Diterpene synthesis in Stevia rebaudiana: recruitment and up-regulation of key enzymes from the gibberellin biosynthetic pathway


*For correspondence (fax +1 519 663 3454; e-mail brandleje@em.agr.ca).


Stevia rebaudiana Bertoni leaves accumulate a mixture of at least eight different glycosides derived from the tetracyclic diterpene steviol. These natural products taste intensely sweet and have similar biosynthetic origins to those of gibberellic acid (GA). The initial steps leading to the formation of GA result from the two-step cyclization of geranylgeranyl diphosphate (GGDP) to (–)-kaurene via the action of two terpene cyclases (–)-copalyl diphosphate synthase (CPS) and (–)-kaurene synthase (KS). Steviol biosynthesis probably uses the same mechanism although the genes and enzymes from S. rebaudiana that are involved in the cyclization of GGDP have not been characterized. We have isolated both the CPS and KS genes from S. rebaudiana and found that recombinant CPS and KS were catalytically active, suggesting that the CPS and KS genes participate in steviol biosynthesis. The genes coding for CPS and KS are usually present in single copies in most plant species and their expression is normally low and limited to rapidly growing tissues. The KS gene has been duplicated in the S. rebaudiana genome and both the KS and CPS genes are highly expressed in mature leaves, a pattern opposite to that found with GA biosynthesis. This pattern may, at least in part, lead to temporal and spatial separation of GA and steviol biosynthesis and probably helps to prevent over-expression from interfering with normal GA metabolism. Our results show that CPS and KS are part of the steviol glycoside biosynthetic pathway and that Stevia rebaudiana has recruited two genes to secondary metabolism from a highly regulated pathway involved in hormone biosynthesis.


The Paraguayan perennial herb Stevia rebaudiana belongs to the Asteraceae family; its leaves accumulate a mixture of at least eight different glycosides derived from the tetracyclic diterpene steviol ( Brandle et al. 1998 ). These products taste intensely sweet; for example, rebaudioside A has been shown to be up to 320 times sweeter than sucrose on a weight basis ( Phillips 1987). This property has fostered interest in their biosynthesis and metabolism. The rosaceous shrub Rubus sauvissimus and two Stevia species, S. rebaudiana and S. phlebophylla, are the only plant species known to produce steviol glycosides ( Kinghorn et al. 1984 ; Ohtani et al. 1992 ). In S. rebaudiana, these compounds occur almost exclusively in the leaves, with small amounts in the stem and none detectable in the roots ( Brandle & Rosa 1992). Eight sweet steviol glycosides have been isolated and identified from S. rebaudiana leaves ( Kinghorn & Soejarto 1985).

All diterpenes are derived from geranylgeranyl diphosphate (GGDP) and are prevalent throughout the plant kingdom ( Bohlmann et al. 1998 ). Of the polycyclic diterpenes, gibberellic acid (GA) is perhaps the best known and GA biosynthesis is certainly ubiquitous among plants. Steviol is also a polycyclic diterpene and it shares many structural features with GA. The tetracyclic diterpene skeletons of steviol and GA are nearly identical. This observation has led to the prediction that the first steps leading to steviol from GGDP are identical to those described for GA biosynthesis ( Fig. 1) ( Bennett et al. 1967 ; Hanson & White 1968). For GA, GGDP is first converted by protonation-initiated cyclization to (–)-copalyl diphosphate (CDP) by CDP synthase (CPS). The gene encoding CPS has been identified and characterized in Arabidopsis, maize (Zea mays) and pea (Pisum sativum) ( Ait-Ali et al. 1997 ; Bensen et al. 1995 ; Sun & Kamiya 1994). The CPS gene occurs as a single copy in the diploid genomes of Arabidopsis and maize. In pea, a single functional copy and a non-functional pseudogene were found. The fact that CPS mutants in maize, Arabidopsis and pea are not completely deficient in GA has led to speculation that other CPS alleles or CPS-like genes exist in genomes of these species ( Hedden & Kamiya 1997). While the existence of other CPS-like loci is possible in diploid species, none have been identified so far. In polyploid species, gene duplication is common and recent work has demonstrated the presence of at least two differentially regulated CPS genes in tetraploid pumpkin (Cucurbita maxima) ( Smith et al. 1998 ). An Arabidopsis CPS promoter–GUS fusion was used to demonstrate that CPS expression is restricted to certain developmental stages and tissues ( Silverstone et al. 1997 ). In Arabidopsis, expression levels are generally low, but in developing pea seeds expression appears to be higher ( Ait-Ali et al. 1997 ; Silverstone et al. 1997 ). CPS expression may be crucial because the conversion of GGDP to CDP may be the control point for the flow of metabolites into the GA biosynthetic pathway ( Silverstone et al. 1997 ). Next, (–)-kaurene is produced from CDP by an ionization-dependent cyclization catalysed by (–)-kaurene synthase (KS), which is also thought to be a key regulatory step. The genes encoding (–)-kaurene synthase from pumpkin and Arabidopsis have also been isolated and characterized and shown to occur as single copies ( Yamaguchi et al. 1996 , 1998). In pumpkin, KS transcripts were shown to be most abundant in young growing tissues, but were detected in all seedling organs tested ( Yamaguchi et al. 1996 ). In addition to the angiosperm CPS and KS enzymes, two terpene synthases, abietadiene synthase from grand fir (Abies grandis) and kaurene synthase from the fungus Phaeosphaeria have been postulated to catalyse the cyclization of GGDP via CDP. However, the conifer and fungal enzymes are bifunctional and perform both the protonation- and ionization-dependent cyclization steps that are catalysed by separate enzymes in angiosperms ( Kawaide et al. 1997 ; Vogel et al. 1996 ).

Figure 1.

Outline of the elements common to the biosynthesis of GA and steviol from GGDP.

The Arabidopsis gene (GA3), which is responsible for the subsequent oxidation of (–)-kaurene at the C-19 position to (–)-kaurenoic acid, is a novel P450 mono-oxygenase ( Helliwell et al. 1998 ). The GA3 gene is also developmentally regulated in a pattern that resembles CPS and KS. Following the production of (–)-kaurenoic acid, the pathways leading to the steviol glycosides and GA diverge. Steviol is produced by hydroxylation of (–)-kaurenoic acid at the C-13 position ( Kim et al. 1996 ) and GAs by hydroxylation at the C-7 position ( Hedden & Kamiya 1997). Although it has been suggested that steviol can function as a precursor to the C-13 hydroxylated gibberellins and that it has inherent gibberellin-like activity, steviol is rapidly inactivated by glycosylation ( Alves & Ruddat 1979). The two oxygenated functional groups of steviol, the C-19 carboxylate and the C-13 alcohol, provide attachment points for the sugar side chains that determine the identity of the eight different glycosides.

Like other plant secondary metabolites, steviol glycosides probably function in a defensive capacity, perhaps as insect feeding deterrents ( Nanayakkara et al. 1987 ). These glycosides accumulate in leaves at concentrations ranging from 10–30% of the leaf dry weight, and thus a very large part of total metabolism in S. rebaudiana is committed to the synthesis of these structurally complex molecules. In contrast, gibberellins such as GA20 are present in S. rebaudiana leaves at concentrations of 1.2 μg kg–1 fresh weight, over 10 000 times lower than steviol glycosides ( Alves & Ruddat 1979).

The highly active nature of the steviol glycoside biosynthetic pathway and its close relationship to the intermediate steps in gibberellin biosynthesis make it a good system to study the evolution of functional diversity in terpenoid synthases. In this study we have examined two steps common to the synthesis of steviol glycosides and the GAs in order to understand how the biosynthesis of the two compounds might be regulated at the genetic level. We describe the nucleotide sequences of the CPS and KS genes, their catalytic activity, expression patterns and the relative abundance of their transcripts in S. rebaudiana. We have used Southern analysis to examine their genomic organization, in situ hybridization to evaluate expression patterns, and Northern blots and cDNA library representation to estimate expression levels. We found that CPS and KS expression was very high in mature leaves and that this pattern was opposite to that found in most plants. Our results suggest that CPS and KS participate in the synthesis of steviol glycosides and that Stevia rebaudiana has developed a secondary metabolic pathway from a highly regulated primary route previously committed to hormone biosynthesis.


S. rebaudiana CPS and KS genes are homologous to other plant terpene cyclases

To isolate the CPS gene, the degenerate oligonucleotide primers CDPKS and CDPS1 were used to amplify a 660 bp probe from the S. rebaudiana leaf cDNA library. The PCR product was isolated, cloned and five random clones were sequenced. The clones were identical to each other and were similar (> 40% identity), at the amino acid level, to the Arabidopsis and maize CPS genes. One of them, pEKA660, was used as a probe to screen the cDNA library. For KS, the CDPKS and KS1 degenerate oligonucleotide primers gave discrete, multiple PCR products of three different sizes, which were sampled and subjected to an additional round of PCR. Seven clones were sequenced from the second round of PCR and four of those showed similarity (> 40% identity) at the amino acid level to pumpkin KS. The insert from the clone pKS325 was used as a probe to screen the cDNA library.

Approximately 400 positive clones were identified from among 3 × 105 plaques (approximately 0.13% of the total transcripts) when the CPS fragment of pEKA660 was used as a probe to screen the leaf cDNA library. Twenty-two of those plaques were sampled and sizes estimated by PCR. One plaque with a 5′ PCR product approximating the expected size of a full-length sequence was subjected to excision in vivo and the nucleotide sequence of the cDNA was determined (pBK-CDPS1). Restriction enzyme and DNA sequence analysis indicated that the other 21 clones were indentical to pBK-CDPS1. The cDNA was 2590 bp in length, with an open reading frame of 2360 bp (787 amino acids). The CPS from S. rebaudiana is more closely related to plant CPSs than it is to other terpene cyclases and it is also a Tpsc gene family member ( Fig. 2) ( Bohlmann et al. 1998 ). The predicted amino acid sequence from the CPS cDNA was 51% identical to the LS gene from pea, 46% identical to GA1 from Arabidopsis and 41% identical to the AN1 gene from maize ( Fig. 3). The aspartate-rich ‘DXDD’ box that is thought to be involved in protonation and common to all CPS proteins is also conserved in S. rebaudiana CPS ( Wendt et al. 1997 ). One unique feature of S. rebaudiana CPS was the three equally spaced ‘SPAT’ repeats that occur in the N-terminus.

Figure 2.

Dendrogram illustrating the relationships between CPS, KS1-1, KS22-1 and other known terpene cyclases from plants.

Scale bar indicates 0.1 amino acid substitutions per site.

Figure 3.

Sequence alignment of plant CPS protein sequences predicted from cDNA sequences.

The S. rebaudiana CPS protein sequence is compared to the protein sequences derived from the CPS genes from pea (LS), Arabidopsis (GA1) and maize (AN1). Identical residues are shown in black boxes with white letters and the ‘DXDD’ motif is boxed. Sequences used to design degenerate primers are shown by arrows.

The KS fragment of pKS325 was employed to screen the leaf cDNA library. Thirty-three positive clones were found among 3 × 105 plaques (approximately 0.01% of the total transcript) which were further screened using the T3 and CDPKS or T7 and KS1 primers. Four size categories were found and representatives from each size category were partially sequenced from the 5′ end. The second largest, a 3.1 kb cDNA, was found to be full-length (pBK-KS1-1) and was sequenced completely. The deduced amino acid sequence from the 3.1 kb cDNA (KS1-1) was found to share 48% identity with pumpkin KS and 43% identity with a maize EST (csu186) ( Fig. 4). Later searches revealed 42% identity with GA2 of Arabidopsis, which is also a KS gene. The clones were re-evaluated following Northern analysis and a second shorter transcript (KS22-1, 2.8 kb) was recovered. The two KS clones share only 31% identity within the 5′ UTR, but are 99% identical, at the DNA level, within the reading frame and the 3′ UTR. The first 24 bp upstream of the transcriptional start site of KS22-1 are identical to those of KS1-1. The remaining 475 bp of the KS1-1 5′ UTR have no identitity to the remaining 147 bp of the KS22-1 5′ UTR. Identity between KS22-1 and other KS genes was similar to that found with KS1-1. Both KS1-1 and KS22-1 were more closely related to other KSs than to other terpene cyclases, and are members of the Tpse family of terpene cyclases ( Fig. 2) ( Bohlmann et al. 1998 ). The conserved ‘DDXXD’ sequence found in other KS proteins is also present in both KS1-1 and KS22-1 ( Yamaguchi et al. 1996 ).

Figure 4.

Sequence alignment of plant KS protein sequences predicted from cDNA sequences.

The two S. rebaudiana KS sequences, KS1-1 and KS22-1, are compared to those for the KS genes from cucumber (CmKS), Arabidopsis (GA2) and a partial cDNA from maize (csu186). Identical residues are shown in black boxes with white letters and the ‘DDXXD’ motif is boxed. Sequences used to design degenerate primers are shown by arrows.

Recombinant CPS and KS proteins are catalytically active

Translation products of CPS, KS1-1 (data not shown) and KS22-1 cDNAs were generated in E. coli to test for catalytic activity. To confirm CPS activity, the protein extract was incubated with GGDP and the product hydrolysed with alkaline phosphatase. Analysis of a hexane extract of the reaction mixture by gas chromatography–mass spectrometry (GC–MS) with single ion monitoring at m/z 290 (molecular ion of copalol) revealed a peak at the retention time of copalol ( Fig. 5). Identity was confirmed by comparison of mass spectral fragmentation data for the product with that of authentic copalol. Parallel reactions conducted with lysate from cells not carrying the CPS gene showed a peak only at the retention time of geranylgeraniol. A mixture of cell lysates from the CPS and KS1-1 or KS22-1 expression cultures (equal amounts based on total protein content) was similarly incubated with GGDP and then extracted with hexane. GC–MS analysis with selected ion monitoring at m/z 272 (molecular ion of (–)-kaurene) showed a peak at the retention time of (–)-kaurene in both cases ( Fig. 5). Comparison of mass spectral fragmentation data for this product with that of a standard confirmed its identity as (–)-kaurene. In the control, using only the pET30 protein and GGDP, a peak in the mass chromatogram for this product was not observed. On this basis, we conclude that both of the KS genes from S. rebaudiana code for an enzyme that converts CDP to (–)-kaurene. Together, CPS and KS sequentially catalyse the two-step cyclization of GGDP to (–)-kaurene in stevia. This cyclization constitutes the first committed step in the synthesis of both GA and steviol glycosides.

Figure 5.

In vitro analysis of (–)-copalyl diphosphate synthase (CPS) and (–)-kaurene synthase (KS).

GC–MS analysis with selected ion monitoring of diterpene products: (a) geranylgeraniol (m/z 290); (b) copalol (m/z 290); (c) (–)-kaurene (m/z 272); (d) product of incubation of GGDP with CPS (m/z 290); (e) product of incubation of GGDP with CPS and KS22-1 in combination (m/z 272). Similar results were obtained with KS1-1.

KS genes are duplicated in the S. rebaudiana genome

The CPS cDNA was used to probe a Southern blot of genomic DNA digested with three restriction enzymes that did not recognize sites within the probe and two that did. The blot was washed under high-stringency conditions and the number of hybridizing bands from each digest conformed to expectations based on the cDNA sequence. Although this suggests that there is a single CPS gene in the S. rebaudiana genome, the presence of similar loci in the genome cannot be eliminated ( Fig. 6a). The genomic blot was stripped and re-probed with the KS1-1 cDNA, then washed under high-stringency conditions. The Southern analysis showed more fragments than expected from a single gene and the pattern indicated that at least two closely related copies of the KS gene were present in the diploid S. rebaudiana genome ( Fig. 6a).

Figure 6.

Genomic organization and expres- sion analysis of copalyl diphosphate synth- ase (CDPs) and kaurene synthase (KS).

(a) Southern blot analyses of 10 μg of genomic DNA digested with BamHI (B), EcoRI (E), HindIII (H), KpnI (K), and SacI (S) and probed with the CPS and KS1-1 cDNAs.

(b) Northern blot analysis of 10 μg of total RNA from S. rebaudiana leaves and root tips probed with the CPS and KS1-1 cDNAs and a 732 bp actin PCR product amplified from the cDNA library.

(c) Over-exposure of (b).

(d) Northern blot analysis of 10 μg of total RNA from S. rebaudiana from 2-week-old seedling leaves, 6-week-old plant leaves and 21-week-old plant leaves.

S. rebaudiana CPS and KS genes are highly expressed in leaves

Northern analysis of total mRNA was used to compare the relative amount of CPS transcript in leaf tissue to the background levels of CPS expression in root tips ( Fig. 6b). The CPS transcript is readily apparent in the leaf tissue, but could only be detected in the root tips after a long exposure ( Fig. 6c). A Northern blot of total RNA from leaves from 2-week-old seedlings, 6-week-old seedlings and 21-week-old plants was also probed with CPS, revealing a marked increase in transcript levels as the tissues aged ( Fig. 6d). For the Northern analysis of KS expression, identical blots were used as for the CPS expression analyses. The blots of leaf and root total RNA were probed with the KS1-1 cDNA ( Fig. 6b). It was apparent that two different transcripts, one 2.8 kb and the other 3.1 kb, were present. No 3.1 kb KS1-1 expression was detected in the root tips, but the 2.8 kb KS22-1 transcript was visible in the root tip lane following long exposure ( Fig. 6c).

We also examined the expression patterns of CPS and KS1-1 by in situ hybridization. Sense and antisense RNA probes representing the full-length CPS and KS1-1 genes were hybridized to root and leaf thin sections. Both the CPS and KS1-1 genes were expressed throughout the mesophyll and palisade parenchyma of leaf tissues ( Fig. 7). No signal was visible in the epidermis, trichomes or vascular tissue. The signal was not as strong for the KS1-1 antisense probe and the amount of background was higher for the KS1-1 sense probe than was observed for the CPS hybridizations. In root tips there was no difference in hybridization between the sense and antisense RNA probes (data not shown).

Figure 7.

In situ hybridization analyses of S. rebaudiana leaf cross-sections probed with sense (S) and antisense (AS) RNA probes from the CPS (a) and KS1-1 (b) cDNAs.

The epidermis (E), vascular tissue (V), palisade (P) and mesophyll (M) are indicated. Bar = 0.2 mm.


Despite its close relationship to GA, little is known of steviol glycoside biosynthesis at the genetic level. Isolation and sequencing of CPS and the two KS gene family members demonstrates that, at the amino acid level, these enzymes are similar to other diterpene cyclases found in plants. Functional analysis of the proteins shows that they are responsible for the synthesis of (–)-kaurene from GGDP in S. rebaudiana.

The protein sequence predicted by the S. rebaudiana CPS gene is similar to those characterized in other plant species. The ‘DXDD’ motif that is involved in the protonation of the first double bond in GGDP is conserved. The CPS protein sequence is highly conserved across plant species, probably because its role in gibberellin metabolism cannot support changes in sequence that alter product outcome ( Bohlmann et al. 1998 ). Since the pathway for steviol biosynthesis is shared with that of gibberellin, the stevia CPS sequence is also highly conserved relative to other plant CPS proteins. The protein sequences predicted by the KS1-1 and KS22-1 genes are similar to those found in other plant species. The ‘DDXXD’ motif that co-ordinates Mg2+ and acts as a phosphate binding site in 5-epi-aristolochene synthase is conserved in both of the KS family members ( Starks et al. 1997 ). Amino acids near to the active site of 5-EAS are conserved and similar to those occurring in other plant KS enzymes ( Yamaguchi et al. 1998 ). Differences in amino acid sequence between KS1-1 and KS22-1 do not appear to involve any residues used in the formation of the active site, which reflects the conservative type of change expected when genes with essential functions are involved in a developing gene family. The two KS genes are 98% identical at the amino acid level, suggesting that the family is of relatively recent origin compared to the 70% homology found in families thought to be ancient in origin ( Clegg et al. 1997 ; Dornleas et al. 1998 ). Gene families were not anticipated for the diterpene cyclases CPS or KS because of the highly regulated nature of GA metabolism ( Bohlmann et al. 1998 ), although recent evidence has been presented that demonstrates the presence of two quite divergent copies of CPS in pumpkin, not an unexpected outcome in a tetraploid ( Smith et al. 1998 ). The presence of multi-gene families has been reported for the sesquiterpene cyclases, cadinene synthase and 5-EAS, although none of these genes are involved in the synthesis of essential metabolites ( Chen et al. 1995 ; Facchini & Chappell 1992). Duplication of KS in diploid S. rebaudiana is unique thus far. The evolution of a specialized pattern of expression that developed as a consequence of duplication was probably a significant component of the changes that led to the increases in steviol glycoside production that have developed over time.

When CPS and either KS1-1 or KS22-1 synthesized in a bacterial expression system were assayed in the presence of GGDP, only one product was observed: (–)-kaurene. Many monoterpene synthases such as pinene synthase, on the other hand, produce multiple products ( Bohlmann et al. 1997 ). Other enzymes from secondary metabolism such as chalcone synthase have been duplicated and recruited to new biosynthetic roles, which also results in product diversification ( Clegg et al. 1997 ). In cases such as these however, the appearance of new products does not affect basic cellular processes and these changes are less likely to be subjected to the negative selection that would most certainly result when GA metabolism is altered. Although over-expression of CPS and KS probably leads to substantial increases in the concentrations of intermediates, it does not seem to affect growth and development. GA synthesis is probably regulated further along the biosynthetic pathway, which helps to prevent leakage of intermediates into the GA synthesis pathway ( Sun & Kamiya 1994; Yamaguchi et al. 1998 ). Given the level of (–)-kaurene synthesis that occurs in S. rebaudiana leaves, it must also be accompanied by a very strict separation of GA and steviol glycoside biosynthesis via either a developmental or compartmentalization mechanism.

In plants, CPS and KS genes code for enzymes that are usually involved in the biosynthesis of the hormone gibberellin ( Hedden & Kamiya 1997). When they are functioning in hormone biosynthesis, the expression of both the CPS and KS genes is generally low, tissue-specific and restricted to certain developmental stages ( Ait-Ali et al. 1997 ; Yamaguchi et al. 1998 ). In fact, Sun et al. (1992) were unable to use conventional RNA blots for the analysis of GA1 expression because of extremely low transcript abundance. The expression levels of CPS in S. rebaudiana leaves are very high relative to the levels assumed to be associated with GA biosynthesis found in root tips. This expression pattern differs from previously reported CPS and KS expression patterns, but is consistent with the assumption that expression in root tips in S. rebaudiana is limited to that involved with GA biosynthesis ( Silverstone et al. 1997 ; Yamaguchi et al. 1998 ). The presence of high levels of KS and CPS transcript in leaves, but not in root tips, also correlates well with the presence of steviol glycosides in leaves and their absence in roots. Unlike other plant species, CPS expression in fully expanded S. rebaudiana leaves is readily apparent in Northern blots using total RNA. The duplication of the KS gene and the specialized expression pattern that has developed may have helped to further elevate levels of that transcript in leaves.

A CPS promoter–GUS fusion study found most of the GUS activity to reside in rapidly growing young tissues ( Silverstone et al. 1997 ). Similarly, expression of (–)-kaurene oxidase (GA3) in Arabidopsis was highest in young seedling leaves and lowest in mature leaves when it is involved in GA biosynthesis ( Helliwell et al. 1998 ). The pattern of CPS expression in S. rebaudiana is highest in mature leaves rather than young leaves. This is opposite to what is found when the gene is active in GA biosynthesis and clearly demonstrates the extent to which this gene has been recruited for secondary metabolism. In S. rebaudiana, GA biosynthesis must be gated at a later stage in the pathway to control excessive synthesis. That expression levels are highest in mature tissues instead of young rapidly growing tissue raises the possibility of temporal and spatial separation, at least in part, preventing overlap of steviol and GA biosynthesis.

In situ hybridizations revealed that CPS expression was limited to mesophyll tissue, a result consistent with a plastidal localization of these enzymes and a ‘leaves only’ model of glycoside biosynthesis. This demonstrates that expression of CPS and KS in S. rebaudiana is not located in specialized structures such as glandular trichomes or secretory cavities that are common sites for the synthesis of many terpenoid secondary metabolites ( McCaskill & Croteau 1999). The CPS-promoter–GUS fusion study found the CPS promoter to be active in the vascular tissue of fully expanded leaves ( Silverstone et al. 1997 ). We found no expression of CPS in vascular tissue by in situ hybridization, reflecting differences in sensitivity between the two assays and perhaps the spatial separation between the GA and steviol biosynthesis. Transcript levels of both KS1-1 and KS22-1 are also high. Only the KS22-1 transcript is visible in root RNA after over-exposure of total RNA Northern blots, demonstrating the type of specialization and regulation that is common during the development of a gene family. In situ hybridizations revealed that, like CPS, expression was limited to mesophyll tissue. The loss of part of the 5′ UTR as may be the case with KS22-1 may have contributed to the deregulation this gene, since cis-acting elements contained within the UTRs can affect translation substantially ( Boado & Pardridge 1997). However, differences in the size of the 5′ UTR in the two KS genes may be the result of an aberrant splicing event ( Cai et al. 1998 ). Regardless of the mechanism, profound changes in regulation of CPS and KS expression in S. rebaudiana leaves have enabled the synthesis and accumulation of high concentrations of steviol glycosides.

Experimental procedures

Plant material

Fully expanded leaves and roots from five field-grown S. rebaudiana plants were collected, frozen in liquid N2, and stored at –70°C. Root tips were collected from the same plants by removing the plants from the field, washing the roots and harvesting 1 cm of the root tip. The root tips were also quick frozen in liquid N2 and stored at –70°C.

RNA and DNA isolation

Total RNA was isolated from frozen leaf and root tissue (1 cm of adventitious root tips) using Trizol reagent (Life Technologies) as directed by the supplier. Root RNA was stored at –70°C for later use in Northern analysis. Leaf messenger RNA, for cDNA synthesis, was poly(A)+-selected using an oligo(dT)–cellulose mRNA purification kit (Pharmacia) as directed by the supplier. Genomic DNA was isolated from frozen leaf tissue using a modified maize DNA mini-prep ( Dellaporta et al. 1983 ).

Probe generation for library screening

Degenerate oligonucleotide primers were designed based on identical or highly conserved amino acid sequences of known (–)-copalyl diphosphate (Zea mays and Arabidopsis thaliana) and (–)-kaurene synthase (Cucurbita maxima) genes. The downstream facing primer, CDPKS (5′-GCNTAYGAYACNGCNTGGGT-3′), was designed against a common region for both genes and was used in conjunction with either gene-specific upstream facing primer, CPS1 (5′-TCNACNAYCCADATRTGYTCRAA-3′) or KS1 (5′-TCDATCATNCCNGGRAADATDAT-3′). Only the C. maxima KS sequence was available so several KS degenerate primers were designed against regions that were highly conserved but not identical to the published CPS genes. Of these, only KS1 amplified discrete products. The polymerase chain reaction (PCR) was used to amplify fragments from the cDNA library (see below), which were then blunt-end ligated into the cloning vector pCR-Script (Stratagene) and sequenced. Sequence data was submitted to NCBI and searched using the BLASTX algorithm and the non-redundant amino acid database. The inserts with strong homology to known CPS and KS genes were isolated, labelled and used to screen the cDNA library to obtain full-length clones.

cDNA library construction and screening

A leaf cDNA library was made using the ZAP Express cDNA synthesis and predigested vector kits (Stratagene) as directed by the supplier. Hybridizations were carried out in 5× SSC, 2% blocking reagent (Boehringer Mannheim), and 0.5% SDS for 16 h at 65°C followed by three washes with 0.2× SSC and 0.1% SDS at 65°C. Positive plaques were picked and characterized by PCR using the degenerate oligonucleotides that generated the original probe fragment. Plaques containing potential full-length sequences were used in a second round of screening under the same conditions as above. Clones pBK-CDPS1 and pBK-KS1-1 were identified, in vivo excised and sequenced using an ABI377 automated sequencer. Further sequencing of positive plaques identified a second KS-like clone termed pBK-KS22-1.

In vitro functional assay

The BamHI/NotI fragment of pBK-CDPS1 was inserted into the pET30b expression vector to create pET30CDPS1. Due to an in-frame stop codon in the 5′ untranslated region, both KS open reading frames were re-amplified with Pfu DNA polymerase using oligonucleotides including a BglII or a XhoI restriction endonuclease site. The PCR products were digested and inserted into the pET30a expression vector to create pET30KS1-1 and pET30KS22-1. LB cultures (50 ml) were grown to an OD600 of 0.6 at 37°C and induced by adding IPTG to a final concentration of 0.1 m m for each expression construct and the pET30 control plasmid. Expression was carried out for 2 h at 30°C. The cells were pelleted, resuspended in GGDP assay buffer (50 m m potassium phosphate pH 8, 10% glycerol, 2 m m DTT, 5 m m MgCl2) with lysozyme (100 μg ml–1), incubated at 37°C for 15 min and sonicated on ice (3 × 20 sec). Lysate from the samples was centrifuged at 12 000 g and the supernatant removed. The protein concentration of the supernatant was measured using the Bio-Rad microassay system against a bovine serum albumin standard.

The conversion of GGDP to CDP was carried out by incubating 500 μg of pET30CDPS1 protein extract with 10 μg of GGDP (Sigma, St Louis, Missouri, USA) in a final volume of 250 μl GGDP assay buffer for 2 h at 30°C. In order to measure CDP using gas chromatography–mass spectrometry (GC–MS), CDP was hydrolysed with 75 units bacterial alkaline phosphatase at pH 8 for 16 h at 37°C to produce hexane-soluble copalol. GGDP was converted to (–)-kaurene by adding 250 μg of pET30CDPS1 protein extract to 250 μg of pET30KS1-1 or pET30KS22-1 protein extract and 10 μg of GGDP and incubating for 2 h at 30°C. As a control, 500 μg of the pET30 protein extract was incubated with 10 μg of GGDP for 2 h at 30°C and extracted with hexane.

Analysis of geranylgeraniol, copalol and (–)-kaurene

In vitro assay samples were extracted three times with hexane and then centrifuged. The pooled hexane was dried under a stream of N2 at room temperature and re-dissolved in hexane for GC–MS analysis using a Hewlett-Packard 5890 Series II GC with a model 5971A mass selective detector (ionization voltage 70 ev). The GC system was equipped with a J & W DB-1 capillary column, 30 m × 0.25 mm, 0.25 μm film thickness. The GC conditions were: injector temperature, 250°C; flow rate of carrier gas (helium), 1.2 ml min–1. The column temperature was maintained at 80°C for 1 min, then increased at a rate of 15°C min–1 to 245°C, and then at a rate of 5°C min–1 to 300°C. The MS transfer line temperature was set at 290°C. Selected ion monitoring was used for analysis of the diterpenes: geranylgeraniol and copalol, m/z 290; (–)-kaurene m/z 272. Electron impact spectra were also recorded. The samples were compared to authentic standards of geranylgeraniol (American Radiolabelled Chemicals, St Louis, Missouri, USA), and copalol and (–)-kaurene which were gifts from Dr Rodney Croteau.

Southern and Northern analysis

Genomic DNA was digested with BamHI, EcoRI, HindIII, KpnI and SacI according to the manufacturer′s instructions. DNA fragments and glyoxal-treated RNA were separated by agarose gel electrophoresis, transferred to a positively charged nylon membrane and UV cross-linked. Probe hybridizations and detections were carried out following the method of Engler-Blum et al. (1993) .

In situ hybridization

The tips (1 cm long) of adventitious roots and 2–3 mm cross-sections of young leaves were harvested from the field-grown S. rebaudiana plants, and fixed in FAA (36.5–38% formaldehyde:acetic acid:95% ethanol:H2O (10:5:50:35) (v/v)) 12–15 h, and transferred to 70% ETOH for 1 h at room temperature. The fixed materials were dehydrated by treatment with 50% tertiary butyl alcohol (TBA) for 2 h, followed by 70% TBA overnight, 85% TBA, 95% TBA and 100% TBA for 1 h at each concentration, followed by three changes of undiluted TBA. The dehydrated materials were added to 5 ml of TBA:paraffin oil (1:1) and held for 3 h at room temperature, then infiltrated with three changes of melted paraffin at 60°C in an electric oven and finally embedded in wax following procedures outlined by Johansen (1940). Leaf and root sections (10 μm thick) were cut with a microtome and two or three adjacent sections mounted on a slide with a drop of H2O. The slide was then placed on a warming tray (40°C) for 2–3 min to allow expansion. Sections were then dried at 40°C for 24 h and then stored at 4°C.

The template for in vitro transcription was made by linearizing the plasmid carrying the cDNA probe using restriction enzyme digestion. Approximately 1 μg of template DNA, 4 μl transcription buffer, 4 μl DIG-RNA labelling mixture (Boehringer Mannheim), 7 μl DEPC-H2O and 2 μl T7 or T3 RNA polymerase (100 U) were mixed and incubated at 37°C for 2–3 h to synthesize either the sense or the antisense RNA probe. The newly synthesized samples were then digested with DNAase I for 15 min at 37°C. The RNA probe was precipitated with isopropanol at –20°C overnight, centrifuged, washed, and dried under vacuum. Samples were resuspended in 50 μl of NaHCO3 incubated at 60°C for 25 min to chop the RNA to about 200 bp, neutralized with 5 μl of 5% HOAC, then precipitated with 1/10 volume of NaOAC and 2 volume of 95% ETOH and dried.

In situ hybridizations were conducted essentially as described by Langdale (1994). An aliquot of 25 μl of DIG-labelled sense or antisense probe (10 ng μl–1 in hybridization solution) was applied to each section. The hybridization solution contained 50% formamide, 1× salts (0.3 m NaCl, 1 m m Tris–HCl, pH 6.8, 10 m m sodium phosphate pH 6.8, 5 m m EDTA), 10% dextran sulphate, 1 mg ml–1 yeast tRNA, 0.5 mg ml–1 polyA (Gibco BRL) and 10 μg ml–1 of sense or antisense RNA probe. The sections were then covered and incubated at 50°C overnight on the Omnislide apparatus (Hybaid). The slides were washed with fresh wash buffer again for 5 h, incubated with RNAase A (20 μg ml–1) in NTE (500 m m NaCl, 10 m m Tris–HCl pH 8.0) at 37°C for 30 min, washed five times in NTE at 37°C for a total of 1 h and transferred to 1× PBS and stored at 4°C overnight. Detection of hybridization was conducted by first blocking slides in 0.5% blocking reagent (Boehringer Mannheim) in PBS and BSA/Triton/PBS (0.1% BSA, 0.3% Triton X-100, in PBS) each for 45 min, incubating with anti-DIG antibody in BSA/Triton/PBS (1/2000 dilution) at room temperature for 1 h, washing three times (20 min each) in BSA/Triton/PBS, equilibrating in AP buffer (100 m m Tris–HCl pH 9.5, 100 m m NaCl, 50 m m MgCl2) for 5 min, putting 200 μl of substrate solution (45 μl NBT solution, 35 μl X-phosphate solution in 10 ml (AP buffer, 10% PVA)) on each slide, and developing for 20 h.


The authors wish to thank Dr Rod Croteau for the (–)-kaurene and copalol standards and Bob Pocs for help with the GC–MS analysis.


  1. GenBank database accession numbers AF034545 (CPS cDNA); AF097310 ( KS1-1 cDNA); AF097311 ( KS22-1); AF105149, (csu186).