Previous studies have shown that significant differences exist in the abundance and composition of terpenes in flowers and fruit across various Actinidia species (Crowhurst et al. 2008; Nieuwenhuizen et al. 2015) and among different cultivars and accessions within the same species (Wang et al. 2021). The availability of a high-quality reference genome for diploid A. chinensis var. chinensis Red5 (Pilkington et al. 2018) has made this variety a key tool in gene discovery efforts. Therefore, a detailed analysis of terpene composition in a range of Red5 tissues and fruit developmental stages was undertaken by GC-MS after solid phase micro extraction (SPME) (Fig. 1, Table S1). In whole flowers, linalool was highly abundant (2800 ng∙g−1 fresh weight - FW) along with farnesol (470 ng∙g−1 FW) and (E)-nerolidol (123 ng∙g−1 FW). These compounds were also the major compounds detected in flowers of A. chinensis ‘Hort16A’ (Green et al. 2012), but the lilac compounds present in A. arguta flowers were not observed (Chen et al. 2010). In buds, linalool, geraniol and limonene were the major compounds detected. In fruit, 1,8-cineole was the most abundant terpene detected (max. 130 ng∙g−1 FW), confirming the results from Zeng et al. (2020). A range of other terpenes including β-caryophyllene and bornylene were detected in developing fruit, but generally at relatively low concentration (< 30 ng∙g−1 FW). Linalool was the major compound detected in young expanding leaves.

The Red5 genome (Pilkington et al. 2018) and the corresponding annotated gene models (designated ‘-R5’) were interrogated using BLASTP with known TPS sequences from kiwifruit, Arabidopsis, and tomato as queries. Twenty-two unique TPS gene models were identified (Table S2A) and systematically named as AcTPS1-R5-AcTPS22-R5. Eleven TPS genes and alleles previously reported and functionally characterised in various kiwifruit species (Table S2B) were identified in the Red5 genome (Table 1): AcTPS8-R5 and AcTPS11-R5 correspond to AdAFS1 (α-farnesene synthase) and AdGDS (germacrene-D synthase) from A. deliciosa (Nieuwenhuizen et al. 2009); AcTPS13-R5 corresponds to linalool synthases AaLS1 and ApLS1 from A. arguta and A. polygama (Chen et al. 2010) and AcTPS14-R5 corresponds to AcNES1 (nerolidol synthase) previously isolated from A. chinensis ‘Hort16A’ (Green et al. 2012). Five Red5 gene models located on chromosome 29 (AcTPS16-R5-AcTPS-20-R5) have been resolved to four functional TPS genes (AcTPS1a-d) in Red5 (Zeng et al. 2020) and are likely alleles/orthologues of AcTPS1 and AaTPS1 from ‘Hort16A’ and A. arguta (Nieuwenhuizen et al. 2015). Five Red5 gene models were truncated and likely represent non-functional pseudogenes (AcTPS2-, 4-, 6-, 12-, 22-R5, Table S2A).

Of the remaining eight gene models, two represent putative full-length genes encoding ent-copalyl diphosphate (AcCPS) and ent-kaurene synthases (AcEKS) involved in the biosynthesis of the non-volatile gibberellins (Keeling et al. 2009). Three gene models (AcTPS1-R5, 3-R5 and 5-R5) were full-length and showed highest homology to TPS genes encoding α-farnesene, linalool and germacrene-D synthases (Table 1). Primers designed to these three gene models amplified TPS proteins of 776, 546 and 561 amino acid residues respectively (Table 1). Two gene models (AcTPS9-R5 and AcTPS10-R5) appeared to be incorrectly annotated as separate genes. A full-length TPS gene with homology to CsRLIS (Camellia sinensis (R)-linalool synthase, QNI69163.1) was amplified using primers to the N-terminus of AcTPS9-R5 combined with the C-terminus of AcTPS10-R5 (Figure S1) resulting in a 579 amino acid (aa) open reading frame (ORF). AcTPS15-R5 encodes a truncated gene model of 1410 bp (470 aa). The N-terminus of AcTPS15-R5 was extended by studying RNAseq reads upstream of the existing gene model’s start codon (Figure S1). The full-length AcTPS15-R5 encodes a protein of 575 aa. Alignment of the predicted amino acids of Red5 TPS genes showed several conserved motifs (Figure S2). AcTPS3-R5, AcTPS5-R5 and AcTPS9/10-R5 contained both the RRX(8)W, DDXXD and NSE/DTE motifs, while AcTPS1-R5 and AcTPS15-R5 only contained the DDXXD and NSE/DTE motif (Figure S2).

To investigate the evolutionarily relationships of Actinidia TPS genes, a phylogenetic tree was generated comparing previously characterised functional kiwifruit TPS genes, the fifteen full-length AcTPS-R5 genes and selected TPS from other species (Fig. 2; Table S3). The results showed that all the Actinidia TPS proteins clustered into angiosperm-specific clades (Fig. 2). The TPS-b and TPS-g subclades include the highest number of AcTPS genes, with five and eight members respectively and include members that function as either mono- or sesquiterpene synthases. AcTPS5-R5 clusters with other TPS-a subfamily members including the previously characterised AdGDS (germacrene-D synthase) (Nieuwenhuizen et al. 2009), whilst AcTPS1-R5 clusters with AdAFS1 (α-farnesene synthase) into TPS-e/f.

In order to gain better insight into the evolution of TPS in kiwifruit, the chromosomal location was determined for each of the Red5 TPS genes (Fig. 3). The largest number of TPS genes occurred on chromosome 29 (n = 6) which contains the tandemly repeated AcTPS1a-d cluster previously characterised by Zeng et al. (2020), which suggests the occurrence of multiple duplication and recombination events may have occurred on chromosome 29. As the diploid kiwifruit genome is the result of several ancient duplication events (Pilkington et al. 2018), homeologous pairs of duplicate of TPS can also be expected. Two pairs of TPS genes were identified on homeologous chromosomes 12 and 17: AcTPS3-R5::AcTPS9/10-R5 and AcTPS5-R5::AcTPS11-R5. AcTPS1-R5::AcTPS8-R5 are located on homeologous chromosomes 8 and 15 and AcTPS13-R5::AcTPS14-R5 are located on homeologous chromosomes 18 and 22 (Fig. 3). A number of the TPS pseudogenes identified in the Red5 genome appear to be non-functional paralogues of neighbouring genes e.g. AcTPS4-R5 and AcTPS5-R5 on chromosome 12 and AcTPS12-R5 and AcTPS13-R5 on chromosome 18.

To determine the temporal changes and tissue distribution of TPS transcripts, expression was measured in a total of 17 kiwifruit tissues (i.e. buds, young leaves, mature leaves, cane, flowers, fruit and roots). Relative quantification real-time PCR (qRT-PCR) was used with gene-specific primers designed to all thirteen TPS genes likely involved in producing volatile terpene products (i.e. excluding AcCPS and AcEKS involved in non-volatile gibberellin biosynthesis). The expression levels are presented as a heat map in Fig. 4. Distinct tempo-spatial expression patterns were observed for all TPS genes. Several genes such as AcGDS, AcAFS1 and AcNES1 were expressed in the majority of tested samples, apart from very low levels of expression in fruit flesh and peel at 150 d. In contrast, the expression of other genes was undetectable in most of the tested tissues (e.g. AcTPS1c, d). Of the newly identified volatile TPS genes, AcTPS15-R5 was predominantly expressed in young peel samples and root. AcTPS5-R5 was expressed strongly in developing fruit and bud tissues. AcTPS1-R5 and AcTPS3-R5 showed similar expression patterns with relatively high expression in roots, bud and ripe fruit (150 d) peel tissues. AcTPS9/10-R5 showed the highest expression in flower and 60 d fruit peel, but was expressed at lower levels in other organs such as mature fruit and leaves.

Of the thirteen AcTPS genes from Red5 likely involved in producing volatile terpene products, AcTPS1a-d have previously been biochemically characterised (Zeng et al. 2020) and four (AcAFS1, AcGDS, AcLS1, AcNES1) were assigned likely functions based on high sequence similarity (> 90%) to genes previously assigned activities in other Actinidia species (Table 1). Transient over-expression in Nicotiana benthamiana was used to determine the functional characteristics of the five new volatile-related kiwifruit TPS genes in planta. Binary vector constructs for each TPS were co-infiltrated into N. benthamiana leaves together with pHEX2_AcDXS (1-deoxy-D-xylulose-5-phosphate synthase) to increase monoterpene substrate production by the MEP pathway or with pEAQ-tHMGR-2A-BCCP1 (3-hydroxy-3-methylglutaryl-coenzymeA reductase) to increase sesquiterpene production by the mevalonate pathway (Zeng et al. 2020). The results (Fig. 5A/B; Table S4) showed that AcTPS15-R5 mainly catalysed the formation of the monoterpene geraniol and therefore it was named A. chinensis geraniol synthase (AcGES). AcTPS5-R5 (AcBCS) produced the sesquiterpene β-caryophyllene. AcTPS3-R5 (AcNES2) and AcTPS1-R5 (AcNES3) both produced mainly (E)-nerolidol, while AcTPS9/10-R5 (AcLIS/NES) made mostly linalool and its oxidised derivatives plus a small amount of nerolidol (Fig. 5A/B; Table S4).

The sesquiterpene synthases AcNES2, AcNES3 and AcBCS showed between 2-20 fold increases in sesquiterpene production by increasing the mevalonate-derived substrate pool by up-regulating HMGR (Table S4). AcGES produced ~ tenfold higher levels of geraniol in combination with DXS (Table S4). AcLIS/NES showed a ~ 70-fold increase in (E)-nerolidol production in combination with HMGR, but still produced larger absolute amounts of linalool (> 4-fold) and derivatives in the presence of HMGR. When combined with DXS, a 20-fold increase in linalool and derivatives was observed. Overall, (E)-nerolidol levels with HMGR were > 100-fold lower than linalool with DXS. These results suggest that AcLIS/NES is primarily a monoterpene synthase in planta supported by the MEP substrate pathway and produces much smaller amounts of (E)-nerolidol supported by the mevalonate pathway (Fig. 5A/B; Table S4).

To determine the activity of the five new volatile-related TPS kiwifruit TPS genes in vitro, recombinant HIS-tagged proteins were expressed in the Escherichia coli and proteins were purified by Ni2+ affinity purification. The products formed using GDP or FDP as substrates were analysed by SPME GC-MS. The in vitro results for the five active recombinant TPS enzymes (Fig. 6; Table S5) were very similar to those obtained for the transient in planta expression analysis.

AcLIS/NES produced the monoterpene linalool (87%) and the sesquiterpenes (E)-nerolidol (64%) as major products from GDP and FDP respectively (Fig. 6; Table S5). The reaction products formed by AcNES2 and AcNES3 with FDP were (E)-nerolidol (94% and 83% respectively). AcGES and AcBCS produced geraniol (81%) and β-caryophyllene (100%) in the presence of GDP or FDP respectively (Fig. 6; Table S5).

Enzyme kinetic analysis (Table 2) of the purified recombinant AcTPS proteins showed that the monoterpene synthase AcGES showed higher affinity towards GDP compared to AcLIS/NES (0.8 vs 31 µM), whilst AcBCS showed the highest affinity towards FDP (1.5 µM) with AcNES3 the lowest (85 µM). Unexpectedly, the dual function enzyme AcLIS/NES showed a higher Km with GDP (31 µM) versus FDP (7 µM) and also a lower catalytic efficiency (Kcat/Km) value with GDP versus FDP (0.00678 vs. 0.1103 s−1∙mM−1, respectively). In planta, the enzyme mainly produced linalool. Comparing the catalytic efficiency of the three (E)-nerolidol biosynthesis related proteins (AcLIS/NES, AcNES2 and AcNES3) indicated that AcLIS/NES showed the highest efficiency with FDP (0.1103 s−1∙mM−1) and the highest turnover rate (Kcat of 0.00082 s−1).

To determine the subcellular localisation of the five new volatile-related AcTPS genes, full-length ORFs were fused at the C-terminus to the GFP reporter gene and infiltrated into N. benthamiana leaves. Transient expression of GFP in protoplasts was then analysed by confocal laser scanning microscopy. Strong GFP fluorescence signals for AcNES3-GFP were detected in the cytoplasm (Fig. 7A, panel 1), whilst non-overlapping punctate red spots were observed in the corresponding chlorophyll autofluorescence (CA) image (Fig. 7A, panel 2).

For AcNES2-GFP, relatively weak GFP fluorescence signals were observed in the cytoplasm where ChloroP (Emanuelsson et al. 2000) predicted the gene to be localised (Fig. 7B1), but also some in chloroplasts (Fig. 7B2). GFP fluorescence signals for the sesquiterpene synthase AcBCS-GFP construct were detected in the cytoplasm (Fig. 7C, panel 1) coincident with fluorescence of RFP from the vector (pMDC43) localised to the cytoplasm (Fig. 7C, panel 2). For AcLIS/NES-GFP the GFP fluorescence signal was localised to the chloroplasts, suggesting that the primary function of AcLIS/NES in planta would be as a monoterpene (linalool) synthase utilising the chloroplastic pool of GDP (Fig. 7D) consistent with the in planta transient expression data (Fig. 5). The merged GFP signal from AcGES-GFP and chloroplast autofluorescence indicated that the monoterpene synthase AcGES was localised to the chloroplast (Fig. 7E). No GFP or RFP fluorescence was detected in untransfected protoplasts.

Defence-related terpene emissions from plants are commonly mediated by JA-dependent signalling pathways and those of other phytohormones (Ponzio et al. 2013). To investigate this process during vegetative development in kiwifruit leaves, tissue-cultured plants of two commercially available kiwifruit cultivars ‘Hayward’ and ‘Hort16A’ were treated with MeJA. Considerable knowledge on the susceptibility of both cultivars to pathogens and insects is available. Both cultivars are routinely propagated commercially in tissue culture, providing access to plants for replicated analyses. Terpene accumulation inside the leaf tissue, as well as the emission/release of monoterpenes into the headspace, were both measured by GC-MS. High levels of linalool (and derivative oxides) and lower levels of other terpenes such as β-ocimene, α-terpineol and geraniol (Fig. 8A, Table S6A) were present in the tissue of both control and treated leaves. In ‘Hort16A’, significantly higher levels of linalool, geraniol and β-ocimene accumulated after MeJA treatment compared to control leaves, while the levels were unchanged in ‘Hayward’. Different patterns of volatile release were also observed in the headspace samples (Fig. 8B, Table S6B). In ‘Hort16A’, β-ocimene and β-caryophyllene showed increased emission. Linalool and very high levels of DMNT were released after MeJA treatment in both cultivars. DMNT is derived from nerolidol by the action of a P450 enzyme (Lee et al. 2010).

The expression of TPS family members upon MeJA treatment was also investigated in leaves to see if there were any correlations with terpene accumulation and/or release (Fig. 8C). Transcript levels for eight AcTPS were up-regulated by MeJA treatment in one or both cultivars including transcripts for four sesquiterpene synthesis genes: AcNES1, -2, AcAFS1 and AcBCS, that produce (E)-nerolidol, α-farnesene and β-caryophyllene respectively. α-farnesene was not detected at significantly higher levels in the treated samples, but the increased AcNES1, -2 expression correlated well with the release of DMNT after MeJA treatmentwhile AcNES3 was more lowly expressed. Increased β-caryophyllene also correlated with the increased AcBCS expression in ‘Hort16A’ tissue. For the monoterpene synthases, AcLIS/NES (linalool) expression was up-regulated in both cultivars upon MeJA treatment and correlated with the increased linalool accumulation and release, while AcLS1 was only detected at low expression levels in ‘Hort16A’. In contrast AcGES expression was only significantly elevated in ‘Hort16A’ and correlated with increased geraniol accumulation in this cultivar.

To test the effect of defense-related phytohormones on terpene production and AcTPS gene expression in young fruit, 45 d old kiwifruit berries of ‘Red5’ were subjected to methyl jasmonate (MeJA), salicylic acid (SA) and a combined treatment (Figure S5). As shown in Figure S5, exogenous application of SA induced visible damage to the berry surface. With MeJA treatment no obvious visible phenotype was observed. When the berries were treated with the mixed solution of MeJA and SA, the SA-related damage to the surface of the fruit appeared to be largely alleviated. These results indicate that MeJA can effectively antagonize the damage induced by SA.

GC-MS results showed that MeJA-treated kiwifruit berries presented a significantly different terpene profile compared to buffer-treated controls. The terpene profiles of SA and SA + MeJA treated fruit were similar to the controls (Fig. 8D). The production of monoterpenes α-pinene, 1,8-cineole, linalool and geraniol were all significantly increased in the MeJA-treated berries. The sesquiterpenes β-caryophyllene and α-cubebene were also elevated. As in tissue-cultured leaves, (E)-nerolidol did not accumulate (< 1 ng∙g−1) in the tissue. These results indicate that MeJA treatment can activate terpene production in kiwifruit berries. In addition, transcript accumulation of multiple AcTPS genes were up-regulated by MeJA treatment, with six of them representing sesquiterpene synthesis genes: (E)-nerolidol (AcNES1, AcNES2, AcNES3, AcNES/LIS,), β-caryophyllene (AcBCS), α-farnesene (AcAFS) and germacrene synthase (AcGDS) (Fig. 8E). A subset of monoterpene biosynthesis related genes: AcLS1, AcTPS1a, AcTPS1b, were also induced after MeJA treatment (Fig. 8E). Of all the AcTPS genes tested, only AcNES3 showed down-regulation in the MeJA-treated Red5 berries (Fig. 8E). Overall, exogenous application of MeJA on young fruit increased biosynthesis of a broad range of mono- and sesquiterpenes and in parallel induced expression of several AcTPS genes, while SA was able to antagonize terpene induction.

MeJA treatments of leaves and fruit indicated that terpenes and TPS gene transcripts could be induced via the JA-dependent signalling pathway. To establish if this observation could be extended to induction by economically important pests or pathogens of kiwifruit, experiments were undertaken to test the response of tissue-cultured leaves to herbivory by brown-headed leaf roller (BHLR, Ctenopseustis obliquana), a primary polyphagous pest of horticultural crops in New Zealand. Terpene accumulation inside the leaf tissue, as well as the emission/release of monoterpenes into the headspace, were both measured by GC-MS.

For both commercial cultivars tested (‘Hayward’ and ‘Hort16A’), linalool was again the predominant terpene accumulating in tissue. Slightly increased amounts of linalool accumulated in the herbivory samples of both cultivars versus controls (not significant at P <  0.05) and > 50-fold smaller amounts of other monoterpenes such as α-terpineol and geraniol were present (Fig. 9A, Table S6C). When terpene emission was investigated, only very low levels of any terpenes were observed in healthy control plants (< 1 ng∙g−1 FW), and no terpenes were detected in the insect-only samples (Fig. 9B, Table S6D). However, upon insect herbivory, emission of linalool, (E)-nerolidol, and (Z)-nerolidol were significantly induced in insect-damaged plants compared to controls (Fig. 9B). DMNT was also consistently emitted from the herbivory samples (data not quantified) but at much lower levels (< 1 ng∙g−1 FW) compared to the MeJA treated samples and with no DMNT detected in control plants.

Similar to the MeJA treatments, herbivory significantly induced expression of some AcTPS genes (Fig. 9C). Consistent with increased (E)-nerolidol and (Z)-nerolidol release in both cultivars, AcNES1, -2, -3 gene expression was also induced significantly in both cultivars upon herbivory. The increased linalool accumulation and emission upon herbivory correlated with increased TPS expression by the AcLIS/NES gene, whereas AcGES induction in both cultivars and AcBCS induction in ‘Hayward’ did not result in any significant increases in geraniol and/or β-caryophyllene production upon herbivory.

A comparison of active TPS family members in kiwifruit with other horticultural crop plants such as tomato which contains 29 TPS (Zhou and Pichersky 2020), grape - 69 TPS (Martin et al. 2010) and apple - 10 TPS (Nieuwenhuizen et al. 2013) shows that the Red5 kiwifruit genome contains only a small-sized volatile TPS gene family. The present study identified 22 TPS gene models and a total of 15 predicted full-length and functional AcTPS genes, dominated by members of the TPS-b and -g subfamilies (Fig. 2). A single cluster of TPS-b gene models (AcTPS16-20, -22) assigned to chromosome 29 represents > 25% of TPS gene models identified in kiwifruit. Other TPS genes were present as homeologous pairs likely resulting from the ancient genome duplication/polyploidisation events in the Actinidia genome described by Wang et al. (2018).

The Actinidia genus has a reticulate polyploidy structure with diploids, tetraploids and hexaploids in diminishing frequency. All species have a haploid chromosome number of n = 29. The sequenced A. chinensis genomes from Red5 and ‘Hongyang’ and the sequenced A. eriantha genome from ‘Midao 31’ (Wang et al. 2023) are derived from diploid cultivars. Evidence for TPS tandem gene duplication and neofunctionalisation in the Red5 genome was seen at the AcTPS1 locus on chromosome 29 where four closely related TPS genes produced different terpene products. Evidence for gene duplication and loss of function was suggested by the presence of TPS pseudogenes in the Red5 genome that appeared to be nonfunctional paralogues of neighbouring genes (Fig. 3). Orthologs of TPS genes found in other Actinidia species and cultivars were present in the Red5 genome and shared high sequence identity (> 90%). The terpene products produced by the five novel Red5 AcTPS genes described in this paper, the four previously characterised Red5 AcTPS1a-d genes, plus four TPS genes (AaLS1, AcNES1, AdAFS1, AcGDS) functionally characterised in other Actinidia species and cultivars can account for all the major volatile terpenes in Red5 fruit, flowers as well as vegetative tissues. Knowledge of Red5 TPS gene variation should allow identification of further allelic diversity within the A. chinensis gene space and functional homologs of Red5 TPS genes across the > 60 species in the Actinidia genus. However, the number of TPS genes and alleles present in kiwifruit cultivars with higher ploidy (e.g. hexaploid ‘Hayward’, tetraploid ‘Hortgem Tahi’) will be greater and likely more complex to study than in Red5, but their neofunctionalisation may serve as a source of novel genes allowing the plants to respond to different insects and pathogens.

Volatile analysis showed that terpene abundance and composition in kiwifruit is subject to spatial and temporal regulation, with big differences observed between different organs and tissues and during fruit development. This regulation is also reflected in the spatial and temporal gene expression patterns observed for each of the AcTPS-R5 members present in the genome. For example, high levels of gene expression were observed for five AcTPS-R5 genes in the peel and flesh of unripe fruit, often peaking at 75-90 days post anthesis. Other family members showed quite specific expression in leaves (AcTPS1d), roots (AcTPS15) or buds (AcTPS1c). The in planta function of terpene synthases is driven by their access to GDP from the MEP pathway in the plastids versus FDP derived from the mevalonate pathway in the cytosol/ER/peroxisome. Subcellular location of the AcTPS enzymes was used to verify which substrate pool was available to each TPS enzyme (Fig. 7). This analysis was particularly helpful in clarifying the function of AcLIS/NES as a linalool synthase in planta, despite the enzyme producing nerolidol more efficiently in kinetic studies.

One role of terpenes in kiwifruit biology can be found in the flowers, where they form part of the floral bouquet that helps to attract pollinators (Fig. 10). Several TPS have been identified previously in Actinidia that are highly expressed in floral organs and produce specific blends of terpenes. For example AdAFS1 and AdGDS1 in ‘Hayward’ produce α-farnesene and germacrene D respectively (Nieuwenhuizen et al. 2009) whilst AcNES1 and AaLS1 produce (E)-nerolidol and linalool respectively in ‘Hort16A’ and ‘Hortgem Tahi’ (Chen et al. 2010; Green et al. 2012). In Red5, (E)-nerolidol and linalool are abundant volatiles produced in flowers, but interestingly the TPS gene correlated with linalool production in Red5 flowers (AcLIS/NES) is different to LS1 that produces floral linalool in at least two other Actinidia species - A. arguta and A. polygama (Chen et al. 2010).

A second role of terpenes in kiwifruit biology is the attraction of frugivores (seed dispersal animals) by contributing to the taste and aroma blend of ripe fruit (Fig. 10). The TPS1 locus on chromosome 29 seems to fulfil a particularly dominant role in this function with several genes and alleles producing different terpenes in ripe fruit having now been characterised from multiple species (Davidson et al. 2023; Nieuwenhuizen et al. 2015; Wang et al. 2021; Zeng et al. 2020). The diversity of compounds produced from this single locus may serve as a source of novelty to breed kiwifruit with new aromas and flavours.

The third role for terpenes in kiwifruit examined in this paper is in plant defence (Fig. 10). From the MeJA treatment experiments it is apparent that jasmonate signalling forms an important part of the defence arsenal in kiwifruit. Jasmonate is known to be very rapidly produced upon herbivory (Kessler and Baldwin 2002) and we show that a subset of terpene synthase genes is induced by MeJA treatment of leaves and young fruit (Fig. 8). Interestingly DMNT emission was much higher in MeJA treated tissue-cultured plants, while its (E)-nerolidol precursor was much lower, compared to BHLR infested plants where the inverse was observed. This observation points to other factors being important too in the biotic interactions, such as actual physical leaf damage and insect saliva factors that may suppress or modify plant defences.

When considering all the TPS expression profiles and functional data together, it becomes apparent that roles in pollinator attraction, seed dispersal and plant defence are not mutually exclusive and a subset of terpene synthases likely fulfil overlapping functions (Fig. 10). For example, nerolidol synthase (NES1) and α-farnesene synthase (AFS1), first identified as important contributors to the floral aroma in ‘Hort16A’ and ‘Hayward’, respectively are also highly expressed in Red5 cane and young fruit (Fig. 4), as well as being induced by MeJA and/or BHLR infestation. Similarly, while the TPS1 genes are part of the complex ripe fruit aroma locus (Zeng et al. 2020), AcTPSd also shows high expression in leaves (Fig. 4). Several examples of functional redundancy/convergence were also observed. Three nerolidol synthases that are part of distinct TPS subfamily lineages were all induced by herbivory (NES1-3), while two linalool synthases from different TPS subfamilies were associated with floral roles (LS1, LIS/NES). The comprehensive TPS gene identification, expression and functional analysis presented in this paper allows the wider picture of overlapping roles and functional redundancy to be observed in Actinidia for the first time.

Apart from basal levels of defence, induced defences are an important mechanism for plants to minimise the cost of defence, while maximising the return. In kiwifruit, until now, little was known about defence-related terpenes and their synthases. In vegetative tissues, we observed several terpenes and their synthases to be induced upon herbivory by the economically relevant pest BHLR. This included linalool, (E)- and (Z)-nerolidol and derivatives such as linalool oxides, and the homoterpene DMNT and the concomitant induction of AcTPS genes such as AcLIS/NES, and several AcNES genes. AcNES1 has previously been shown to produce both nerolidol isomers in vitro but ‘Hort16A’ flower extracts contained only the (E)-isomer (Green et al. 2012).

By comparing ‘in tissue’ accumulation versus headspace emission, unique distributions of volatiles were uncovered. For example linalool was identified to both accumulate and to be released into the headspace, while (E)- and (Z)-nerolidol and DMNT were exclusively found in the headspace. This indicates that these latter compounds are likely produced mainly de novo upon herbivory and only at the site of feeding by up-regulation of TPS expression. Linalool on the other hand is already present at significant levels pre-herbivory and appears to be further induced by up-regulation of linalool TPS expression.

Fleshy fruit have evolved from their dry fruit ancestors and the fleshy parts serve functions in protecting, nurturing and dispersing of seeds. Fleshy fruit types are thought to have co-evolved with endozoochory (seed dispersal through ingestion by animals) and are often associated with shaded habitats (Lorts et al. 2008). The aroma of ripe kiwifruit may have evolved as an adaptation to attract mammals such as bears, monkeys and martens to consume the fruits and consequently disperse their seeds (Naoe et al. 2019; Nevo et al. 2015). These dispersal functions of terpenes were preceded by pollination attraction in most flowering insect-pollinated plants and by defence against pest and diseases in all plants. Identifying the full set of Red5 TPS genes and their overlapping functions provides a framework to better understand the volatile terpene language within the Actinidia genus and in other horticultural crops and how the multiple functions of terpene synthases have evolved to support the various biological and ecological roles. Armed with this knowledge we may be able to target specific terpenes in breeding to simultaneously improve kiwifruit flavour/aroma and resistance to pests.

Red5 fruit for analysis through development were obtained from Plant & Food Research (PFR) orchards in Motueka, New Zealand. A detailed description of these samples including the size, weight, dry matter content, and a representative photograph is given in Zeng et al. (2020). Additional trays of young Red5 fruit were harvested at 45 d post anthesis from PFR orchards in Motueka for hormone treatment. Other Red5 tissues were collected from plants grown under ambient conditions in PFR glasshouses in Auckland, New Zealand. Tissue-cultured plants ~ 3-5 cm tall of ‘Hort16A’ and ‘Hayward’ were provided by Multiflora Laboratories (Auckland, New Zealand). N. benthamiana plants were grown in a PFR glasshouse at 25 °C with a 16 h light/8 h dark photoperiod. Brown-headed leaf rollers (BHLR, 3rd/4th instar) (Ctenopseustis obliquana) were raised by as described in Barrington et al. 1993.

Frozen kiwifruit tissues from developing fruit (2-3 g), flower (3-4 g), leaves (1-2 g), bud (1-2 g) and tobacco leaves (1-2 g) were ground in triplicate with liquid nitrogen and 30% (w/w) NaCl was added. Volatile analysis of the tissues by SPME GC-MS was conducted according to Zeng et al. (2020) with the minor modifications. Volatiles were collected at 40 °C for 10 min with agitation, controlled by a multipurpose sampler injection system (Gerstel Mülheim, Germany). SPME fibres (1 cm) coated with 50/30 μm DVB/CAR/PDMS were used for the volatile collection. Separation was effected using a 30 m × 0.25 mm internal diameter × 0.25 μm film thickness DB-WAX UI (Agilent, Santa Clara, CA) capillary GC column in an Agilent7890 GC coupled to a Leco BT time-of-flight mass spectrometer (Leco Corp., St Josephs, MI). Terpenes were semi-quantified using the single m/z 93 ion and identified by comparison with the National Institute of Standards and Technology (NIST) database (version V2.3, 2017; in-house mass spectral libraries and confirmed by comparison of retention indices with those of authentic standards and literature values. The sample peak areas were converted into ng∙g−1 fresh weight (FW) by comparison with the internal standard cyclohexanone (2.025 μg per sample) or hexadecane standard added in each sample. All the treatments consisted of at least three replicates.

For dynamic headspace analysis, the emitted volatiles were trapped onto Tenax TA matrix (100 mg of Tenax TA 80-100 mesh, Sigma-Aldrich, St. Louis, MO) (See M&M sections ‘MeJA treatment’ and ‘Herbivore treatment’ respectively for details) and samples were eluted with 1 mL of pentane/ether(1:1 v/v). Afterwards, 50 µg of cyclohexanone was added as the internal standard and the eluted volatile solvent extract was concentrated tenfold under a stream of N2 gas. Extracts (1 µL) were loaded onto the GC-MS and analysed as described in Zeng et al. 2020. Terpene peaks were semi-quantified using single m/z 93 ion according to the method described in the above paragraph.

TPS sequences were identified using BLASTP against the manually annotated A. chinensis var. chinensis Red5 genome available from ftp://ftp.ncbi.nlm.nih.gov/genomes/all/GCA/ 003/024/255/GCA_003024255.1_ Red5_PS1_1.69.0. Sequences were manually curated, aligned using ClustalW, and analysed for known conserved motifs and features using the Geneious software (v. R10). Phylogenetic trees were constructed using MEGA program v7.0.14.

Total RNA was isolated using the Spectrum plant total RNA extraction Kit (Sigma-Aldrich, St. Louis, MO). The RNA was qualified by NanoDrop 1000 UV spectrometry and the quality verified by agarose gel electrophoresis. 10 U DNase I (Roche Applied Science, Mannheim, Germany) was applied to each sample to remove the genomic DNA, and for each sample cDNA was made from 1 µg total RNA using a QuantiTect Reverse Transcription kit (QIAGEN, Hilden, Germany), according to the manufacturer’s protocol. The cDNA used for qRT-PCR was diluted 50-fold prior to amplification. qRT-PCR gene expression analysis for TPS genes was performed alongside the combined reference genes EF1α (Nieuwenhuizen et al. 2009) and Actin (Walton et al. 2009) on a LightCycler 480 platform using the LightCycler 480 SYBR Green master mix (Roche Applied Science, Mannheim, Germany) in a reaction volume of 5 µL. Results were analysed using the LightCycler 480 software (Roche Applied Science, Mannheim, Germany). The amplification program was: 5 min at 96 °C; 40 cycles of 10 s at 95 °C, 10 s at 60 °C, and 20 s at 72 °C; followed by melting curve analysis: 95 °C 5 s, 65 °C 60 s, then ramping at 0.18 °C∙s−1 to 95 °C. The results were analysed using the ΔΔCt method with primer amplification efficiency corrections and using the geometrical mean of the two reference genes (Vandesompele et al. 2002). Primer sequences and efficiencies (determined by the serial dilution method) are listed in Table S7. Three biological and four technical replicates were conducted in each real-time PCR analysis.

Full-length clones of AcBCS, AcNES2 and AcNES3 were amplified from Red5 cDNA (75 + 150 days peel sample) using primers to the full-length ORFs predicted by the gene models for these genes given in Table 1. The 3’ primer for AcGES was obtained at the C-terminus of the gene model Acc26061. The 5’ primer was designed after extending the predicted ORF by 105 aa at the N-terminus using publicly available RNAseq data (Figure S1, http://kiwifruitgenome.org/jbrowse/). N- and C-terminal primers for AcLIS/NES primers were designed to the truncated gene models of Acc19057 and Acc19058 combined (Figure S1). For AcGES and AcLIS/NES, cDNA from Red5 peel (60 d) were used as template.

Full-length AcTPS ORFs were recombined into the Gateway sequencing vector pDONR221 (Invitrogen) using the primers listed in Table S7. After full-length sequence confirmation, genes were cloned into the pHEX2 binary vector (Hellens et al. 2005) by Gateway LR reactions to generate the over-expression constructs pHEX2_AcTPS (CaMV 35S:ORF:ocs-3’). Transient overexpression of AcTPS genes in N. benthamiana leaves was performed as previously described (Zeng et al. 2020), but using a final concentration of bacteria of OD600 = 0.5. Leaves infiltrated with a pHEX2_GUS construct were used as negative controls (Nieuwenhuizen et al. 2009). pHEX2_AcDXS (Nieuwenhuizen et al. 2015) or pEAQ-tHMGR-2A-BCCP1 (Lee et al. 2019) were co-infiltrated with each TPS construct and pHEX2_GUS to increase substrate production for synthesis of terpenoids. N. benthamiana leaves (1-2 g) were harvested at 7 d post infiltration, ground in liquid nitrogen, and stored at -80 °C. Terpene volatiles were collected by SPME and analysed by GC-MS. All the experiments were conducted with at least three to six biological replicates.

The complete ORF of AcNES2, AcNES3, AcBCS and AcGES, including any potential ChloroP-predicted plastid targeting signal peptides (Emanuelsson et al. 2000), were amplified by PCR and cloned into the pET300 vector (N-terminal HIS-tag, Invitrogen). AcLIS/NES was cloned into the pMAL-c6T vector (NEB) to express Mal-BP-TPS recombinant protein. Primers used for PCR amplification are shown in Table S7. Protein expression and purification was done as described previously (Zeng et al. 2020). The purity of the recombinant proteins was confirmed by SDS-PAGE (Figure S3) and protein concentration was calculated by NanoDrop 2000. Samples (2.5 mL volume) were desalted using a PD10 column (GE-Pharmacia) eluted in 3.5 mL buffer (containing 20 mM HEPES, 150 mM NaCl, 1 mM DTT, pH 7.5). Samples were stored in 10% (v/v) glycerol in small aliquots at -80 °C until further use.

TPS enzyme assays were performed as described previously (Zeng et al. 2020). In short, purified recombinant enzyme (2-50 μg protein) was added to 2 mL of an optimised TPS assay buffer (50 mM Bis-Tris Propane pH 7.5, 10 mM MgCl2, 10% glycerol (v/v) and 5 mM DTT). Samples were incubated (30 °C for 1 h, whilst shaking) using 50 μM GDP or FDP as substrates.

Kinetic parameters Vmax, Km and Kcat of each AcTPS were calculated in the presence of GDP and FDP by non-linear least square fitting of the data to the Michaelis-Menten equation using the GraphPad prism 8.0 software (San Diego, CA, U.S.A). Kinetic parameters for each TPS were calculated using the major product peak. Reactions were initiated by adding 1 µg of affinity-purified HIS-tagged or MAL-MBP-tagged enzyme in a final volume of 1 mL of optimised assay buffer (pH 7.5) and varying concentrations (0.5 μM-50 μM) of substrates (FDP or GDP). The assays were incubated at 30 °C for 1 h, then samples were snap frozen at -80 °C until analysis. For both TPS enzyme assays and kinetic analysis, headspace terpenes were collected by SPME and analysed by GC-MS. Hexadecane was used as the internal standard.

Complete AcTPS ORFs were amplified without a stop codon (using primers in Table S7) and inserted into the pMDC83-GFP vector (Curtis and Grossniklaus 2003) by Gateway cloning. Translational fusions were infiltrated into N. benthamiana leaves and protoplasts produced as described in Zeng et al. (2020). Fluorescence images were acquired on a confocal laser-scanning microscope (TCS SP8; Leica, Wetzlar, Germany) using previously described excitation wavelengths and detection windows (Zeng et al. 2020). All fluorescence experiments were repeated independently at least three times. AcTPS-GFP signals were merged with chlorophyll autofluorescence signals or red fluorescence signals from pMDC43-RFP to verify chloroplastic and cytoplasmic subcellular localisation respectively.

Tissue-cultured kiwifruit plants ‘Hort16A’/’Hayward’ were grown in ½ MS medium + 0.8% agar at 21 °C with a 14 h photoperiod. When 4-8 cm tall, the plants were transferred to 1 L headspace sampling jars containing ½ MS + agar 0.8% (w/v). Leaves were lightly sprayed with 200 µM solution of MeJA or control solution (0.125% Triton X) and allowed to air dry. Dried air was provided to the jar at a constant flow rate of 30 mL∙min−1 for 6 d and the exit air was sampled continuously using a Tenax TA 80/100 mesh (100 mg) cartridge. At completion, terpenes were eluted with 1 mL pentane:ether (1:1) and analysed by GC-MS as described above in the M&M section ‘Tissue and headspace volatile analysis by GC-MS’. At completion of the sampling, leaf tissues were also collected for SPME GC-MS tissue analysis as described above and for qRT-PCR gene expression analysis.

Twenty young Red5 fruit (45 d after flowering) were either dipped for 10 min in 200 µM MeJA solution (including 0.125% Triton X-100), 10 mM salicylic acid or a combination of both. Control fruit were dipped into Triton solution (0.125% Triton X-100 only) and fruit were allowed to air dry. Twenty-four hours post-treatment, fruit were photographed and peel tissue of treated and control samples was snap frozen immediately in liquid nitrogen and stored at − 80 °C for further qRT PCR gene expression and tissue volatile accumulation analysis (SPME GC-MS).

Four tissue-cultured kiwifruit plants ‘Hort16A’/’Hayward’ (4-8 cm as described in ‘MeJA treatment’) per 1 L pot were infested with 10 BHLR (3rd or 4th instar). Dried air was provided to the headspace sampling jars as describe for MeJA treatment and samples collected on Tenax TA (100 mg) cartridges for 6 d. Terpenes were eluted with 1 mL pentane:ether (1:1) and analysed by GC-MS as described above. At completion of the sampling, tissues were also collected for SPME GC-MS analysis and qRT-PCR as described above. Details of the experimental set-up are shown in Figure S4.

Local, national and international guidelines and legislation were followed and appropriate permissions and/or licences for this study were obtained.

Not applicable.

The authors declare that they have no competing interests.