Skip to main content

Plant Terpenoids for Human Health: a chemical and genomic approach to identify and produce bioactive compounds

Final Report Summary - TERPMED (Plant Terpenoids for Human Health: a chemical and genomic approach to identify and produce bioactive compounds)


Executive Summary:

The TERPMED project focuses on two classes of compounds: sesquiterpene lactones (SLs) and phenolic diterpenes (PDs). A range of analytical methods based on HPLC-PDA, GC-MS and LC-MS have been developed and optimized to specifically detect and quantify these compounds in samples from the plant species targeted in the project: Tanacetum parthenium, Inula britannica, Rosmarinus officinalis and Salvia fruticosa. An HPLC-PDA-MS method coupled to an on-line antioxidant detection system has been set up to specifically detect PDs with antioxidant activity. Unbiased metabolomic approaches based on accurate mass-LC-MS methods have also been developed and optimized for large-scale metabolite profiling of tissue samples. By using such untargeted approaches followed by multivariate analysis techniques, a comprehensive and global view of the metabolome of plant extracts has been obtained to establish differences between species, genotypes, tissues and organs from the collection of plants generated in the project.

The above analytical methods have allowed to determine that parthenolide (SL) is primarily stored in the ovary trichomes of T. parthenium disc florets, SLs accumulate preferentially in the trichomes of I. britannica flower corolla, carnosic acid (PD) mainly accumulates in leaf glandular trichomes of R. officinalis and S. fruticosa, and carnosol (PD) is primarily found in other leaf cells of these two species. High quality RNA preparations obtained from ovaries and ovary trichomes of T. parthenium disc florets, ovaries of I. britannica flowers, and leaf trichomes of R. officinalis and S. fruticosa have been subjected to high throughput RNA sequencing and the resulting sequence data have been assembled, annotated and made available via an interactive EST database that is accessible through the TERPMED website (http://www.terpmed.eu).

One of the key objectives of the TERPMED is the elucidation of the biosynthetic pathways leading to parthenolide and carnosic acid. The four T. parthenium enzymes responsible for converting farnesyl diphosphate into parthenolide have been cloned and functionally validated in yeast. The whole parthenolide pathway has been successfully reconstituted in leaves of N. benthamiana plants. The terpene synthases that catalyze the first two steps of the carnosic acid biosynthetic pathway from geranylgeranyl diphosphate have also been cloned and functionally validated in vitro by enzyme activity assays and in vivo by expression in both yeast and N. benthamiana. These terpene synthases catalyze the two-step synthesis of miltiradiene, which is the diterpene backbone precursor of carnosic acid. The miltiradiene pathway has been reconstructed in leaf trichomes of N. tabacum. Identification of the cytochrome P450s mediating successive oxidation reactions leading to carnosic acid is in progress and some strong candidates have already been identified.

Purification procedures have been developed to obtain highly purified preparations of parthenolide and carnosic acid at the gram scale, whereas a range of minor SLs and PDs have been purified in yields ranging from μg to mg. All these compounds have been tested for their ability to activate the Nrf2-ARE antioxidant pathway. Despite all SLs with an exocyclic methylene group activate this pathway, they are not promising lead compounds for further development of drugs targeting the Nrf2-ARE pathway due to their neurotoxicity. On the contrary, carnosol is a promising lead for development of drugs to treat neurodegenerative diseases since it activates the Nrf2-ARE pathway with a low level of toxicity. Production of carnosol through chemical oxidation of carnosic acid has already proven feasible.

A generic metabolite database has been developed to store and make available information about compounds identified in the project. This database is linked to the sequence database through an integrated database for which a pathway representation is used. Information for the biosynthetic pathways can be visualized in pathway maps representing the knowledge on reaction networks obtained in the project. For compounds in the pathways that are present in the compound database, a link from the pathway representation to the compound database is provided. For selected enzymes of the pathways links to the protein sequence are provided.

Project Context and Objectives:

Plant secondary metabolites are one of the most important sources of therapeutic drugs. In fact, many drugs currently in use are derived from plants or lead compounds of plant origin. Terpenoids are the most numerous and chemically diverse class of plant metabolites with over 25,000 compounds identified so far. This huge chemical diversity is however hardly exploited for the development of new drugs. This is due to several reasons such as poor availability of the source plant material, too low concentrations in the plant material and difficulties to obtain pure compounds. The TERPMED project was aimed at providing solutions to overcome these difficulties for two classes of plant terpenoids showing promise as potential drugs for treating cancer and central nervous system disorders: sesquiterpene lactones and phenolic diterpenes. Within these two classes of compounds, the project focuses primarily on parthenolide, carnosic acid and carnosol. Parthenolide is the principal bioactive sesquiterpene lactone and the presumed active ingredient of Tanacetum parthenium (feverfew) extracts, which have been approved in Europe as an herbal drug sold without prescription for the treatment of migraine. More recently, parthenolide has been the subject of several studies suggesting its use as a novel treatment for cancer. As for the phenolic diterpenes, carnosic acid and its derivative carnosol are the most abundant compounds of this class in the leaves of Rosmarinus officinalis (rosemary) and Salvia species. These products are also relevant for human health. Rosemary extracts are already commercialized as food preservatives due to their anti-oxidant properties and some recent studies indicate that carnosic acid and carnosol have potential therapeutic properties which could make these compounds useful for the treatment and/or prevention of diseases such as some neurodegenerative disorders and cancer.

The overall scientific objective of the TERPMED project was to gain understanding of the metabolism of these compounds and other closely related compounds identified as potentially interesting during project development, and use the knowledge generated for designing metabolic engineering strategies aimed at establishing efficient and sustainable plant production platforms for the most promising bioactive molecules.

The specific aims of the project were:

- To develop dedicated analytical methods for detection and quantification of the compounds of interest in tissues and organs of the model and related species targeted in the project.
- To develop unbiased analytical methods for the establishment of comprehensive metabolite profiles of selected tissues and organs of the species of interest.
- To generate a large collection of plants samples and seeds of model and related species and accessions.
- To determine growth conditions, developmental stages, and tissues and organs of model and related species where the highest levels of target compounds accumulate, and to establish unbiased metabolite profiles in different species, accessions, tissues and organs, developmental stages and growth conditions.
- To develop semi-preparative purification methods yielding milligrams of highly purified preparations of the target compounds, including their precursors and downstream products, as well as of other potentially interesting related compounds.
- To establish optimized protocols for obtaining high-quality RNA preparations suited for high-throughput sequencing from the tissues and organs where target compounds accumulate.
- To develop dedicated transcriptome and compound databases linked through an integrated database.
- To generate a list of candidate genes coding for enzymes potentially involved in the synthesis of the compounds of interest and to functionally validate the candidate enzymes in order to elucidate the biosynthetic pathway leading to the production of these compounds.
- To reconstruct the biosynthetic pathways in innovative plant-based production platforms.
- To screen the collection of compounds generated in the project for biological activity in order to identify potential lead compounds for further development of drugs to treat central nervous system disorders and cancer.
- To identify the biological material suited for preparative production of the most promising compounds and to scale-up purification procedures to isolate these compounds at the gram level.
- To establish an intranet/internet web interface to disseminate information and data from the project to partners within the consortium and to the general public.

Project Results:

DESCRIPTION OF THE MAIN SCIENTIFIC & TECHNICAL RESULTS AND FOREGROUND

1) Development of qualitative and quantitative analytical methods to screen plant material for sesquiterpene lactones and phenolic diterpenes

To investigate the tissue/organ specificity of target SLs (parthenolide) and PDs (carnosic acid and carnosol) production, different in situ staining methods for detection of these compounds have been tested. However, all tested methods are not sufficiently specific to determine the site of accumulation of the compounds of interest. Thus, alternative analytical approaches based on HPLC-PDA, GC-MS and LC-MS have been developed and optimized to specifically detect and quantify SLs in samples of T. parthenium (feverfew) and Inula species as well as PDs in samples of R. officinalis (rosemary) and Salvia species. It should also be noted that a large-scale metabolite profiling technique based on the LC-PDA-Ion Trap MS-Orbitrap FTMS system (see below) has been successfully applied to specifically detect and quantify SLs and PDs in a range of samples from plants of interest. An HPLC-PDA-MS method coupled to an on-line antioxidant detection system has also been developed to specifically detect PDs showing antioxidant activity in rosemary and salvia samples.

Analysis of SLs and PDs levels in different organs and tissues of the above plant species, including manually and mechanically isolated glandular trichomes, have demonstrated that parthenolide is primarily stored in the ovary trichomes of T. parthenium disc florets (Majdi et al., 2011) whereas I. britannica SLs accumulate in the trichomes of flower corolla. With regard to carnosic acid and carnosol, these PDs have been detected in both leaf glandular trichomes and the remaining leaf tissue of rosemary and salvia plants, though the ratio carnosic acid vs carnosol is always much higher in the trichomes than in other leaf tissue. Thus, carnosic acid accumulates preferentially in leaf glandular trichomes whereas the remaining leaf tissue is the primary site of carnosol storage.

Methods for untargeted comparative metabolomics of plant samples have also been developed. Methods and protocols previously developed in other metabolomics projects have been adapted for large-scale LC-MS analysis of the plant species targeted in the project. Extraction methods, the conditions for LC and QTOF-MS, and the methods for untargeted processing of large numbers of LC-MS derived raw data files have been optimized for the specific plant samples analyzed in the project. In addition, methods for comparative metabolite profiling using a new Thermo LC-PDA-Ion Trap MS-Orbitrap FTMS system have been implemented. Special attention has been paid to ensure proper extraction, separation, and detection of target SLs and PDs. Compared to the LC-QTOF-MS system, the new LC-MS system has higher mass accuracy, higher dynamic range of metabolite level, and better possibilities for metabolite identification purposes through MSn. The protocols and chromatographic conditions developed for the LC-QTOF-MS system have been adapted for specific characteristics of the LC-Orbitrap system. Apart from high accurate mass detection, this system also enables rapid switching of the Ion Trap MS online, in order to detect ions in both positive and negative modes. This positive-negative switching of the Ion Trap MS is independent of the detection by the Orbitrap FTMS (accurate mass in only one specific mode). Although the Ion Trap generated nominal masses only, ionization mode switching is helpful in metabolite identification and to determine whether major metabolites will be missed upon analyzing samples in only one mode. Metalign software and in-house developed scripts have also been tested and settings optimized for untargeted extraction of mass signals from detection mode (Ion Trap positive mode, Ion Trap negative mode, FTMS positive or negative mode) per analysis run. In addition, new protocols have been developed for on-line MS/MS and MSn fragmentation, in order to better annotate detected metabolites. By using such untargeted metabolomics approach followed by multivariate analysis techniques, such as hierarchical clustering and principal component analysis, a comprehensive and global view of the metabolome of plant extracts (1226 and 1800 reconstructed putative metabolites in I. britannica and R. officinalis extracts, respectively) has been obtained to establish differences between species, genotypes, tissues and organs.

2) Screening metabolite natural variation and effect of developmental and growth conditions

To screen metabolite variation between plant species and accessions, tissues, developmental stages and growth conditions, to support elucidation of biosynthetic pathways and to define favourable materials for purification of compounds of interest, the targeted and untargeted analytical methods developed in the project have been used to analyze a range of tissue samples obtained from the large collection of plants generated in the project. This collection of germplasm consists of plant material from 61 species of interest with 190 accessions. Plants for the collection have been obtained from natural populations, different botanical gardens and purchased from seed companies, and sampled. A seed bank has also been created for collected species and accessions.

To determine growth conditions that induce accumulation of SLs and PDs, the levels of target SLs and PDs have been determined in samples from plants grown in the greenhouse under various experimental conditions, including different soil composition, illumination regime and relative humidity. Analysis of parthenolide content in various samples of feverfew plants has shown that parthenolide production is significantly higher in plants grown in the greenhouse compared to plants grown in vitro. Regardless of growth conditions, parthenolide is present only in aerial parts. As the highest levels of parthenolide are present in the trichomes of disc floret ovary, the most critical factor affecting parthenolide accumulation in flowers is the illumination regime, so that light conditions that promote flowering (long day) also promote parthenolide accumulation. No parthenolide has been detected in plants grown in the shade. On the contrary, no relevant effect of soil composition on parthenolide content during the flowering stage has been observed. The highest level of parthenolide has been detected in flowers of T. parthenium plants grown under long day conditions and parthenolide content varies depending on the developmental stage of flowers. Comparison of parthenolide content in ovaries of flowers at 7 developmental stages has shown that it increases from stage 2 to stage 5 of flower development and then progressively decreases (Majdi et al., 2011). It should also be noted that the relative content of parthenolide in flowers at the same developmental stage might vary up to a factor of 3 among different accessions of T. parthenium. In Inula plants, the highest levels of SLs have also been detected in flowers, and comparison of SLs content in disc florets from flowers at 7 different developmental stages has shown that highest levels of SLs are found in the first 3 stages of flower development and then decrease.

With regard to the effect of growth conditions on PDs production, neither relevant change in the total amount of carnosic acid and carnosol nor in the carnosic acid vs carnosol ratio has been observed in rosemary plants grown on different soils. In vitro cultures of target and related species have also been established using culture media with optimized mineral, hormone, and carbon source composition, but none of the tested conditions has proven to significantly increase production of target compounds compared to that in material from plants grown on soil either in the greenhouse or from natural populations. Analysis of carnosic acid and carnosol levels in leaves at 7 successive developmental stages of R. officinalis and S. fruticosa plants has shown that rosemary leaves contain higher levels of both PDs compared to salvia leaves regardless of the leaf developmental stage. Young leaves from 16 accessions of R. officinalis plants grown in the greenhouse have been tested for carnosic acid and carnosol contents. Both PDs have been detected in all samples tested, albeit at significantly different levels, and some accessions display highly contrasting ratios of carnosic acid vs carnosol. Carnosic acid has been found to be the most abundant PD in some accessions whereas carnosol is the most abundant PD in others. As indicated above, carnosic acid accumulates preferentially in the glandular trichomes of leaves. Thus, young leaves from 29 Salvia species and accessions grown in the greenhouse have been analyzed for their PDs content and for the number and type of glandular trichomes. Correlation analysis between these parameters has shown that the amount of trichomes does not necessarily correlate with PDs production, since some Salvia species whose leaves contain abundant glandular trichomes accumulate very low levels of carnosic acid and carnosol. In fact, the analyzed Salvia species can be grouped into species that are characterized by a high carnosic acid and carnosol production, species that accumulate significantly lower levels of both PDs, and species that have very low levels of these compounds. Among the species tested, S. fruticosa has been found to be the best producer of both carnosic acid and carnosol.

The large-scale metabolite profiling technique LC-PDA-Ion Trap MS-Orbitrap FTMS system has been used to compare the metabolite profiles of trichomes isolated from the abaxial (bottom) and adaxial (upper) side of R. officinalis and S. officinalis leaves. As expected, carnosic acid is the main PD found in the trichomes, though the levels of this compound vary depending on the leaf side from which trichomes have been collected. Rosemary trichomes from the abaxial side contain more than two-fold carnosic acid than those from the adaxial side. Such a difference in PDs levels has not been observed in salvia trichomes. In addition to the major PDs, a number of other minor peaks have been specifically detected in the adaxial trichomes, thus suggesting differences in the PD biosynthesis pathway. A number of other as yet unknown compounds are also significantly different between both leaf sides in both plant species.

3) Purification and characterization of sesquiterpene lactones and phenolic diterpenes from target species

To generate a collection of purified SLs and PDs that could be used as reference compounds, substrates for enzymatic analysis and biological assays, extraction methods and purification protocols based on Centrifugal Partitioning Chromatography (CPC) and preparative HPLC have been developed and used for purification of parthenolide from flower heads of T. parthenium plants grown in the field. The purification procedure yields grams of parthenolide with purity greater than 98%. A CPC-based protocol followed by preparative HPLC has also been developed for purification of carnosic acid and carnosol from leaves of S. officinalis plants. A clean-up step of crude extracts prior to CPC has been introduced to neutralize the interference of as yet unknown compounds. The purification procedure yields hundreds of milligrams of carnosic acid with a purity of 90% and tens of milligrams of carnosol with a purity of 98%. CPC fractions enriched in SLs and PDs generated in scaled-up procedures for purification of parthenolide, carnosic acid and carnosol have been subjected to additional purification steps based on normal and reverse phase semi-preparative HPLC. Minor SLs and PDs in these fractions have been obtained in yields ranging from μg to mg and their identity has been established by LC-MS and NMR analysis. Overall 10 minor SLs have been purified from T. parthenium, one of which is an as yet unreported SL (Fischedick et al., 2012), and 10 different SLs have been purified from I. britannica, 5 of which are new compounds (Fischedick et al., 2013a). A total of 5 minor PDs have been isolated from S. officinalis (Fishedick et al., 2013b).

4) Elucidation of the pathways for parthenolide and carnosic acid biosynthesis. Search for candidate genes

As a first step towards the identification of candidate genes for the biosynthetic pathways of target compounds, optimized protocols for RNA extraction from producing tissues have been established. As indicated above, parthenolide and carnosic acid have been found to accumulate preferentially in disc floret ovary trichomes of T. parthenium and leaf trichomes of R. officinalis and S. fruticosa, respectively. Carnosol is also present in the leaf trichomes but at lower levels than carnosic acid. Thus, protocols for isolation of glandular trichomes from these plant tissues have been developed. For T. parthenium trichome isolation, disc floret ovaries of feverfew flowers at stages of development 3 and 4 were collected. At these stages and active synthesis of parthenolide is likely to occur (Majdi et al. 2011) and therefore the corresponding biosynthetic genes should be actively expressed. In brief, feverfew ovaries were frozen in liquid nitrogen, vortexed in a Greiner tube and the resulting mixture was subjected to successive filtration steps through meshes of different pore size. For R. officinalis and S. fruticosa trichome isolation, trichome glands were separated from leaves by mechanical abrasion using glass beads and several filtration steps were subsequently applied to purify the glandular trichomes from cell debris and non-glandular trichomes. Both protocols yield trichomes in sufficient quantity and purity for RNA isolation.

High quality RNA preparations suitable for high throughput sequencing have been obtained from ovaries of T. parthenium and I. britannica, ovary trichomes of T. parthenium, and leaf trichomes of R. officinalis and S. fruticosa, using commercial kits with only slight modifications of the protocols recommended by the manufacturers. Libraries of cDNA from RNA of T. parthenium and I. britannica ovaries have been constructed and sequenced using the Illumina platform, whereas those generated from RNA of ovary trichomes of T. parthenium and leaf trichomes of rosemary and salvia have been sequenced using the Roche 454 platform. The resulting transcriptomic data have been assembled, annotated and made available to project partners via an interactive EST database that is accessible through the project website. The EST databases have been searched using different criteria, which has enabled to create a first list of candidate genes for each pathway.

Conversion of farnesyl diphosphate into parthenolide had been anticipated to involve four successive enzymatic steps catalyzed by germacrene A synthase (GAS), germacrene A oxidase (GAO), costunolide synthase (COS) and parthenolide synthase (PTS). Sequences of cDNA for candidates TpGAS and TpGAO cDNA have been assembled from contigs in the feverfew EST database that showed the highest homology to sequences of GASs and GAOs reported from other species of the Asteraceae family. Based on the assumption that COS is a cytochrome P450 that evolved from the P450 GAO, five COS candidates showing high sequence similarity to GAOs from different Asteraceae species have been identified. A first list of 59 TpPTS has also been generated using as a search criteria amino acid sequence proximity to CYP71 P450s, COSs, epoxidases, and C10, C15 and C20 epoxidases. Similarity to epoxidases was included as search criteria because conversion of costunolide into parthenolide involves an epoxidation of the double bond at carbon positions C4 and C5 in the costunolide backbone that is presumably catalyzed by a P450 mono-oxygenase.

The first list of TpPTS gene candidates has been narrowed down to a final list of 9 candidates on the basis of their trichome-specific expression and their pattern of expression throughout ovary development. The pattern of expression of TpGAS, TpGAO and TpCOS candidates has been analyzed in ovaries of flowers at 6 developmental stages by quantitative real-time PCR. These three genes display highly similar patterns of expression, which is highest in ovaries at stages 2 and 3, and then decrease from stage 4 until virtually zero at stage 6. The expression pattern of TpGAS, TpGAO, and TpCOS is therefore fully consistent with the accumulation profile of parthenolide in ovaries during flower development. The pattern of expression of TpPTS candidate genes has also been analyzed in ovaries at the same developmental stages. Candidates Tp2116, Tp4149, and Tp9025 show maximum expression at ovary developmental stages 2 to 4, much like TpGAS, TpGAO and TpCOS, and have therefore been selected as the strongest TpPTS candidate genes. Other candidates that display increased expression after stage 4 of ovary development, such as Tp8878 and Tp8879, have also been considered interesting candidates because they could encode parthenolide pathway side branch enzymes using costunolide and/or parthenolide as substrates.

To generate a first list of candidates for the PDs biosynthetic pathway, the EST databases from R. officinalis and S. fruticosa leaf trichomes has been searched for sequences related to terpene synthases, dioxygenases and cytochrome P450 enzymes of specific clades considered to be relevant to terpene biosynthesis. According to the current view of labdanoid biosynthesis, two successive terpene synthases should catalyze the biosynthesis of the diterpene backbone precursor of carnosic acid: a copalyl diphosphate synthase (CPS) and a kaurene synthase like (KSL) enzyme. Sequences of cDNA coding for putative CPSs and KSLs have been assembled from contigs identified in the rosemary and salvia EST databases that showed highest homology to sequences of CPS and a miltiradiene synthase (KSL) from S. miltiorrhiza. The first list of putative terpene synthases consisted of one SfCPS, one RoCPS, two SfKSLs and two RoKSLs. Repeated oxidation at position C20 of the diterpene backbone precursors should create the carboxylic acid group of carnosic acid. Phylogenetic analysis of the various contigs in the EST databases that showed similarity to 2-oxoglutarate-dependent dioxygenases (2OG-dioxs) potentially involved in the oxidation of the diterpene skeleton at this position with the complete set of 2OG-dioxs sequences of Arabidopsis identified a promising C20 oxidase candidate from S. fruticosa. Cytochrome P450s constitute one of the largest gene families coding for enzymes that are involved in many different biosynthetic pathways in plants. Although it is almost impossible to predict their biochemical function on the basis of sequence similarity, some P450 clades have been shown to contain terpene-oxidizing enzymes (e.g. the CYP71 clade). Thus, combining phylogenetic analysis and assignment of the P450s to particular clades, a first list of 48 P450 candidates (24 from each species) has been generated.

To reduce the list of candidates, quantitative real-time PCR expression analysis of the PDs candidate genes has been performed using RNA obtained from trichomes of young leaves of R. officinalis and S. fruticosa, their corresponding leaves without trichomes and trichomes from aged leaves. Genes coding for enzymes involved in the carnosic acid pathway should display both high trichome expression specificity ratio (trichomes vs leaves without trichomes) and high young vs aged trichome expression ratio. To further narrow down the list of candidates a second round of expression analysis has been performed using RNA from young and aged whole leaves of 4 different rosemary varieties showing highly contrasting ratios of carnosic acid vs carnosol, and from trichomes of S. fruticosa leaves at 4 developmental stages and their corresponding leaves without trichomes. Comparison of the expression profile of the P450 candidates with that of the functionally validated CPS and KSL candidates RoCPS, RoKSL1, RoKSL2, SfCPS and SfKSL1 (see below) has allowed generating a subset of strong candidates consisting of 9 rosemary P450s from the CYP716, CYP76, and CYP96 clades, and 9 salvia P450s from the CYP71 and CYP96 clades.

5) Elucidation of the pathways for parthenolide and carnosic acid biosynthesis. Functional characterization of candidate genes and pathway reconstruction

The full-length coding sequences for putative TpGAS, TpGAO, Cichorium intybus COS (5 sequences), TpCOS, and TpPTS (9 sequences) have been cloned from T. parthenium and C. intybus cDNA, and transferred into the appropriate yeast and/or plant expression vectors. The full-length coding sequences for all PDs biosynthetic pathway candidates (putative CPSs and KSLs, SfC20 oxidase, and P450 enzymes) have been cloned from R. officinalis and S. fruticosa cDNA, and then transferred to the appropriate vectors for expression in E.coli yeast and/or in planta.

Putative TpGAS and TpGAO genes have been co-expressed in yeast. GC-MS analysis of crude yeast extracts has shown that cells co-expressing TpGAS and TpGAO produce germacrene A acid, which is not detected in extracts from cells expressing only TpGAS. This result has demonstrated that TpGAO is able to catalyze oxidation of germacrene A to germacrene A acid. To test the catalytic function of the 5 putative CiCOS, the candidates have been separately co-expressed with TpGAS and CiGAO in yeast. Costunolide production has been detected only when the cDNA corresponding to CiCOS candidate 3368 is expressed. The same result has been obtained in yeast cells co-expressing TpGAS, TpGAO and TpCOS, thus confirming that the putative TpCOS can catalyze the conversion of germacrene A acid into costunolide (Qing et al., 2011). To test the catalytic activity of TpPTS candidates, microsomes of yeast cells expressing the nine TpTPS candidates separately have been obtained and incubated in vitro with costunolide. Microsomes from yeast cells expressing TpPTS candidate 2116 produce a new compound that has been unambiguously identified as parthenolide, thus demonstrating that Tp2116 encodes a true PTS. Microsomes from yeast expressing candidate Tp8878, which had been assumed to be involved in a side branch of parthenolide biosynthesis on the basis of its pattern of expression throughout flower ovary development, indeed has been found to produce a new compound identified as 3β-hydroxycostunolide. It is interesting to note that candidate Tp8878 produce 3β-hydroxyparthenolide when parthenolide is used as substrate instead of costunolide. Thus, Tp8878 encodes a cytochrome P450 enzyme that can oxidize both costunolide and parthenolide to produce the more polar derivatives 3β-hydroxycostunolide and 3β-hydroxyparthenolide, respectively. Finally, candidate Tp8879 has been identified as a kauniolide synthase due to its ability to transform costunolide into kauniolide in vitro.

Mature forms of SfCPS and SfKSL1 lacking the N-terminal chloroplast transit peptide have been expressed in E. coli and purified by affinity chromatography. In vitro enzyme activity assays have show that recombinant SfCPS is able to transform GGDP into the reaction product copalyl-PP, and that recombinant SfKSL1 co-incubated with recombinant purified SfCPS and its substrate GGDP is able to transform copalyl-PP into a levopimaradiene-like compound that has been identified as miltiradiene. The identity of this compound has been unequivocally established by NMR analysis of the SfKSL1 reaction product purified in large amounts from yeast cells engineered to co-express the mature forms of both SfCPS and SfKSL1. Further confirmation of the terpene synthase activity of these candidates has been obtained by transient co-expression of full-length forms of these two enzymes in leaves of N. benthamiana plants. GC-MS analysis has allowed detection of a compound identified as miltiradiene based on its retention time and its mass spectrum. An additional minor peak whose mass spectrum shows high similarity to abietatriene accompanies always the miltiradiene peak in both transient N. benthamiana and yeast expression assays. It remains to be established whether this product results from an enzymatic or a non-enzymatic reaction. The functional identity of terpene synthase candidates from R. officinalis has been established using the same experimental approaches based of yeast and N. benthamiana expression. Miltiradiene has been identified as the major reaction product when RoKSL1 is co-expressed with RoCPS in both yeast cells and N. benthamiana leaves. Interestingly, co-expression of RoCPS and RoKSL2 in N. benthamiana leaves also leads to miltiradiene formation. Furthermore, both RoCPS and RoKSL1 have been combined in a T-DNA binary vector under the control of different trichome specific promoters. This construct has been transformed into N. tabacum lines that do not produce endogenous tobacco labdanoids and cembranoids. Significant amounts of miltiradiene have been detected in solvent dipping extracts from leaves of the resulting transgenic plants. The system is now established as a robust platform for substrate production, assaying enzymes downstream the PDs pathway and pathway reconstitution. Overall, these results point to a conservation of the first steps of PDs biosynthetic pathway between R. officinalis and S. fruticosa, and demonstrate that CPS and KSL candidates from R. officinalis and S. fruticosa are true terpene synthases that can transform GGDP into copalyl-PP and this intermediate into miltiradiene, the diterpene backbone precursor of carnosic acid, respectively.

To test the catalytic activity of the P450 candidates, some of them have already been transiently expressed in leaves of N. benthamiana along with CPS and KSL from the appropriate species and the cytochrome P450 reductase 1 (AtCPR1) from A. thaliana. So far, co-expression of candidates Ro034450 and Sf000850 in leaves of N. benthamiana has resulted in the production of a new compound. Even though the structural identification of this product is not yet completed, GC-MS analysis suggests it is a miltiradiene derivative with an aldehyde group, most likely at position C20, where the carboxyl group of carnosic acid is present. LC-Orbitrap FTMS analysis indicates a miltiradiene-like derivative which may have a hydroxyl, an aldehyde residue or a carboxylic acid group. All these results favour the hypothesis that this P450 enzyme from rosemary and salvia can introduce a carboxyl group at position C20, which would represent a key step in the biosynthesis of PDs. Putative P450s have also been co-expressed in yeast with SfCPS and SfKSL, and a cytochrome P450 reductase from poplar (CPR2). Co-expression of candidate Sf016156/21415 yields a new compound having a mass spectrum highly similar to ferruginol. Co-expression of candidates Ro021640 and Ro021643, which show high amino acid sequence similarity to each other and to the Sf016156/21415 candidate, with RoCPS, RoKSL1, and CPR2, also results in the production of ferruginol. Thus, these results demonstrate that both species contain a P450 enzyme that is able to transform miltiradiene into ferruginol, which appears to be the first hydroxylated intermediate in the carnosic acid pathway.

With all genes of the parthenolide biosynthetic pathway from FPP available, the entire biosynthetic pathway for this compound has been reconstituted in N. benthamiana plants through transient co-expression of TpGAS, TpGAO, TpCOS and TpPTS. The catalytic soluble domain of Arabidopsis HMG-CoA reductase has also been introduced into plants to boost the supply of the parthenolide precursor FPP. Four days after infiltration, total parthenolide yield is about 1.4 mg·g-1 FW. However, the vast majority of parthenolide is found conjugated to cysteine (1368.4 ng·g-1 FW parthenolide-Cys) and glutathione (87.5 ng·g-1 FW parthenolide-GSH) and only traces of free parthenolide are detected (2.05 ng·g-1 FW). As costunolide, parthenolide and the corresponding hydroxylated derivatives are cytotoxic, conjugation to GSH or cysteine may be part of a detoxification reaction in the N. benthamiana host cells. These parthenolide conjugates are probably delivered to vacuoles for detoxification. As the conjugation to GSH is reversible at physiological pH and the conjugation to cysteine is not, this would explain the relatively high levels of cysteine conjugated products.

So far, the carnosic acid biosynthetic pathway has been partially reconstructed in yeast cells and N. benthamiana leaves, which have been engineered to co-express CPS and miltiradiene synthase (KSL) from both R. officinalis and S. fruticosa for efficient production of miltiradiene, the structural precursor of carnosic acid and carnosol. Miltiradiene production increases significantly when GGDP synthase is co-expressed with these two terpene synthases to boost production of the miltiradiene precursor GGDP. These two miltiradiene production platforms are being used as tools for functional characterization of P450 enzymes acting downstream in the pathway to carnosic acid and carnosol. Besides transient reconstruction of the miltiradiene pathway in N. benthamiana leaves, the first two steps of the pathway have also been stably introduced in the trichomes of N. tabacum leaves, resulting also in the production of miltiradiene. This system has been established as a robust platform for production of miltiradiene, which can be used as substrate for in vitro enzyme activity assays.

6) Biological assays

To identify the most promising SLs and PDs, in terms of bioactivity activity, for subsequent scaling-up production, the whole set of SLs isolated from T. parthenium flowers and PDs obtained from S. officinalis leaves (see above) has been tested for both the ability to activate the Nrf2-antioxidant response element (ARE) pathway in primary mouse cortical neuronal cultures and the general cellular toxicity. In general, all SLs with an exocyclic methylene group at position 11 can activate the Nrf2-ARE pathway. However, all active SLs have also been found to be toxic in MTS cytotoxicity assays. Therefore, none of these SLs can be considered as a promising lead compound for further development of drugs targeting the Nrf2-ARE pathway (Fischedick et al., 2012a). As regard to PDs, all isolated compounds except 12-methoxy-carnosic acid can activate the Nrf2-ARE pathway in mouse primary cortical neurones, albeit with varied potency, thus suggesting that the catechol moiety is essential for this activity. The most potent Nrf2-ARE pathway activators are carnosol and carnosaldehyde, which also display neuroprotective activity in mouse primary cortical neuronal cultures against oxidative (H2O2) stress induced cell death. As carnosol shows lower toxicity than carnosaldehyde against the primary cortical cultures, this PD is considered a promising compound to study as an agent that may be useful in treating neurodegenerative diseases (Fischedick et al., 2013b).

Previous studies have demonstrated the anti-cancer activity of parthenolide in vitro through induction of apoptotic cell death in a number of human cancer cell lines. It has been suggested that depletion of intracellular pool of GSH by parthenolide probably contributes to its apoptotic activity. As more than 90% of parthenolide produced in leaves of N. benthamiana plants expressing the feverefew parthenolide biosynthetic pathway is conjugated to either GSH or cysteine, the anti-cancer activity of parthenolide GSH and cysteine conjugates has been examined in different human cell lines. Both sensitive and multi-drug resistant lines of non-small cell lung carcinoma, glioblastoma and colon carcinoma cells as well as normal human keratinocytes have been used for these assays. Parthenolide-cysteine and parthenolide-GSH conjugates are significantly less potent than free parthenolide in all tested cancer cell lines and normal human keratinocytes. Nevertheless, the concentrations necessary to inhibit cell growth by 50% (IC50 values) values of the conjugates for colon cancer cells are substantially lower than those for normal cells, indicating selectivity of both parthenolide conjugates towards colon carcinoma cells. The parthenolide-cysteine and parthenolide-GSH conjugates exert the highest bioactivity in HT-29 cells (colon adenocarcinoma) with IC50s of 17.3 and 10.7μM, respectively. The sensitivity of cancer cells to free or conjugated parthenolide is not affected by multi-drug resistance as the inhibitory profiles of the compounds are similar in both sensitive and resistant colon carcinoma cell lines. Cysteine and GSH had no influence on cell growth. Thus, even though less effective in most cell lines, these conjugates show quite high and selective activity against colon carcinoma cells and this feature could be an advantage in colon cancer treatment. Perhaps parthenolide conjugates act as a pro-drug in these cells, requiring biotransformation into free parthenolide to exert their anti-cancer effect. Considering that poor water-solubility of free parthenolide is a significant limitation for its application in cancer treatment and that parthenolide conjugates are selectively active against colon cancer cells, the conjugation of parthenolide could be a new strategic tool in drug development for cancer treatment.

Although not initially foreseen under the project proposal, due to serendipity, one of the SLs isolated from T. parthenium, identified as tanaparthin β-peroxide, has been screened for antimalarial activity and found positive with an IC50 of 8,9 μM. The antimalarial activity testing has been conducted using both a parasite lactate dehydrogenase (pLDH) assay and SYBR-green coloration on human blood cells infected with the parasite in the asexual phase. Tanaparthin β-peroxide is therefore a potential new antimalarial agent, though further investigation is needed before it can be considered as a drug candidate. Previously purified SLs and PDs from T. parthenium, S. officinalis and I. britannica have been added to the infected blood cells to determine inhibition of parasite proliferation against dihydro-artemisinine as a control. Good to moderate activities with IC50 values ranging from 20-70 μM have been observed for parthenolide, costunolide diepoxide and a still unidentified SL. Regarding the I. britannica SLs, IC50 values of all tested compounds were above 100 μM, which indicates low activity. Good activity was found for the PDs carnosol and sageone, with IC50 values of 19,4 μM. This indicates that carnosol would indeed be an interesting product for further assessment of bioactivities as it is also the most promising compound in the neuroprotective assays. From results with the purified compounds, it has been investigated if antimalarial activity could be confirmed in crude extracts. Exhaustive extraction methods using different organic solvents and hot water have been applied to plant material from the three above-mentioned plant species and R. officinalis. The obtained crude extracts have been tested for antimalarial activity and the composition of the extracts has been determined to assess which known compounds might be responsible for the antimalarial activity. Results of organic and water crude extracts testing indicate that additional as yet unidentified compounds with antimalarial activity are present in rosemary and salvia extracts. Further fractionation and bioassay testing would be required to identify all positive active products.

7) Scaled-up production of the most promising compounds

No heterologous plant material for large-scale production of selected compounds has been generated so far in the project. Thus, the extraction and purification methods developed in the project have been adapted for large-scale (gram level) compound purification from homologous plant material, i.e. feverfew, rosemary and salvia plant material. As indicated above, grams of parthenolide with purity greater than 98% have been obtained from flower heads of T. parthenium. Carnosic acid has also been produced at the gram scale levels with purity greater than 95%. Regarding the production of carnosol, a chemical conversion from carnosic acid using Ag2O as oxydizing agent has proved to be feasible for reaching the same production levels as carnosic acid. As biological assays have revealed that tanaparthin β-peroxide isolated from feverfew shows promising antimalarial activity, the purification procedure used for small-scale purification of this compound has been used as a template for further scaling-up. However, all attempts made so far have been unsuccessful, probably because tanaparthin β-peroxide is suffering from decomposition during the last few purification steps or it is lost in the insoluble residue. Thus it is concluded that further optimization of the process is required to be able to obtain sufficient amount of highly pure tanaparthin β-peroxide for further analytical purposes.

8) Project database and dissemination

An internet/intranet web interface for communication within the consortium and to disseminate information and data from the project to the public has been established. The TERPMED project website has been set up in the first year of the project and made available through http://www.terpmed.eu/. The website features both a public section for dissemination of project information towards the general public and a secure private section for sharing of confidential data, resources, results and documents between the project partners. The database includes a sequence section and a compound section that are linked through an integrated database.

Transcriptome (EST) data from T. parthenium and I. britannica ovaries, T. parthenium ovary trichomes, R. officinalis and S. fruticosa leaf trichomes generated in the project have been assembled, annotated and made available via a sequence database that is accessible only to project partners through the TERPMED website. Publicly available transcriptome datasets from S. sclarea leaf trichomes and mixed samples from S. miltiorrhiza have also been assembled, annotated and added to the database. Functionality of the database includes ORF detection, BLASTX alignments to the NCBI non-redundant protein database, InterProScan protein domain predictions and BLASTN alignments between the various transcriptome assemblies. A web interface to search the annotated transcriptome data based on annotation keywords has been implemented. A browse interface has been added to the search interface to visualize and browse the BLAST and protein domain annotations on the sequence. Moreover, a BLAST interface has been implemented that allows partners to search the transcriptome databases using known gene and protein sequences.

A metabolite database has been developed that stores and makes available information about compounds identified in the project. This includes quantitative information about the experiment in which the compound has been identified such as plant species, tissue and the amount of the selected compound that was present in plant materials analyzed. For uploading of new data, a template is available which can be downloaded and into which the required information can be copied, which ensures that uploading is easily performed.

For the integrated database, a pathway representation is used. Information related to parthenolide, carnosic acid and carnosol can be visualized in pathway maps representing the knowledge on reaction networks obtained in the project. For compounds in the pathways that are present in the metabolite database, a link from the pathway representation to the compound database is provided. The user can access the information in the compound database by clicking on the compound in the pathway figure. Similarly, for selected enzymes of the pathways, links to the protein sequence are provided.

References

- Majdi, M., Liu, Q., Karimzadeh, G., Malboobi, M. A., Beekwilder, J., Cankar, K., De Vos, P., Todorovic, S., Simonovic, A., and Bouwmeester, H.J. (2011). Biosynthesis and localization of parthenolide in glandular trichomes of feverfew (Tanacetum parthenium L. Schulz Bip.). Phytochemistry, 72:1739–1750. doi:10.1016/j.phytochem.2011.04.021.
- Liu, Q., Majdi, M., Cankar, K., Goedbloed, M., Charnikhova, T., Verstappen, F. W. A., De Vos, R., Beekwilder, J., van der Krol, S., and Bouwmeester, H.J. (2011). Reconstitution of the costunolide biosynthetic pathway in yeast and Nicotiana benthamiana. PLoS ONE, 6, e23255. doi:10.1371/journal.pone.0023255.
- Fischedick, J., Standiford, M., Johnson, D., De Vos, R., Todorović, S., Banjanac, T., Verpoorte, R., and Johnson, J.A. (2012). Activation of antioxidant response element in mouse primary cortical cultures with sesquiterpene lactones isolated from Tanacetum parthenium. Planta Medica. 78:1725-1730. doi:10.1055/s-0032-1315241.
- Fischedick, J.T. Pesic, M., Podolski-Renic, A., Bankovic, J., de Vos, R.C.H. Peric, M., Todorovic, S., and Tanic, N., (2013a) Cytotoxic activity of sesquiterpene lactones from Inula britannica on human cancer cell lines. Phytochem. Lett. 6:246-252.
- Fischedick, J.T. Standiford, M., Johnson, D.A. and Johnson, J.A. (2013b) Structure activity relationship of phenolic diterpenes from Salvia officinalis as activators of the nuclear factor E2-related factor 2 pathway. Bioorg. Med. Chem. 21:2618-2622. doi: 10.1016/j.bmc.2013.02.019.

Potential Impact:

Plant natural products are an important source of compounds that can be used as drugs or as lead compounds to develop new molecules with improved therapeutic properties. Advanced technologies offer the possibility to use rational design for new plant-derived drug discovery. Progress by the European industry towards innovation and competitiveness in this field is essential to maintain and consolidate its leading position in the global market. The TERPMED project has been designed with the general objective of developing knowledge and technology that may help to achieve this goal. During its lifetime, the project has developed a range of protocols and analytical tools that target specifically two classes of plant terpenoids showing promise for treating cancer and central nervous system disorders: sesquiterpene lactones and phenolic diterpenes. Chemical diversity for the functional molecules has been screened in a collection of germplasm generated in the project and the most suitable species and varieties for gene discovery have been selected. Gene discovery has been achieved through a combination of transcriptomics and metabolomics approaches. The genes identified and characterized have been used to reconstitute totally or in part the biosynthetic pathways leading to the synthesis of the main target compounds, parthenolide and carnosic acid, in heterologous plant hosts which can be used as green factories for sustainable large-scale production of these compounds and as platforms to generate and produce new functional molecules through combinatorial biosynthesis approaches. The project has also generated the necessary knowledge and technology to isolate these compounds from plant material. Extraction and purification procedures at laboratory scale have been developed and used to isolate a range of sesquiterpene lactones and phenolic diterpenes, including the target compounds, from their natural plant sources. This has enabled identification of some compounds of theses two classes that had not been reported yet. The resulting library of compounds has been tested for biological activity and the most promising ones have been used in a pilot up-scaling purification study for semi-industrial production. In summary, the TERPMED project has laid the foundation for improved production of known biologically active molecules and has enabled to identify new bioactive molecules that can be considered as promising lead compounds for further development of drugs targeting cancer, neurological disorders and malaria. And last but not least, the project has generated a pool of highly trained researchers in this interdisciplinary area. Therefore, though it is difficult to estimate in quantitative terms the impact of exploitation of the results the TERPMED has produced, it is evident that they will support the European Union’s ambition of becoming one of the world’s most competitive knowledge-based economies.

Main dissemination activities

To give visibility to the TERPMED, a project website available through the weblink http://www.terpmed.eu has been developed. The website features both a public section for dissemination of relevant project information towards the general public and a secure private section which is only accessible to consortium members. As the website will remain online, it will continue to be one of the main dissemination tools for the project’s results also in the future. The TERPMED weblink is posted on the websites of each of the consortium members’ institutions.

The TERPMED project (i.e. background, major research goals, experimental approaches and strategies, main results obtained, etc.) has been presented in various national and international workshops and scientific meetings by different members of the consortium.

Results obtained in the project’s lifetime have been disseminated to the scientific community through a number of oral and poster presentations in regional, national and international meetings, and through publications in peer-reviewed specialized journals, both as research articles and reviews, and as book chapters. In all cases the EC financial support to the TERPMED has been duly acknowledged.

Other relevant dissemination activities include a summary of the project consisting of a description of project background and major objectives, which has been contributed to an EC report on “The main projects related to genetically modified organisms accomplished over the last decade 2000-2010 as well as ongoing FP7 projects”, a volume of a book “EC-sponsored research on safety of GMOs” (http://ec.europa.eu/research/quality-of-life/gmo/) and a file with some key facts from the TERPMED for the general public, including a list of scientific publications, reviews and book chapters resulting from the TERPMED, which has been sent to the Directorate General of Research and Innovation.

List of Websites:

http://www.terpmed.eu/