The family of terpene synthases in plants: a mid-size family of genes for specialized metabolism that is highly diversified throughout the kingdom


(fax +1 865 974 1947; e-mail


Some plant terpenes such as sterols and carotenes are part of primary metabolism and found essentially in all plants. However, the majority of the terpenes found in plants are classified as ‘secondary’ compounds, those chemicals whose synthesis has evolved in plants as a result of selection for increased fitness via better adaptation to the local ecological niche of each species. Thousands of such terpenes have been found in the plant kingdom, but each species is capable of synthesizing only a small fraction of this total. In plants, a family of terpene synthases (TPSs) is responsible for the synthesis of the various terpene molecules from two isomeric 5-carbon precursor ‘building blocks’, leading to 5-carbon isoprene, 10-carbon monoterpenes, 15-carbon sesquiterpenes and 20-carbon diterpenes. The bryophyte Physcomitrella patens has a single TPS gene, copalyl synthase/kaurene synthase (CPS/KS), encoding a bifunctional enzyme producing ent-kaurene, which is a precursor of gibberellins. The genome of the lycophyte Selaginella moellendorffii contains 18 TPS genes, and the genomes of some model angiosperms and gymnosperms contain 40–152 TPS genes, not all of them functional and most of the functional ones having lost activity in either the CPS- or KS-type domains. TPS genes are generally divided into seven clades, with some plant lineages having a majority of their TPS genes in one or two clades, indicating lineage-specific expansion of specific types of genes. Evolutionary plasticity is evident in the TPS family, with closely related enzymes differing in their product profiles, subcellular localization, or the in planta substrates they use.


Many plants are truly autotrophic organisms. Using only carbon dioxide, minerals, water, and sunlight, they are able to synthesize all of the metabolites necessary for their growth and development, often referred to as ‘primary’ metabolites. The very existence of autotrophic plants has opened up ecological niches for organisms – animals, fungi, bacteria, and even parasitic plants – that do not need to synthesize all the components of their own primary metabolism but instead obtain them by consuming plants (or other organisms further up the food chain). Conversely, plants often depend on other organisms for optimal growth and completion of their life cycle. Selection on plants for greater fitness is likely to have played a major role in the evolution of a suite of biochemical pathways in plants, the so-called ‘secondary’ metabolism, whereby plants synthesize chemicals that can be harmful or beneficial to other organisms (Fraenkel, 1959). For example, plants produce an arsenal of toxic compounds that serve as defense against pathogens or herbivores, and plants also produce an array of compounds that attract other organisms which, while feeding on the plants, provide benefits to the plants (e.g., pollinators, nitrogen-fixing bacteria, and mycorrhizal fungi) (Pichersky and Gang, 2000; Pichersky and Gershenzon, 2002; Hartmann, 2007).

Since each plant species occupies a unique ecological niche, it is not surprising that it typically evolves the ability to synthesize a different set of such compounds to help it interact with its biotic and even its abiotic environment. This part of plant metabolism, now referred to as ‘specialized’ metabolism because it represents specific solutions (i.e., adaptations) to specific ecological problems, is therefore not wholly shared by all plant lineages. By definition, specialized metabolism is the most diverse part of plant metabolism. Furthermore, because other organisms evolve new responses to plant specialized metabolites in the process of adaptive selection, the same process of selection for higher fitness in the plant leads in turn to the appearance of yet new specialized metabolites, a veritable arms race. The result is that plants, in the aggregate, are known to be able to synthesize several hundred thousand different specialized metabolites, and the actual number might be much higher (Hartmann, 2007). However, it is clear that any given plant species can synthesize only a small fraction of such compounds.

Often, plant genomes contain sets of related genes (i.e., ‘gene families’) that encode enzymes that use similar substrates and give similar products, but have clearly diverged in different lineages. The terpene synthase (TPS) family in plants is such an example (Bohlmann et al., 1998a). Some estimates suggest that more than 25 000 terpene structures may exist in plants (McCaskill and Croteau, 1997). Analysis of the several plant genomes that have been sequenced and annotated indicates that, with the exception of the moss Physcomitrella patens, which has a single functional TPS gene, the TPS gene family is a mid-size family, with gene numbers ranging from approximately 20 to 150 in sequenced plant genomes analyzed in this article (Table 1). Here we discuss present knowledge of the TPS family in plants–the number and organization of the genes in each genome, function of specific genes, and the evolution of novel biochemical functions in different lineages.

Table 1.   Sizes of the TPS family and subfamilies in seven model plant genomes
SpeciesGenome size (Mb)Chromosome number (1N)Total TPS gene modelsaPutative full length TPSsbTPS subfamilyc
  1. aTotal number of TPS gene models identified in this analysis as described in Data S1 unless noted otherwise.

  2. bThe number of gene models encoding proteins larger than 500 amino acids and therefore likely to encode full length TPS proteins.

  3. cThe number of putative full-length TPS genes in each subfamily. The columns ‘(e)’ and ‘(f)’ list the number of TPS genes in each species according to the original classification. The subfamilies TPS-e and TPS-f are merged to form a new subfamily designated as TPS-e/f.

  4. dThese may contain allelic variants of same genes.

  5. eThese numbers were based on the analysis presented by Martin et al. (2010).

  6. fThese numbers were based on the analysis presented by Aubourg et al. (2002).

P. patens48027410010(0)(0)000
S. moellendorffii1062718d14d0030(3)(0)308
V. vinifera48719152e69e301920(1)(0)1170
P. trichocarpa410196832131220(2)(1)320
A. thaliana125540f32f22610(1)(1)210
O. sativa38912573418030(9)(0)920
S. bicolor73010482415210(3)(0)330

Substrates and products of TERPENE SYNTHASES

Terpenoids (isoprenoids) are any compounds that are derived from the isomeric 5-carbon building blocks isopentenyl diphosphate (IPP) and dimethylallyl diphosphate (DMAPP). Two independent pathways in plants, the MEP pathway operating in plastids and the mevalonic acid pathway operating in the cytosol/peroxisomes, lead to the synthesis of both IPP and DMAPP (Lichtenthaler, 1999; Sapir-Mir et al., 2008). These building blocks are combined with each other by the activity of prenyl transferases, which are also known as isoprenyl diphosphate synthases (Wang and Ohnuma, 2000). The condensation of one DMAPP molecule with one IPP molecule in what is called a head-to-tail condensation gives geranyl diphosphate (GPP), if in the trans configuration, or neryl diphosphate (NPP), if in the cis-configuration. Adding another IPP unit to GPP will give farnesyl diphosphate (FPP) in the trans configuration or in the cis configuration, and another IPP unit will give geranylgeranyl diphosphate (GGPP) (Takahashi and Koyama, 2006). Longer prenyl diphosphates are also synthesized and some are ultimately added (with the loss of the diphosphate moiety) as components to form such compounds as chlorophylls, phylloquinones, and ubiquinones (McGarvey and Croteau, 1995).

Some terpenoids produced by the activity of prenyl transferases and TPSs are found in all plants and therefore belong to ‘primary,’ or general, metabolism. For example, head-to-head condensation of two FPP molecules leads to the production of squalene (with the loss of both diphosphate groups), the precursor of sterols. A similar condensation of two GGPP molecules gives phytoene, the precursor of carotenoids. GGPP is also a precursor of gibberellins; a pair of structurally related TPS enzymes in angiosperms and gymnosperms, copalyl diphosphate synthase (CPS) and kaurene synthase (KS), convert GGPP first to copalyl diphosphate (CPP), then to ent-kaurene, the precursor of all plant gibberellins (Yamaguchi, 2008) (Figure 1).

Figure 1.

 An outline on the formation of terpenoids catalyzed by various types of terpene synthases.
Isopentenyl diphosphate (IPP) is the common precursor of all terpenes. It is synthesized by both the cytosol-localized mevalonic acid (MVA) pathway and the MEP pathway in plastids. IPP is isomerized to give dimethylallyl pyrophosphate (DMAPP). DMAPP either serves as the substrate for hemiterpene biosynthesis or fuses with one IPP unit to form geranyl diphosphate (GPP). The condensation of one GPP molecule with one IPP molecule gives farnesyl diphosphate (FPP), and the condensation of one FPP molecule with one IPP molecule will give geranylgeranyl diphosphate (GGPP). GPP, FPP and GGPP are the precursor for monoterpenes, sesquiterpenes and diterpenes, respectively. While these prenyl diphosphates in the trans-configuration have been believed to be the ubiquitous natural substrates for terpene synthases, recent studies showed that two prenyl diphosphates in the cis-configuration, neryl diphosphate (NPP) and Z,Z-FPP, are also the naturally occurring substrates of terpene synthases. Isoprene synthase, monoterpene synthases, sesquiterpene synthases, and diterpene synthases convert DMAPP, GPP (or NPP), FPP (or Z,Z-FPP), and GGPP to isoprene, monoterpenes, sesquiterpenes and diterpenes, respectively. In general, the biosynthesis of isoprene, monoterpenes, and diterpenes occurs in the plastid and the biosynthesis of sesquiterpenes occurs in the cytosol. One exception exists for sesquiterpene synthases: one plastid-localized TPS in tomato produces sesquiterpenes derived from Z,Z-FPP (Sallaud et al., 2009). Diterpene synthases can be divided into two groups: bifunctional (bi) and monofunctional (mono). Bifunctional diterpene synthases catalyze the consecutive conversion of GGPP to copalyl diphospate (CPP) then CPP to stable diterpenes. Monofunctional diterpene synthases catalyze the formation of diterpenes from CPP or directly from GGPP without a CPP intermediate.

In addition to CPS and KS, the plant TPS family consists of enzymes that use DMAPP to make isoprene (C5), GPP or NPP to make various cyclic and acyclic monoterpene hydrocarbons (C10) and monoterpene alcohols, FPP to make various cyclic and acyclic sesquiterpene hydrocarbons (C15) and sesquiterpene alcohols, and GGPP, CPP, or hydroxyl-CPP to make various cyclic and acyclic diterpene hydrocarbons (C20) and diterpene alcohols (Tholl, 2006) (Figure 1). Based on the reaction mechanism and products formed, plant TPSs can be classified into two groups: class I and class II. CPS is a representative of class II TPSs. It catalyzes the formation of CPP through protonation-induced cyclization of GGPP. However, most known plant TPSs are class I TPSs. In the initial step of the enzymatic reactions catalyzed by class I TPSs, the prenyl diphosphate is ionized and carbocation intermediates are formed. A striking feature of class I TPS enzymes is that, because of the stochastic nature of bond rearrangements that follow the creation of the unusual carbocation intermediates, a single TPS enzyme using a single substrate often gives rise to multiple products (Steele et al., 1998; Christianson, 2006; Degenhardt et al., 2009a). For example, AtTPS-Cin from Arabidopsis thaliana catalyzes the formation of 10 monoterpenes with 1,8-cineole being the most abundant product (Chen et al., 2004). The reaction that converts GPP to the final terpene products catalyzed by AtTPS-Cin starts with the metal ion-dependent ionization of GPP. The resulting carbocation can undergo cyclizations, hydride shifts and other rearrangement before the reaction is terminated via deprotonation or nucleophile capture (often water) (Figure 2).

Figure 2.

 The reaction mechanism of a representative monoterpene synthase, AtTPS-Cin from Arabidopsis (Chen et al., 2004).
The multiple products of AtTPS-Cin are enclosed in dashed boxes. In the reaction, the production of the initial cyclic species, the α-terpinyl cation, opens the door for the secondary cyclizations. The production of acyclic monoterpenes may proceed via either the geranyl cation or the linalyl cation. OPP represents the diphosphate moiety.

Many of the direct products of TPS enzymes in specialized metabolism, namely isoprene, mono- and sesquiterpenes, and some diterpenes, are volatile compounds under the temperature and atmospheric conditions of the environment in which plants live. Modification of the TPS products by oxidation, peroxidation, methylation, acylation, or cleavage changes their physical properties and may alter their biological activities. The terpenes produced directly by TPS enzymes and those obtained after additional modification reactions may serve various ecological roles such as pollinator attraction (Pichersky and Gershenzon, 2002), attraction of insect predators of herbivores (Unsicker et al., 2009), and chemical and physical barriers to feeding or ovipositing insects (Keeling and Bohlmann, 2006a; Heiling et al., 2010), among others. In addition, specialized products of TPSs and their derivatives are widely used as pharmaceuticals (e.g., taxol derived from taxadiene), flavors (e.g., menthol derived from limonene) and fragrances (e.g., santalols derived from santalene), or as industrial materials (e.g., diterpene resin acids), and some are being explored as biofuel precursors (e.g., farnesene) (Bohlmann and Keeling, 2008).

Basic structure of TPS genes and enzymes: TPS genes are likely derived from a duplication of an ancestral gene encoding a bifunctional kaurene synthase

The genome of the moss Physcomitrella patens contains a single functional TPS gene, a bifunctional CPS/KS, with homology to CPS, KS, and TPS genes from gymnosperms and angiosperms (Hayashi et al., 2006). This TPS catalyzes the formation of ent-kaurene and 16-hydroxykaurene in vitro (Hayashi et al., 2006; Anterola et al., 2009). The N-terminal domain of the protein has features of the active site of a CPS, (i.e., features of a class II diterpene synthase), while the KS activity (a class I diterpene synthase) seems to be located in the C-terminal half of the protein (Hayashi et al., 2006). P. patens utilizes a diterpene metabolite derived from ent-kaurene as an endogenous developmental regulator, which provides insights into the evolution of gibberellin functions in land plants (Hayashi et al., 2010). Sequence analysis indicates that CPS and KS in both gymnosperms and angiosperms are derived from a duplication of an ancestral CPS/KS gene followed by subfunctionalization that involved the loss of activity in one of these domains in one gene and the loss of the other domain in the other duplicate gene (Keeling et al., 2010). In addition to plants, some fungi and bacteria are known to produce gibberellins. In fungi, as in the moss P. patens, the biosynthesis of ent-kaurene, the gibberellin precursor from GGPP, is catalyzed by bifunctional enzymes (Toyomasu et al., 2000). In contrast, in the legume-associated bacterium Bradyrhizobium japonicum, the production of gibberellin precursor ent-kaurene from GGPP is catalyzed by two separate monofunctional diterpene synthases, CPS and KS (Morrone et al., 2009), resembling gibberellin biosynthesis in higher plants. Sequence analysis suggests that the diterpene synthases associated with gibberellin biosynthesis, as well as those of related diterpenoids, in plants, fungi, and bacteria may share a common evolutionary origin (Morrone et al., 2009).

Similar scenarios of gene duplication and subfunctionalization, as well as neofunctionalization, can be invoked for the evolution of TPS genes and enzymes of specialized metabolism from an ancestral CPS/KS gene prototype (Trapp and Croteau, 2001; Keeling et al., 2010). Some gymnosperm diterpene synthases, such as abietadiene synthase from grand fir (Abies grandis), retain both class I (KS-type) and class II (CPS-type) functional domains and bifunctional properties, catalyzing the formation of an enzyme-bound CPP from GGPP and then converting CPP to a diterpene (Peters et al., 2000). In angiosperms, all diterpene synthases that have been characterized to date are monofunctional, with loss of activity in one domain or the other. Some of these monofunctional diterpene synthases (class II) catalyze the formation of ent-CPP, syn-CPP, or hydroxy-CPP from GGPP, while others (class I) use one of these intermediates to make a diterpene hydrocarbon (Cho et al., 2004; Otomo et al., 2004a; Prisic et al., 2004; Falara et al., 2010). However, in both gymnosperms and angiosperms some diterpene synthases have been found that have retained class I activity only (with or without the actual loss of the N-terminal class II domain) and use GGPP to directly produce a diterpene without a CPP intermediate (Mau and West, 1994; Wildung and Croteau, 1996; Herde et al., 2008; Köksal et al., 2010b).

In contrast to the diterpene synthases, which exist both in the form of bifunctional and monofunctional enzymes, all known mono- and sesquiterpene synthases are believed to be monofunctional, having retained only one active site corresponding to the KS-type domain or class I activity of the CPS/KS gene prototype. In most cases, the actual CPS-type domain has been substantially deleted (and not just rendered inactive), but some exceptions have been found, which probably represent recent evolution from diterpene synthases (Dudareva et al., 1996; Sallaud et al., 2009; Schilmiller et al., 2009; and see below).

The topic of the structure and function of plant TPS proteins has been examined in a number of excellent recent reviews (e.g., Christianson, 2006, 2008; Degenhardt et al., 2009a). For details on the three-dimensional structures, reaction mechanisms, catalytic plasticity, and diversity of product profiles of plant TPSs, with emphasis on class I TPS from angiosperms and gymnopserms, we refer the reader to the representative work by Starks et al. (1997), Whittington et al. (2002), Greenhagen et al. (2006), Yoshikuni et al. (2006), Hyatt et al. (2007), Kampranis et al. (2007), O’Maille et al. (2008), and Tantillo (2010). X-ray structural analyses of plant TPSs have been performed with angiosperm and gymnosperm class I TPS proteins (see below). Although a three-dimensional structure of a functional class II plant diterpene synthase has not yet been reported, work on the gymnosperm class I/class II multi-product abietadiene synthase (Peters et al., 2001, 2003) provided substantial insights into the structural organization and interactions of the two active sites and reaction mechanisms of a plant TPS that closely resembles an ancestral bifunctional enzyme. Of particular interest in the context of the evolution of the plant TPS family is a model for the origin of class I/class II diterpene synthase structural and functional domains recently presented by Cao et al. (2010) and supported by X-ray structure analysis of taxadiene synthase, a gymnosperm class I diterpene synthase (Köksal et al., 2010b). Since there is much evidence that gymnosperm and angiosperm mono- and sesquiterpene synthases evolved from an ancestral diterpene synthase, the model presented by Cao et al. (2010) serves as a unifying, general framework, which can explain the evolution of protein domain structures of all plant TPSs. Briefly, in agreement with their common origin, all gymnosperm and angiosperm TPSs share a common overall structure derived from a three-domain ancestral TPS resembled by the gymnosperm abietadiene synthase (which in turn resembles P. patens PpCPS/KS). The three domains of an ancestral plant diterpene synthase can be related further back to domains of bacterial and fungal TPSs and trans prenyl transferases (Cao et al., 2010). Variations of the three-domain structure in many of the well characterized enzymes of the plant TPS family can be explained by loss of a particular domain such as the KS- or CPS-type domain or by loss of activity associated with a particular domain in different lineages of TPS evolution. These variations also account for the variable length of plant TPSs of approximately 600–900 amino acids.

The class I activity, which is found in all plant TPSs, except for monofunctional CPS, resides in the C-terminal domain and includes the ‘DDXXD’ and ‘NSE/DTE’ motifs for metal-dependent ionization of the prenyl diphosphate substrate. The class II activity resides in a separate domain and contains, in its active form, a ‘DXDD’ motif for the protonation-initiated cyclization of GGPP to CPP. The class II active site domain is non-functional in those TPSs for which three-dimensional structures have been obtained (Starks et al., 1997; Whittington et al., 2002; Hyatt et al., 2007; Kampranis et al., 2007; Gennadios et al., 2009; Köksal et al., 2010a,b). A third, functionally less well characterized TPS domain (Cao et al., 2010; Köksal et al., 2010b), was first identified in plants with a sequence of approximately 200 amino acids near the N-terminus (Bohlmann et al., 1998a). This third domain contains variations of an ‘EDXXD’ motif, which in its active form contributes to class II diterpene synthase activity (Cao et al., 2010; Köksal et al., 2010b). All TPSs containing this third domain are members of the TPS-d3, -c, -e, -f clades of the TPS family (Martin et al., 2004; also see below). In agreement with an ancestral origin of this domain (Bohlmann et al., 1998a; Cao et al., 2010; Köksal et al., 2010b), these clades represent a diverse spectrum of angiosperm and gymnosperm TPSs including diterpene, sesquiterpene, and monoterpene synthases.

In contrast to the cytosolic sequiterpene synthases, most mono- and diterpene synthases have obvious N-terminal plastid transit peptides. Transit peptides are presumably cleaved off in the mature TPS upstream of the ‘RRX8W’ motif, which is essential for catalysis of monoterpene cyclization (Whittington et al., 2002; Hyatt et al., 2007) and is also conserved with variations in most sesquiterpene- and diterpene synthases. The model of common ancestry of all plant TPSs, as reconstructed on the level protein domain structure by Cao et al. (2010) and supported by recent x-ray structure analysis of a gymnosperm diterpene synthase (Köksal et al., 2010b), is also supported on the level of genomic TPS sequences with patterns of conserved intron positions across many TPSs from bryophytes, gymnosperms and angiosperms representing all major TPS subfamilies (Trapp and Croteau, 2001; Aubourg et al., 2002; Keeling et al., 2010; Martin et al., 2010), although exceptions from conserved intron-exon structures have been reported (Lee and Chappell, 2008).

Comparative genomics of the TPS family among diverse plant taxa

We present here an analysis of the TPS family (which formally includes bona fide CPS and KS genes) in five angiosperm species whose genomes have been most extensively sequenced and annotated – the dicot species grapevine (Vitis vinifera) (Jaillon et al., 2007), poplar (Populus trichocarpa) (Tuskan et al., 2006) and A. thaliana (Arabidopsis Genome Initiative 2000), and the monocot species rice (Oryza sativa) (Goff et al., 2002) and sorghum (Sorghum bicolor) (Paterson et al., 2009). We also include in the analysis, for comparison purposes, the genomes of the bryophyte Physcomitrella patens (Rensing et al., 2008) and the lycophyte Selaginella moellendorffii ( As terpenes play an important role in the physiology of various gymnosperms but there is currently no complete sequence of a gymnosperm genome, we also included, in some comparisons, data on TPS gene composition from various EST and full-length cDNA databases from a number of gymnosperm species, with the majority of the annotated TPS sequences coming from several spruce (Picea) species (Martin et al., 2004; Ralph et al., 2008) and grand fir (Abies grandis) (Bohlmann et al., 1999).

TPS family size

The P. patens genome contains four TPS gene models (Table 1), one of which has been shown to be the bifunctional PpCPS/KS (Hayashi et al., 2006). A second TPS gene encodes a protein that aligns well with PpCPS/KS, except that a number of stop codons are present in frame, suggesting it is a pseudogene. The other two TPS gene models are short gene fragments. While these shorter sequences may be due to the poor genome sequence and/or automated annotation, our manual annotation suggests that they are more likely to be true pseudogenes, i.e., the degraded products of duplicated PpCPS/KS genes.

The number of TPS gene models in the five angiosperm plant species ranges from 40 in Arabidopsis (Aubourg et al., 2002) to 152 in V. vinifera (Martin et al., 2010), counting both full-length genes, clearly non-functional genes, and, in the less well annotated genomes, short gene fragments (i.e., incompletely sequenced) (Table 1). Arabidopsis, with the smallest number of total TPS gene models, has 32 full-length TPS genes (Aubourg et al., 2002), and it is likely that the genomes of the other four angiosperm plants analyzed here have at least as many functional TPS genes and probably considerably more. Because of the lack of a completely sequenced and annotated gymnosperm genome, it is not currently possible to estimate the number of TPS genes in such genomes, but analysis of data from EST and full length cDNA databases for conifers suggests that the number of expressed functional TPS genes – at least 70 in white spruce (Picea glauca) – is similar to the number of TPS genes found in angiosperms (Keeling et al., 2011). The S. moellendorffii genome contains 18 TPS gene models (Table 1), but to the best of our knowledge none of them has been functionally characterized. Among these 18 TPS proteins, six contain both the class II ‘DXDD’ and the class I ‘DDXXD’ motifs; four contain only the ‘DXDD’ motif and five contain only the ‘DXXDD’ motif. Neither motif can be found in the remaining S. moellendorffii TPSs, which may represent pseudogenes. The majority of the S. moellendorffii TPS genes contain more than 10 introns, similar to the diterpene synthases in gymnosperms and in the moss P. patens (Hayashi et al., 2006; Keeling et al., 2010). Several pairs of TPS genes in S. moellendorffii are highly homologous, implying recent gene duplication. In our analysis of the S. moellendorffii genome sequence we noted eight TPS gene models that encode proteins with very limited homology to plant TPSs and high similarity to microbial TPSs. Most of these genes do not contain introns. The proteins they encode are approximately 200–400 amino acids and contain the conserved ‘DDXXD’ and ‘NSE’ motifs. While it is possible that these TPSs are authentic S. moellendorffii genes, it is also possible that they represent DNA from endophytic organisms.

TPS gene arrangement on chromosomes

The space occupied by TPS genes in the P. patens genome is about 0.008 gene/Mb. In the other six plants analyzed here, the content of TPS genes ranges from 0.07 gene/Mb in sorghum to approximately 0.3 gene/Mb in Arabidopsis and grapevine. These data suggest that the TPS family has undergone significant expansion during the evolution of land plants. Chromosomal locations of the TPS genes in P. patens and S. moellendorffii have not yet been determined. TPS genes are found on all five chromosomes in Arabidopsis, and in all but one chromosome in both rice and sorghum, whose chromosome number is 12 and 10, respectively (Table 1). In poplar, with 19 chromosomes, 12 chromosomes contain TPS genes. However, this number needs to be viewed with caution because 11 TPS genes in poplar have not yet been mapped to any specific chromosomes. In grapevine, a recent analysis of the 12-fold coverage genome sequence of a highly inbred Pinot noir variety identified 69 putatively functional VvTPS genes, 20 partial VvTPS, and 63 VvTPS probable pseudogenes (Martin et al., 2010). These VvTPS gene models are localized on seven of the 19 grapevine chromosomes, but 18 VvTPSs remained unmapped.

A significant number of TPS genes in the genome of the five angiosperm plants occur in tandem arrays of two or more genes (sometime separated by one or a few unrelated genes). In Arabidopsis, rice, poplar, grapevine and sorghum, 42, 64, 59, 85 and 66% of TPS genes, respectively, occur in such tandem arrays. These tandem arrays are likely the consequence of local gene duplication by unequal crossover. Consistent with this hypothesis, the genes in the tandem arrays are typically highly homologous to each other. For example, AtTPS23 and AtTPS27 are two Arabidopsis TPS genes located in such a cluster, and they are identical to each other both in the coding region and intron sequences, and thus represent a very recent TPS gene duplication (Chen et al., 2004). The tandem arrays of TPS genes are in some cases quite extensive, as in grapevine, where 45 VvTPS genes (almost one third of all VvTPSs) are organized as an extremely dense TPS gene cluster across a stretch of 690 kb on chromosome 18 (Martin et al., 2010) and in rice, where 14 TPS genes (approximately one-quarter of the total rice TPS genes) occur in a 480-kb stretch on chromosome 4.

Functions of TPS genes

The enzyme encoded by the P. patens PpCPS/KS gene catalyzes the formation of ent-kaurene and 16-hydroxy-ent-kaurene (Hayashi et al., 2006). ent-Kaurene is a precursor to ent-kaurenoic acid, which in P. patens serves the general physiological role in development that the gibberellins play in angiosperms (Hayashi et al., 2010). A specialized ecological role for PpCPS/KS is not presently known; however, the 16-hydroxy-ent-kaurene product of PpCPS/KS accumulates in P. patens with substantial concentrations of up to 0.5–1 mm and is also being released as a volatile compound (von Schwartzenberg et al., 2004). These levels of accumulation and release might be indicative of a role of 16-hydroxy-ent-kaurene in interactions with other organisms, but this hypothesis remains to be tested. No terpenes have yet been reported in S. moellendorffii, so possible ecological roles for TPS genes in this lycophyte species are not yet known either. In contrast, the ecological roles of terpenes in gymnosperms and angiosperms are many (Gershenzon and Dudareva, 2007), and much has been learned about the genes and enzymes that specify their production. In reviewing the information on the function of the TPS genes in the model systems discussed here, it is worth noting that the gymnosperm species studied are extremely long-lived trees with wind-pollinated cones, the two monocot angiosperm species, rice and sorghum, are grasses whose flowers are wind-pollinated, and among the three dicot angiosperm species, Arabidopsis is a small annual with scented flowers, poplar is a tree with wind-pollinated flowers, and grapevine is a cultivated perennial plant with scented flowers and highly aromatic fruits.

In gymnosperms, the largest number of TPS genes of specialized metabolism has been identified and functionally characterized in species of spruce (Picea) and in grand fir (Abies grandis), with a few individual TPS genes identified and characterized in other species such as loblolly pine (Pinus taeda), Douglas fir (Pseudotsuga menziesii), Ginkgo biloba and Taxus (Keeling and Bohlmann, 2006a; Keeling et al., 2011). The interest in diterpene synthases from Ginkgo and Taxus is due to their importance for the biosynthesis of medicinally useful metabolites, of which the anticancer drug taxol is perhaps the most prominent example of a high value terpenoid (Croteau et al., 2006). For the exploration of the TPS gene family in conifers (order Coniferales), and in particular members of the pine family (Pinacea), the major driving force has been to understand the biochemical, molecular, genomic and evolutionary underpinnings of the great chemical diversity of terpenoids in conifer defense against insects and pathogens, and to apply this knowledge for tree breeding and forest health protection. Conifers produce large quantities of oleoresin composed of complex mixtures of dozens of different acyclic and cyclic monoterpenes, sesquiterpenes, and diterpenes, the latter predominantly in the form of diterpene resin acids (Keeling and Bohlmann, 2006a,b). In addition, conifers emit volatile terpenoids in response to insect attack (Miller et al., 2005; Mumm and Hilker, 2006). Oleoresin terpenoid defenses act directly against invading insects and pathogens, as well as larger herbivores, while induced volatile emissions may function in indirect defense to attract the natural enemies of the attacking herbivores. The size of the TPS gene family and the diversity of biochemical functions of TPSs in any one conifer species mirror the complexity of oleoresin composition and volatile emissions. Published reports describe functions for 10 different TPS genes in Norway spruce (P. abies) (Fäldt et al., 2003a; Martin et al., 2004) and 11 different TPS genes in grand fir (Vogel et al., 1996; Bohlmann et al., 1997, 1998b, 1999; Steele et al., 1998). A recent large-scale transcriptome analysis identified 69 TPS genes in white spruce (P. glauca), 55 TPS genes in Sitka spruce (P. sitchensis), and 20 TPS genes in a hybrid white spruce (P. glauca × P. engelmannii), of which 21 have been functionally characterized as mono-, sesqui- and diterpene synthases of specialized metabolism (Keeling et al., 2011).

The majority of the conifer TPSs are multi-product enzymes, and it was in grand fir that the phenomenon of highly multi-product TPS enzymes was first highlighted with the discovery of the sequiterpene synthase genes encoding δ-selinene synthase (34 products) and γ-humulene synthase (52 products) (Steele et al., 1998). However, a few single-product TPSs have also been described in conifers. In the case of the Norway spruce ditepene synthases levopimaradiene/abietadiene synthase (PaTPS-LAS) and isopimaradiene synthase (PaTPS-Iso), multi-product and single-product enzymes exist as closely related (>90% identical) paralogous TPSs in the same species (Martin et al., 2004). Reciprocal site-directed mutagenesis revealed that the functional bifurcation of these two enzymes is due to only four amino acid residues (Keeling et al., 2008), thus providing an example of TPS gene duplication and neofunctionalization and the plasticity of TPS function. Recent studies confirmed that a critical mutation in the isopimaradiene synthase leading to change of product profiles in site-directed mutagenesis also occurs as a natural TPS variation in Sitka spruce (Keeling et al., 2011). This example is perhaps representative of the many events of TPS gene duplication, followed by sub- and neofunctionalization that have shaped the functional diversity of large families of active TPS genes of specialized metabolism in conifers. In addition to the more than 30 functionally characterized TPS genes of specialized metabolism functionally characterized in spruce, primary metabolism CPS and KS genes have also been characterized in white spruce and in Sitka spruce, where these genes appear to be actively expressed as single-copy genes (Keeling et al., 2010).

Expression of many of the conifer TPS genes involved in oleoresin defenses and volatile emissions is induced by herbivore or pathogen attack, and this effect can be mimicked to some extend by mechanical wounding of trees or treatment with methyl jasmonate (Miller et al., 2005; Zulak et al., 2009). Analysis of TPSs at the proteome and transcriptome levels in Norway spruce and Sitka spruce showed that methyl jasmonate-induced TPS transcription lead to increased protein abundance and enzyme activity and correlated with changes in the terpenoid metabolite profiles (Zulak et al., 2009; Hall et al., 2011). Spatially, the constitutive and induced expression of TPS genes and proteins in spruce is localized, at least in part, to the epithelial cells of resin ducts in stem tissues as determined by immunofluorescence localization (Zulak et al., 2010) and laser-capture microdissection (Abbott et al., 2010). In most conifers, constitutive and induced traumatic resin ducts are the primary site of massive accumulation of TPS products. Most of the products of mono- and sesquiterpene synthase activity accumulate without further modification. In contrast, most of the products of the conifer diterpene synthases are oxidized to diterpene resin acids by the activity of CYP720B cytochrome P450 enzymes, before being sequestered into resin ducts (Ro et al., 2005).

In Arabidopsis, 14 TPS genes have been functionally characterized, which include two diterpene synthase genes that encode CPS and KS respectively for the biosynthesis of gibberellins (Sun and Kamiya, 1994; Yamaguchi et al., 1998) and 12 for the biosynthesis of specialized metabolites. Because of its comprehensive molecular and genetic resources, Arabidopsis has become a particularly useful model for in depth studies on the spatial organization and temporal regulation of terpene metabolism in relation to the various biological functions of terpenes. Also, comparisons between ecotype- and genus-specific genomes allow investigating basic molecular mechanisms that control the intra- and inter-specific natural variation of terpene biosynthesis. Volatile terpenoids are emitted from several Arabidopsis tissues under constitutive and/or induced conditions (van Poecke et al., 2001; Aharoni et al., 2003; Chen et al., 2003; Fäldt et al., 2003b; Rohloff and Bones, 2005; Snoeren et al., 2010).

Most of the terpenoids detected so far in Arabidopsis are synthesized by approximately one-third of the enzymes of the Arabidopsis TPS family, with the enzymatic activities and products of the others yet to be determined. Microarray and RT-PCR analyses have indicated distinct or overlapping tissue-specificity of AtTPS gene expression, which has been investigated in greater detail for 10 genes (Figure 3). Flowers of the Arabidopsis ecotype Columbia (Col) emit a complex mixture of monoterpenes and over 20 sesquiterpene hydrocarbons with (E)-β-caryophyllene as the predominant compound (Chen et al., 2003; Tholl et al., 2005). Nearly all of the sesquiterpenes are produced by two flower-specific enzymes, the (E)-β-caryophyllene/α-humulene synthase (At5g23960, AtTPS21) and the multi-product sesquiterpene synthase AtTPS11 (At5g44630), and the other terpene compounds are synthesized by four additional enzymes (At4g16740, AtTPS03; At2g24210, AtTPS10; At1g61680, AtTPS14; At3g25810, AtTPS24) (Figure 3a). None of the floral terpenes is synthesized in flower petals, instead their formation occurs particularly in the stigma and sepals (AtTPS21), intrafloral nectaries and ovules (AtTPS11), and in pollen (AtTPS03) (Chen et al., 2003; Tholl et al., 2005; Huang et al., 2010). The terpenes that are emitted from these floral tissues at comparatively low rates can serve multiple functions: they may attract small pollinating insects at short range that contribute to cross-fertilization in natural populations. More recently, floral volatiles have been associated with defensive activities (Raguso, 2009) and Arabidopsis floral terpenes are likely to exhibit such functions to specifically protect floral organs such as the stigma against pathogen invasion. Differences in life histories and breeding systems influence floral terpene volatile emissions in the genus Arabidopsis. For example, in the highly scented flowers of the outcrossing perennial A. lyrata, formation of (E)-β-caryophyllene is substituted by the biosynthesis of benzenoid volatiles but instead has evolved as an insect-inducible trait in leaves catalyzed by a TPS21 orthologue (Abel et al., 2009).

Figure 3.

 Tissue- and cell type-specific expression patterns of Arabidopsis TPS genes in flowers (a), roots (b), and leaves (c) according to transcript and/or promoter-reporter gene analyses.
Numbers indicate different TPS genes. Arrows and dots in (b) indicate root-specific expression. Expression profiles of two root-specific TPS genes with predicted function as sesquiterpene and diterpene synthases are shown. In (c), herbivore-induced AtTPSs are shown. 3, AtTPS03, α-farnesene synthase; 4, AtTPS04, (E,E)-geranyllinalool synthase; 8, AtTPS08, putative diterpene synthase; 11, AtTPS11, multi-sesquiterpene synthase; 12 and 13, AtTPS12, AtTPS13, (Z)-γ-bisabolene synthases; 21, AtTPS21, (E)-β-caryophyllene synthase; 22, AtTPS22, putative sesquiterpene synthase; 24, AtTPS24, multi-product monoterpene synthase; 23 and 27, AtTPS23, AtTPS27, 1,8-cineole synthases.

Similar defensive functions as suggested for flower-specific TPS genes are predicted for 14 AtTPS genes with primary or exclusive expression in Arabidopsis roots. Four of these genes have been functionally characterized to encode monoterpene and sesquiterpene synthases (Chen et al., 2004; Ro et al., 2005) and diterpene synthase or sesquiterpene synthase activities have been predicted for others. According to high resolution transcriptome maps (Birnbaum et al., 2003; Brady et al., 2007) and promoter-reporter gene studies, many root-expressed genes exhibit cell type- and root zone-specific expression patterns (Figure 3b) that may emerge in defense against soil-borne, root-attacking organisms with different feeding or infection strategies. For example, promoter-GUS analyses demonstrated that the identical, duplicated genes AtTPS23 and AtTPS27 (At3g25820 and At3g25830), which are responsible for the formation of the monoterpene 1,8-cineole (Chen et al., 2004), are expressed in the vascular tissue of the root elongation zone and in epidermal cells and root hairs of older root-growth zones (Chen et al., 2004) (Figure 3b), where 1,8-cineole can directly be released into the rhizophere. Similarly, the two (Z)-γ-bisabolene synthase genes AtTPS12 (At4g13280) and AtTPS13 (At4g13300) are expressed in the stele of younger roots and in the cortex and endodermis of mature roots (Ro et al., 2006) (Figure 3b). Both genes have emerged by gene duplication and are found in pairs together with the co-expressed P450 genes CYP71A19 and CYP71A20 suggesting that these gene pairs form cell type-specific biosynthetic modules.

Leaves of the Arabidopsis ecotype Col release a mixture of volatiles containing the C16-homoterpene TMTT, the sesquiterpene (E,E)-α-farnesene, and the monoterpene β-myrcene in an induced response to herbivory, elicitor treatment, and the application of jasmonic acid or jasmonate mimics (van Poecke et al., 2001; Chen et al., 2003; Herde et al., 2008; Huang et al., 2010; Snoeren et al., 2010). AtTPS04 (At1g61120) catalyzes the formation of (E,E)-geranyllinalool as the first dedicated step in the biosynthesis of TMTT (Herde et al., 2008; Lee et al., 2010), while (E,E)-α-farnesene and β-myrcene are produced by the (E,E)-α-farnesene synthase AtTPS03 and presumably the β-myrcene/ocimene synthase AtTPS10, respectively (Bohlmann et al., 2000; Huang et al., 2010). Both AtTPS03 and AtTPS04 genes are induced locally at wound sites (Herde et al., 2008; Huang et al., 2010) (Figure 3c). The herbivore-induced volatile mixture was shown to attract parasitoids of herbivores (van Poecke et al., 2001; Loivamäki et al., 2008) and contributes to increased plant fitness (van Loon et al., 2000). Parasitoid attraction varies depending on the volatile blend emitted from different Arabidopsis ecotypes (Snoeren et al., 2010). For example, the ecotypes Col and Ws differ in their emission of (E)-β-ocimene, and this difference is due to a mutation that inactivates the (E)-β-ocimene-producing AtTPS02 in the Col ecotype (Huang et al., 2010).

Although the genome sequence for Populus was published in 2006, the functions of only three poplar TPS cDNAs have been characterized so far, including two encoding isoprene synthase and one encoding a sesquiterpene synthase. All of these genes are involved in the formation of volatile terpenes. Genes encoding isoprene synthase have been isolated from P. alba × tremula (Miller et al., 2001) and P. alba (Sasaki et al., 2005). The proteins encoded by these two isoprene synthase genes are 99% identical, suggesting that they are variants of the same gene. The release of volatile isoprene, produced by isoprene synthase, from Populus accounts for a large amount of the biogenic carbon found in the atmosphere. It is believed that isoprene emission has a role in the ecological response of Populus to abiotic stress (Sharkey and Yeh, 2001; Behnke et al., 2007). A full-length cDNA (PtdTPS1) was isolated from P. trichocarpa × deltoides and shown to encode a sesquiterpene synthase responsible for the synthesis of insect-induced (−)-germacrene D (Arimura et al., 2004). The herbivore-induced local and systemic mono- and sequiterpenoid emissions produced by poplar TPS genes may serve functions in multitrophic defence (Arimura et al., 2004) and in within-plant signaling (Frost et al., 2008).

In contrast to the poplar system, a substantial effort has been made to comprehensively analyze the genome organization and functions of the TPS gene family in grapevine (Martin et al., 2010). Terpenoids, in the form of free volatiles and as glycoside conjugates of monoterpene alcohols, are among the most important flavor compounds of grape berries and wine bouquet (Swiegers et al., 2005; Lund and Bohlmann, 2006). In addition, grapevine flowers produce a terpenoid-rich floral scent (Martin et al., 2009). In total, 43 VvTPS FLcDNAs from several grapevine varieties have been functionally characterized, representing mostly monoterpene synthases and sesquiterpene synthases (Lücker et al., 2004; Martin and Bohlmann, 2004; Martin et al., 2010). The majority of the functionally characterized VvTPS are multi-product enzymes. The products of these TPSs account for many of the major and minor terpenoids of berry and wine flavor, including a suite of monoterpene alcohols. A functionally characterized sesquiterpene synthase from Cabernet Sauvignon (VvValCS) and Gewürztraminer (VvValGw) produces (–)-valencene and (–)-7-epi-α-selinene as the main products (Lücker et al., 2004; Martin et al., 2009). This TPS gene is expressed in anthers and developing pollen grains and contributes to diurnal floral scent emission (Martin et al., 2009).

The rice TPS family has also been relatively well studied. Rice plants produce and emit a mixture of volatile terpenoids, including monoterpenes and sesquiterpenes, after insect herbivory (Yuan et al., 2008). By correlating terpenoid emission and expression of TPS genes analyzed by microarrays, three TPS genes were identified to be responsible for the production of the majority of insect-induced volatile terpenoids in rice. One of these genes encodes a monoterpene synthase making linalool as the single product, which is the most abundant insect-induced volatile from nipponbare rice plants (Yuan et al., 2008). The other two characterized TPS genes encode sesquiterpene synthases. Both of them make multiple sesquiterpenoids (Yuan et al., 2008).

Rice plants also produce a large number of labdane type diterpenoids. The known rice diterpenoids fall into five structurally related families: gibberellins, phytocassanes A–E, oryzalexins A–F, momilactones A and B, and oryzalexin S, which are derived from ent-kaurene, ent-cassa-12,15-diene, ent-sandaracopimaradiene, syn-pimara-7,15-diene and syn-stemar-13-ene, respectively (Peters, 2006; Toyomasu, 2008). Three rice TPSs have been shown to function as CPS, including OsCPS1 involved in gibberellin biosynthesis, OsCPS2 involved in the biosynthesis of non-gibberellin diterpenoids (Prisic et al., 2004) and OsCPSsyn producing syn-CPP for the biosynthesis of syn-labdane-related diterpenoids (Xu et al., 2004). A number of rice TPSs that use ent-CPP or syn-CPP as the direct substrate have also been characterized. These include OsKS1, which is a bona fide KS involved in gibberellin biosynthesis to produce ent-kaurene (Sakamoto et al., 2004), OsKSL4 (OsDTS2) for producing syn-pimaradiene (Wilderman et al., 2004), OsKS5 for producing ent-pimara-8(14),15-diene (Kanno et al., 2006), OsKS6 for producing ent-kaur-15-ene (ent-isokaurene) (Kanno et al., 2006; Xu et al., 2007), OsKSL7 (OsDTC1) for producing ent-cassadiene (Cho et al., 2004), OsKSL8 (OsDTC2) for producing syn-stemarene (Nemoto et al., 2004), OsKSL10 for producing ent-sandaracopimaradiene (Otomo et al., 2004a), and OsKSL11 for producing stemod-13(17)-ene (Morrone et al., 2006). These non-gibberellin diterpenoids function as phytoalexins to defend rice plants against microbial infection (Kato et al., 1994). Some of these compounds are produced by the rice plants then released into the environment to inhibit the germination and growth of neighboring plants, therefore acting as allelochemicals (Kato-Noguchi and Ino, 2003).

In contrast to rice, little is known about terpenes and TPS genes in sorghum. A recent study found that like rice sorghum plants emit volatile terpenoids, predominated by sesquiterpenes, upon insect herbivory. Colinearity analysis based on the synteny between the rice and sorghum genomes led to the identification of sorghum TPS genes, which are the orthologues of key rice TPS genes for producing insect-induced volatile terpenoids (X. Zhuang and F. Chen, unpublished results). Although not included in the analysis in this review, maize has served as another important monocot model for investigating terpenoid metabolism, especially in the context of plant-insect interactions. The majority of the TPS genes isolated from maize are involved in making herbivory-induced volatile terpenoids (Schnee et al., 2002, 2006; Köllner et al., 2008, 2009). Transgenic studies using some of these maize TPS genes convincingly showed that their products have a role in attracting parasitoids (Schnee et al., 2006) and entomopathogenic nematodes (Degenhardt et al., 2009b), the natural enemies of maize herbivores.

Evolution of TERPENE SYNTHASES: molecular phylogeny versus function

Topography of the TPS family tree

Previous phylogentic analyses of TPS protein sequences from gymnosperms and angiosperms recognized seven major clades (or subfamilies), designated TPS-a through TPS-g (Bohlmann et al., 1998a; Dudareva et al., 2003; Martin et al., 2004). We have now substantially extended this analysis to include new TPS sequences from the sequenced genomes of several plant species. The new analysis also recognizes seven TPS subfamilies – the original a, b, c, d, and g, a merged clade of the original e and f subfamilies now designated as e/f, and a new subfamily h (Figure 4). Function and taxonomic distribution of plant TPS subfamilies are summarized in Table 2.

Figure 4.

 Phylogeny of putative full-length TPSs from the seven sequenced plant genomes and representative characterized TPSs from gymnosperms (Table S1).
Based on the phylogeny and functions of known TPSs, seven subfamilies of TPSs are recognized. These include subfamily TPS-c (most conserved among land plants), subfamily TPS-e/f (conserved among vascular plants), subfamily TPS-h (Selaginella moellendorffii specific), subfamily TPS-d (gymnosperm specific), and three angiosperm-specific subfamilies TPS-b, TPS-g and TPS-a. The TPS-a subfamily is further divided into two groups, a-1 being dicot-specifc and a-2 being monocot-specific. The TPS-d subfamily is further divided into three groups, d-1, d-2 and d-3, which show distinction in function of TPSs in each group. The TPS-e/f subfamily is merged from the previously separate TPS-e and TPS-f subfamilies, which are also shown on the phylogenetic tree.

Table 2.   Function and taxonomic distribution of plant TPS subfamilies
  1. aDiTPS, diterpene synthase; IspS, isoprene synthase; MonoTPS, monoterpene synthase; SesquiTPS, sesquiterpene synthase.

TPS-b MonoTPS, IspSAngiosperms
TPS-c CPS/KS, CPS, other DiTPSLand plants
TPS-dTPS-d-1Primarily MonoTPS, SesquiTPSGymnosperms
TPS-d-3Primarily DiTPS, SesquiTPSGymnosperms
TPS-e/f KS, other DiTPS, monoTPS, SesquiTPSVascular plants
TPS-g MonoTPS, SesquiTPS, DiTPSAngiosperms
TPS-h Putative bifunctional DiTPSSelaginella moellendorffii

PpCPS/KS, the single member of the TPS family in P. patens, is placed in the TPS-c clade. Other members of this clade are all bona fide CPS proteins from gymnosperms (spruce PgCPS and PsCPS) and angiosperms (Arabidopsis CPS, and rice OsCPS1, OsCPS2 and OsCPSsyn) as well as three S. moellendorffii TPSs that contain only the ‘DXDD’ motif but not the ‘DDXXD’ motif, suggesting that they are monofunctional CPS. Since, as discussed above, evidence suggests that all other TPS genes evolved from a prototype bifunctional CS/KS gene, it is likely that the TPS-c clade represents the base of the tree (the tree in Figure 4 is unrooted, see Figure S1 for a rooted tree using bacteria CPS as the outgroup), and, furthermore, that mono-functionalization occurred in the TPS family very early in land plant evolution.

Closely related to the TPS-c subfamily are the TPS-e and TPS-f subfamilies. TPS-e contains all bona fide KS proteins from gymnosperms (PgKS and PsKS) and angiosperms (Arabidopsis KS and rice OsKS1). As in the TPS-c subfamily, several species analyzed have more than one TPS belonging to the TPS-e subfamily. Three S. moellendorffii TPSs form a subclade that is located near the bifurcation node of the TPS-c and the TPS-e clades (Figure 4). The presence of the ‘DDXXD’ motif but not the ‘DXDD’ motif in the proteins encoded by these S. moellendorffii genes suggests they function as class I TPS, probably KS. Therefore, this branch of S. moellendorffii TPSs are placed in the TPS-e subfamily. When the TPS-f subfamily (Figure 4), which contains AtTPS04, a diterpene synthase making geranyllinalool in Arabidopsis (Herde et al., 2008), uncharacterized TPSs from poplar and grapevine (Martin et al., 2010) as well as CbLIS, a monoterpene synthase producing S-linalool from C. breweri flowers (Dudareva et al., 1996) and two unusual TPSs from Solanum (see below), was first defined, it split from the base of the TPS-e subfamily (i.e., it appeared to be a sister clade). However, with the additional sequences in the present analysis, it is clear that TPS-f is derived from TPS-e (Figure 4), and we have therefore combined the two subfamilies into one clade, designated as the TPS-e/f subfamily. Evidently, genes in subclade f represent evolutionary novelties, but we note that none of the rice and sorghum TPSs belongs to the TPS-f subclade, indicating that it is probably dicot specific.

Eight S. moellendorffii TPSs that do not belong to the previously defined subfamilies of TPS-c and TPS-e/f form a new clade, which is introduced here as the new TPS-h subfamily (Figure 4). Except for one TPS that is presently missing the sequence at the region containing the ‘DDXXD’ motif, all S. moellendorffii TPSs in the TPS-h subfamily contain both ‘DXDD’ and ‘DDXXD’ motifs. Interestingly, among the more than 300 TPS genes identified from the five angiosperm species analyzed in the work described here, none contains both ‘DXDD’ and ‘DDXXD’ motifs. Conversely, all functionally characterized bifunctional diterpene synthases of specialized metabolism in the gymnosperms contain both the ‘DXDD’ and the ‘DDXXD’ motif, similar to the bifunctional PpCPS/KS (Keeling et al., 2010). These gymnosperm bifunctional diterpene synthases belong to the TPS-d subfamily (Figure 4). It is likely that gymnosperm bifunctional diterpene synthases evolved from a CPS/KS prototype TPS (probably before the gymnosperm-angiosperm split, since neither lineage appears to contain a CPS/KS) (Keeling et al., 2010). Similarly, it is likely that the putative bifunctional TPSs in the newly defined subfamily TPS-h evolved from PpCPS/KS and may be involved in specialized metabolism in S. moellendorffii.

In the gymnosperms we see a clear phylogenetic separation of TPS genes of specialized metabolism and TPS genes of primary gibberellin metabolism. Known gymnosperm CPS and KS of gibberellin metabolism belong, respectively, to the TPS-c and TPS-e/f clades like their counterparts in the angiosperms. Consistent with previous analyses, all gymnosperm TPSs for specialized metabolism belong to the gymnosperm-specific subfamily TPS-d. The TPS-d subfamily can be further divided into TPS-d-1, TPS-d-2 and TPSd-3 (Martin et al., 2004). TPS-d-1 contains all known gymnosperm monoterpene synthases for the production of a large array of conifer defense compounds, in addition to a few TPSs which produce the simple acyclic sesquiterpene (E,E)-α-farnesene. Most known gymnosperm sesquiterpene synthases, including those enzymes that produce large arrays of multiple terpenoids (Steele et al., 1998) belong to the TPS-d-2 group. The TPS-d-3 contains primarily diterpene synthases and several known sesquiterpene synthases. Members of the gymnosperm group TPS-d-3 contain an ancestral domain of approximately 200 amino acids which is also present in TPSs of the clades TPS-c, TPS-e/f, and TPS-h. Well characterized diterpene synthases in the TPS-d-3 group are the single-product and multi-product enzymes of conifer diterpene resin acid biosynthesis (Peters et al., 2000; Keeling et al., 2008) and taxadiene synthase from Taxus (Wildung and Croteau, 1996). Although sesquiterpene synthases are found in all three groups, they can be distinguished by gene structures: the sesquiterpene synthases in TPS-d-1 and TPS-d-2 are approximately 600 amino acids in length and those in TPS-d-3 are approximately 800 amino acids.

The angiosperm-specific TPS-a, TPS-b and TPS-g clades have substantially diverged from the other TPS clades (Figure 4). Many of the genes in these three clades have been functionally characterized in model- and non-model systems. Based on current knowledge, these three clades comprise entirely of genes of specialized mono-, sesqui- or diterpene biosynthesis with roles in ecological plant interactions, rather than roles in primary plant metabolism. For the five flowering plants for which genome sequences have been analyzed in this review, the TPS genes in TPS-a account for more than half of their TPS genes. TPS-a can be further divided into two groups, TPS-a-1 and TPS-a-2, with the former being monocot-specific and the latter dicot-specific. Most of the characterized TPSs in the TPS-a clade are sesquiterpene synthases. The TPS-a clade includes the first ever published TPS gene of specialized plant metabolism, the 5-epi-aristolochene synthase from tobacco which has since served as a model TPS for much of our understanding of TPS genes and enzymes in plants (Facchini and Chappell, 1992; Starks et al., 1997; Greenhagen et al., 2006; O’Maille et al., 2008). Overall, the sesquiterpene synthase TPS-a clade appears to be highly divergent in all seed plants charatcterized to date.

The TPS-b clade is also specific to flowering plants, and all characterized TPSs in this subfamily are either monoterpene synthases (including all Arabidopsis monoterpene synthase except linalool synthase) or isoprene synthases. The first member of this clade was discovered as (–)-limonene synthase in Mentha spicata (Colby et al., 1993). While the majority of TPS-b genes are from dicots, two TPSs from sorghum also belong to this group. However, none of the rice TPS genes fall into the TPS-b clade. Many of the enzymes of the TPS-b group produce cyclic monoterpenes, and many of the specific monoterpene synthase functions represented in the distantly related angiosperm-specific TPS-b and the gymnosperm-specific TPS-d-1 clades appear to have evolved convergently in the angiosperms and gymnosperms.

TPS-g is a clade closely related to TPS-b and was first defined by monoterpene synthases that produce the acyclic floral scent compounds myrcene and ocimene in snapdragon (Dudareva et al., 2003). Two members of the TPS-g group identified and characterized in Arabidopsis (At1g61680; Chen et al., 2003) and rice (Os02g02930; Yuan et al., 2008), also produce an acyclic monoterpene, namely linalool synthases. Similarly, all members of the TPS-g group identified in grapevine also only produce acyclic mono-, sesqui-, and diterpene alcohols (Martin et al., 2010). Thus a prominent feature of the members of the TPS-g group is the prevalence of acyclic products. A common structural feature of the members of the TPS-g subfamily is the lack of the RRX8W motif which is highly conserved near the N-terminus of monoterpene synthases (mostly cyclases) of the angiosperm TPS-b clade and the gymnosperm TPS-d-1 clade.

Lineage-specific expansion in the TPS family and changes in subcellular localization and substrate specificity

The presence of the S. moellendorffii-specific TPS-h subfamily, the gymnosperm-specific TPS-d subfamily and the angiosperm-specific subfamilies of TPS-a, TPS-d and TPS-g indicates lineage-specific expansion of the TPS family (Figure 4). In angiosperms, the TPS-a subfamily is the major determinant of the size of the TPS family of individual species. Apparently, the expansion of the TPS-a family occurred after the split of the monocot and dicot lineages (Figure 4). Moreover, the positions of Arabidopsis TPS genes on the branches of clade TPS-a-1 indicate that many of them arose by gene duplications that occurred after the divergence of the Arabidopsis lineage from the V. vinifera and P. trichocarpa lineages. Another example of species-specific expansion of the TPS genes can be observed within the TPS-e/f subfamily. Among the six plant species that contain TPS-e/f genes, the rice genome contains nine members while the other five species have 1–3 genes (Table 1). The relatively large number of TPS-e/f genes in rice leads to the production of a suite of labdane-type diterpenoids that provide defense against pathogens (Otomo et al., 2004b).

Building on the pioneering work of Joe Chappell, Rodney Croteau, and Charles A. West since the early nineties, the last decade has witnessed an enormous proliferation of functional characterization of TPS genes of plant specialized metabolism. The most recent acceleration of TPS gene discovery in well-travelled and remote places of the plant kingdom has been fueled by access to comprehensive genome and transcriptome sequence resources for a number of model and non-model systems. Just over 10 years ago our understanding of the plant TPS gene family was based on just over 30 genes, without a comprehensive coverage of a representative set of TPS genes in any one species (Bohlmann et al., 1998a). However, even with these few genes, a general model for the description of the TPS family became apparent. For example, all monoterpene synthases were thought to be localized to plastids and to use GPP as substrate. Sesquiterpene synthases were thought to be generally localized to the cytosol where they use all-trans FPP as their susbtrate with the use of GPP occurring only in vitro. And finally, the larger diterpene synthases had their place in plastids with all-trans GGPP as their substrate.

While much of this general model still holds true, new discoveries continue to reveal new facets of the TPS gene family. Even with this initial model, it was clear that the TPS family possesses a remarkable flexibility to evolve enzymes with new subcellular localization and substrate specificity. For example, the topology of the phylogenetic tree (Figure 4) suggests that mono- and sesquiterpene synthases evolved from diterpene synthases independently in gymnosperms and angiosperms. A change from diterpene synthase to sesquiterpene synthase involves not only a change in substrate specificity but also in subcellular localization (by the loss of a transit peptide). A relative recent example of such an event may exist with the closely related diterpene synthases and sesquiterpene synthases of the gymnosperm TPS-d-3 clade. Here, the bona fide sesquiterpene synthase genes encoding (E)-α-bisabolene synthases in grand fir and Norway spruce (Bohlmann et al., 1998b; Martin et al., 2004) still share with the conifer diterpene synthases the sequence encoding the ancestral 200 amino acid motif, but have apparently lost the region encoding the N-terminal transit peptide. Similary, Clarkia breweri linalool synthase (Dudareva et al., 1996), a monoterpene synthase that belongs to the TPS-f subfamily, evolved from a diterpene synthase, possibly a geranyllinalool synthase such as Arabidopsis AtTPS04, a change that in this case involved only a switch in substrate specificity. This enzyme also retained the 200 amino acid motif, linking it closely to a diterpene synthase ancestor.

Another good example of neofunctionalization of duplicated TPS genes involving a change in subcellular localization comes from the two TPS genes of the TPS-g subfamily that were isolated from snapdragon (Nagegowda et al., 2008). These two genes, AmNES/LIS-1 and AmNES/LIS-2, encode two TPS proteins that are nearly identical, suggesting that they are the consequence of recent gene duplication. The two proteins showed similar catalytic properties in vitro, both synthesize linalool and nerolidol as specific products using GPP and FPP as substrate, respectively. However, AmNES/LIS-2 protein has an additional 30 amino acids at the N-terminus, causing the protein to be transported into the plastid, where it uses the available GPP to make linalool. In contrast, AmNES/LIS-1 lacks a transit peptide, and is localized in the cytosol where it uses the available FPP to make nerolidol. These results indicate that in planta AmNES/LIS-1 functions as a sesquiterpene synthase while AmNES/LIS-2 acts as a monoterpene synthase, suggesting that subcellular targeting of TPSs controls the in planta substrate they use and the type of terpene products formed. As described above, the recently duplicated Arabidopsis AtTPS02 and AtTPS03 genes represent a very similar case, where the former produces (E)-β-ocimene in the plastid and the transit peptide-lacking AtTPS03 produces (E,E)-α-farnesene in the cytosol, although both enzymes can synthesize both compounds (from the respective substrates) in vitro (Huang et al., 2010).

Recent work based on the characterization of trichome transcriptomes in tomato revealed other surprising evolutionary diversifications in the TPS family. A Solanum lycopersicum (cultivated tomato) TPS enzyme, β-phellandrene synthase (PHS1), belongs to the diterpene synthase clade TPS-e/f but it uses the unusual substrate neryl diphosphate, the cis-isomer of GPP, to make mostly β-phellandrene as well as a few other monoterpenes in the plastids of the glandular trichomes on the leaf and stem surfaces (Schilmiller et al., 2009). Even more surprisingly, its closely related enzyme in Solanum habrochaites (89% identical) uses Z,Z-FPP to make two sesquiterpenes, bergamotene and santalene, in the plastids of the trichomes (Sallaud et al., 2009).

The TPS-a clade also appears to be more diverse with regard to subcellular targeting and substrates of its enzymes than what was initially thought. Several angiosperm species have substantial proportions of their TPS genes in this subfamily, some with apparent transit peptide (including 18 uncharacterized Arabidopsis TPS proteins) and some without. Most biochemically tested enzymes of this subfamily are sesquiterpene synthases but the diterpene synthase casbene synthase (Mau and West, 1994) was an early recognized member, suggesting that alternative substrate specificities may have evolved within the TPS-a subfamily (Bohlmann et al., 1998a). This pattern has been substantiated with the molecular characterization of several Nicotiana sylvestris TPS genes of the TPS-a subfamily. While direct biochemical characterization for the N. sylvestris proteins remains to be done, Ennajdaoui et al. (2010) showed that when the trichome-specific expression of these genes is blocked, the production of the diterpene cembratrien-ol is diminished. Consistent with a role in diterpene formation, the predicted TPS proteins appear to have plastid transit peptides.

The above mentioned new findings highlight the fascinating potential of the TPS gene family to evolve surprising variations of biochemical functions and subcellular localization, and strongly establish that substrate specificity, product profiles and localization of a TPS gene cannot be predicted based on association with a specific TPS subfamily (i.e., gross structure indicating evolutionary relationships) or general sequence similarity. As different subfamilies expand in different lineages by gene duplication and divergence, as has happened for example in the TPS-a, TPS-b, and TPS-g in angiosperms, TPS-d in gymnosperms, and TPS-h in S. moellendorffii, it can be expected that proteins with altered subcellular localization and new substrate specificities would have evolved.

Future directions

Along with other types of plant specialized metabolites, terpenoids play important roles in plant interactions with the environment (Gershenzon and Dudareva, 2007). Understanding the function and evolution of the TPS family can provide important insights into species-specific adaptations to unique niches. To better understand the physiological and ecological roles of specific TPS genes and enzymes will require research in several areas. Roles of specific terpenes or general roles of classes of terpenes, a topic only slightly covered here, must be examined in planta and ideally in the natural environments of the plants that produce these terpenes. The manipulation of TPS gene expression in model and non-model plants will be critical to this end. Despite the many major accomplishments since the first elucidation of the function of a plant TPS gene (Facchini and Chappell, 1992) and the three-dimensional structure of a plant TPS protein (Starks et al., 1997), the ongoing and future structural and biochemical investigation of TPS enzymes will continue to be a field of exciting new discoveries. At present, functional characterization has been completed only for subsets of the TPS families in any one species, including Arabidopsis, grapevine and spruce, the three species for which, relative to other species, the biochemical functions of a substantial number of TPS proteins are already known, and there are only a few experimentally determined three-dimensional structures for TPS of plant origin (Gennadios et al., 2009; Hyatt et al., 2007; Kampranis et al., 2007; Köksal et al., 2010a,b; Starks et al., 1997; Whittington et al., 2002). Functional and structural characterization of complete TPS families in selected model plant species such as Arabidopsis and rice, as well as plants occupying key position in the plant phylogeny, such as the conifers and S. moellendorffii, will provide insights into the array of terpenes that a single species can synthesize, and how its enzymes evolved the capacity to do so. At the same time, large transcriptome sequencing projects targeted at plant species that produce interesting terpenoid metabolites will accelerate the identification of comprehensive sets of TPS genes in a large variety of non-model systems. An improved knowledge base on plant terpenoid metabolism will facilitate the manipulation of the terpene pathway for changing agronomic traits such as fruit flavors (Lewinsohn et al., 2001; Davidovich-Rikanati et al., 2007), floral scent (Lücker et al., 2001), plant defense against insects (Schnee et al., 2006) and high level production of known and novel biochemicals (Bohlmann and Keeling, 2008; Kirby and Keasling, 2009).


Work in the laboratories of the authors has been funded by the Department of Energy Office of Biological and Environmental Research – Genome to Life Program through the BioEnergy Science Center (BESC) (FC), the Department of Energy grant DE-FG02-08ER64667 (to Janice Zale and FC), the Sun Grant Initiative (FC), the Natural Sciences and Engineering Research Council of Canada (JB), Genome British Columbia (JB), and Genome Canada (JB), the National Research Initiative Competitive grant 2007-35318-18384 from the USDA National Institute of Food and Agriculture (DT), a National Science Foundation AdvanceVT research seed grant (DT), the Max Planck Society (DT), by National Science Foundation award DBI-0604336 (EP), and by National Research Initiative Competitive grant 2008-35318-04541 from the USDA National Institute of Food and Agriculture (EP). We thank Dr Guanglin Li for his assistance on database search and phylogenetic analysis.