Monoterpenes and sesquiterpenes, like all other terpenes, are derived from the C₅ isoprene precursors isopentenyl diphosphate (IPP) and dimethyl allyl diphosphate (DMAPP) (Tholl, 2015). Plants employ two independent pathways to produce these precursors, namely the mevalonate pathway (MVA) and the methylerythritol phosphate (MEP) pathway which are compartmentalised into the cytoplasm and plastids, respectively ( Figure 1 ). Compartmentalisation of MVA and MEP intermediates is not strict, and it has been shown that intermediates can be exchanged across the plastidial membrane in a process termed “metabolic crosstalk” (Gutensohn et al., 2013; Schwab and Wüst, 2015). Metabolites which are exchanged between these pathways include IPP itself, as well as the prenyldiphosphate precursors of terpene biosynthesis, geranyl diphosphate (GPP), farnesyl diphosphate (FPP) and geranylgeranyl diphosphate (GGPP) (reviewed in Hemmerlin et al., 2012; Gutensohn et al., 2013; Liao et al., 2016). Table 1 shows enzymes of the MVA and MEP pathway which have been characterised in grapevine.

The first step in the MEP pathway is catalysed by 1-deoxy-D-xylylose-5-phosphate synthase (DXS), an enzyme which plays a major contribution to metabolic flux control in plastidial terpene biosynthesis (Tholl, 2015). Grapevine DXS (VvDXS1) has been established as an important contributor to the aroma of Muscat cultivars (Doligez et al., 2006; Battilana et al., 2009). Battilana et al. (2009) reported that VvDXS1 co-localizes with a major QTL on linkage group 5 which associates with three monoterpenes: linalool, nerol, and geraniol, which are responsible for the distinct floral and citrus aromas of Muscat cultivars. Further studies of VvDXS1 found a single nucleotide polymorphism (SNP) at position 1822 (G substituting a T) that was hypothesised to be a “gain of function” mutation (Emanuelli et al., 2010). VvDXS1 genes that were heterozygous (GT) at position 1822 caused a non-synonymous substitution of a lysine (K) with an asparagine (N) at position 284 of the VvDXS1 protein. Functional characterisation of VvDXS1 showed that the non-synonymous amino acid substitution influences enzyme kinetics by increasing the catalytic efficiency of VvDXS1, thereby increasing the total monoterpene content of cultivars carrying this SNP (Battilana et al., 2011). This was further supported by transgenic tobacco lines overexpressing the K284N SNP allele of VvDXS1 showing up to 20 times higher levels of glycosylated monoterpenes than lines expressing the neutral allele (Battilana et al., 2011). Additionally, microvine lines overexpressing the neutral and muscat allele of VvDXS1 had a 1.7- and 4.4-fold increase in total monoterpene content compared to the wild type, respectively (Dalla Costa et al., 2018). The K284N SNP of VvDXS1 appears to be a reliable marker for muscat-aroma in grapevine cultivars. A recent study looking at the association between VvDXS1 and aromatic substance content in different flavour types (muscat-like, aromatic and neutral aroma) of grapevine varieties also associated the K284N SNP with increased monoterpene content in grapevine (Yang et al., 2017b).

Dalla Costa et al. (2018) investigated the effect of the K284N SNP on terpene content in 90 grapevine germplasms. Predictably, cultivars that were homozygous (TT) or heterozygous (GT) for the Muscat-allele had a significantly higher level of monoterpenes than cultivars homozygous (GG) for the neutral allele. Interestingly, the authors also reported a similar trend, albeit to a lesser extent, in sesquiterpene content. Furthermore, overexpression of VvDXS1 in combination with VvLinNer (see Table 1 ), a linalool/nerolidol synthase, led to a significant increase in linalool (a monoterpene) and nerolidol (a sesquiterpene) content (Wang et al., 2021). The increase in sesquiterpene content associated with VvDXS1 overexpression may be explained by the phenomenon of “metabolic cross-talk” between the MEP and MVA pathways. The increased flux towards plastid-bound MEP pathway precursors, due to overexpression of VvDXS1, potentially leads to an increase in transport of these precursors to the cytosol where they are incorporated in the MVA pathway resulting in an increase in sesquiterpene biosynthesis.

While VvDXS1 is an effective marker for muscat-aroma, it is not the sole determinant of monoterpene biosynthesis. Emanuelli et al. (2010) found that several cultivars which are characterised as aromatic show no presence of the K284N SNP. Indeed, out of 20 aromatic cultivars, it was reported that 75% are homozygous for the neutral allele (Emanuelli et al., 2010). Therefore, the monoterpene content of aromatic cultivars is likely influenced by enzymes other than VvDXS1. Furthermore, three cultivars with muscat-like aroma but no Muscat parentage were also shown to be homozygous for the neutral allele. However, these three cultivars (Gewürztraminer, Chardonnay musqué clone 44-60 Dijon, and Chasselas musqué) each had unique heterozygous SNPs in VvDXS1 located close to the K284N SNP. Further investigation of these SNPs is necessary to determine whether they are associated with increased monoterpene accumulation in Muscat-like aromatic cultivars.

While apparently predominantly controlled by DXS, the metabolic flux through the MEP pathway in plants is further regulated by other enzymes which include hydroxymethylbutenyl diphosphate reductase, HDR (Vranová et al., 2013). In grapevine, the expression of VvHDR has been shown to correlate with the veraison-initiated accumulation of monoterpenes in certain cultivars (Martin et al., 2012; Wen et al., 2015; Costantini et al., 2017; Yue et al., 2020), indicating the potentially regulatory role of VvHDR in grapevine monoterpene biosynthesis.

A major contributor to metabolic flux control of the MVA pathway is HMG-CoA reductase (HMGR) (Rodríguez-Concepción, 2006). Three HMGRs have been identified in grapevine, VvHMGR1-3 (Zheng et al., 2020; Zheng et al., 2021). The three genes are differentially expressed in grapevine organs and during berry development. Interestingly, it was found that VvHMGR3 plays a role in fruit colour formation. Heterologous suppression of VvHMGR3 in strawberry increased the rate of colour formation and increased anthocyanin formation and inversely, overexpression of VvHMGR3 suppressed colour (Zheng et al., 2020). Furthermore, the authors found that brassinosteroids (BRs) (which are produced via the MVA pathway) inhibit VvHMGR expression. A BR-HMGR model is proposed whereby VvHMGR expression leads to an increase in BR accumulation and in turn BRs have negative feedback on HMGR activity. Additionally, BRs increases anthocyanin content.

The final products of the MEP and MVA pathways, IPP and DMAPP, are fused through consecutive head-to tail condensation reactions, catalysed by short chain prenyltransferases, to form prenyl diphosphates which serve as the precursor backbones for terpenoids (Vranová et al., 2013; Tholl, 2015).

C₁₀ Geranyl diphosphate (GPP) is the precursor for monoterpene biosynthesis and is formed through the activity of GPP synthases (GPPS). Plant GPPSs exist as either hetero- or homodimeric enzymes (Nagegowda & Gupta, 2020). Heterodimeric GPPS consists of a large subunit (LSU) and a catalytically inactive small subunit (SSU-I). GPPS-LSU shares high homology with geranylgeranyl diphosphate synthase (GGPPS) and in some instances has been shown to possess GGPPS activity as a homodimer (Tholl, 2015). GGPP is the precursor molecule to many other primary and specialised terpenes such as carotenoids, abscisic acid, chlorophylls, phytol tocopherols, gibberellins, plastoquinones, polyprenols, and diterpenoids.

To date, very little research has been done on grapevine GPPSs (VvGPPSs). Early reports indicate that VvGPPSs are localised to the plastids, as is the case with plant GPPSs in general (Feron et al., 1990; Soler et al., 1992). More recently, gene expression and transcriptomic studies investigated the expression of VvGPPS with respect to terpene accumulation. Transcript abundance levels of the VvGPPS gene were shown to parallel the veraison-initiated accumulation of monoterpenes and is potentially contributes to flux control in monoterpene biosynthesis (Martin et al., 2012; Wang et al., 2021). However, these studies do not differentiate between GPPS-LSU and GPPS-SSU. An integrated transcriptomic and metabolomic study of ripening Moscato Bianco berries showed no correlation between VvGPPS-LSU and terpene accumulation (Costantini et al., 2017).

C₁₅ farnesyl diphosphate (FPP), produced by FPP synthase (FPPS) is the precursor for sesquiterpenes, triterpenes and primary metabolites such as phytosterols, brassinosteroids, dolichols and ubiquinones (Tholl, 2015). Plant FPPSs are homodimeric enzymes and have been reported to localise to cytosol, mitochondria, or peroxisomes in different plant species (Tholl, 2015). As with VvGPPSs, the molecular functional characterisation of grapevine FPPSs (VvFPPSs) is limited. The importance of VvFPPS in grapevine sesquiterpene biosynthesis is highlighted in a study analysing the transcription of genes related to the biosynthesis of rotundone, the sesquiterpene responsible for the peppery aroma of Syrah cultivars. This work indicated that VvFPPS potentially plays a vital role in the accumulation level of rotundone in Syrah cultivars by increasing the substrate pool available for rotundone precursor synthesis (Takase et al., 2016a).

Plant GPPSs and FPPSs were generally accepted to be trans prenyltransferases, i.e., they synthesise the trans (E) conformation of GPP and FPP, respectively, however, short-chain cis prenyltransferases have been identified in several plants (Akhtar et al., 2013; Demissie et al., 2013). For example, a cis FPPS (zFPPS) was demonstrated to produce Z,Z-FPP in the glandular trichomes of tomatoes (Sallaud et al., 2009). zFPPS is localised to the plastids, unlike cytosolic trans FPPS, and is therefore theorised to use precursors from the MEP pathway. Furthermore, Z,Z-FPP was shown to be used as a substrate for sesquiterpene synthase. To date, tomato is the only species where cis prenyldiphosphates have been reported as terpene synthase (TPS) substrates.

Another non “traditional” route for terpene biosynthesis was shown through the function of isopentenyl phosphate kinases (IPKs). IPS catalyses the conversion of isopentenyl phosphate (IP) and possibly dimethylallyl phosphate (DMAP) to IPP and DMAPP, respectively (Henry et al., 2018). The presence of genes encoding IPKs in all sequenced plant genomes indicate a possible regulatory role for terpene biosynthesis via IPK via IPP/IP and DMAPP/DMAP ratio modulation (Nagegowda & Gupta, 2020). IP and DMAP was shown to be produced through dephosphorylation of IPP (and DMAPP) by members of the Nudix hydrolase super family (AtNudx1 and AtNudx3) (Henry et al., 2018). Nudix hydrolyses in rose has also recently been reported to provide a TPS-independent path for monoterpene production (Magnard et al., 2015). Rose Nudix hydrolase (RhNudx1) catalyses the formation of geranyl monophosphate (GP) from GPP; GP is then further converted to geraniol by an unidentified phosphatase. With regards to grapevine, a nudix hydrolase (VIT_10s0003g00880), whose expression increased along berry development and correlated with linalool content in Moscato Bianco was proposed as a candidate gene for an alternative route of monoterpene production in grapevine (Costantini et al., 2017).

The plant TPS gene family is mid-sized with TPS genes typically ranging from 30-170 per plant species. The gene family has been divided into seven subfamilies (TPS-a, -b, -c, -d, -e/f, -g and -h) based on their sequence similarity and proposed function (Bohlmann et al., 1998; Chen et al., 2011). The TPS-c and TPS-e/f subfamilies are involved in primary metabolism, encoding copalyl diphosphate synthases (CPSs) and kaurene synthase (KSs) which are involved in gibberellin biosynthesis (Chen et al., 2011). The TPS-d family produce gymnosperm-specific terpene synthases, while TPS-h genes encode putative bifunctional diterpene synthases in the lycophyte Selaginella moellendorffii (Chen et al., 2011). TPS-a, b, and g families are angiosperm-specific and produce mono-, sesqui-, and diterpenes involved predominantly in specialised metabolism (Chen et al., 2011). TPS-a subfamily members typically produce sesquiterpenes and diterpenes, while members of the TPS-b and TPS-g subfamilies produce monoterpenes, although the TPSs from the TPS-g subfamily exclusively produce acyclic terpene alcohols. Grapevine TPSs (VvTPSs) fall within every subfamily except TPS-d and TPS-h (Martin et al., 2010; Smit et al., 2020). More than half of the specialised VvTPSs are sesquiterpene synthases (TPS-a), with the rest making up the TPS-b and -g subfamilies. Several VvTPSs from a number of cultivars have been functionally characterised and are summarised in Table 1 . Additionally, Figure 2 shows a phylogenetic tree of functionally characterised VvTPSs.

The expansion of the TPS gene family within plants is thought to occur primarily through tandem or segmental gene duplication. The highly inbred and homozygous Pinot Noir genome has served as the reference genome (PN40024) for grapevine for the last decade (Jaillon et al., 2007). In a keystone study for grapevine TPSs by (Martin et al., 2010), using the reference genome, the grapevine TPS gene family was functionally annotated which revealed that this gene family is substantial with 69 putatively functional TPSs. The increasing number of available genome sequences for different grapevine cultivars shows that due to its homozygosity PN40024 is limiting to our understanding of VvTPSs. Indeed, a comparison of the draft diploid genomes of Cabernet Sauvignon, Carménère, and Chardonnay to PN40024 revealed that there is a larger number of VvTPSs in these cultivars than the canonically accepted 69 (Smit et al., 2020). Furthermore, the number of VvTPSs varies significantly between cultivars, ranging from approximately 80 to 200. Grapevine genomes show extensive duplication events, which led to the expansion of the VvTPS family and the fact that 69-90% of VvTPSs in grapevine are related to gene duplication events (Jiang et al., 2019; Smit et al., 2020). Additionally, annotation of VvTPSs revealed that the majority of sesquiterpene synthases cluster on chromosome 18, while monoterpene synthases cluster on chromosome 13 (Martin et al., 2010; Smit et al., 2020). Another interesting feature of VvTPSs revealed by the sequencing of diploid genomes is that approximately 30% of VvTPSs are hemizygous, which can play a vital role in understanding the inheritance of these genes for molecular breeding. Taken together, these factors including the large gene family size, cultivar variation, extensive duplication, and hemizygosity of VvTPSs goes a long way in explaining the large variation in terpene composition observed in different grapevine cultivars and highlights the immense potential of grapevines to produce novel and diverse terpenes.

Extensive gene duplication of TPSs has also been associated with the variation in plant terpene composition in different organs and at different developmental stages. Duplicated TPS genes serve the same enzymatic function, but often show divergent temporal and spatial expression patterns resulting in tissue or time specific terpene profiles. These sub-functionalisation events have also been demonstrated in grapevine. Expression analysis of gene paralogs of (E)-β-caryophyllene synthases, β-ocimene synthases, and linalool synthases showed that all gene paralogs were differentially expressed during Moscato Bianco grape berry development (Matarese et al., 2013). Beyond the grape berry, these gene paralogs were also differentially expressed in different grapevine organs (Matarese et al., 2013; Matarese et al., 2014). Differential gene expression patterns for paralogs of β-ocimene synthases and linalool synthases were also shown in Sauvignon Blanc, Riesling, and Hamburg Muscat berry development (Yue et al., 2020). Moreover, the expression pattern of gene paralogs differed between cultivars, further showcasing the high level of variation of VvTPSs and terpene accumulation between grapevine cultivars.

Terpene synthase enzymes can be divided into two classes (type I TPSs or type II TPSs) based on their mechanism of catalysis. Class II TPSs catalyse the ionisation of GGPP via protonation. More common are the class I TPSs which contain all mono- and sesquiterpenes. Class I TPSs catalyse the ionisation of the phosphate group on the prenyldiphosphate substrate (GPP, FPP, or GGPP), forming a highly reactive carbocation intermediate which can undergo various reactions such cyclisations or hydride shifts until the reaction ends with proton loss or the addition of a nucleophile (Degenhardt et al., 2009).

TPSs have various protein motifs that play an important role in their enzyme function. Class I TPSs contains two aspartate-rich motifs, DDxxD and NSE/DTE in the C-terminal domain which flank the active site. DDxxD and NSE/DTE both bind a trinuclear magnesium cluster which is involved in the positioning of the substrate (Degenhardt et al., 2009). Unlike the DDxxD motif, which is highly conserved through all plant TPSs, the NSE/DTE motif is less conserved with a consensus sequence of (L,V)(V,L,A)-(N,D)D(L,I,V)x(S,T)xxxE (Degenhardt et al., 2009). Upstream of the DDxxD motif is a highly conserved RxR motif which prevents nucleophilic attack on any of the carbocationic intermediates (Degenhardt et al., 2009). An altered RxQ motif appears in sesquiterpene synthases which produce nerolidol, an acyclic terpene. This altered motif may be less effective at preventing nucleophilic attack of the carbocationic intermediate leading to the termination of the enzyme reaction before cyclisation can occur (Durairaj et al., 2019). On the N-terminal end, TPSs contain an RRx₈W motif which has been predicted to play a role in terpene cyclization. Mono- and diterpene synthases contain an N-terminal plastid transit peptide upstream of the RRx8W motif resulting in their localization to plasmids.

In the grapevine TPS-g subfamily the RRx8W motif is not well conserved (Martin et al., 2010) which supports the proposed involvement of the RRx8W motif in cyclisation as TPSs from the TPS-g subfamily primarily produce acyclic monoterpene alcohols. Interestingly, the NSE/DTE motif of the TPS-g subfamily in grapevine has a modified and highly conserved sequence LWDDLx(S,T)xxxE (Martin et al., 2010). The NSE/DTE motif may thus play a role in determining the cyclisation function in grapevine TPSs.

In the plant kingdom, several multi-substrate TPSs which can use GPP, FPP, and GGPP in vitro to produce monoterpenes, sesquiterpenes, and diterpenes respectively, have been identified (reviewed in (Pazouki and Niinemetst, 2016). Three multi-substrate TPSs have been characterized in grapevine, namely, VvPNLinNer1, VvPNLinNer2, and VvCSLinNer, capable of producing linalool (a monoterpene) and nerolidol (a sesquiterpene) from GPP and FPP, respectively ( Table 1 ) (Martin et al., 2010). Additionally, four TPSs, VvPNLNGl1-4, also accepted GGPP to produce (E,E)-geranyl linalool (Martin et al., 2010). VvGwbOciF and VvCSbOciF could also accept both GPP and FPP to produce (E)-β-ocimene or (E,E)-α-farnesene, respectively. Lastly, VvCSENerGl and VvPNENerGl accepted either FPP or GGPP to produce E-nerolidol or (E,E)-geranyl linalool, respectively (Martin et al., 2010). These enzymes were characterised in vivo using metabolically engineered E. coli. Subcellular localisation of Riesling VvLinNer (VvRiLinNer) showed that the enzyme is localized to the chloroplasts and the authors proposed that due to its localisation, VvRiLinNer could only produce linalool in planta (Zhu et al., 2014). This inference is supported by a previous study that demonstrated grape derived monoterpenes are almost exclusively synthesised via the plastid-localised MEP pathway (Luan, 2002), while cytosolic localised sesquiterpenes are produced from both the cytosolic MVA and plastidial MEP pathway intermediates (May et al., 2013). Contrarily, a recent study reported that N. benthimiana leaves transiently overexpressing VvLinNer (isolated from the cultivar Shine Muscat) had elevated levels of both linalool and nerolidol, with linalool being predominant (Wang et al., 2021). Seeing as VvLinNer is localised to the plastids, this recent finding potentially demonstrates that the substrate pool in plastids may be derived from both the MEP and MVA pathways in grapevine. However, the authors do not state whether the overexpressed VvLinNer was efficiently taken up by the plastids therefore it is unclear whether the increased nerolidol is due to FPP production within the plastids or that the heterologously expressed VvLinNer may be act within the cytosol. Studies in other plants have indicated the potential for FPP presence in plastids. For example, targeting of FaNES1, a cytosolically localised linalool/nerolidol synthase from strawberry, to plastids in Arabidopsis resulted in an increase in nerolidol abundance, albeit at lower levels than linalool (Aharoni et al., 2004).

Terpene synthases are also able to produce multiple products from a single substrate, a trait that greatly increases terpenoid diversity. Nearly half of the identified mono- and sesquiterpene synthases generate more than one product (Vattekkatte et al., 2018). The ability of TPSs to generate such a wide variety of products is not yet fully understood. One contribution may be the highly reactive carbocationic intermediate that can undergo various reactions to be stabilised. However, single product TPSs do exist, therefore it’s likely that a structural feature of the enzyme contributes to its ability to produce multiple products. No common feature has been identified in TPSs that contribute to their ability to produce multiple products; however, several studies suggest that the conformation of the active site influences this ability (reviewed by Degenhardt et al., 2009; Vattekkatte et al., 2018). For example, the ability of γ-humulene from A. grandis to produce 52 different sesquiterpenes was associated with the presence of two DDxxD motifs flanking the active site (Steele et al., 1998; Little & Croteau, 2002). Furthermore, through modelling studies it was shown that TPS4 from Zea mays can produce multiple products due to two pockets in the active site which control the conformational change of the carbocationic intermediate (Kollner et al., 2004).

It is important to note that several characterisation studies infer the function of the gene through in vitro analysis and heterologous gene expression. A recent study by Salvagnin et al. (2016) highlighted the importance of studying TPS function within its native plant. The authors analysed grapevine E-(β)-caryophyllene synthase (VvGwECar2) under three conditions: in vitro, in a heterologous plant system (Arabidopsis) and in a homologous plant system (Vitis vinifera). While the enzyme still produced E-(β)-caryophyllene and α-humulene as its major products in all systems, the ratio of these compounds was different in each system. Furthermore, the composition and abundance of secondary products were different in each system. For instance, in the Arabidopsis system thujopsene was also produced, but this was not detected in grapevine.

Another major contribution to terpene diversity is the functional plasticity of TPS active sites, i.e., a single nucleotide substitution can lead to a change in the enzyme function. While TPSs share conserved motifs (e.g., DDxxD and NSE/DTE) that are vital to enzyme function, several structure-function studies have demonstrated that small amino acid substitutions can lead to TPSs producing entirely new products.

Various examples of functional plasticity have been reported for grapevine TPSs. Recent studies into the sesquiterpene rotundone revealed genotypic variation in the cultivars that have high levels of this terpene. Rotundone is responsible for the peppery aroma associated with cultivars such as Shiraz, Cagnulari, Schioppettino, Vespolina, Graciano, and Gruene Veltliner (Mattivi, 2016). A novel allele of the VvTPS24 gene model, VvGuaS, a sesquiterpene synthase whose main product is the rotundone precursor α-guaiene, was identified in Shiraz berries (Drew et al., 2016). Previously, TPS24 was shown to encode for VvPnSeInt, which produces selina-4,11-diene as its main product (Martin et al., 2010). Drew et al. (2016) also identified two polymorphisms in the TPS24 gene of Shiraz which is responsible for two non-synonymous amino acid substitutions in the active site of the enzyme resulting in functional conversion of the enzyme from VvPnSeInt to VvGuaS. This is an example of how small genetic variations (single nucleotides) in TPS genes can lead to a complete functional change of the enzymes for which they encode, further increasing the diversity of terpene profiles observed across grapevine cultivars. Additionally, an association study between VvTer, an α-terpineol synthase gene, and α-terpineol content in the grape berries derived from 61 cultivars identified two SNPs that associated with higher α-terpineol content. However further study is necessary to ascertain the functional effects of these polymorphisms (Yang et al., 2017a). Another example of the cultivar specific nature of grape TPSs is the recently characterised (E)-β-farnesene synthase (VvMATPS10) from Muscat of Alexandria flowers (Smit et al., 2019). This gene had been previously characterised to code for a bergamotene synthase (VvGWaBer) in Gewürztraminer (Martin et al., 2010), however when isolated from Muscat of Alexandria it showed a unique sequence and function, producing (E)-β-farnesene, as opposed to bergamotene, as a single product. These studies highlight the high functional plasticity of VvTPSs and how this plasticity results in the cultivar-specific functions of VvTPS.