The Mevalonic Acid Pathway

In plants, the universal C₅-building blocks of all terpenes, IPP and DMAPP, are synthesized by two independent pathways, the mevalonic acid (MVA) pathway and the methylerythritol phosphate (MEP) pathway. The MVA pathway occurs in prokaryotes and in almost all eukaryotes including higher plants. In plants, all enzymatic steps of the MVA pathway are thought to be mainly located in the endoplasmic reticulum (ER)/cytosol compartment of the cell, where they provide the precursors for the biosynthesis of sesquiterpenes, phytosterols, brassinosteroids, and triterpenes (Newman and Chappell, 1999). The cytosolic localization of the MVA pathway was recently challenged by Sapir-Mir et al. (2008) based on the bioinformatic prediction of peroxisomal targeting sites for several enzymes of the MVA pathway in Arabidopsis and evidence for the peroxisomal localization of MVA pathway enzymes in mammals (Kovacs et al., 2007). Furthermore, Leivar et al. (2005) demonstrated the localization of Arabidopsis MVA pathway enzymes in small vesicles.

The pathway begins with two consecutive condensation reactions of three molecules of acetyl-CoA to form the C₆-compound 3-hydroxy-3-methylglutaryl-CoA (HMG-CoA) (Fig. 1). In animals and plants, both steps are catalyzed by two separate enzymes, acetoacetyl-CoA thiolase (AACT, EC 2.3.1.9), which is responsible for the condensation of two acetyl-CoA molecules to form the intermediate acetoacetyl-CoA (AcAc-CoA), and HMG-CoA synthase (HMGS, EC 2.3.3.10) forming S-HMG-CoA by condensation of one molecule of AcAc-CoA with a third molecule of acetyl-CoA. The Arabidopsis genome contains two AACT genes (ACT1 and ACT2), which encode the closely related functional AACT isoforms AACT1 and AACT2 (Ahumada et al., 2008) (Table 1). While ACT2 is highly expressed throughout the plant, low levels of ACT1 transcript occur in roots and inflorescences. Interestingly, in contrast to a cytosolic localization of the AACT2 enzyme, AACT1 appears to be targeted to peroxisomes based on two alternative peroxisomal targeting sequences (the N-terminal PTS1 or the C-terminal PTS2) resulting from different splicing events (Carrie et al., 2007; Ahumada et al., 2008). Act2 T-DNA insertion mutants are embryo lethal, which supports an essential role of AACT2 in isoprenoid biosynthesis, while act1 null mutants are viable with no growth defects, indicating a possibly different metabolic role of AACT1 in peroxisomes (Jin and Nikolau, 2007).

A cDNA encoding HMGS from Arabidopsis was described to functionally complement erg11 and erg13 yeast mutants defective for HMGS (Montamat et al., 1995), and an HMGS from Brassica juncea was enzymatically characterized (Nagegowda et al., 2004). In contrast to the single HMGS gene copy in Arabidopsis (Table 1), HMGS in Brassica and other plants is encoded by small gene families with differential tissue-specific expression and wound- or stress hormone-induced responses (Wegener et al., 1997; Alex et al., 2000; Nagegowda et al., 2005).

In the following irreversible step of the MVA pathway, S-HMG-CoA is converted into R-mevalonate by two reduction steps, each requiring NADPH as the reducing equivalent (Fig. 1). The reaction is catalyzed by the enzyme HMG-CoA reductase (HMGR, EC 1.1.1.34), which has been under intensive investigation due to its key regulatory role in sterol biosynthesis (Goldstein and Brown, 1990). Several genes encoding HMGR isoforms have been cloned from different plant species, providing more detailed information about the structure, location, and regulation of this enzyme (Nelson et al., 1994; Weissenborn et al., 1995; Ha et al., 2001; Ha et al., 2003). The Arabidopsis genome contains two HMGR genes (HMG1, HMG2) encoding two functionally active HMGR isoforms (Caelles et al., 1989; Learned and Fink, 1989; Enjuto et al., 1994; Suzuki et al., 2009) (Table 1).

All plant HMGR genes known to date encode membrane-bound class I HMGR proteins (in contrast to class II HMGR occurring in some eubacteria), which share a similar domain structure: a highly variable N-terminal region, a more conserved membrane domain with two membrane-spanning sequences, and a highly conserved catalytic C-terminal region (Enjuto et al., 1994; Campos and Boronat, 1995; Denbow et al., 1995; Re et al., 1997). A refined in vivo analysis of HMG1 indicated a localization of this enzyme in vesicle-like structures, which may be derived from segments of the ER (Leivar et al., 2005).

Members of plant HMGR gene families exhibit distinct patterns of expression in different plant organs, during development, and in response to phytohormones (Chye et al., 1992) and external stimuli such as light, wounding, elicitor treatment, and pathogen attack (Choi et al., 1992; Genschiketal., 1992; Nelson et al., 1994; Stermer et al., 1994; Weissenborn et al., 1995; Kondo et al., 2003). Arabidopsis HMG1 is expressed throughout the plant and is induced in darkness in aerial organs, whereas HMG2 seems to be expressed only in meristematic and floral tissues (Enjuto et al., 1994; Enjuto et al., 1995). Consistent with an expected housekeeping function of HMG1, hmg1 null mutants show a pleiotropic phenotype of early senescence, sterility, and dwarfing due to the suppression of cell elongation (Suzuki et al., 2004). Sterol levels in hmg1 were found to be reduced and the mutant phenotype could be rescued by the application of the sterol precursor squalene (Suzuki et al., 2004). Double hmg1 hmg2 mutants are not viable because of the requirement of both genes for gametophyte development (Suzuki et al., 2009). HMG1 and HMG2 also play a major role in the biosynthesis of triterpene specialized metabolites (Fig. 1) since levels of triterpenes are reduced by 65 % and 25% in hmg1 and hmg2 mutants, respectively, in comparison to those in wild type plants (Ohyama et al., 2007). In contrast to other plant HMGR genes, neither of the Arabidopsis HMGRs appears to show major stress-induced responses.

It is generally accepted that HMGR catalyzes the main rate-limiting step in the MVA pathway and downstream terpene biosynthetic pathways. This is evident from reduced sterol and triterpene levels in the Arabidopsis hmg mutants and an increase in sterol levels in transgenic Arabidopsis overexpressing HMGR (Manzano et al., 2004). Moreover, positive correlations between HMGR gene expression and enzyme activity and the induced production of sesquiterpene phytoalexins have been reported in many plants although HMGR activity might not always be the sole rate-controlling step (e.g., Stermer and Bostock, 1987; Chappell et al., 1991; Choi et al., 1994; Chappell et al., 1995).

Besides the regulation of plant HMGRs at the transcriptional level in response to external cues and by feedback regulation from downstream metabolites such as sterols (Jelesko et al., 1999; Holmberg et al., 2002; Wentzinger et al., 2002), a developmental and light-regulated post-translational control of HMGR protein levels has been demonstrated (Korth et al., 2000). In Arabidopsis, metabolic perturbation by enhancing or depleting the flux through the sterol pathway causes posttranslational down- or upregulation of HMGR activity, respectively, without changes in transcript or protein levels (Nieto et al., 2009). In addition, the inhibition of Arabidopsis HMGR activity by the SNF1 (sucrose nonfermenting)-related protein kinase 1 (SnRK1) was recently supported in mutants of PRL1 (Pleiotropic Regulatory Locus 1), a global regulator of sugar, stress, and hormone responses, which inhibits SnRK1 (Fores-Perez et al., 2010).

Figure 1.

Terpene biosynthesis pathways and their subcellular organization in Arabidopsis.

The dashed line indicates a possible partial or complete location of MVA pathway enzymes in peroxisomes as predicted by Sapir-Mir et al. (2008). Dashed arrows indicate more than one enzymatic step. Enzymes are in red (cytosol/peroxisome), orange (plastid), or yellow (mitochondrion). Abbreviations for enzymes and metabolites are as described in the text.

In the last three steps of the MVA pathway, mevalonic acid is converted into isopentenyl diphosphate (IPP) by two sequential ATP-dependent phosphorylation steps, catalyzed by the enzymes mevalonate kinase (MK) and phosphomevalonate kinase (PMK), and a third ATP-driven decarboxylative elimination catalyzed by mevalonate diphosphate decarboxylase (MVD). While crystal structures of mevalonate kinase and phosphomevalonate kinase have been solved from yeast and bacteria (Bonanno et al., 2001; Romanowski et al., 2002), comparatively little is known about these enzymes in plants. A single copy of the MK gene with relatively high homology to genes from yeast and mammals was characterized from Arabidopsis (Riou et al., 1994; Lluch et al., 2000) (Table 1). The MK gene is transcriptionally expressed in all parts of the plant with highest expression in root meristematic and floral tissue. Interestingly, Arabidopsis HMG2 shows a similar expression pattern, which suggests a common regulatory mechanism of these genes in tissues with high requirements for mevalonate-derived terpenoid products (Lluch et al., 2000). Similar to yeast and mammals, plant MKs respond to feed-back inhibition by prenyl diphosphates like geranyl diphosphate (GPP) and farnesyl diphosphate (FPP), which act as competitive inhibitors of ATP (Schulte et al., 2000). In contrast to MK and PMK, Arabidopsis MVD is encoded by two loci, of which MVD1 was functionally characterized by complementation in yeast (Cordier et al., 1999) (Table 1). Indications that the last three enzymatic steps of the mevalonate pathway may not be important control points in plant terpene biosynthesis, came from earlier studies showing that these enzyme activities were higher or similar to that of HMGR or they did not correlate with the rate of terpene formation (Ji et al., 1993; Sandmann and Albrecht, 1994; Bianchini et al., 1996).

The Methylerythritol Phosphate (MEP) Pathway

Geranyl Diphosphate Synthase

Geranyl diphosphate (GPP) synthases (GPSs, EC 2.5.1.1) produce GPP (Fig. 2a), which is generally considered to be the precursor of most C₁₀ monoterpenes (Fig. 1). In plants, different classes of homodimeric and heterodimeric GPSs have been identified (Burke and Croteau, 2002; Schmidt and Gershenzon, 2008; Hsieh et al., 2011). In Arabidopsis, the single gene GPS1 (Table 1) was originally characterized by Bouvier et al. (2000) to encode a homodimeric GPS. However, a recent study by Hsieh et al. (2011) questioned this finding since the enzyme appears to have a multi-product medium/long chain prenyl diphosphate synthase activity when IPP is supplied in excess to the allylic substrates DMAPP, GPP and FPP. The result was further supported by a structural analysis of the GPS1 protein (renamed by Hsieh et al. as polyprenyl di(pyro)phosphate synthase, PPPS), which indicated an active-site cavity sufficient to accommodate the medium/long-chain products (Hsieh et al., 2011). Since the PPPS protein is targeted to plastids (Bouvier et al., 2000) where IPP and DMAPP are produced at ratios of approximately 5:1 by the MEP pathway, it is possible that this enzyme indeed exhibits a polyprenyl diphosphate synthase activity in vivo. Van Schie et al. (2007) demonstrated that RNAi-silencing of PPPS (GPS1) resulted in a dwarfed phenotype and a PPPS knock-out line appeared to be embryo-lethal. While the exact function of PPPS remains to be determined, the enzyme might play a role in the biosynthesis of plastoquinones and possibly gibberellins since it produces some GGPP in vitro (Hsieh et al., 2011). Several proteins with homology to Arabidopsis PPPS/GPS1 have been found in other plants and designated as homodimeric GPSs. Van Schie demonstrated GPS activity for a protein in tomato, although at a low IPP/DMAPP ratio; thus, it remains to be determined whether the proteins in this clade, which are closer related to long chain prenyltransferases (Hsieh et al., 2011), may also have polyprenyl diphosphate synthase activity. In gymnosperms, homodimeric prenyl diphosphate synthases of a separate clade were found to have GPS activity (Schmidt and Gershenzon, 2008; Schmidt et al., 2010). These enzymes produce substantial amounts of the longer prenyl diphosphates FPP and/or GGPP, indicating their possible role in other terpene biosynthesis pathways. Interestingly, a flower-specific homodimeric GPS has recently been reported from orchids, which is closer related to the small subunit of heterodimeric GPSs (see below) (Hsiao et al., 2008).

Table 1.

Arabidopsis genes encoding enzymes of the mevalonic acid and MEP pathways, IPP isomerases, and prenyltransferases.

Figure 2.

Reaction mechanisms and enzyme products of selected Arabidopsis monoterpene synthases (a), sesquiterpene synthases (b), and diterpene synthases (c). TPS enzymes are marked by red boxes.

Since the presence of a homodimeric GPS in Arabidopsis is questionable, GPP might instead be produced by a heterodimeric type GPS. Heterodimeric or heterotetrameric GPSs identified from Mentha × piperita, Anthirrinum majus, Ciarkia breweri, and Humulus lupulus (Burke et al., 1999; Tholl et al., 2004; Wang and Dixon, 2009) are composed of a large and a small subunit. The large subunit (LSU) has significant homology (∼50%) to GGPP synthases (see below) and can exhibit GGPP synthase activity as a recombinant protein, while the small subunit (SSU) shares only ∼20% sequence similarity with homomeric prenyltransferases and is functionally inactive. Based on structural evidence obtained from heterotetrameric GPS from peppermint (Chang et al., 2010), physical interaction of both subunits is necessary to produce GPP. In Arabidopsis, a gene encoding a GGPP synthase-related protein (GGR) was identified (Table 1), which contains 2 conserved CxxxC motifs found to be essential for the interaction of GPS.SSU with GPS.LSU (Wang and Dixon, 2009). GGR modifies Arabidopsis GGPP synthase11 (see below) in vitro by increasing its GPS activity. Phylogenetically, GGR belongs to a separate lineage of GPS.SSU (SSU II). Whether proteins in this subfamily are essential for monoterpene biosynthesis is not clear since their expression is not tightly correlated with the formation of monoterpenes in contrast to SSU I proteins in monoterpenerich plants such as snapdragon and hops (Tholl et al., 2004; Wang and Dixon, 2009).

In summary, it remains somewhat unclear which enzymes specifically contribute to the formation of GPP in Arabidopsis. The biosynthesis of GPP seems to be restricted to plastids and the presence of a substantial cytosolic GPP pool is unlikely as shown at the example of the bifunctional Arabidopsis monoterpene/sesquiterpene synthase TPS02, which is located in the cytosol and only produces sesquiterpenes but no monoterpenes in vivo (see below).

Farnesyl Diphosphate Synthase

(E,E)-Farnesyl diphosphate (FPP) (Fig. 2b) synthesized by FPP synthases (FPS, EC 2.5.1.10) is a central precursor in primary terpene metabolism (biosynthesis of sterols, brassinosteroids, dolichols, and ubiquinones; protein prenylation) and specialized metabolism in the formation of sesquiterpenes and triterpenes (Fig. 1). Plant trans-FPSs are homodimeric enzymes, which belong to type I (eukaryotic) FPSs. Three Arabidopsis FPS isoforms can be distinguished, which are encoded by two genes of high sequence similarity, FPS1 and FPS2 (Table 1). FPS1 is expressed in all tissues; however, differential transcription of FPS1 leads to the formation of a long version of FPS1 mRNA (FPS1L) with primary accumulation in flowers and a short FPS1 transcript (FPS1S) having highest expression in flowers and roots (Cunillera et al., 1997, 2000a). The isoform derived from FPS1L, which carries an N-terminal extension of 41 amino acids, is targeted to mitochondria, while FPS1S most likely remains in the cytosol (Fig. 1). FPP synthesized in mitochondria may function as a precursor of sesquiterpenes. A plastidial pool of FPP has been identified based on the formation of sesquiterpenes by bi-functional monoterpene/sesquiterpene synthases targeted to plastids (Aharoni et al., 2003; Huang et al., 2010). However, plastidial FPP is most likely a by-product of other prenyltransferases such as GGPP synthase located in plastids.

Expression of FPS2, which encodes the third cytosolic FPS isoform (Fig. 1), occurs primarily in floral organs, specifically in pollen grains, at the sites of lateral root initiation, and at the junctions between primary and secondary roots and stems (Cunillera et al., 2000a). Despite these reported expression patterns, studies with FPS1 and FPS2 knock-out lines demonstrated that a single functional FPS isozyme is sufficient to support growth and development (Closa et al., 2010). However, the functions of both enzymes are not completely redundant: FPS1 is required during several stages of plant development, while FPS2 plays a major role during embryo and seedling development. Double mutations of FPS1 and FPS2 severely impair male genetic transmission and cause an arrest of early embryo development (Closa et al., 2010).

Overexpression of FPS1 leads to a metabolic imbalance resulting in necrosis coupled with increased levels of H₂O₂ and premature senescence (Masferrer et al., 2002; Manzano et al., 2004). Moreover, increased FPS activity appears to reduce the availability of IPP and DMAPP for cytokinin biosynthesis (Masferrer et al., 2002). Overexpression of HMGR together with FPS1 can rescue the necrosis phenotype indicating an insufficient supply of metabolites derived from the MVA pathway in FPS1-expressing plants (Manzano et al., 2004).

Geranylgeranyl Diphosphate Synthase

Trans-geranylgeranyl diphosphate (GGPP) synthases (GGPS, EC 2.5.1.29) represent important branch point enzymes since their enzymatic product (E,E,E)-GGPP (Fig. 2c) is a central precursor for a diverse group of primary and specialized isoprenoid compounds: carotenoids and carotenoid breakdown products including the hormones ABA and sthgolactones, chlorophylls, tocopherols, gibberellins, plastoquinones, and diterpenes (all synthesized in plastids), geranylgeranylated proteins and polyprenols (synthesized in the cytosol), and polyterpenes (synthesized in mitochondria) (Fig. 1). Consistent with the subcellular compartmentation of these diverse pathways, different GGPS isoforms are encoded by small gene families. The Arabidopsis genome contains a gene family of 11 predicted GGPSs (GGPS111, Lange and Ghassemian, 2003), of which six (GGPS1, 2, 3, 4, 7, 11) have been functionally characterized in vitro and/or by genetic complementation (Zhu et al., 1997a; Zhu et al., 1997b; Okada et al., 2000; Wang and Dixon, 2009) (Table 1). All characterized enzymes are most likely homodimers and produce GGPP as the primary or sole (GGPS2) product. GGPS1 synthesizes GGPP from IPP and FPP as the allylic substrate (Zhu et al., 1997b). Wang and Dixon (2009) recently demonstrated that GGPS6 is probably not a functional GGPS enzyme since the recombinant GGPS6 protein produces a polyprenyl diphosphate of more than 20 carbon atoms in vitro. GFP-fusion experiments indicated that GGPS7 and 11 are located in plastids, GGPS1 is targeted to mitochondria, and GGPS3 and 4 are located in the ER/cytosol (Okada et al., 2000) (Fig. 1). Further analysis of putative targeting sequences using different algorithms (Lange and Ghassemian, 2003) suggested five other GGPS isoforms to be plastidial proteins (GGPS2, 5, 8–10, 11). Besides the subcellular segregation of GGPSs, different spatial expression patterns have been observed. According to Northern and GUS reporter experiments, GGPS11 is expressed throughout the plant, while GGPS4 is expressed in flowers and root tips (Okada et al., 2000). GGPS3 promoter activity was detected in the vascular tissue and in flowers, and GGPS7 promoter activity was observed in the hypocotyl and the vascular tissue of roots (Okada et al., 2000). Microarray analyses further support different expression profiles of the other GGPS genes in the root (lateral root cap, epidermis and stele), the embryo, flowers and vascular tissue (The Bio-Array Resource for Plant Functional Genomics,  http://www.bar.utoronto.ca/). In summary, differential subcellular and tissue-specific regulation of the various Arabidopsis GGPS isoforms suggest a functional specialization of GGPS enzymes by allocating GGPP precursors and controlling metabolic flux to distinct terpene biosynthetic pathways.

Chain Length Regulation of Prenyltransferases

The mechanism that regulates the length of short-chain prenyltransferase products has been investigated by a combination of structural analyses and random or site-directed mutagenesis (Vandermoten et al., 2009). Crystal structures of homodimeric FPSs and GGPSs reported from chicken, bacteria, and protozoa, yeast, and human (Tarshis et al., 1994; Koyama et al., 2000; Hosfield et al., 2004; Chang et al., 2006; Gabelli et al., 2006; Kavanagh et al., 2006) have shown that short chain prenyltransferases are principally composed of 13 helices, ten of which surround the active site cavity. Two highly conserved aspartate-rich regions, a first DDx_2–4 motif (FARM) and a second DDxxD motif (SARM), which are essential for substrate binding and catalytic activity, are positioned on opposite walls of the cavity. Amino acid residues upstream of the FARM motif determine the chain length of the prenyl diphosphate product by changing the size of the hydrophobic binding pocket (Ohnuma et al., 1996a; Ohnuma et al., 1996b; Tarshis et al., 1996). “Bulkier” aromatic amino acid residues cause a smaller binding pocket in type I FPP synthases such as Arabidopsis FPS1 and 2 but are replaced by smaller residues (alanine, serine, methionine) in type II GGPSs, which comprise eubacterial and plant GGPSs including all Arabidopsis GGPSs. Termination of chain elongation at C₂₀ seems to be dependent on residues located deeper in the catalytic cavity as demonstrated for GGPS from yeast (Chang et al., 2006). The recently renamed Arabidopsis PPPS (former GPS1) does not possess aromatic amino acids near the FARM supporting the observed formation of longer chain products. Chain length termination in prenyltransferases is not absolute since the reactions of several enzymes such as GGPS11 show promiscuity by synthesizing shorter products in small or sometimes equal amounts. These reactions may allow evolutionary switches or transition between different chain-length prenyltransferases. Modifications of prenyltransferase activities are also obvious in heterodimeric GPSs. The recently elucidated structure of peppermint heterodimeric GPS shows that interaction of the small subunit, which does not contain a FARM motif, with the large GGPS-type subunit limits access to an elongation cavity and restricts the enzyme's specificity to the formation of a C₁₀-product (Chang et al., 2010).

Cis-lsoprenyl Diphosphate Synthases

Most cis-prenyltransferases generate prenyl diphosphate products longer than C₅₀ such as dolichols by using all-trans short chain prenyl diphosphates as allylic primer substrates (Takahashi and Koyama, 2006). However, recently a (Z,Z)-FPP synthase was identified from wild tomato, which provides (Z,Z)-FPP as substrate for the trichome-specific formation of sesquiterpenes (Sallaud et al., 2009). Likewise, the cis-prenyl diphosphate neryl diphosphate (NPP) and not GPP was demonstrated to function specifically as the substrate for a tomato monoterpene synthase (Schilmiller et al., 2009). These studies indicate that the role of short chain (Z)-prenyl diphosphates in terpene specialized metabolism has been underestimated. The Arabidopsis genome contains a total of nine genes with sequence similarity to cis-prenyltransferases. With the exception of two genes involved in dolichol biosynthesis (Cunillera et al., 2000b; Zhang et al., 2008), the function of these enzymes is largely unknown.

Arabidopsis TPSs of the TPS-a Family

Within the plant TPS superfamily, which is comprised of seven subfamilies (TPS-a, b, c, d, e/f, g, and h) (Chen et al., 2011), 22 Arabidopsis TPS proteins are most closely related to other angiosperm terpene synthases of the TPS-a subfamily (Aubourg et al., 2002; Tholl et al., 2005) (Fig. 4a). In this subfamily, Arabidopsis TPSs exhibit higher similarity to one another than to TPS proteins of other species, which indicates rapid evolution of a species-specific paralogous gene cluster, a common concept in plant TPS gene evolution. The cluster contains several “sub”-clusters of closely related TPSs (Fig. 4a), whose corresponding genes are not all located on the same chromosome indicating recombination and translocation events of these genes. The expansion of a TPS subfamily after speciation may be caused by ecological adaptations of the plant in attraction of or defense against other organisms. Because of this species-specific radiation, TPS genes of the same function can arise repeatedly in different plant species and families (Bohlmann et al., 1998; Martin et al., 2004).

All type-a Arabidopsis TPSs are class I proteins, which lack a γ-domain and carry a non-functional N-terminal domain (Fig. 3). Four type-a Arabidopsis TPS genes (TPS11, TPS12, TPS13, and TPS 21) encode proteins without plastidial transit peptides and have been characterized as sesquiterpene synthases. The compounds produced by the respective recombinant enzymes are listed in Table 2. TPS12 and its closely related enzyme TPS13 both catalyze the formation of (Z)-Y-bisabolene as a main product and α-bisabolol and (E)-nerolidol as minor products (Ro et al., 2006). The TPS12 and TPS13 genes are positioned in close proximity to one another on chromosome 4 and most likely emerged by recent tandem gene duplication (Fig. 4b). Of the remaining 18 type a genes, six (TPS01, TPS05, TPS07, TPS16, TPS17, TPS28) were found not to be expressed in flowers, leaves, and roots of Arabidopsis accession Columbia-0 (Col-0) according to RT-PCR analysis in our laboratory. However, expression of these genes at mostly low transcript levels were reported in microarray datasets of siliques, seeds, embryo development, and roots (Table 2) (The Bio-Array Resource for Plant Functional Genomics,  http://www.bar.utoronto.ca/). Proteins encoded by all 18 genes carry predicted plastidial or mitochondrial targeting sequences (Table 2) and are assumed to be diterpene synthases (or monoterpene synthases) located in plastids or sesquiterpene synthases possibly targeted to mitochondria. Biochemical characterization of TPS08 and TPS20 has revealed that these enzymes indeed function as diterpene synthases (Vaughan, Tholl et al., unpublished results), while characterization of TPS22 and its related protein TPS25 suggests that both enzymes are sesquiterpene synthases that appear to be localized to mitochondria (Huh, Tholl et al., unpublished results). Whether any of the other type-a TPS genes are functionally active in vivo remains to be determined. Recently, Shah has reported the detection of the diterpene aldehyde dehydroabietinal, which appears to function as a systemic signal at picomolar concentrations when induced by infection with an avirulent strain of Pseudomonas syringae (Shah, 2009). It is likely that the precursor of this diterpene is produced by a TPS in the type-a family. Hence, it is possible that some of the uncharacterized type-a TPSs produce volatile or non-volatile terpenes at very low amounts or concentrations below detectable levels. In addition, some type-a TPS genes might only be functionally active in specific accessions and inactive in others even though they are still transcribed.

Arabidopsis TPSs of the TPS-b Family

Phylogenetic amino acid sequence comparison places six Arabidopsis TPSs in the TPS-b subfamily, in which they form, similar to the Arabidopsis type-a TPSs, a species-specific clade (Aubourg et al., 2002; Tholl et al., 2005) (Fig 4a). The Arabidopsis TPS-b clade contains seven class I-type monoterpene synthases, all of which have been biochemically characterized: TPS02 (Huang et al., 2010), TPS03 (Fäldt et al., 2003; Huang et al., 2010), TPS10 (Bohlmann et al., 2000), TPS24 (Chen et al., 2003), and TPS-Cin encoded by the identical genes TPS23 and TPS27 (Chen et al., 2004). The terpene products synthesized by these enzymes in vitro are listed in Table 2. The genes TPS23, TPS24, and TPS27 cluster together on chromosome 3 (Fig. 4b) and have emerged from a recent gene duplication event, which is particularly evident for TPS23 and TPS27, whose coding sequences and promoter regions are 100% identical. The TPS-Cin protein encoded by TPS23 and TPS27 and the closely related TPS24 enzyme synthesize similar groups of monoterpene products but differ in their major products 1,8-cineole and (E)-β-ocimene/myrcene, respectively. Moreover, differences in the promoter sequences of the TPS24 and TPS23/27 genes lead to different expression patterns with TPS-Cin being expressed in roots and TPS24 in flowers (Chen et al., 2003; Chen et al., 2004) (see below). The closely related genes TPS02 and TPS03 encoding bifunctional (E)-βocimene/(E,E)-α-farnesene synthases represent another gene pair as a result of gene duplication (Fig. 4b). In contrast to all the other monoterpene synthases of clade I (including TPS02), which contain predicted chloroplast transit peptide sequences and are assumed to be located in plastids (confirmed for TPS02), TPS03 lacks a functional plastidial transit peptide and is localized to the cytosol (Huang et al., 2010).

Monoterpene synthases in the type-b subfamily commonly contain a conserved RR(x)₈W motif downstream of their 40 to 70 amino acid N-terminal plastidial transit peptide (Whittington et al., 2002b; Hyatt et al., 2007). Although the RR motif is thought to be required for the isomehzation of GPP to a linalyl cation in the formation of cyclic monoterpenes (Williams et al., 1998), this motif is highly conserved in all TPS type-b Arabidopsis monoterpene synthases including those that produce acyclic terpenes (Aubourg et al., 2002). In comparison to the type-b enzymes, Arabidopsis type-a TPS proteins contain a modified R(x)₉W motif, of which only the highly conserved W residue is present in the sesquiterpene synthases TPS11, TPS12, and TPS13. Interestingly, TPS06, TPS17, and TPS30 proteins carry modified versions of the highly conserved DDxxD motif (TPS06: DNTFD, TPS17: NDTCD, TPS30: NDVCD). To what extent these modifications interfere with substrate binding or impair enzyme activity is currently not known. However, it was demonstrated in goldenrod, that a naturally occurring variant of the DDxxD motif (NDxxD) in a germacrene D synthase had no impact on catalytic activity (Prosser et al., 2004).

Enzymes of Other TPS Subfamilies

Four Arabidopsis TPS genes have not, in contrast to the type-a and type-b genes, undergone duplication and belong to different TPS subfamilies. TPS14 is a (+)-S-linalool synthase that lacks the RR(x)₈W motif and is related to acyclic monoterpene synthases from snapdragon in the TPS-g subfamily (Dudareva et al., 2003) (Fig. 4a). The remaining three TPSs are all diterpene synthases and form a separate clade within the Arabidopsis TPS family (Fig. 4a): the class II-type CPP synthase, TPSGA1, the class I-type kaurene synthase, TPSGA2, both of which provide the precursors in gibberellins biosynthesis (Sun and Kamiya, 1994; Yamaguchi et al., 1998), and the class I-type geranyllinalool synthase TPS04, which produces the alcohol precursor in the biosynthesis of the volatile C₁₆-homoterpene 4,8,12-trimethyltridecatetra-1,3,7,11-ene (TMTT) (Herde et al., 2008). Recently, a cytochrome P450 monooxygenase (P450) of the CYP82 family has been identified, which converts (E,E)-geranyllinalool into TMTT via an oxidative C-C cleavage reaction (Lee et al., 2010) (Fig. 1). All three diterpene synthases contain the insertional or γ-domain of approximately 200 amino acids with an EDxxD-like motif (Fig. 3) in class II-type TPSGA1. TPSGA1 is highly similar to other CPSs in the plant TPS-c subfamily, while TPSGA2 is related to ent-KS in the recently merged TPS-e/f subfamily (Chen et al., 2011), and TPS04 shows highest similarity to S-linalool synthases from Clarkia breweri and Clarkia concinna in the TPS-e/f family (Herde et al., 2008).

Table 2.

Properties of Arabidopsis TPS genes and proteins and TPS enzymatic products.

The TPSGA1, TPSGA2, and TPS04 genes contain 12 to 15 exons (Table 2) similarly to the gene structure of the presumed progenitor gene of all plant TPS (Bohlmann et al., 1998; Trapp and Croteau, 2001; Aubourg et al., 2002). Four exons interrupted by three introns code for the insertional (γ) domain. Consecutive loss of exons and introns including those of the insertional domain most likely gave rise to the emergence of modern class I type TPSs, which contain seven exons (all Arabidopsis type-a and type-b TPS except for TPS22 with 6 exons) (Trapp and Croteau, 2001; Aubourg et al., 2002) (Table 2). Exon/intron positions and the order of intron phases are conserved among the paralogous type-a and type-b Arabidopsis genes. The larger gene sizes of TPSGA1, TPSGA2, and TPS04 correspond to molecular weights between 90 and 100 kDa in contrast to type-a and type-b proteins with sizes between 60 and 70 kDa. The fact that all three diterpene synthase genes are single-copy genes and have not emerged from paralogous gene clusters indicates their function in evolutionary early terpene biosynthetic pathways, which is supported by the assumed emergence of angiosperm and gymnosperm CPS and KS from a bi-functional CPS/KS prototype as found in P. patens (Hayashi et al., 2006).

Figure 4.

Clades of the Arabidopsis TPS family (a) and chromosomal Arabidopsis TPS gene clusters (b).

(a) Arabidopsis TPS amino acid sequences were aligned by the MUSCLE algorithm and a tree was built with the GENEIOUS program (Biomatters Ltd.). Bootstrap values larger than 50% are shown. Colors indicate enzymes of TPS subfamilies a (green), b (purple), c (dark blue), e/f (light blue), and g (orange). Functionally characterized TPS proteins are shaded, (b) TPS pseudogenes (with AGI number) or gene fragments (non-annotated) are marked with grey arrows. Numbers indicate distances in kb between genes.

While TPSGA1, TPSGA2, and other diterpene synthases are targeted to the plastid, the TPS04 protein does not contain an obvious plastid transit peptide sequence. Fusion studies with yellow fluorescent protein (YFP) indicated that TPS04 is localized in the cytosol, where it uses GGPP either produced by a cytosolic GGPS or exported from chloroplasts as the substrate (Fig. 1) (Herde et al., 2008). The latter assumption is supported by the severely reduced formation of TMTT in the presence of the MEP pathway inhibitor fosmidomycin (Lee, Tholl et al., unpublished results).

Gene Clusters and Biosynthetic Modules

Repeated gene duplications in the Arabidopsis TPS gene family gave rise to the formation of six small gene clusters each containing two to five genes positioned head to tail to each other by a close distance (0.04 to 6 kb) (Fig 4b). Three of these clusters contain pseudogenes or TPS gene fragments (Fig 4b). A total of eight pseudogenes (plus the TPS02 pseudogene in the Col-0 accession, see below) have been identified in the Arabidopsis genome (Aubourg et al., 2002), all of which exhibit disruptions of the coding sequence by deletions or insertions with four genes consisting of only a partial number of exons. Pseudogenes in the TPS17/TPS18/TPS19 cluster are very similar in sequence to the intact TPS genes indicating their origin by gene duplication. Frequent pseudogenization of Arabidopsis TPS genes may have arisen by genomic instability caused by insertion of transposon-like elements, several of which are positioned in close proximity to TPS genes or occur in a gene cluster (Aubourg et al., 2002).

Besides their tendency to form TPS gene clusters, several TPS genes also appear to cluster with GGPS genes as well as genes encoding P450s or glycosyltransferases that may catalyze secondary transformations of the primary terpene compounds. For example, the (Z)-Y-bisabolene synthase genes TPS12 and TPS13 each cluster head-to-head with a P450 gene located on the opposite strand sharing the same promoter region (Fig 4b). Both P450 genes CYP71A19 (At3g13290) and CYP71A20 (At4g13310) share 88% amino acid sequence identity and were duplicated in pair with the respective TPS gene. According to GUS-reporter gene analyses and microarray expression profiles (Birnbaum et al., 2003), CYP71A19 and CYP71A20 are co-expressed with TPS12 and TPS13 in the root cortex/endodermis suggesting that the corresponding enzymes form modules in the biosynthesis and possible hydroxylation of (Z)-Y-bisabolene. Other functional TPS-P450 gene clusters have been suggested for genes TPS05, TPS09, TPS11, TPS22 (Aubourg et al., 2002). The seed-specific transcription of TPS05 correlates well with that of the clustering gene CYP81D6 but is different from expression patterns of neighboring glucosyltransferase genes. Another gene pair with organ-specific co-expression in flower ovaries consists of the multi-sesquiterpene synthase gene TPS11 and the clustering P450 gene CYP706A3, which may be involved in the oxidation of one or several of the TPS11 sesquiterpene products. The other P450 genes that cluster with the root-expressed genes TPS22 and TPS09 show no or only low expression in root tissue.

The GGPP synthase genes GGPS5, GGPS6, and GGPS7, each of which has apparently been duplicated with genes TPS22, TPS24, and TPS25, respectively, share overlapping expression with these genes in siliques and developing seeds but have highest transcript levels in roots, where the clustering TPS genes are not expressed (The Bio-Array Resource for Plant Functional Genomics,  http://www.bar.utoronto.ca/). Moreover, the recombinant TPS22 and 25 proteins have been shown to have no diterpene synthase activity in vitro (Huh, Tholl, unpublished results) indicating that these enzymes are most likely not part of a functional biosynthetic module. Finally, the root-specific expression of genes GGPS9 and GGPS10 correlates to some extent with that of their clustering genes TPS19 and TPS20 although expression of the genes of each pair does not synchronize in the same cell type (The Bio-Array Resource for Plant Functional Genomics,  http://www.bar.utoronto.ca/).

Constitutive Formation of Volatile Terpenes in Flowers

By far, the largest number of terpene compounds produced by Arabidopsis has been detected in floral tissues. Flowers of the Col-0 accession emit a complex mixture of over 20 sesquiterpene hydrocarbons with (E)-β-caryophyllene as the predominant compound (Chen et al., 2003; Tholl et al., 2005). Two enzymes, TPS11 and TPS21, are responsible for the formation of nearly all of the flower-specific sesquiterpenes: TPS21 synthesizes primarily (E)-β-caryophyllene and α-humulene, and TPS11 catalyzes the formation of essentially all of the remaining sesquiterpenes (Tholl et al., 2005) (Table 2, Fig. 5a). In contrast to floral scent emissions from many other angiosperms, Arabidopsis floral volatiles are not produced in flower petals. Promoter-GUS reporter gene experiments and microarray datasets show that TPS11 is primarily expressed in intrafloral nectaries and ovules, while expression of TPS21 occurs largely in the stigma of open flowers and to some extent in sepals (Fig. 5a). Formation of volatile terpenes in these flower organs can principally serve two different functions: Volatiles may attract small pollinating insects although only at a short distance since the emission of the floral volatile blend is fairly low compared to highly scented flowers (Chen et al., 2003). Cross-fertilization leading to more than 1% outcrossing events in natural populations may stabilize populations by reducing inbreeding depression (Abbott and Gomes, 1989). Second, the specific expression pattern of both TPS genes indicates possible defensive activities of terpenes in these organs. Especially, nectaries and the stigma represent sugar-containing and/or moist areas that can promote bacterial or fungal growth (Buban et al., 2003). Evidence for the role of (E)-β-caryophyllene in inhibiting the growth and invasion of microbial pathogens in the floral stigma has been obtained by the analysis of TPS21 gene knock-out plants impaired in (E)-β-caryophyllene emission (Huang, Tholl et al., unpublished results).

Figure 5.

Tissue- and cell-type specific Arabidopsis TPS gene expression patters according to RT-PCR and promoter-GUS reporter gene analyses.

Spatial expression is shown for flowers (a), insect-damaged leaves (b), and roots (c). Arrows in (c) mark gene expression in specific root cell types. Numbers indicate different TPS genes. Only those genes are shown, for which enzymatic products have been detected in vivo.

Besides the emission of sesquiterpenes, small amounts of the monoterpenes β-myrcene, limonene, and linalool have been detected in Col-0 flowers. Three monoterpene synthases, the multiproduct monoterpene synthase TPS24 (Fig. 5a), the β-myrcene/ (E)-β-ocimene TPS10, and the linalool synthase TPS14, all of which are expressed in Arabidopsis flowers (Chen et al., 2003), contribute to some degree to the formation of these compounds (Table 2). Other terpene volatiles emitted by Col-0 flowers in low amounts are (E,E)-α-farnesene and the homoterpene TMTT (Tholl et al., 2005; Huang et al., 2010), which are produced by TPS03 (Fig. 5a) and via the homoterpene-specific pathway, respectively (see below). The low emission of these compounds and of the detected monoterpenes, in most cases, does not correlate with the transcript levels of the corresponding TPS genes and may be related to substrate limitation and/or posttranslational protein modifications.

Induced Formation of Terpenes in Leaves

Most of the terpenes detected in Arabidopsis leaves in response to biotic stress have been of volatile nature although recent findings by Shah indicate the formation of a diterpene aldehyde as a potential systemic signal in microbial pathogen infection (Shah, 2009). This finding resembles that of a diterpene alcohol as an endogenous defense signal in tobacco in response to infection by tobacco mosaic virus (Seo et al., 2003).

Under physiologically normal growth conditions, leaves of Arabidopsis Col-0 plants release only traces of volatile terpenes. However, treatment with the fungal peptide elicitor alamethicin, application of jasmonate and the jasmonate-isoleucine mimic coronalon, and feeding by the crucifer specialist insects Pieris rapae and Plutella xylostella induce the emission of a volatile blend consisting of green leaf volatiles, the benzenoid compound methyl salicylate, and the C₁₆-homoterpene TMTT and the sesquiterpene (E,E)-α-farnesene as the primary terpene constituents (Van Poecke et al., 2001; Herde et al., 2008; Snoeren et al., 2010). Snoeren et al. (2010) observed the additional release of five monoterpenes (3-carene, linalool, β-myrcene [also detected by Van Poecke et al., 2001], (Z)-β-ocimene, α-phellandrene), the sesquiterpene (E)-nerolidol, and the C₁₁-homoterpene 4,8-dimethylnona-1,3,7-triene (DMNT) from pools of Col-0 plants treated with P. rapae and/or jasmonate, although no specific emission rates of these compounds were reported. Also, the apocarotenoids β-ionone and cyclocitral were detected in response to insect feeding (Van Poecke et al., 2001). The observed composition of the induced volatile mixtures depends on the type of stressor (type of herbivore, chemical treatment) but can also vary with conditions of plant growth and volatile collection (Tholl, 2006). Furthermore, depending on the accession, volatile profiles vary qualitatively and quantitatively. For example, (E)-β-ocimene represents a common induced volatile in other accessions (see below). Van Poecke et al. (2001) and Loivamäki et al. (2008) demonstrated a possible role of herbivore-induced volatiles in plant indirect defense via the attraction of parasitoids such as Cotesia rubecula and Diadegmasemiclausum, which parasitize P. rapae larvae. In the laboratory, attack by C. rubecula can increase plant fitness in terms of seed production (van Loon et al., 2000). A possible role of homoterpenes such as TMTT in the attraction of herbivore predators has been suggested by several studies (de Boer et al., 2004; Mumm et al., 2008). In lima bean, TMTT has been characterized for its ability to mediate plant-plant interactions by inducing the transcription of defense genes such as LOX2 and PDF2 in un-attacked neighboring plants (Arimura et al., 2000).

The first committed step in the biosynthesis of TMTT is catalyzed by the (E,E)-geranyllinalool synthase TPS04 (GES) of the TPS-f subfamily (Herde et al., 2008). Expression of the TPS04 gene is induced locally by insect feeding (Fig. 5b) and elicitor treatment and its expression is dependent on the jasmonate signaling pathway but independent of salicylic acid and ethylene. TPS04 is tightly co-expressed with the recently identified P450 gene CYP82G1 (At3g25180) responsible for the oxidative breakdown of (E,E)-geranyllinalool (Lee et al., 2010) resulting in an efficient conversion of the alcohol intermediate to the homoterpene end product.

In the Col-0 accession, insect-induced formation of (E,E)-α farnesene is catalyzed by the bi-functional mono-/sesquiterpene synthase TPS03 (see under variation of terpene biosynthesis) (Huang et al., 2010). TPS03 is expressed locally at wound sites (Fig. 5b) but also constitutively in flowers, where it contributes to low emissions of (E,E)-α-farnesene. Expression of TPS03 occurs in sepals, anthers (particularly in pollen), and in the stigma of immature and mature flowers (Fig. 5a) (Huang et al., 2010). The flower-specific expression profile of TPS03 suggests a role of (E,E)-α-farnesene in pollinator attraction or florivore/antimicrobial defense similar to that discussed for the other flower-expressed genes TPS21 and TPS11. (E)-β-Ocimene, the monoterpene analog of (E,E)-α-farnesene, which is emitted in response to herbivory by other Arabidopsis accessions such as Wassilewskija (Ws), was shown to be produced by TPS02, the enzyme closely related to TPS03 (see variation of terpene biosynthesis) (Huang et al., 2010). Similar to TPS03, the TPS02 gene shows a dual expression pattern - insect-induced expression in leaves and constitutive expression in floral tissues - consistent with the induced formation and floral release of (E)-β-ocimene in accessions emitting this terpene.

The formation of β-myrcene induced by feeding of P. rapae (Van Poecke et al., 2001; Snoeren et al., 2010) can most likely be attributed to the β-myrcene/(E)-β-ocimene synthase TPS10, which is induced under these conditions (Van Poecke etal., 2001) (Fig. 5b). The other monoterpenes detected by Snoeren et al. (2010) might be produced by uncharacterized TPS enzymes of the TPS-a family or by TPS-b monoterpene synthases, although the reported compounds are not primary products of these enzymes and/or the respective genes have not been found to be induced by insect feeding.

Root-Specific Formation of Terpenes

RT-PCR expression profiles (Chen et al., 2003) and detailed root cell transcriptome analyses by Birnbaum et al. (2003) and Brady et al. (2007) revealed a considerable number of 14 TPS genes with primary or exclusive expression in Arabidopsis roots (Table 2). This group of root-specific genes includes the two identical 1,8-cineole synthase (TPS-Cin) genes TPS23 and 7PS27and 12 genes of the TPS-a family. The CPP synthase TPSGA1 and the kaurene synthase TPSGA2 of the gibberellin biosynthetic pathway are also expressed in roots with highest transcript signals in the quiescent center, but their expression is not limited to root tissues. According to high resolution spatio-temporal maps of the root transcriptome (Brady et al., 2007) supported by promoter-GUS reporter gene studies, several of the root-specific TPS genes exhibit cell type-specific expression patterns in root zones of different developmental stages. For example, the TPS-Cin genes TPS23 and TPS27 are expressed in the stele (vascular tissue) of the root elongation zone and in epidermal cells and root hairs of older root-growth zones suggesting a direct diffusion or secretion of the synthesized monoterpene into the rhizosphere (Chen et al., 2004) (Fig. 5c). A similar dual-expression pattern was observed in promoter-GUS analyses of the (Z)-Y-bisabolene synthase genes TPS12 and TPS13 with expression in the stele of younger roots and in the cortex and endodermis of mature roots (Ro et al., 2006) (Fig. 5c). Furthermore, transcripts of the predicted diterpene synthase gene TPS08 were reported in the root vasculature, more specifically in the xylem pole pericycle and the procambium (Fig. 5c). Gene-coexpression profiles suggest that root-specific TPS genes form biosynthetic modules with up- or downstream enzymes expressed in the same cell type such as specific GGPS isoenzymes or P450 enzymes catalyzing secondary transformations. One such example is the duplicated gene clusters of TPS12 and TPS13 and the P450 genes CYP71A19 and CYP71A20 described above.

The primary products of the described root-specific TPSs have been detected in different culture systems (Steeghs et al., 2004; Vaughan, Tholl et al., unpublished results) and provide grounds for testing terpene chemical defenses in Arabidopsis roots. The cell-specific nature of terpene metabolism in Arabidopsis roots may emerge in defense against soil-borne, root-attacking organisms such as nematodes, microbial pathogens, and insect larvae, which have different feeding or infection strategies that target different cells and tissues. These constitutive defenses are enhanced or supported by additional induced responses depending on the root-attacker (Sohrabi and Tholl, unpublished results).