Terpene synthases in cucumber (Cucumis sativus) and their contribution to herbivore‐induced volatile terpenoid emission

Summary Terpenoids play important roles in flavour, pollinator attraction and defence of plants. In cucumber (Cucumis sativus) they are important components of the herbivore‐induced plant volatile blend that attracts natural enemies of herbivores. We annotated the cucumber TERPENE SYNTHASE gene (CsTPS) family and characterized their involvement in the response towards herbivores with different feeding guilds using a combined molecular and biochemical approach. Transcripts of multiple CsTPS genes were upregulated in leaves upon herbivory and the products generated by the expressed proteins match the terpenoids recorded in the volatile blend released by herbivore‐damaged leaves. Spatial and temporal analysis of the promoter activity of CsTPS genes showed that cell content‐feeding spider mites (Tetranychus urticae) and thrips (Frankliniella occidentalis) induced promoter activity of CsTPS9 and CsTPS19 within hours after initiation of infestation, while phloem‐feeding aphids (Myzus persicae) induced CsTPS2 promoter activity. Our findings offer detailed insights into the involvement of the TPS gene family in the dynamics and fine‐tuning of the emission of herbivore‐induced plant volatiles in cucumber, and open a new avenue to understand molecular mechanisms that affect plant–herbivore interactions.


Introduction
Specialized metabolites modulate interactions of plants with their biotic environment. Numerous endogenous compounds function in direct defence as toxins and repellents towards herbivores and pathogens (Schoonhoven et al., 2005;Hopkins et al., 2009). Volatile compounds have additional functions as attractants for pollinators and carnivorous enemies of herbivores, as well as in inter-and intra-plant communication (Pichersky & Gershenzon, 2002;Degenhardt et al., 2003;Kappers et al., 2005;Dudareva & Pichersky, 2020). Upon herbivory, the plant's specialized metabolome changes depending on the feeding habit of the infesting herbivore. For example, chewing caterpillars inflict significant damage, while aphids cause only little tissue damage, manoeuvring their flexible stylet intercellularly through the epidermis and mesophyll to reach the phloem . Other herbivores inflict moderate damage, including spider mite and thrips that pierce mesophyll plant cells and feed on their contents. In addition to the mechanical wounding inflicted, cues in the herbivore's oral secretion trigger a cascade of reactions including early Ca 2+ signalling and a burst of reactive oxygen species (Maffei et al., 2007), followed by changes in the concentrations of phytohormones (Wu & Baldwin, 2010). The synthesis, perception and crosstalk of these hormones, the transcription factors involved and their target genes together constitute a complicated signaltransduction network through which the plant metabolome and therefore the defensive state of the plant is rearranged.
Terpenoids represent the most diverse group of plant specialized metabolites  and many have roles in the interaction between plants and their environment. Terpenoids are the main constituents of the blend of leaf-emitted volatiles after oviposition, herbivory and wounding that induce endogenous jasmonic acid (JA) (Bohlmann et al., 2000;Herde et al., 2008;Cao et al., 2010;Hilker & Fatouros, 2015), and nonvolatile terpenoids increase in plant organs upon exposure to (a)biotic stresses (Bohlmann et al., 2000;Balkema-Boomstra et al., 2003;Nagegowda, 2010).
Plant defences against biotic stressors can be affected by internal and external factors including light and the circadian clock. In Arabidopsis, the expression of more than 40% of the genes induced by mechanical damage peaks at dusk and over 80% of the genes is suppressed at dawn (Walley et al., 2007). Arabidopsis plants grown under a similar light : dark rhythm as the cabbage looper Trichoplusia ni, which has rhythmic feeding behaviour, had increased resistance against this herbivore, while plants grown under an opposite light : dark rhythm as the insect were more susceptible (Goodspeed et al., 2012). Both the circadian clock and jasmonates were shown to be essential in maintaining this rhythmic defence.
Here, we investigated the abundance and composition of the volatile blend of cucumber plants upon feeding by different types of herbivores and characterized the genes encoding the TPSs and their transcriptional regulation responsible for the specificity of the response to herbivores with different feeding guilds.

Plants and arthropods
Cucumis sativus plants (genotype 'Corona') were grown in potting soil in a glasshouse (16 h 22°C : 8 h 18 AE 2°C, light : dark) for 3 wk until five true leaves had developed. Arabidopsis thaliana Col-0 (N1092) and p35S::LUC in Col-0 background (N9966) seeds were obtained from the Nottingham Arabidopsis Stock Centre (NASC) and grown in a climate chamber (12 h : 12 h, light : dark, 150 µmol m À2 s À1 , 22°C) for 4 wk. Female adult spider mites (T. urticae) were selected from a mass-rearing on lima beans. Aphids (Myzus persicae) were reared on radishes and wingless adults were used for experiments. Thrips (Frankliniella occidentalis) were reared on pods of broad bean and 5-d-old larvae were used for experiments.

Assessment of leaf damage
Herbivory damage was ass after 3 d. For mite damage, visual observation of chlorotic spots was supported by trypan blue staining (Keogh et al., 1980). Quantification of TSSM and thripsinduced damage was performed using IMAGEJ software (imagej. nih.gov/ij) as described by Visschers et al. (2018).

Identification of CsTPS genes
The cucumber genome (v.2 assembly; www.icugi.org) was screened for genes related to the terpenoid biosynthetic module using INTERPROSCAN (www.ebi.ac.uk/interpro/) according to the method described by Hofberger et al. (2015). Genomic regions containing candidate genes and their flanking 4 kb sequence were extracted, re-annotated and confirmed by FGENESH (www. softberry.com) and GENEWISE (www.ebi.ac.uk/) according to the structure of previously reported TPS proteins (Chen et al., 2011). The TARGETP 1.1 server (Emanuelsson et al., 1999) was used for signal peptide prediction, and amino acid alignment of fulllength CsTPS enzymes was constructed using CLUSTALW (www. genome.jp/tools-bin/clustalw) and MUSCLE (www.ebi.ac.uk/ Tools/msa/muscle/). A phylogenetic tree was constructed using the maximum-likelihood method in MEGA5 (Tamura et al., 2011). Using the previously obtained RNA-Seq dataset comparing two genotypes that differ in TSSM susceptibility (He et al., 2020), reads of genes assigned to the terpenoid biosynthetic module were mapped to assembled sequences to calculate read counts for each unigene. Differentially expressed genes (DEGs) between different experimental conditions were filtered using a Benjamini-Hochberg false discovery rate of 0.05 and a threshold of log 2 -transformed fold-changes (treatment/control) > |1.5|.

Putative cis-element analysis
The 2000 bp intergenic sequences upstream from the initiation start of CsTPS2, CsTPS9 and CsTPS19 were analysed for the presence of cis-acting elements using the PlantCARE database (http://bioinformatics.psb.ugent.be/webtools/plantcare/). Aligned motifs for each promoter were listed as their distances to the start codon of the gene.

Volatile collection and analysis
The second fully expanded leaves of 3-wk-old cucumber plants were infested with 50 adult TSSM, 10 thrips or 10 aphids, or were left uninfested. Volatile emissions of herbivore-infested and nontreated plants were collected on Tenax absorbent using dynamic headspace sampling as described by Zhang et al. (2020).
For semiquantification of volatiles, 1 µl of carvone in 10 ll MeOH was added to each Tenax liner before analysis and the areas under the curve (AUC) were normalized to that of the internal standard. For each experimental condition, volatile emissions were collected from five independent plants.

RNA isolation and gene expression analysis
Total RNA from cucumber leaves was extracted and reversetranscribed for quantitative reverse transcription polymerase chain reaction (qRT-PCR) analysis as described previously (He et al., 2020). Expression levels were normalized to cucumber b-Actin (Csa6M484600) and a-Tubulin (Csa4G000580) using the DDC t method (Livak & Schmittgen, 2001). Every measurement was performed with five biological replicates. Primer sequences are listed in Supporting Information Table S1.

Generation and expression of recombinant CsTPSs
Full-length cDNA sequences of CsTPS genes were cloned into the expression vector pACYCDuet (Novagen, Birmingham, UK) and transformed into the Escherichia coli strain BL21 (DE3). Primers used to obtain open reading frames are listed in Table S1. Production of heterologous protein was induced using 10 µM farnesyl pyrophosphate (FPP) or geranyl pyrophosphate (GPP) in a 1 ml vial as described previously (Mercke et al., 2004). As a negative control, raw protein extracts from E. coli expressing the empty pACYCDuet vector with substrates (FPP, GPP or GGPP) were incubated as described earlier. To collect terpenoid products, a 10 mm polydimethylsiloxane (PDMS, film thickness 1 mm) stir bar (Gerstel, M€ ulheim, Germany) was enclosed in each assay vial for 60 min incubation at 30°C with 250 rpm shaking. Subsequently, the stir bar was briefly rinsed in water, dried under a stream of nitrogen and enclosed in a glass liner for GC-MS analysis. The PDMS stir bars in between the measurements were cleaned by heating them to 310°C for 40 min with a helium flow. The tentative identification of enzyme-derived compounds was based on the comparison of mass spectra with those in the NIST 2005, Adams (2007) and Wageningen Mass Spectral Database of Natural Products, as well as experimentally obtained linear retention indices (LRIs). Essential oil of basil (Ocimum basilicum) was used to characterize cadinol. For each TPS enzyme and substrate combination, assays were repeated at least twice (n = 3) but most often three times (n = 4). For all replicates, the major products were similar and in the same order of relative magnitude. Terpenoids were semiquantified by calculating the area under the curve (AUC). To determine efficient mono-(sesqui-) TPS activity, the ratio of the sum of the AUC of all mono-(sesqui-) terpene products to that of all mono-(sesqui-) terpenes, including the nonspecific geraniol (farnesols), was set to be > 50%.

Induction and quantification of ffLUC activity
Four-week-old Arabidopsis reporter plants were screened for temporal dynamic imaging of bioluminescence under a diurnal light regime with light ramping to mimic natural light conditions. Plants were acclimatized 24 h before imaging started. The first day of imaging was always under noninduced conditions. Plants were sprayed with 1 mM D-luciferin (Promega) twice a day. Thirty-minute interval imaging of firefly-LUCIFERASE (ffLUC) activity and the determination of relative luminescence profiles were done as described by Van Hoogdalem (2020). Photon emission was depicted with false colour scales, with blue indicating low activity and red indicating high activity.
Transgenic Arabidopsis plants were monitored for reporter activity after various (a)biotic stresses, including mechanical damage (leaf puncturing using a 0.2-mm-diameter needle), JA, SA or abscisic acid (ABA) (all 5 ll 1 mM + 0.01% Tween-20), or individual TSSM, thrips or aphids. Leaves were visually checked for whether herbivores stayed on the leaf where they were introduced. After luminescence measurements were finished, plants were visually checked as to whether herbivores were alive. Experiments were performed with three independent lines per construct and five plants per experimental condition (n = 15). Arabidopsis p35S::ffLUC reporter plants were used as controls.

Results
Cell-content feeding TSSM and thrips, and phloem-feeding aphids induce different terpenoid-enriched volatile profiles After 3 d of TSSM feeding, damage as chlorotic spots was clearly visible by eye and total volatile emission increased 11-fold and 16-fold upon thrips feeding compared with the emission of noninfested plants (Fig. 1a). By contrast, after aphid feeding volatile emission increased less than two-fold. The volatile blend consisted of green leaf volatiles, benzoates, oximes and terpenoids, of which the latter comprised 30% of the total blend released by noninfested plants (Fig. 1b; Table S2). More than half of all terpenoids emitted by noninfested plants were sesquiterpenoids, of which a-copaene was the most dominant. The presence and abundance of terpenoids changed depending on the herbivore. The contribution of terpenoids increased to 38% and 43% after 3 d of thrips and TSSM infestation, respectively (Fig. 1b). In both cases, (E,E)-a-farnesene was the dominant terpenoid followed by (E)-b-ocimene, linalool, myrcene and (E)-4,8dimethylnona-1,3,7-triene (DMNT). Interestingly, infestation www.newphytologist.com by both cell-content feeders increased the contribution of terpene alcohols and aldehydes when compared to the blend of noninfested plants. While aphid infestation resulted in an increased contribution of terpenoids to the total volatile blend, the composition of the induced terpenoid blend differed from that of cellcontent feeders with predominantly monoterpenes and monoterpene alcohols but a lower proportion of sesquiterpenes compared with the blend of noninfested plants (Fig. 1c). Three days after the onset of feeding, limonene was the dominant terpenoid in aphid-infested plants (Fig. 1d).

Identification of the terpene biosynthetic module
To study the regulation of terpene biosynthesis by herbivory, the cucumber genome was analysed for putative gene models associated with terpene biosynthesis (Fig. 2). Seventy genes could be assigned to one of the six functional modules in terpenoid biosynthesis, including eight prenyl-transferases, two IPP isomerases and 10 MEP and MVA pathway-associated genes (Table S3). Additionally, 12 triterpene synthases were detected, and 34 gene models were identified as putative TPS genes. From these, 24 full-length gene models encoding putative proteins with 313 to 813 amino acids and at least five exons were renamed as CsTPS1-24 according to their chromosomal position, while three other ones that are too short to encode a functional TPS protein were renamed as CsTPS25-27 (Table S4). The majority of CsTPS genes were found in clusters located on chromosomes I, II and III, suggesting multiple duplication and neofunctionalization events on these chromosomes (Fig. S1). Phylogenetic analysis classified most cucumber TPSs as TPS-a (11 members), TPS-b (eight members) and TPSg (three members) (Fig. S2). Chromosomes VI and VII both contain a single full-length TPS gene, CsTPS23 and CsTPS24 respectively, and the partial CsTPS27 is located on chromosome VII. CsTPS23 and CsTPS24 are classified as TPS-c and TPS-e/f, respectively, and most likely encode COPALYL DIPHOSPHATE SYNTHASE and KAURENE SYNTHASE. Gene models associated with the MEP and MVA pathways or encoding IPP isomerases and prenyl-transferases were found to be located across all chromosomes and not specifically in the proximity of TPSs.
Based on amino acid sequence similarity to representative TPSs from Arabidopsis and tomato, CsTPS1-11 were tentatively classified as monoterpene synthases, CsTPS12-22 as sesquiterpene synthases and CsTPS23 and CsTPS24 as diterpene synthases. Most TPSs contain elements known to be conserved in TPS, such as RRX8W, RXR, DDXXD and NSE/DTE motifs (Table S4). Plastid transit peptides were predicted for CsTPS1-3, CsTPS9-11 and CsTPS24supporting their putative role as monoterpene synthases or diterpene synthase (CsTPS24)while a mitochondrial targeting peptide was predicted for CsTPS4. A secretory pathway signal peptide was predicted for CsTPS5. Remarkably, a plastid transit peptide was predicted for the putative sesquiterpene synthase CsTPS15, while no signal peptides were predicted for putative monoterpene synthases CsTPS6-8 and diterpene synthase CsTPS23.

Transcriptional responses of the terpenoid biosynthetic module upon herbivory
The effect of herbivory on the expression of the terpenoid biosynthetic genes was analysed using the spider mite-induced leaf transcriptome dataset presented by He et al. (2020). Transcripts of most of the annotated genes involved in the MEP and MVA pathways were present in leaves of genotypes differing in TSSM susceptibility (Fig. 2). Multiple genes in both pathways producing the precursors for terpene biosynthesis were regulated during early TSSM infestation. 3-HYDROXY 3-METHYLGLUTARYL-CoA REDUCTASE (HMGR) in the MVA pathway and 1-DEOXY-D-XYLULOSE-5-PHOSPHATE SYNTHASE (DXS) in the MEP pathway were strongly induced in genotype 'Chinese Long', which was least susceptible to TSSM, while they were repressed in the susceptible genotype 'Corona' (Figs 2, S3). A similar transcriptional response was found for two of the four GERANYLGERANYL DIPHOSPHATE SYNTHASEs (GGPPSs) while GERANYL DIPHOSPHATE SYNTHASE (GPPS) and FARNESYL DIPHOSPHATE SYNTASE (FPPS) were regulated similarly in both genotypes.
In control leaves, TPS5 showed the highest expression in both genotypes, but overall expression of the TPS genes was low, at 2.5-5.8% of the average overall gene expression (Table S5). TSSM feeding increased the expression of several TPSs (Fig. 2). Quantitative PCR analysis confirmed that TSSM induced CsTPS9 and CsTPS19, and to a lesser extent CsTPS2-5 and CsTPS21 (Figs 3, S3). Thrips feeding induced higher TPS expression than TSSM, with CsTPS9, CsTPS19 and CsTPS21 as the most strongly induced genes. By contrast, aphid feeding resulted only in some induction of CsTPS2 expression and, to a lesser extent, of CsTPS3-4, CsTPS19 and CsTPS21.

Characterization of TPS
Nineteen TPS, including all TSSM-induced TPSs, except for CsTPS4, were successfully cloned and heterologously expressed in E. coli ( Fig. 4; Table S6). All heterologous TPSs, except CsTPS13 and CsTPS18, accepted substrates GPP and FPP, resulting in the formation of various mono-and sesquiterpenes, respectively. CsTPS proteins with a predicted chloroplast-target peptide efficiently produced one or multiple monoterpenoids. These enzymes also catalysed the formation of minor amounts of sesquiterpenes from FPP.
CsTPS1-3 have predicted chloroplast-target peptides and produced predominantly linalool when GPP was supplied as substrate. Furthermore, CsTPS9 catalysed the formation of (E)-bocimene and myrcene, and small amounts of sabinene and (Z)-bocimene from GPP. CsTPS11 and CsTPS15 both catalysed the formation of myrcene, limonene, (E)-b-ocimene, linalool and aterpineol from GPP in different amounts and/or ratios, and CsTPS11, in addition, catalysed the formation of an unidentified monoterpene (LRI 1169). The product profile of CsTPS10 was distinct from the other chloroplast-targeted enzymes as it produced a-pinene, a-phellandrene, sabinene, b-pinene, myrcene, linalool and a-terpineol from GPP. Despite the presence of a  predicted chloroplast-target peptide, CsTPS1-3, CsTPS9-11 and CsTPS15 catalysed the formation of minor amounts of sesquiterpenes from FPP. We were not able to clone CsTPS6-8, CsTPS16 and CsTPS20, all genes without a targeting sequence. CsTPS13 and CsTPS18 were not active in any of the assays that we performed. Other proteins without a predicted targeting peptide were CsTPS12, CsTPS14 and CsTPS17, which all catalysed the formation of (E)-b-farnesene, and to a lesser extent (E,E)-afarnesene and (E)-nerolidol from FPP. CsTPS19 predominantly catalysed the formation of (E,E)-a-farnesene from FPP, consistent with Mercke et al. (2004), but produced also traces of (E)-bfarnesene, (Z,E)-a-farnesene, bisabolene and an unknown sesquiterpene (LRI 1485). CsTPS21 catalysed the formation of (E)-caryophyllene and a-humulene from FPP, and the major products of CsTPS22 were (E)-nerolidol and cadinol. When cytosolic TPSs were supplemented with GPP, most of the enzymes produced small amounts of myrcene, limonene and linalool. An exception was CsTPS19 which efficiently catalysed the formation of (E)-b-ocimene, myrcene and linalool from GPP. CsTPS19 also accepted GGPP to produce the diterpenoid geranyl linalool (Fig. S4), confirming CsTPS19 to be an efficient mono-, sesqui-and diterpene synthase. Both CsTPS23 and CsTPS24 were predicted to encode a diterpene synthase, but only CsTPS24 accepted GGPP to produce geranyl linalool. Both enzymes accepted GPP as a substrate to produce (E)-b-ocimene, linalool and myrcene in minor amounts. CsTP24 produced a small amount of cadinol and both enzymes produced (E)nerolidol from FPP.

Induction of herbivore-inducible TPS results in circadian enzymatic activity
CsTPS2, CsTPS9 and CsTPS19 were selected for further analysis of the regulation of terpene biosynthesis upon herbivory. Multiple cis-acting regulatory elements (CAREs) located in the 2000 bp sequences upstream of the initiation start of these genes, considered to represent the promoter (pCsTPS), were identified as responsive to stress-related phytohormones JA, SA and ABA ( Fig. 5a; Table S7). The number of these motifs in pCsTPS19 was about half of those of pCsTPS2 and pCsTPS9 (Table 1). Furthermore, the promoters contained multiple motifs related to light responsiveness and circadian rhythmicity (Table 1; Fig. 5a), suggesting that these pCsTPSs may be regulated by photoperiod in addition to JA, SA and ABA.
The herbivore species used in our study all accept Arabidopsis as host (Zhurov et al., 2014;Kloth et al., 2015;Thoen et al., 2016). Visual damage caused by TSSM feeding could be observed after 2 d as white spots, mostly near the veins, and the occurrence of dead cells was confirmed by trypan blue staining (Fig. 5b). Thrips feeding resulted in silver damage as a result of collapsed cells, first visible at 2 d after the onset of feeding. Aphid infestation did not inflict visual damage but infestation was considered to be successful as offspring were present at 3 d after introduction of the aphids.
b-Glucuronidase staining of pCsTPS9::GUS/ffLUC reporter plants showed that expression of the reporter gene was absent in noninfested plants, except for the cotyledons, which in some plants stained blue (Fig. S5). Upon TSSM feeding, leaves stained blue in a patchy pattern corresponding to the damage spots inflicted by the mites. Stained cells in these infested areas were mostly located in the mesophyll layer. Some of the younger leaves that were not damaged by TSSM showed minor staining in the petioles and the veins. By contrast, TSSM-infested pCsTPS19 reporter plants showed only blue colouring in local infested leaves, and no blue colouring was observed in TSSM-infested pCsTPS2 reporter plants. Thrips infestation resulted in a stronger response of pCsTPS9 and pCsTPS19 reporter plants compared with TSSM, but, similar to TSSM feeding, pCsTPS9 reporter To better visualize the dynamics of promoter activation, we used luminescence monitoring. Luminescence increased in pCsTPS9 reporter plants within 1 h after the introduction of a single thrips, and a 73-fold increase was observed at the end of the second light period (Fig. 5e). After recording luminescence, we observed that the originally infested leaf was seriously damaged by thrips and a number of other leaves showed silver damage spots as well. By contrast, aphids did not cause any detectable induction of luminescence in the pCsTPS9 reporter plants (Fig. 5e). Visual observation showed that aphids walked around for c. 2 h and then remained in the same position, implying they were probing/feeding (Kloth et al., 2015). After 96 h no damage was visible on the aphid-infested plants.
Luminescence in pCsTPS19 reporter plants increased upon TSSM and thrips feeding but not aphid feeding (Fig. 6), similar to pCsTPS9. Also, the greater damage inflicted by thrips infestation resulted in stronger luminescence than as a result of TSSM feeding. Damage-induced luminescence was only visible locally, at positions where thrips and TSSM had been feeding in pCsTPS19 reporter plants, while in pCsTPS9 some systemic luminescence was observed. By contrast, pCsTPS2 reporter plants only displayed a minor increase in luminescence upon thrips infestation, mechanical damage or JA treatment but were responsive to TSSM and aphid feeding and SA and ABA treatment (Fig. 6a).

Discussion
The cucumber TPS gene family is relatively small The TPSs constitute a mid-sized gene family in plants (Chen et al., 2011). The CsTPS gene family consists of 27 gene models  Fig. S4 shows the GC-MS chromatograms for CsTPS2, CsTPS9 and CsTPS19 incubated with different substrates. Product profiles of other enzymes that were characterized can be found in Table S6.   Wei et al. (2016), and hence form a relatively small TPS family compared with other flowering plant species such as Arabidopsis (40 putative TPS gene models; 32 putatively full length; Aubourg et al., 2002), tomato (44;29;Falara et al., 2011), rice (57;34;Chen et al., 2011) and grape (152;69;Martin et al., 2010). Remarkably, the TPS family in apple consists of 55 gene models of which only 10 are functional (Nieuwenhuizen et al., 2013). The majority of the cucumber TPS genes are organized into four clusters located on three chromosomes, consistent with the clustering of TPS genes in other plant species, including Arabidopsis (Aubourg et al., 2002), tomato, (Falara et al., 2011) and grape (Martin et al., 2010). Clustering of metabolism-associated genes is relatively common, possibly ensuring co-inheritance to keep biosynthetic pathways complete (Nutzmann & Osbourn, 2014). Furthermore, clustered genes could share similar regulation mechanisms such as through chromatin modification (Wegel et al., 2009). TPSs were reported to frequently colocalize with P450 genes (Boutanaev et al., 2015). Remarkably, in cucumber only a single P450 gene and no members of other classes of genes such as glycosyl transferases were found located within or near any of the TPS clusters. Just as reported for other species, CsTPSs located within the same cluster in the genome were assigned to similar clades in the phylogenetic tree and are hence more homologous to each other, probably as a consequence of tandem duplication. Evolutionary analysis of terpenoid biosynthesis-related genes and supergene clusters of 17 genomes demonstrated that genes encoding TPSs are more enriched for tandem duplications than genes encoding enzymes involved in the upstream MVA pathway and IPP isomerases (Hofberger et al., 2015).
Like many TPSs characterized in other plant species, CsTPSs can accept different substrates. Most CsTPSs that were tested in vitro catalysed the formation of multiple terpenes from the same precursor, a common phenomenon in plant TPSs. For example, 10 different monoterpenes were formed by a single Arabidopsis TPS (Chen et al., 2004). Most of the characterized tomato TPSs catalysed the formation of more than one terpene (Falara et al., 2011). At the same time, some of the terpenes we detected were synthesized by multiple CsTPSs. For example, linalool was the major product of TPS-g clade CsTPS1, CsTPS2 and CsTPS3, and also a minor product of most of the other CsTPSs. Linalool is a common floral and foliar volatile with two distinct enantiomers that have distinct roles in pollinator attraction and plant defence (Raguso, 2016;He et al., 2019). Linalool enantiospecific enzymes have been identified in, for example, Arabidopsis, producing (R)-(À) and (S)-(+)-linalool, respectively, as their major products (Ginglinger et al., 2013). Whether cucumber leaves and flowers emit specific isomers is unknown and the present study did not allow us to distinguish between both enantiomers. Further studies might also investigate whether CsTPS1-3 contribute to enantiomeric-specific linalool formation, if any, and its specific role in plant-arthropod interactions.
Most cucumber TPSs convert GPP and FPP to acyclic monoand sesquiterpenes while the formation of cyclic terpenes was catalysed by a limited number of CsTPSs only, including CsTPS10 which catalyses the formation of a pinyl cation en route  to the formation of b-pinene, a-pinene, sabinene and aphellandrene. The root-specific CsTPS11 was also demonstrated to use a pinyl cation as intermediate in the formation of cyclic terpenes (Wei et al., 2016). Other cyclic sesquiterpenes, including (E)-caryophyllene, a-humulene and cadinol, were produced by just a few CsTPSs. Depending on the presence of terpenoid precursors in different cell compartments, the product profile of CsTPSs in planta may differ from those in vitro. Previously, we demonstrated that targeting a nerolidol synthase from strawberry to different cell compartments in Arabidopsis determined the abundance and ratio of mono-and sesquiterpenoid products, confirming the importance of precursor availability for product formation (Aharoni et al., 2004;Kappers et al., 2005;Houshyani et al., 2013). Accordingly, the predicted subcellular localization of cucumber TPSs coincides with the presence of the precursors which they effectively use. An exception is CsTPS19, which accepts GPP, FPP and GGPP efficiently to produce (E)-b-ocimene, (E,E)-a-farnesene and geranyl linalool, respectively, supporting its role as a genuine multiplefunction TPS. Thus, the enzymatic activity of the CsTPSs in combination with their subcellular localization and their expression together determine which terpene compounds are produced in cucumber under which conditions.
Although thrips infestation resulted in more damage and a higher total volatile emission compared with that of TSSM, the composition of the terpenoid blend was comparable. By contrast, upon feeding by aphids, which mainly induces SA signalling (Moran & Thompson, 2001), the terpenoid blend differed both quantitatively and qualitatively from that of TSSM-and thripsdamaged plants.
Considering the multiple minor products that are produced by CsTPSs besides their major products, induction of these genes enables the plants to produce a wide spectrum of volatiles and fine-tune their volatile signature in response to herbivory. Most of the terpenoids emitted by noninfested and infested leaves correlated well with the product profiles of the CsTPSs and the expression of the corresponding genes. An exception was the increased emission of a-pinene, a-phellandrene and sabinene by leaves infested with cell-content feeding herbivores, while the gene that encodes the most likely corresponding terpene synthase (CsTPS10) was not upregulated. Genes associated with the biosynthesis of terpenoid precursors upstream of the TPSs were also found to be differentially regulated, and this might have implications for the availability of precursors for constitutively expressed TPSs in different cell organelles. Hence, the final terpenoid metabolite profile will be determined by TPSs that are induced upon herbivory as well as those that are constitutively expressed. Cucumber genotypes previously characterized for their herbivore-induced plant volatiles emit mostly similar compounds with different abundances that consequently affected the level of indirect defence (Kappers et al., 2010(Kappers et al., , 2011. Further fine-tuning of the volatile signature in response to different herbivore feeding will have consequences for multitrophic interactions. Although we did not compare the different herbivore-induced volatile blends regarding the attractiveness of these odours towards natural enemies, natural enemies can distinguish different blends of terpenoid volatiles upon infestation by different herbivorous arthropods. For instance, lima bean plants emitted different volatile blends as a result of feeding by Spodoptera exigua and T. urticae and, consequently, P. persimilis predators were more attracted to plants infested by their prey, T. urticae (de Boer et al., 2004). When lima bean and cucumber plants were infested by both herbivores separately or together, the plants emitted different amounts of volatile compounds, including several terpenes, and the dual-infested plants were more attractive to predatory mites than those damaged by only a single herbivore species (De Boer et al., 2008).
Involvement of stress-related phytohormones in the response of CsTPS to herbivores with different feeding guilds Two-spotted spider mites and thrips activated transgenic Arabidopsis pCsTPS9 and pCsTPS19 reporter plants, while aphids and TSSM activated pCsTPS2 reporter plants, indicating that the promoters of CsTPS9 and CsTPS19 respond to cell-content feeders. As reporter activity of these plants was also activated by mechanical damage, a shared upregulation in the response to TSSM and thrips could be the result of the fact that both herbivores cause mechanical wounding. Indeed, thrips inflicted more damage than TSSM and, correspondingly, thrips induced stronger luminescence in reporter plants. Limited, one-time, mechanical damage quickly activated the promoter, which then decreased to the control level within 1 d. Repetitive mechanical damage of lima bean plants using an artificial caterpillar resulted in an induced volatile blend that was strikingly similar in quality to the blend induced by herbivore feeding (Mithofer et al., 2005), suggesting that repetitive mechanical damage inflicted by herbivory is sufficient to trigger the biosynthesis of herbivoreinducible volatiles in plants. Phloem

Research
New Phytologist damage, as they navigate their stylets between the cell walls to reach phloem vessels with limited harming of cell integrity (Tjallingii & Hogen Esch, 1993;Kloth et al., 2016). Comparison of the up-and downregulated genes in Arabidopsis infested by herbivores with different feeding habits showed that similar transcriptional responses were induced by chewing generalist species Plutella xylostella and Spodoptera litoralis, while generalist F. occidentalis and phloem-feeding generalists Bemisia tabaci and M. persicae caused more and different transcriptional changes compared with P. xylostella (Reymond et al., 2004;de Vos et al., 2005;Kempema et al., 2007;Kusnierczyk et al., 2007;Little et al., 2007;Ehlting et al., 2008).
Both pCsTPS9 and pCsTPS19 reporter plants were responsive to JA but not to SA and ABA. The JA/ethylene pathway is activated in response to thrips feeding (Steenbergen et al., 2018) and TSSM infestation in multiple species, including lima bean , tomato  and cotton (Miyazaki et al., 2014), although a recent study showed that, unlike adults, juvenile TSSM induce SA but not JA defences in tomato (Liu et al., 2020). Endogenous JA and SA increased within hours after the onset of TSSM infestation in Arabidopsis (Zhurov et al., 2014) and Capsicum (Zhang et al., 2020). In cucumber, JA induces a blend of volatiles that is qualitatively similar to the blend induced by TSSM (Kappers et al., 2010). Furthermore, methyl-SA was emitted upon TSSM herbivory by multiple plant species, including lima bean (Dicke et al., 1990a,b), tomato  and cucumber (Kappers et al., 2011). Neither SA nor ABA application triggered any response of the reporters driven by pCsTPS9 or pCsTPS19, and although the SA-regulation network may play a role, it appears that JA dominates the regulation of these TPSs that are part of the inducible defence to cell-content feeders.
Interestingly, the promoter activity of pCsTPS2 reporter plants was triggered by aphids and TSSM, and by SA and ABA application, whereas JA only provoked minimal promoter activity in these plants. Aphids are known to induce formation of ABA, and ABAregulated genes are over-represented among genes that are induced by M. persicae saliva infiltration into Arabidopsis leaves (Hillwig et al., 2016). Feeding by the carmine spider mite T. cinnabarinus altered ABA content in tomato plants (Gawro nska & Kiełkiewicz, 1999). Furthermore, SA-regulated transcripts increase upon aphid feeding (Moran & Thompson, 2001), although this response is very local (de Vos et al., 2005). The induction of the CsTPS2 promoter by TSSM and aphids could be explained via the presence of ABA-and SA-responsive elements in this promoter, suggesting that ABA and SA are important for regulation of CsTPS2 and its contribution to the aphid feeding-induced volatile blend. Multiple CAREs present in the promoter sequences of CsTPS2, CsTPS9 and CsTPS19 were identified as possibly involved in JA, SA or ABA responsiveness. For example, G-boxes (CACGTG), required for JA-mediated expression regulation (Kim et al., 1992;Endt et al., 2007), W-boxes (TTGACC) associated with responsiveness to SA (Li et al., 2006) and ABRE motifs, related to ABA responsiveness (Lenka et al., 2009), were present in all three promoters. The wound-responsive WUN motif was found in pCsTPS9 and pCsTPS19 but not in pCsTPS2, and this might play an as-yetunknown role in the different responses to the mechanical damage inflicted by herbivores from different feeding guilds. Whether motifs in these promoters really function as binding sites to potential transcription factors, and which conditions render specific CAREs indispensable for promoter activity are still unclear and were not the purpose of our study. However, the presence of these motifs probably allows promoters to be bound by transcription factors induced through JA, SA or ABA signalling, hence regulating the volatile blend resulting from hormonal crosstalk.

Regulation of CsTPS by light and the circadian clock
The observed rhythmic oscillation in luminescence might be the result of herbivore behaviour, as they often display rhythmic feeding. For example, T. ni caterpillars show diurnal feeding behaviour (Goodspeed et al., 2012). However, regardless of whether feeding behaviour of the herbivores in this study was circadian, the nocturnal maximum activity of reporter plants upon JA treatment demonstrates that the expression of the studied CsTPSs displays circadian rhythmicity.
Rhythmic emission of volatiles and expression of genes involved in their biosynthesis have been reported in multiple species. Methyl-JA-induced emission of terpenes and methyl-SA from Norway spruce displayed a diurnal rhythm (Martin et al., 2003). Lima bean leaves mechanically damaged during the day emitted maximum amounts of (E)-b-ocimene and (Z)-3-hexenyl acetate in the late photo-phase, while nocturnally applied mechanical damage triggered nocturnal emission of (Z)-3-hexenyl acetate but only minor amounts of (E)-b-ocimene, which burst after onset of the photo-phase . Phaseolus vulgaris plants released trace amounts of volatiles with no obvious rhythm, but upon infestation with Liriomyza huidobrensis larvae, plants released higher amounts of volatiles with a clear rhythm which peaked at the end of the day (Sufang et al., 2013). The expression of Artemisia annua QH6, encoding a pinene synthase is diurnally regulated (Lu et al., 2002) and luciferase activity driven by the QH6 promoter with a mutated G-box showed a rhythm lacking a peak in the early morning which was present when the intact G-box was present (Zhou et al., 2015). Multiple light-associated CAREs are present in the promoters we tested, including the light-responsive element box I (TTTCAAA) (Yamada et al., 1994), and a circadian motif (CAANNNNATC, Piechulla et al., 1998). The G-box present in each of the promoter sequences could be essential for light regulation as well (Lopez-Ochoa et al., 2007).
Our results suggest that the promoters tested induce peak gene expression during the night, while the emission of the corresponding terpenes and green leaf volatiles occurs mainly during the light period (I. F. Kappers, unpublished). Possibly, high nocturnal expression of CsTPS2, CsTPS9 and CsTPS19 results in the accumulation of active enzymes which are 'ready to go' when enough substrate becomes available at the onset of the day to fuel production of energy-costly secondary metabolites only during the photoperiod. This is in agreement with the burst of emission of (E)-b-ocimene upon the onset of light by lima bean plants which were damaged in the previous dark period . In lima bean, the expression of b-OCIMENE SYNTHASE is regulated via JA accumulation at wounded sites and the biosynthesis of (E)-b-ocimene is dependent on CO 2 fixation by photosynthesis in the chloroplasts , where the MEP pathway synthesizing the terpenoid precursors GPP and GGPP is also located. The expression of the genes of the MEP pathway is light-dependent (Hemmerlin et al., 2012) and, indeed, expression of almost all MEP-pathway genes in Arabidopsis seedlings is repressed in darkness (Hsieh & Goodman, 2005). The expression of MEP-pathway genes encoding 1-DEOXY-D-XYLULOSE-5-PHOSPHATE SYNTHASE and two GPP SYNTHASES are upregulated in cucumber leaves upon TSSM infestation (He et al., 2020). Hence, it is not unlikely that the supply of precursors determines the diurnal emission of terpene volatiles in cucumber leaves triggered by herbivory. To verify this, the rhythmicity of expression of the genes encoding the precursor supply pathways should be evaluated.
In conclusion, we identified the cucumber TERPENE SYNTHASE (CsTPS) gene family from the sequenced cucumber genome and characterized their role in the production of volatiles in leaves with and without herbivore feeding. We identified the CsTPS genes that contribute to the volatile terpenoid blend of cucumber leaves upon feeding by important cucumber pest species with dissimilar feeding guilds and revealed the involvement of stress-related phytohormones and circadian rhythmicity in the regulation of the production of this terpenoid blend.

Supporting Information
Additional Supporting Information may be found online in the Supporting Information section at the end of the article.            New Phytologist is an electronic (online-only) journal owned by the New Phytologist Foundation, a not-for-profit organization dedicated to the promotion of plant science, facilitating projects from symposia to free access for our Tansley reviews and Tansley insights.
Regular papers, Letters, Viewpoints, Research reviews, Rapid reports and both Modelling/Theory and Methods papers are encouraged. We are committed to rapid processing, from online submission through to publication 'as ready' via Early Viewour average time to decision is <26 days. There are no page or colour charges and a PDF version will be provided for each article.
The journal is available online at Wiley Online Library. Visit www.newphytologist.com to search the articles and register for