). In particular, Cistus creticus Copal-8-ol diphos-phate synthase (CcCLS) from C. creticus was highlyexpressed in leaf trichomes, as was the case withSrCPS2 shown here (Falara et al., 2010). Although theG. robusta GrTPS1 is placed closest to SrCPS2 on thephylogenetic tree, its expression pattern has not yetbeen investigated (Fig. 6A; Zerbe et al., 2013). Our invitro and in planta coupled enzyme assays with SrCPS2

and SrKSL newly identified SrKSL as a manoyl oxidesynthase. Interestingly, SrKS1 was also capable ofproducing two manoyl oxide stereoisomers in combi-nationwith SrCPS2 in vitro aswell as in planta (Figs. 7Dand 8A). We postulate that the conversion of LPP tothese two manoyl oxide stereoisomers may not requirea highly specific enzyme or SrKS1 might have a dualfunction to use both CPP and LPP as substrates. Thelatter hypothesis is made more likely by the recentfinding that SmKSL2 from Salvia miltiorrhiza reactedwith both CPP and LPP to catalyze the formation of ent-kaurene and epi-manoyl oxide in a sequential reactionwith SmCPS5 and SmCPS4, respectively (Cui et al., 2015).

Another coupled enzyme assay with SrCPS andSrKSL produced an unidentified compound in vitro aswell as in planta, which was identified as ent-copalol bythe NIST 2014 library (Figs. 7E and 8C). GC-MS hasproven to be effective in resolving stereoisomers ofmany terpenoid compounds (Armstrong et al., 1996;Huang and Armstrong, 2009). Because the peak ob-served in assays combining SrCPS and SrKSL has adifferent retention timewith a similar mass spectrum asent-copalol, we postulate that the unidentified com-pound could be a stereoisomer of ent-copalol.

Figure 9. The biosynthetic routes ofSGs and other labdane-type diter-penoids fromGGPP in different partsof Stevia leaves. For these distinctpathways, SrGGPPS1 and SrGGPPS3may function separately in trichomesand leaf mesophyll cells, respectively.UGTs, UDP-glycosyltransferases.


Kim et al.



Distinct Functions of diTPSs in Stevia Leaves

The extensive differences in the function of SteviadiTPSs were also clearly reflected in the differences inmetabolites derived from L-T and T samples (Figs. 2and 9). We were able to detect a range of SGs in the L-Tsamples. This result is in agreement with our tran-scriptome and qRT-PCR data showing that SrCPS andSrKS1, which are involved in SG biosynthesis, arehighly expressed in the L-T sample (Figs. 6B and 9). Italso suggests that SGs are mostly stored in the sametissue where they are produced. Instead, Stevia tri-chomes produced a variety of terpenoids rather thanSGs, most of which were volatiles. Manoyl oxide wasdetected among trichome metabolites (Figs. 2B and 9;Table I), correlating with the high transcript levels ofSrCPS2 and SrKSL in trichomes. This further suggeststhat enzymes encoded by these two genes are relevantin the production ofmanoyl oxide (Figs. 6B, 7C, 8A, and9). It is also possible that SrKS1 may have a redundantrole of SrKSL for manoyl oxide biosynthesis in tri-chomes, because it catalyzed the formation of manoyloxide and epi-manoyl oxide in a sequential reactionwith SrCPS2 and its transcripts could be detected in theT sample aswell (Figs. 6B, 7D, 8A, and 9). InC. forskohlii,manoyl oxide can be found in oil bodies of the rootcork cells and is known to be a precursor to forskolin(Pateraki et al., 2014). We suggest that manoyl oxide inStevia may act as a precursor to other as yet unidenti-fied diterpenoids, because forskolin is not found inStevia trichomes (Fig. 9).Many diterpenoids are further modified by varying

degrees of oxidation. These modifications give rise to adiversity of structural classes, which provide furtherfunctional groups for additional modifications, such asglycosylation, acetylation, and methylation. In plants,the cytochrome P450s are mainly responsible for theoxidation of terpenoids (Boutanaev et al., 2015). InStevia, kaurene is oxidized and hydroxylated to formsteviol by KO and KAH, respectively (Humphrey et al.,2006). The latter enzyme performs the first committedstep in SG biosynthesis, providing an aglycone, steviol,which is sequentially glycosylated by a series of UDP-glycosyltransferases in Stevia leaves. We found thatboth P450 genes, SrKO1 and SrKAH, were abundantlyexpressed in the L-T sample compared with the T sam-ple (Fig. 6B). The results of transcript analysis correlatedwith those of chemical analysis. However, we did notfind a new Stevia KAH homolog in our RNA-seq dataset, suggesting a nonredundant role of SrKAH in SGbiosynthesis.Recently, functional diterpenoids with biological ac-

tivities for treatment of a wide range of medical con-ditions have been reported from diverse medicinalplants (Guo et al., 2013; Zerbe et al., 2013; King et al.,2014; Pateraki et al., 2014). Although manoyl oxideitself is known to be an anticancer compound, it isconsidered as the molecular core for a large series ofbioactive derivatives (Pateraki et al., 2014). In additionto manoyl oxide, we found several labdane-type and

oxidized bicyclic diterpenoids such as oxomanoyl oxideand labda-8(20),13-dien-15-oic acid (agatholic acid)from Stevia trichomes (Fig. 2; Table I); the functions ofall these terpenoids are unknown. Querying the tran-scriptome of Stevia trichomes, we found more than 70candidate P450s that were preferentially expressed inthe T samples with a minimum of 4-fold increase inexpression level compared with the L-T sample. Futurework should be directed toward the identification ofcytochrome P450s involved in the enzymatic conver-sion of manoyl oxide to bioactive labdane diterpenoidsin the Stevia trichomes (Fig. 9).

MATERIALS AND METHODS

Plant Materials

Stevia (Stevia rebaudiana) ‘Bertoni’ seeds were germinated on soil and grownin a greenhouse under natural light condition in Singapore. Four-week-oldNicotiana benthamiana plants grown in a greenhouse were used for subcellularlocalization and in vivo characterization of Stevia genes.

Trichome Isolation

Trichomes were isolated using a glass bead abrasion method (Lange et al.,2000; Jin et al., 2014). One-month-old Stevia plants grown in a greenhouse wereused. Young leaves (20–30 g; 1 to 2 cm in length) were placed in a 50-mL testtube containing ice-cold imbibition buffer (1 mM aurintricarboxylic acid, 5 mM

thiourea, and 2 mM dithiothreitol [DTT], pH 6.6) and soaked for 1 h. After re-moving the imbibition buffer, glass beads (425–600 mm, Sigma-Aldrich) wereadded to the tube with an extraction buffer (25 mM 3-[N-Morpholino]-2-hydroxypropanesulfonic acid, pH 6.6, 200 mM D-sorbitol, 10 mM Suc, 0.5 mM

sodium phosphate, 1% [w/v] polyvinylpyrrolidone-40, 0.6% [w/v] methylcellulose, 1 mM aurintricarboxylic acid, 5 mM thiourea, and 2 mM DTT), and thecontent was vigorously vortexed for 1 min. The mixture was first passedthrough a cell strainer (mesh size, 100 mm) to remove debris, and trichomeswere collected using a 20-mm strainer. The collected trichomes were washedseveral times with the same buffer before being frozen in liquid nitrogen forfurther use.

To obtain leaf tissues fromwhich trichomeswere removed (leaf-trichomes), acold-brushing method was used to completely remove trichomes (Wang et al.,2001). The quality of trichomes or leaf-trichomes was monitored under a dis-secting microscope.

RNA Isolation for RNA Sequencing

Total RNA was extracted from the isolated trichomes and leaf-trichomesusing the Spectrum Plant Total RNA Kit (Sigma-Aldrich), which includedribonuclease-free DNaseI treatment to remove genomic DNA. The RNA quantitywas determined with Nanodrop spectrophotometer ND-1000 (Thermo FisherScientific). The RNA integrity number (RIN)wasmeasured to evaluate the RNAqualityusingAgilent 2100bioanalyzer andRNA6000NanoLabchipKit (AgilentTechnologies). RNA samples with a RNA integrity number value of 7, x, 10were processed for RNA-seq by the Rockefeller University Genomics ResourceCenter using a HiSeq 2000 (Illumina).

RNA-seq de Novo Assembly

The RNA libraries prepared using the TruSeq RNA Sample Preparation Kitsv2 set A (RS-122-2001, Illumina) were qualified by the Agilent 2200 TapeStationsystem (Agilent). The qualified libraries were run on single lanes for 100 cycles(strand-specific single-end) on HiSeq 2000 (Illumina). Raw reads were analyzedby FastQC (http://www.bioinformatics.babraham.ac.uk/projects/fastqc/) forthe quality control. The Trinity method was used for de novo assembly of shortsequence reads due to the absence of a reference Stevia genome (Grabherr et al.,2011). The assembled unigenes were annotated based on sequence similaritiesto sequences in the NCBI nr database and also the protein sequence databasesfromArabidopsis (Arabidopsis thaliana), Vitis vinifera,Glycine max, andOryza sativa.




http://www.bioinformatics.babraham.ac.uk/projects/fastqc/


qRT-PCR

qRT-PCR was performed to investigate gene expression pattern in differentStevia tissues. One microgram of total RNA from each sample was used forcDNA synthesis with M-MLV Superscript II (Promega). cDNA samples werequantified by Applied Biosystems 7900HT Fast Real-Time PCR system andApplied Biosystems Power SYBR Green PCR Master Mix (Life Technologies).Fold change values of target gene transcripts were subsequently normalized bydividing the delta threshold cycle values by the delta threshold cycle values ofcontrol gene transcripts. Gene-specific primers for qRT-PCR were designedusing the Primer3 program (http://bioinfo.ut.ee/primer3-0.4.0/) and are listedin Supplemental Table S1. Each PCR product obtained from regular PCR wasverified by sequencing after cloning into pGEM-T Easy vector (Promega). Amelting curve analysis in Applied Biosystems 7900HT Fast Real-Time PCRsystem was performed to assess the specificity of the amplified PCR product.All reactions were carried out in technical triplicates and biological replicates. Amock reaction containing all the reverse transcription-PCR reagents, except thereverse transcriptase, was used as a negative control. Stevia Actin gene wasused as an internal control for normalization.

Isolation of Full-Length ORF of Stevia Genes and VectorConstruction for Agrobacterium tumefaciens-MediatedGene Expression

Full-length ORFs of four SrDXS genes (SrDXS1–SrDXS4), three SrGPPSgenes (SrGPPS1–SrGPPS3), SrCPS, and SrKS1 were amplified by PCR fromStevia leaf-derived cDNA. Primer sets listed in Supplemental Table S1 wereused to amplify gene products that were flanked by attB sites for Gatewaycloning (Invitrogen). Purified PCR products were cloned into pDONR221(Invitrogen) and verified by sequencing. Partial sequences for SrCPS2 andSrKSLwere obtained from Stevia RNA-seq data. Using these partial sequences,we designed RACE primers (Supplemental Table S1) and performed a 59 RACEexperiment using SMARTer RACE cDNA Amplification Kit (Clontech). Theresulting sequences of SrCPS2 and SrKSL obtained by 59 RACEwere confirmedby sequencing after cloning into pDONR221.

For yellow fluorescent protein (YFP)-fusion construct, the pDONR221 cloneharboring each gene was integrated into the destination vector, pBA-DC-YFPexpression vector, which contained aCauliflowermosaic virus 35S promoter and aC terminus in frame YFP by LR Clonase (Invitrogen). The final constructs weretransformed into A. tumefaciens GV3101 by electroporation (Bio-Rad).

Sequence Identification, Multiple Sequence Alignments,and Phylogenetic Analysis

DNA sequences were edited and assembled using DNASTAR Lasergene 8.Sequence alignments were conducted using Biology WorkBench 3.2 (http://seqtool.sdsc.edu/CGI/BW.cgi). Phylogenetic analysis was performed usingthe maximum likelihood method in the Molecular Evolutionary GeneticsAnalysis (MEGA) version 6 program (Tamura et al., 2013). Abbreviations andGenBank accession numbers of proteins in phylograms are listed inSupplemental Table S2.

Subcellular Localization

To determine the subcellular localization of C-terminal in-frame YFP-taggedproteins, three SrGGPPS genes (SrGGPPS1, SrGGPPS2, and SrGGPPS3), SrCPS,SrCPS2, SrKS1, and SrKSL were transiently expressed in leaves of 4-week-oldN. benthamiana plant as described in Jin et al. (2014). Infiltrated plants wereincubated in a growth chamber at 24°C under long-day (16-h-light/8-h-dark)conditions. After 3 d, the leaves were excised, mounted onto slides, and imagedby confocal laser scanning microscopy (Carl Zeiss LSM5 Exciter). The 488- and514-nm lines of an argon laser were used to excite GFP and YFP, respectively.Band pass was set to 500 to 550 nm, and long pass was set to 560 nm. Imageswere recorded and processed using the Zeiss LSM Image Browser.

In Vitro Functional Characterization of Stevia diTPSs

To determine the function of Stevia diTPSs, we performed single or coupledenzyme assays with recombinant proteins. PCR-amplified SrCPS, SrCPS2,SrKS1, and SrKSL cDNAs were inserted into pET28b plasmid (Novagen) to

construct the vectors for the production of recombinant N-terminal poly-His-tagged proteins. The final constructs were transformed into Escherichiacoli BL21(DE3)pLysS (Invitrogen). E. coli culture was treated with 0.2 mM

isopropyl b-D-1-thiogalactopyranoside at 16°C for 12 h to induce His-taggedprotein expression, and induced cells were then harvested by centrifugation.After adding binding buffer (20mMHEPES, pH 7.5, 0.5 MNaCl, 25mM Imidazole,and 5% [v/v] glycerol), one protease inhibitor cocktail tablet per 50 mL (Roche),and 0.1 mg L–1 lysozyme to the cell pellet, cells were subsequently lysed by soni-cation. The cell lysate was centrifuged for 30 min at 14,000g, and the supernatantwas used for purification of the recombinant proteins using nickel-nitrilotriaceticacid agarose affinity chromatography. Proteins bound to nickel-nitrilotriaceticacid agarose (Qiagen) were eluted from the column with elution buffer (50 mM

HEPES, pH 7.5, 0.5 M NaCl, and 250 mM Imidazole) and dialyzed against a di-alysis buffer (50mMHEPES, pH 7.5, 0.1 M KCl, 1mMDTT, and 5% [v/v] glycerol).

In vitro enzyme assay for CPS activity was performed in a final volume of500 mL of reaction buffer (50 mM HEPES pH 7.5, 100 mM KCl, 7.5 mM MgCl2,5 mM DTT, 5% [v/v] glycerol, and 20 mM GGPP) with about 5 mg of the purifiedprotein. The reaction mixture was incubated at 30°C for 2 h. After incubation,10 units of FastAP Thermosensitive Alkaline Phosphatase (Thermo Scientific)was added to hydrolyze CPP. The reaction mixture was overlaid with 500mL ofhexane to trap hexane-soluble copalol and then incubated at 37°C for 4 h. Thedephosphorylated compounds were then extracted two times with 500 mL ofhexane. The hexane fractions were completely dried under a stream of N2 gas atroom temperature and redissolved in 100 mL of hexane for GC-MS analysis.

Coupled enzyme assay was performed using two recombinant proteins todetermine if they catalyze consecutive steps in a catalytic pathway. The enzy-matic reaction and compound extractionwere as described above except for thealkaline phosphatase treatment. Enzyme reaction with SrCPS and SrKS1 wereused as reference (Richman et al., 1999). In vitro enzyme assays were done atleast in biological triplicates.

In Vivo Functional Characterization of Stevia diTPSs

In vivo characterization of diTPSs was carried out by A. tumefaciens-mediatedtransient gene expression in N. benthamiana leaves. Overnight cultures ofA. tumefaciens strains harboring SrCPS, SrCPS2, SrKS1, and SrKSL constructs asdescribed above were used alone or in combinations. Harvested cultures weresuspended in a solution containing 10mMMgCl2, 10mMMES, pH 5.6, and 100mM

acetosyringone. After 1-h incubation at room temperature, the A. tumefaciensmixture was infiltrated into the underside of N. benthamiana leaves using a nee-dleless syringe. The infiltrated plantswere incubated in the growth chamber at 24°Cfor 3 d. After incubation, four to five infiltrated leaves were frozen in liquidnitrogen and homogenized with a prechilled mortar and pestle. About 500 mgof leaf powder was dissolved in 500 mL of ethyl acetate containing 1 mL (10 mgmL–1) of camphor (Sigma) as an internal standard and incubated on a horizontalshaker at 200 rpm for 2 h. After centrifugation of the slush at 13,000g for 10 min,the ethyl acetate upper layer was transferred into a new Eppendorf tube andmixed with 300 mg of anhydrous Na2SO4 to remove water. Following an ad-ditional centrifugation for 1 min to separate the Na2SO4, the extract wastransferred into a 2-mL amber glass GC sample vial for GC-MS analysis. In vivocharacterization of diTPSs was done at least in biological triplicates.

GC-MS Analysis

GC-MS analysis was performed on an Agilent 7890A GC (Agilent Tech-nologies) system and an Agilent Technologies 5975C Inert XL Mass SelectiveDetector equipped with a HP-5MS UI column (30 m 3 0.25 mm 3 0.25 mm).Helium was used as the carrier gas at a flow rate of 1 mL min–1. Samples (5 mL)were injected onto the column at 250°C in the splitless mode, and a temperaturegradient of 8°C min–1 from 50°C (1-min hold) to 300°C (5-min hold) was ap-plied. Data were processed by an MSD ChemStation Data Analysis (AgilentTechnologies). The chemical components of Stevia trichomes were identified bycomparison of their mass spectra with those inNIST 2014 library data of GC-MSand with literature data (Adams, 2007) along with the retention indices asso-ciatedwith a series of n-alkanes (C7-C30)mix standard (Sigma-Aldrich, 49451-U).Camphor (10 mg mL–1) was used as an internal standard. The amount of eachcompoundwas calculated bymeasuring its peak area related to that of a knownamount of camphor. The chemical analysis of Stevia trichomeswas done at leastin biological triplicates. The identified components along with their retentionindices and relative percentage values are listed in Table I. Compound identi-fication by in vitro and in planta assays was done by comparison to authenticstandards, reference spectra from literature, and databases (NIST 2014 library).


Kim et al.


http://bioinfo.ut.ee/primer3-0.4.0/




http://seqtool.sdsc.edu/CGI/BW.cgi

http://seqtool.sdsc.edu/CGI/BW.cgi



HPLC Analysis of SGs from Stevia Leaves

FreshStevia leaves, leaf-trichomes, or isolated trichomeswere frozen in liquidnitrogen and ground with a prechilled mortar and pestle. About 100 mg of thesamples was extracted in 1 mL of water at 70°C for 3 h. The extract wascentrifuged at 1,500g for 15 min, and the supernatant was filtered through a0.45-mm filter. The filtratewas further extracted using a Finisterre SPEColumnC2(Teknokroma), which was pretreated with methanol followed by water. Twomilliliters of the filtrate was loaded on the cartridge and allowed to flow through.The cartridge was washed with water followed by acetonitrile:water (20:80, v/v)and air dried for 3min. Sampleswere then eluted in 1mL ofmethanol:acetonitrile(50:50, v/v) and filtered using a 0.45-mm nylon centrifuge tube (Corning).

HPLC analysis of the samples was carried out on a Shidmadzu NexeraX2 ultra high performance liquid chromatography system using a Shim-packVP-ODS column (250 mm 3 4.6 mm; i.d., 5 mm; and particle size, 4.6 mm) anddetected by a photodiode array detector (SPD-M30Awith high-sensitivity cell).Five microliters of sample was injected, and the elution was performed over 24 minwith a 30% to 80% acetonitrile gradient at a flow rate of 1.0 mLmin–1 accordingto protocol by Shimadzu. Column oven was maintained at 40°C. Peak assign-ment for the absorbance spectrum was based on comparison with elutionprofile of known standards (complete Stevia standards kit, KIT–00019565–005,ChromaDex) at a wavelength of 210 nm. HPLC analysis of Stevia samples wasdone at least in biological triplicates.

The RNA-seq data supporting the result of this article is available in theDNAData Bank of Japan (http://www.ddbj.nig.ac.jp/) with accession numberDRA003752. Nucleotide sequences from this study were submitted to theNCBI/GenBank using BankIt with accession numbers SrDXS1 (KT276229),SrDXS2 (KT276230), SrDXS3 (KT276231), SrDXS4 (KT276232), SrDXR2(KT276233), SrHDR2 (KT276234), SrIDI1 (KT276235), SrIDI2 (KT276236),SrGGPPS2 (KT276237), SrGGPPS3 (KT276238), SrCPS2 (KT276239), and SrKSL(KT276240).

Supplemental Data

The following supplemental materials are available.

Supplemental Figure S1. Stevia tissues for RNA-seq.

Supplemental Figure S2. Quality of strand-specific RNA-seq derived fromtwo Stevia tissues.

Supplemental Figure S3. Heat map of transcript expression in leaf-trichomes and trichomes.

Supplemental Figure S4. Biosynthesis of SGs via the MEP pathway.

Supplemental Figure S5. Comparison of deduced amino acid sequences ofthe four-SrDXS small gene family.

Supplemental Figure S6. Comparison of deduced amino acid sequences ofSrDXRs.

Supplemental Figure S7. Comparison of deduced amino acid sequences ofSrHDRs.

Supplemental Figure S8. Comparison of deduced amino acid sequences ofSrIDIs.

Supplemental Figure S9. Comparison of deduced amino acid sequences ofSrGGPPSs.

Supplemental Figure S10. Comparison of deduced amino acid sequencesof SrCPS and SrCPS2.

Supplemental Figure S11. Comparison of deduced amino acid sequencesof SrKS1-1 and SrKSL.

Supplemental Figure S12. GC-MS analysis of in vitro assay with diTPSsfrom Stevia.

Supplemental Figure S13. Mass spectra obtained from coupled in vitroassays of recombinant SrCPS2 and SrKSL or SrCPS2 and SrKS1.

Supplemental Table S1. Oligonucleotide primers used in this study.

Supplemental Table S2. GenBank accession numbers of proteins used inphylograms.

Supplemental File S1. Differential gene expression between leaf-trichomesand trichomes.

ACKNOWLEDGMENTS

We thank Dr. Angelos K. Kanellis (Aristotle University of Thessaloniki) forthe purified chromatographic fraction from yeast (Saccharomyces cerevisiae) cul-ture transformed with CcCLS used as a standard, Dr. Hiroshi Kawaide (TokyoUniversity of Agriculture and Technology) for ent-kaurene, Dr. Rajani Sarojamfor help in trichomes isolation, Lay-Peng Tan (Agilent Technologies Singapore)for technical assistance in GC-MS analysis, and the Temasek Life Sciences Lab-oratory central facilities for support on SEM and confocal microscopy.

Received August 27, 2015; accepted October 2, 2015; published October 5, 2015.

LITERATURE CITED

Adam P, Hecht S, Eisenreich W, Kaiser J, Grawert T, Arigoni D, Bacher A,Rohdich F (2002) Biosynthesis of terpenes: studies on 1-hydroxy-2-methyl-2-(E )-butenyl 4-diphosphate reductase. Proc Natl Acad Sci USA99: 12108–12113

Adams RP (2007) Identification of Essential Oil Components by GasChromatography/Mass Spectrometry, Ed 4. Allured Publishing Cor-poration, Carol Stream, IL

Alvarez MA (2014) Plant Biotechnology for Health: From SecondaryMetabolites to Molecular Farming. Springer International Publishing,Cham, Switzerland, pp 33–59

Armstrong DW, Zhou EY, Zukowski J, Kosmowska-Ceranowicz B (1996)Enantiomeric composition and prevalence of some bicyclic monoterpenoidsin amber. Chirality 8: 39–48

Beck G, Coman D, Herren E, Ruiz-Sola MA, Rodríguez-Concepción M,Gruissem W, Vranová E (2013) Characterization of the GGPP synthasegene family in Arabidopsis thaliana. Plant Mol Biol 82: 393–416

Bleeker PM, Spyropoulou EA, Diergaarde PJ, Volpin H, De Both MT,Zerbe P, Bohlmann J, Falara V, Matsuba Y, Pichersky E, et al (2011)RNA-seq discovery, functional characterization, and comparison of sesqui-terpene synthases from Solanum lycopersicum and Solanum habrochaitestrichomes. Plant Mol Biol 77: 323–336

Boutanaev AM, Moses T, Zi J, Nelson DR, Mugford ST, Peters RJ, Osbourn A(2015) Investigation of terpene diversification across multiple sequenced plantgenomes. Proc Natl Acad Sci USA 112: E81–E88

Brandle JE, Richman A, Swanson AK, Chapman BP (2002) Leaf Ests fromStevia rebaudiana: a resource for gene discovery in diterpene synthesis.Plant Mol Biol 50: 613–622

Brandle JE, Rosa N (1992) Heritability for yield, leaf:stem ratio and ste-vioside content estimated from a landrace cultivar of Stevia rebaudiana.Can J Plant Sci 72: 1263–1266

Brandle JE, Telmer PG (2007) Steviol glycoside biosynthesis. Phytochem-istry 68: 1855–1863

Caniard A, Zerbe P, Legrand S, Cohade A, Valot N, Magnard JL, Bohlmann J,Legendre L (2012) Discovery and functional characterization of twoditerpene synthases for sclareol biosynthesis in Salvia sclarea (L.) andtheir relevance for perfume manufacture. BMC Plant Biol 12: 119

Chen F, Tholl D, Bohlmann J, Pichersky E (2011) The family of terpenesynthases in plants: a mid-size family of genes for specialized metabo-lism that is highly diversified throughout the kingdom. Plant J 66: 212–229

Chen J, Hou K, Qin P, Liu H, Yi B, Yang W, Wu W (2014) RNA-seq forgene identification and transcript profiling of three Stevia rebaudianagenotypes. BMC Genomics 15: 571

Cho BO, Ryu HW, So Y, Cho JK, Woo HS, Jin CH, Seo KI, Park JC, JeongIY (2013) Anti-inflammatory effect of austroinulin and 6-O-acetyl-austroinulinfrom Stevia rebaudiana in lipopolysaccharide-stimulated RAW264.7 macro-phages. Food Chem Toxicol 62: 638–644

Cordoba E, Porta H, Arroyo A, San Román C, Medina L, Rodríguez-Concepción M, León P (2011) Functional characterization of the threegenes encoding 1-deoxy-D-xylulose 5-phosphate synthase in maize.J Exp Bot 62: 2023–2038

Cordoba E, Salmi M, León P (2009) Unravelling the regulatory mecha-nisms that modulate the MEP pathway in higher plants. J Exp Bot 60:2933–2943

Cui G, Duan L, Jin B, Qian J, Xue Z, Shen G, Snyder JH, Song J, Chen S,Huang L, et al (2015) Functional divergence of diterpene syntheses inthe medicinal plant Salvia miltiorrhiza. Plant Physiol 169: 1607–1618

Cui H, Zhang ST, Yang HJ, Ji H, Wang XJ (2011) Gene expression profileanalysis of tobacco leaf trichomes. BMC Plant Biol 11: 76–86




http://www.ddbj.nig.ac.jp/


















De S, Mondal S, Banerjee S (2013) Introduction to stevioside. In Stevioside:Technology, Applications and Health. John Wiley & Sons, Oxford, pp 1–18

Falara V, Pichersky E, Kanellis AK (2010) A copal-8-ol diphosphatesynthase from the angiosperm Cistus creticus subsp. creticus is a putativekey enzyme for the formation of pharmacologically active, oxygen-containing labdane-type diterpenes. Plant Physiol 154: 301–310

Gang DR, Beuerle T, Ullmann P, Werck-Reichhart D, Pichersky E (2002)Differential production of meta hydroxylated phenylpropanoids insweet basil peltate glandular trichomes and leaves is controlled by theactivities of specific acyltransferases and hydroxylases. Plant Physiol130: 1536–1544

Geuns JM (2003) Stevioside. Phytochemistry 64: 913–921Glas JJ, Schimmel BC, Alba JM, Escobar-Bravo R, Schuurink RC, Kant

MR (2012) Plant glandular trichomes as targets for breeding or engi-neering of resistance to herbivores. Int J Mol Sci 13: 17077–17103

Grabherr MG, Haas BJ, Yassour M, Levin JZ, Thompson DA, Amit I,Adiconis X, Fan L, Raychowdhury R, Zeng Q, et al (2011) Full-lengthtranscriptome assembly from RNA-seq data without a reference ge-nome. Nat Biotechnol 29: 644–652

Guo J, Zhou YJ, Hillwig ML, Shen Y, Yang L, Wang Y, Zhang X, Liu W,Peters RJ, Chen X, et al (2013) CYP76AH1 catalyzes turnover of mil-tiradiene in tanshinones biosynthesis and enables heterologous productionof ferruginol in yeasts. Proc Natl Acad Sci USA 110: 12108–12113

Guo Z, Severson RF, Wagner GJ (1994) Biosynthesis of the diterpene cis-abienol in cell-free extracts of tobacco trichomes. Arch Biochem Biophys308: 103–108

Harada E, Kim JA, Meyer AJ, Hell R, Clemens S, Choi YE (2010) Ex-pression profiling of tobacco leaf trichomes identifies genes for bioticand abiotic stresses. Plant Cell Physiol 51: 1627–1637

Hedden P, Phillips AL (2000) Gibberellin metabolism: new insights re-vealed by the genes. Trends Plant Sci 5: 523–530

Huang K, Armstrong DW (2009) GC–MS analysis of crocetane, phytaneand some of their stereoisomers using cyclodextrin-based stationaryphases. Org Geochem 40: 283–286

Humphrey TV, Richman AS, Menassa R, Brandle JE (2006) Spatial or-ganisation of four enzymes from Stevia rebaudiana that are involved insteviol glycoside synthesis. Plant Mol Biol 61: 47–62

Hunter WN (2007) The non-mevalonate pathway of isoprenoid precursorbiosynthesis. J Biol Chem 282: 21573–21577

Ibrahim NA, El-Gengaihi S, Motawe H, Riad SA (2007) Phytochemicaland biological investigation of Stevia rebaudiana Bertoni; 1-labdane-typediterpene. Eur Food Res Technol 224: 483–488

Jin J, Kim MJ, Dhandapani S, Tjhang JG, Yin JL, Wong L, Sarojam R, ChuaNH, Jang IC (2015) The floral transcriptome of ylang ylang (Cananga odoratavar. fruticosa) uncovers biosynthetic pathways for volatile organic com-pounds and a multifunctional and novel sesquiterpene synthase. J Exp Bot66: 3959–3975

Jin J, Panicker D, Wang Q, Kim MJ, Liu J, Yin JL, Wong L, Jang IC, ChuaNH, Sarojam R (2014) Next generation sequencing unravels the bio-synthetic ability of spearmint (Mentha spicata) peltate glandular tri-chomes through comparative transcriptomics. BMC Plant Biol 14: 292

Kandra L, Wagner GJ (1988) Studies of the site and mode of biosynthesis oftobacco trichome exudate components. Arch Biochem Biophys 265: 425–432

Keene CK, Wagner GJ (1985) Direct demonstration of duvatrienediolbiosynthesis in glandular heads of tobacco trichomes. Plant Physiol 79:1026–1032

King AJ, Brown GD, Gilday AD, Larson TR, Graham IA (2014) Produc-tion of bioactive diterpenoids in the Euphorbiaceae depends on evolu-tionarily conserved gene clusters. Plant Cell 26: 3286–3298

Kumar H, Kaul K, Bajpai-Gupta S, Kaul VK, Kumar S (2012) A compre-hensive analysis of fifteen genes of steviol glycosides biosynthesispathway in Stevia rebaudiana (Bertoni). Gene 492: 276–284

Lange BM, Wildung MR, Stauber EJ, Sanchez C, Pouchnik D, Croteau R(2000) Probing essential oil biosynthesis and secretion by functionalevaluation of expressed sequence tags from mint glandular trichomes.Proc Natl Acad Sci USA 97: 2934–2939

Langmead B, Salzberg SL (2012) Fast gapped-read alignment with Bowtie2. Nat Methods 9: 357–359

Li B, Dewey CN (2011) RSEM: accurate transcript quantification fromRNA-seq data with or without a reference genome. BMC Bioinformatics12: 323

Liang PH, Ko TP, Wang AHJ (2002) Structure, mechanism and function ofprenyltransferases. Eur J Biochem 269: 3339–3354

Markovi�c IS, Đarmati ZA, Abramovi�c BF (2008) Chemical composition ofleaf extracts of Stevia rebaudiana Bertoni grown experimentally in Voj-vodina. J Serb Chem Soc 73: 283–297

McGarvey DJ, Croteau R (1995) Terpenoid metabolism. Plant Cell 7: 1015–1026

Obtani K, Yamasaki K (2002) Methods to improve the taste of the sweetprinciples of Stevia rebaudiana. In AD Kinghorn, ed, Stevia: The GenusStevia. Taylor and Francis, London and New York, pp 138–159

Pateraki I, Andersen-Ranberg J, Hamberger B, Heskes AM, Martens HJ,Zerbe P, Bach SS, Møller BL, Bohlmann J, Hamberger B (2014) Manoyloxide (13R), the biosynthetic precursor of forskolin, is synthesized inspecialized root cork cells in Coleus forskohlii. Plant Physiol 164: 1222–1236

Peters RJ (2010) Two rings in them all: the labdane-related diterpenoids.Nat Prod Rep 27: 1521–1530

Philippe RN, De Mey M, Anderson J, Ajikumar PK (2014) Biotechno-logical production of natural zero-calorie sweeteners. Curr Opin Bio-technol 26: 155–161

Phillips MA, Walter MH, Ralph SG, Dabrowska P, Luck K, Urós EM,Boland W, Strack D, Rodríguez-Concepción M, Bohlmann J, et al(2007) Functional identification and differential expression of 1-deoxy-D-xylulose 5-phosphate synthase in induced terpenoid resin formation ofNorway spruce (Picea abies). Plant Mol Biol 65: 243–257

Quevillon E, Silventoinen V, Pillai S, Harte N, Mulder N, Apweiler R,Lopez R (2005) InterProScan: protein domains identifier. Nucleic AcidsRes 33: W116–W120

Richman A, Swanson A, Humphrey T, Chapman R, McGarvey B, Pocs R,Brandle J (2005) Functional genomics uncovers three glucosyltransfer-ases involved in the synthesis of the major sweet glucosides of Steviarebaudiana. Plant J 41: 56–67

Richman AS, Gijzen M, Starratt AN, Yang Z, Brandle JE (1999) Diterpenesynthesis in Stevia rebaudiana: recruitment and up-regulation of keyenzymes from the gibberellin biosynthetic pathway. Plant J 19: 411–421

Sallaud C, Giacalone C, Töpfer R, Goepfert S, Bakaher N, Rösti S, TissierA (2012) Characterization of two genes for the biosynthesis of the lab-dane diterpene Z-abienol in tobacco (Nicotiana tabacum) glandular tri-chomes. Plant J 72: 1–17

Starratt AN, Kirby CW, Pocs R, Brandle JE (2002) Rebaudioside F, a di-terpene glycoside from Stevia rebaudiana. Phytochemistry 59: 367–370

Tamura K, Stecher G, Peterson D, Filipski A, Kumar S (2013) MEGA6:Molecular Evolutionary Genetics Analysis version 6.0. Mol Biol Evol 30:2725–2729

Totté N, Charon L, Rohmer M, Compernolle F, Baboeuf I, Geuns JMC(2000) Biosynthesis of the diterpenoid steviol, an ent-kaurene derivativefrom Stevia rebaudiana Bertoni, via the methylerythritol phosphatepathway. Tetrahedron Lett 41: 6407–6410

Totté N, Van den Ende W, Damme EJMV, Compernolle F, Baboeuf I,Geuns JMC (2003) Cloning and heterologous expression of early genesin gibberellin and steviol biosynthesis via the methylerythritol phos-phate pathway in Stevia rebaudiana. Can J Bot 81: 517–522

Toyomasu T, Sassa T (2010) Diterpenes. In L Mander, HW Lui, eds,Comprehensive Natural Products II. Chemistry and Biology, Vol 1.Elsevier, Oxford, pp 643–672

Turner GW, Gershenzon J, Croteau RB (2000) Distribution of peltateglandular trichomes on developing leaves of peppermint. Plant Physiol124: 655–664

Walter MH, Hans J, Strack D (2002) Two distantly related genes encoding1-deoxy-D-xylulose 5-phosphate synthases: differential regulation inshoots and apocarotenoid-accumulating mycorrhizal roots. Plant J 31:243–254

Wang E, Wang R, DeParasis J, Loughrin JH, Gan S, Wagner GJ (2001)Suppression of a P450 hydroxylase gene in plant trichome glands en-hances natural-product-based aphid resistance. Nat Biotechnol 19: 371–374

Zerbe P, Bohlmann J (2015) Plant diterpene synthases: exploring modu-larity and metabolic diversity for bioengineering. Trends Biotechnol 33:419–428

Zerbe P, Hamberger B, Yuen MM, Chiang A, Sandhu HK, Madilao LL,Nguyen A, Hamberger B, Bach SS, Bohlmann J (2013) Gene discoveryof modular diterpene metabolism in nonmodel systems. Plant Physiol162: 1073–1091

Zi J, Mafu S, Peters RJ (2014) To gibberellins and beyond! Surveying theevolution of (di)terpenoid metabolism. Annu Rev Plant Biol 65: 259–286

Zulak KG, Bohlmann J (2010) Terpenoid biosynthesis and specializedvascular cells of conifer defense. J Integr Plant Biol 52: 86–97


Kim et al.



Documents

Comparative Transcriptomics Unravel Biochemical ... · Comparative Transcriptomics Unravel Biochemical Specialization of Leaf Tissues of Stevia for Diterpenoid Production1 Mi Jung