The proanthocyanin-related transcription factors MYBC1 and WRKY44 regulate branch points in the kiwifruit anthocyanin pathway

The groups of plant flavonoid metabolites termed anthocyanins and proanthocyanins (PA) are responsible for pigmentation in seeds, flowers and fruits. Anthocyanins and PAs are produced by a pathway of enzymes which are transcriptionally regulated by transcription factors (TFs) that form the MYB-bHLH-WD40 (MBW) complex. In this study, transcriptomic analysis of purple-pigmented kiwifruit skin and flesh tissues identified MYBC1, from subgroup 5 of the R2R3 MYB family, and WRKY44 (highly similar to Arabidopsis TTG2) as candidate activators of the anthocyanin pathway. Transient over-expression of MYBC1 and WRKY44 induced anthocyanin accumulation in tobacco leaves. Dual luciferase promoter activation assays revealed that both MYBC1 and WRKY44 were able to strongly activate the promoters of the kiwifruit F3′H and F3′5′H genes. These enzymes are branch points of the pathway which specifies the type of anthocyanin accumulated. Stable over-expression of MYBC1 and WRKY44 in kiwifruit calli activated the expression of F3′5′H and PA-related biosynthetic genes as well as increasing levels of PAs. These results suggest that while previously characterised anthocyanin activator MYBs regulate the overall anthocyanin biosynthesis pathway, the PA-related TFs, MYBC1 and WRKY44, more specifically regulate key branch points. This adds a layer of regulatory control that potentially balances anthocyanin and PA levels.


Identification of differentially expressed genes (DEGs) involved in flavonoid biosynthesis pathway. Skin tissue of MaMe
Red was compared to that of MaMe Yellow at mature green stage, colour stage, and ripe stage to identify DEGs that may be involved in the flavonoid and anthocyanin pathways. A total of 152 genes (Supplementary Table 2) were differentially expressed more than 2 Log 2 fold-change (Fig. 1A) between MaMe Red and Yellow, when all three maturity stages were compared. When the same comparisons were made for fruit flesh, there were 141 DEGs (Supplementary Table 3) that showed differential expression greater than 2 Log 2 fold-change between the flesh of MaMe Red and MaMe Yellow (Fig. 1B). When overlapping the DEGs that were common between skin and flesh and present in all three maturity stages, there were 49 genes predicted to encode for biosynthetic enzymes that are potentially involved in the flavonoid and anthocyanin biosynthesis (Table 1). In addition, a shortlist of 9 gene models (skin) and 27 gene models (flesh) encoding TFs from the MYB, bHLH, homeobox and NAC families were potential flavonoid biosynthesis regulators ( Table 2). The reported bHLH partner of the anthocyanin activating MYBs, termed either bHLH5 or bHLH42 (Acc19563.1) 42,45 , was not differentially expressed at higher than the 2 Log 2 fold-change cut-off. To validate these observations against an independent genetic background, a comparison of fruit skin and flesh of the related species A. purpurea was made. Comparison between skin and flesh maturity time points revealed 203 DEGs consistently changing during ripening in skin ( Fig Table 5). Forty gene models and 26 gene models identified in the skin and flesh comparisons of A. purpurea, respectively, were predicted to encode for biosynthetic enzymes in the flavonoid and anthocyanin pathways (Table 3). In the same comparisons, 24 TFs and 94 TFs were identified from the MYB, bHLH, homeobox and NAC families that may be involved in the flavonoid and anthocyanin biosynthesis in skin and flesh of A. purpurea, respectively (Supplementary Table 6).
Candidate TFs MYBC1 and WRKY44 identified from DEGs. The TF MYB110 (Acc10232.1) was identified in all DEG lists confirming previous studies which have identified its role in elevating anthocyanin biosynthesis 16,19,46 . MYB10 (Acc00493.1) appears to have a more important role in the kiwifruit species, A. chinensis 17 , and was identified as a DEG in flesh samples. The MYB TF MYBC1 (Acc12965.1) 43 and the WRKY TF termed WRKY44 (Acc16887.1) appeared in all the DEG lists of comparisons between MaMe Red and MaMe Yellow. The expression of MYBC1 exhibited high fold-changes during the colour change and ripe stage compared to the mature green stage in both the skin and flesh (Table 2). MYBC1 was highly expressed during the colour change and ripe stage in the purple coloured A. purpurea and MaMe Red, but barely detected in MaMe Yellow (Fig. 3). Similarly, WRKY44 was highly expressed in the A. purpurea and MaMe Red during the colour change www.nature.com/scientificreports/ and ripe stage but very lowly expressed in the anthocyanin lacking MaMe Yellow. The expression of WRKY44 exhibited high fold change increases in both the skin and flesh during colour development ( Table 2). The expression of MYBC1 and WRKY44 was significantly higher in coloured fruit, as was MYB110, correlating with anthocyanin accumulation. There was some reduction in expression in ripe fruit. The expression of other subgroup 5 MYBs was very low (< 20 RPKM) and the expression patterns did not correlate with the expression of MYB110 or with anthocyanin accumulation (Supplementary Fig. 2). MYBC1 (closest A. chinensis gene model, Acc12965.1) has been implicated in the control of A. arguta anthocyanin levels 43 . Another TT2-like MYB termed AcMYB123 (closest gene model Acc28234.1) was proposed to regulate the anthocyanin accumulation in the inner pericarp of the red-centred A. chinensis 42 . In our RNA-seq data, the expression of AcMYB123 (Acc28234.1) and  Fig. 2). Also, the low expression of other WRKY44-like gene models did not indicate any correlation with anthocyanin accumulation (Supplementary Fig. 3). Therefore, MYBC1 (Acc12965.1) and WRKY44 (Acc16887.1) from purple kiwifruit A. melanandra, A. purpurea, and MaMe Red were selected as candidate genes for functional characterisation.
MYBC1 is a subgroup 5 R2R3 MYB and WRKY44 belongs to group I of the WRKY family. Phylogenetic analysis of MYBC1 from A. melanandra (purple-skinned and purple-fleshed), A. purpurea, and MaMe Red suggested a close relationship with TT2 TFs from a variety of plant species (Fig. 4A). Using R2R3 MYB TFs from the anthocyanin-related subgroup 6 and subgroup 5 in the phylogenetic tree revealed that kiwifruit MYBC1 clustered with Gossypium TT2 and Camellia sinensis MYB5a, which belong to subgroup 5 of the R2R3 MYB TF  Fig. 4). Although grouped with the PA regulators from subgroup 5, the C1 motif was not observed in the kiwifruit MYBC1 sequences, whereas some amino acids in the C3 motif and the VIRTKAx[K/R]C motif common in the PA regulators were observed in kiwifruit MYBC1 deduced amino acid sequences 31,34 . Similar to previously identified AcMYB123 (Acc28134.1), kiwifruit MYBC1 clustered with the TT2 clade and belongs to the subgroup 5 of the R2R3 MYB TF family in plants.
Phylogenetic analysis of WRKY44 from A. melanandra and A. purpurea revealed similarity with WRKY44/ TTG2 TFs from a range of plant species (Fig. 4B). Kiwifruit WRKY44 grouped into the TTG2 clade and close to the Camellia sinensis WRKY44 which is a regulator of PA biosynthesis 36 . Kiwifruit WRKY44 possesses two WRKY domains composed of the conserved amino acids WRKYGQK toward the N-terminal as well as a zincfinger ligand and a potential nuclear localisation sequence ( Supplementary Fig. 5) 47 . These results suggested that the kiwifruit WRKY44 belongs to the group I of the WRKY superfamily in plants.

MYBC1 and WRKY44 induced anthocyanin patches on the leaves of Nicotiana tabacum.
Transient over-expression of kiwifruit MYBC1 and WRKY44 in the leaves of N. tabacum induced anthocyanin accumulation (Fig. 5A). However, anthocyanin accumulated unevenly around the infiltration sites, as opposed to the complete saturation of anthocyanin at the infiltration site using the positive control, A. purpurea MYB110. The formation of the anthocyanin patches was variable between leaves. In some cases, only small points of anthocyanin accumulation formed near the infiltrated sites. The majority of anthocyanins induced by transient over-expression of MYBC1 are cyanidin-based and the remainder are delphinidin-based anthocyanins, whereas only cyanidin-based anthocyanin was observed from the transient over-expression of WRKY44 (Fig. 5B). These results suggest that both MYBC1 and WRKY44 from the purple kiwifruit species are able to induce anthocyanin accumulation when transiently over-expressed in tobacco leaves. The patchy and variable anthocyanin formation around the infiltration sites may indicate that MYBC1 and WRKY44 may require additional TF partners to fully activate the anthocyanin pathway.

MYB and WRKY binding motifs annotated on kiwifruit F3′H and F3′5′H promoters. Potential
binding motifs for the TFs from MYB, bHLH and WRKY families were predicted to be present on the F3′H and F3′5′H promoters previously cloned from A. melanandra, A. purpurea, MaMe Red and MaMe Yellow by the online tool PlantPAN3.0 19,48 . Further screening of TF binding sites and motifs, by only including HIT score of 1, revealed predicted binding sites for MYB, bHLH, and WRKY TFs that are related to the phenylpropanoid pathway and flavonoid biosynthesis (Fig. 6). The common TF binding motifs found on all F3′H promoters were for bHLH TF, AtMYC2 and a WRKY TF, AtTTG2 (Fig. 6A). In addition, within the A. melanandra and A. purpurea F3′H promoters, a binding site for another bHLH TF, AtPIF3, and a binding site for MYB (MYB, phenyl) that binds to the promoters of phenylpropanoid biosynthetic genes were identified. All four F3′5′H promoters harboured the binding motifs for the bHLH TF, AtMYC2, the WRKY TF, AtTTG2, the MYB (MYB, PZM) which is the core consensus for the anthocyanin-related P gene in maize, and the MYB-binding motif found in the promoters of phenylpropanoid biosynthetic genes (Fig. 6B). The presence of these predicted binding sites within the F3′H and F3′5′H promoters suggests regulation by flavonoid-related TFs. MYBC1 and WRKY44 activated kiwifruit F3′H and F3′5′H promoters. Dual luciferase promoter activation assays in the leaves of N. benthamiana showed that MYBC1 and WRKY44 significantly activated the F3′H promoters from all four species, including MaMe Yellow (Fig. 7A). There were high endogenous promoter readings, again suggesting activation by endogenous tobacco TFs. Previously it has been shown that MYB110 www.nature.com/scientificreports/ only activated the F3′H promoters from A. melanandra and A. purpurea, but not the MaMe kiwifruits 19 . The results here suggested that MYBC1 and WRKY44 were able to activate all F3′H promoters including those cloned from MaMe kiwifruits. However, no activation was observed by MYB110 on the MaMe F3′H promoters with or without MYBC1 and/or WRKY44 co-infiltration. Infiltrations of MYBC1 and WRKY44 significantly activated F3′5′H promoters from all four species, with and without the co-factor bHLH5 (Acc19563.1) (Fig. 7B). As with F3′H promoters, there were high endogenous promoter readings. Consistent with previous findings, MYB110 was unable to activate F3′5′H promoters from all Table 3. Differentially expressed genes (DEGs) encoding biosynthetic enzymes potentially involved in the flavonoid and anthocyanin pathway obtained from the comparison between mature green (MG), colour change (CC), and ripe (RP) stage in Actinidia purpurea skin and flesh. Base means are shown as an indication of expression.

Skin comparison
Acc00260. Over-expression of MYBC1 and WRKY44 in A. arguta increases F3′5′H expression and PA content. Transgenic calli were produced which over-expressed GUS, MYB110, MYBC1, WRKY44 and MYBC1/ WRKY44 co-expression. These were analysed for gene expression and metabolite composition (Fig. 8A). GUS and MYB110 were transformed in A. arguta 'K2D4' as negative and positive controls, respectively. Phenotypically, MYB110 over-expressing calli are intensely purple throughout the tissue due to the high concentration of cyanidin-based anthocyanin, which is the only anthocyanin type detected (Supplementary Fig. 6A). The phe-   www.nature.com/scientificreports/ notype of calli over-expressing GUS showed no distinctive difference to the slightly more pigmented calli overexpressing MYBC1 and WRKY44. However, metabolite analysis revealed that MYBC1 and MYBC1/WRKY44 co-expression calli accumulated significantly more procyanidin B1, B3, B-type procyanidin dimer C, procyanidin C1, C-type procyanidin trimer A and C, and catechin (Fig. 8B) compared to the GUS and MYB110 overexpressing calli (see categories of PA 49 ). WRKY44 over-expressing calli accumulated significantly higher levels of procyanidin B3 and C-type procyanidin trimer A than the control calli but a significant decrease in B-type procyanidin dimer D and C-type procyanidin trimer C (Fig. 8B). The accumulation of epicatechin and catechin were highest in calli co-expressing MYBC1/WRKY44. There were also small amounts of cyanidin-based anthocyanin detected in MYBC1 and MYBC1/WRKY44 calli. Plantlets regenerated from MYB110 over-expressing callus showed purple pigmentation throughout the shoots and roots, whereas the plantlets regenerated from MYBC1 and MYBC1/WRKY44 showed purple pigmentation in the roots only, compared to the green plantlets regenerated from the GUS control ( Supplementary Fig. 7). Gene expression analysis revealed MYB110 over-expressing calli had high expression of the F3GT gene (Acc20132.1) in comparison to other calli ( Supplementary Fig. 6B). Expression levels of DFR, F3′H1 and F3′H2 (Acc01005.1, Acc12813.1 and Acc18331.1) were similar across all transformed calli. Noticeably, the expression of F3′5′H (Acc32390.1) was significantly elevated in the calli over-expressing WRKY44, MYBC1 and a combination of both, compared to the GUS and MYB110 controls. Expression of the PA-related genes, FLS1 and LAR1 was elevated in the WRKY44 calli, whereas FLS2 was elevated in calli over-expressing GUS and MYB110 controls (Fig. 8C, Supplementary Fig. 6B). LAR5 was elevated in calli transformed with MYBC1 and WRKY44. Expression of both ANR1 and ANR2 was elevated in the calli over-expressing WRKY44, MYBC1 and MYBC1/ WRKY44 (Supplementary Fig. 6B).

Discussion
Kiwifruit MYBC1 and WRKY44 are up-regulated during colour development. Anthocyanin accumulation is regulated at the transcriptional levels by genes encoding biosynthetic enzymes and TFs. Previous studies on red and purple kiwifruit species revealed that the key pathway genes CHS, DFR, F3GT and LDOX are responsible for anthocyanin accumulation 19,50,51 . The R2R3 MYB TFs, MYB10 (AcMYB75/AcMYBF110) and MYB110, positively regulate the anthocyanin pathway at the transcriptional level 16,18,19,52 . However, the transcriptional regulation of the cyanidin and delphinidin branch points controlled by the F3′H and F3′5′H has not been fully understood. In this study, transcriptomic analysis of purple kiwifruit at three developmental stages Kiwifruit MYBC1 and WRKY44 are PA-related TFs. Kiwifruit MYBC1 belongs to the subgroup 5 of the MYB TF family and shared amino acid residues found in the highly conserved motif VIRTKAx[K/R]C that is characteristic to the PA-regulating TFs, but the C1 and C3 motifs seen in the anthocyanin modulating VvMYB5 were not present in the c-terminus of kiwifruit MYBC1 (Fig. 4A, Supplementary Fig. 4) 31,34 . Generally, MYB TFs from subgroup 6 contribute to the regulation of the anthocyanin pathway while MYB TFs from subgroup 5 are thought to be involved in PA accumulation. In Arabidopsis, TT2 is responsible for PA accumulation in the endothelium during seed development by regulating flavonoid genes such as DFR and BAN (anthocyanidin reductase) as well as TT8 (bHLH) and TTG1 (WD40) 53 . In grape berries, VvMYB5a and VvMYB5b are capable of inducing anthocyanin and flavonol accumulation when over-expressed in tobacco 31,54 . VvMYB5b was also able to activate the promoters of structural genes involved in the anthocyanin and PA pathways such as VvLAR1, VvANS, VvANR and VvF3′5′H, suggesting a regulatory role in different branches of the phenylpropanoid pathway. Tea CsMYB5a and CsMYB5e and freesia FhMYB5 all demonstrated regulatory roles in the anthocyanin www.nature.com/scientificreports/ and PA biosynthesis pathway 33,34 . Recently, in red-centred kiwifruit A. chinensis 'Hongyang' , a TT2 type R2R3 MYB TF, AcMYB123 and a bHLH TF, AcbHLH42 were identified to be involved in the inner pericarp-specific accumulation of anthocyanins by activating the expression of AcANS and AcF3GT 42 . Although belonging in the same subgroup 5 as the kiwifruit MYBC1 identified in this study, AcMYB123 (Acc28234.1, MH643775.1) only shared 47% sequence identity and the expression was barely detected in sequenced tissues. MYBC1 from A. arguta is highly similar to MYBC1 identified here (98% amino acid sequence identity) and is implicated to be involved in anthocyanin biosynthesis. It is negatively regulated by microRNA858 43 . Expression of other TT2-like MYBs was barely detected in the purple kiwifruit species studied here, making the possibility of a role in regulation unlikely (Supplementary Figs. 2, 4). Kiwifruit WRKY44 (Acc16887.1) belongs to group I of the WRKY superfamily and clusters in the TTG2 clade closely with tea CsWRKY44, which is involved in the catechin regulation (Fig. 4B) 36 . Within the TTG2 clade, Arabidopsis TTG2 encodes for a WRKY TF that participates in trichome formation and tannin production in seed coat endothelium by regulating the vacuolar transport step in PA pathway 35,55 . Moreover, kiwifruit WRKY44 is homologous to PH3 in petunia, which regulates vacuolar acidification for anthocyanin storage 39 . In petunia flower petals, PH3 encodes for a WRKY TF that is highly similar to AtTTG2 and the transcription is

MYBC1 and WRKY44 regulate F3′H and F3′5′H branch points. Functional characterisation of
MYBC1 and WRKY44 revealed that both TFs are able to induce anthocyanin accumulation when transiently over-expressed in tobacco leaves (Fig. 5). The regulatory roles of MYBC1 and WRKY44 were confirmed by the significant activation of all kiwifruit F3′H and F3′5′H promoters, for which the anthocyanin activator, MYB110, showed no regulatory role as previously found (Fig. 7) 19 . This observation was supported by the presence of multiple phenylpropanoid related MYB and WRKY TF binding sites identified on the promoter sequences (Fig. 6). These findings suggest that MYBC1 and WRKY44 transcriptionally regulate anthocyanin biosynthesis by activating the F3′H and F3′5′H branch points, which determines the hydroxylation patterns of the anthocyanin aglycone.
Kiwifruit calli over-expressing MYBC1 and WRKY44 had no obvious visual phenotype but metabolite and gene expression analysis revealed major differences from calli expressing a GUS-control (Fig. 8). The expression of kiwifruit F3′5′H was significantly up-regulated by the over-expression of WRKY44 and MYBC1. Expression of genes encoding for biosynthetic enzymes in the PA pathway such as FLS1, LAR1, LAR5, ANR1 and ANR2 were elevated in the WRKY44 and MYBC1 calli compared to both GUS-control and MYB110 calli. As a result, the amounts of PA accumulated in those calli increased significantly. Correlation analysis suggested linkage between F3′H and F3′5′H and the B-and C-type procyanidin isomers as a result of the over-expression of WRKY44 and MYBC1. As expected, expression of F3GT was significantly elevated by the over-expression of MYB110 in the www.nature.com/scientificreports/ anthocyanin accumulating calli, reiterating its critical role in anthocyanin regulation 50 . Anthocyanin accumulated in the calli over-expressing MYBC1 and MYBC1/WRKY44 at low levels, but with weak correlation data. However, the roots of the plantlets regenerated from the MYBC1 and MYBC1/WRKY44 calli were visibly red, differing from the unpigmented roots of GUS plantlets ( Supplementary Fig. 7). Stable lines expressing these genes, as adult vines, will be important research tools.
MYB110, MYBC1 and WRKY44 are in the regulatory network shared between PA and anthocyanin biosynthesis. Core genes encoding enzymes in the anthocyanin pathway, such as CHS, DFR, F3GT, and LDOX, have been shown to be transcriptionally regulated by the subgroup 6 R2R3 MYB TFs, MYB10 and MYB110 in kiwifruit species [16][17][18][19] . However the F3′H and F3′5′H branch points that decide the production of cyanidin and delphinidin were not principally regulated by MYB10 or MYB110 in purple kiwifruit species 19 . In grape MYBA1 activates F3′5′H promoters whereas two closely related TFs, MYBA6 and MYBA7 from subgroup 6, showed no activation of the F3′5′H promoters 56 . However, VvMYB5a and VvMYB5b from subgroup 5 activated the F3′5′H promoter by 12 fold as well as other genes involved in the PA and anthocyanin pathways 31 . In transgenic purple tomatoes, the control of flavonoid biosynthesis showed a specialised regulatory mechanism where the over-expression of TFs Del/Ros1 activated a broader spectrum of the genes including the F3′5′H in the flavonoid pathway but the anthocyanin-related TFs LC/C1 did not 57,58 . These results suggest different specificities of TFs controlling anthocyanin biosynthesis. Both MYBC1 and WRKY44 could regulate the anthocyanin pathway at the F3′H and F3′5′H branch points via the MBW complex in kiwifruit (Fig. 10). Functional characterisation showed that MYB110, MYBC1 and WRKY44 regulate different points of the pathway. We propose a model where, when MYB110 is incorporated into the MBW complex, the complex activates core genes such as CHS, DFR, and F3GT (Fig. 10). When MYBC1 forms a MBW complex with bHLH and WD40, it activates the F3′H and F3′5′H branch points which hydroxylate substrates to pass down the anthocyanin pathway. WRKY44 may activate the transcription by engaging with the MBW complex to form MBW-WRKY complex (Fig. 10). Petunia PH3 and Arabidopsis TTG2, homologous to kiwifruit WRKY44, were able to bind to the WD40 protein of the MBW complex and formed a MBW-W complex required to transcriptionally activate the target genes involved in vacuolar acidification (petunia) and hair development (Arabidopsis 39 ). In grapevine, VvWRKY26, a homologue of petunia PH3 and AtTTG2, is recruited specifically by the VvMYB5a driven MBW complex to enhance the expression of target genes involved in vacuolar acidification, probably via the formation of a MBW-WRKY complex 40 . In kiwifruit, WRKY motifs are located in close proximity to the MYB and bHLH binding motifs on F3′H and F3′5′H promoters, suggesting the possible binding of the MBW-WRKY complex on the promoters for transcriptional activation (Fig. 6). Infiltration of WRKY44 (and MYBC1) was able to significantly increase the promoter activation.
The F3′H and F3′5′H enzymes are control points for the accumulation of cyanidin or delphinidin-based anthocyanins. However, they also hydroxylate immediate substrates that feed into the PA biosynthesis by the actions of the downstream enzymes ( Supplementary Fig. 8). Activation of the F3′H and F3′5′H will increase the accumulation of dihydroquercetin and dihydromyricetin, which can be converted into anthocyanin precursors by DFR in the anthocyanin pathway or can be converted in quercetin and myricetin by FLS in the proanthocyanin pathway to form flavonols. The anthocyanin precursors can also be intercepted by LAR and ANR enzymes to generate PAs. While MYB110 may increase the general flux through the anthocyanin pathway by activating the core genes, MYBC1 and WRKY44 may increase intermediate substrates via regulation of F3′H and F3′5′H branch points and subsequent competition between enzymes. Hence, MYB110, MYBC1 and WRKY44 are proposed to participate in the same regulatory network shared by anthocyanin and proanthocyanin biosynthesis.

Conclusion
The anthocyanin biosynthesis pathway and the associated transcriptional regulators have been well studied in a variety of plant species. In this study, the kiwifruit MYBC1 and WRKY44 transcriptionally regulate the key F3′H and F3′5′H branch points that have potential for controlling the types of flavonoids accumulated. The www.nature.com/scientificreports/ involvement of TFs shared by the anthocyanin and PA biosynthesis pathways adds to the potential to fine-tune the balance of the total metabolites in the fruit. These specific hydroxylation steps determine pigmentation as well as increasing nutritional values of fruit. Anthocyanin-enriched extracts from purple kiwifruit showed reduction of key inflammatory signals involved in lung inflammation 59 . Dietary intake of di-and tri-hydroxylated anthocyanins, flavonols, flavanols and PAs is linked to reductions in markers of cardiovascular disease risk 60,61 as well as reducing the incidence of cardiovascular and metabolic diseases and cancers [62][63][64] . Understanding the transcriptional regulation of the metabolites will maximise the value of fruit crops as metabolites form a large basis of their nutritional benefits.

Materials and methods
Plant material. Actinidia melanandra, A. purpurea (sometimes referred to as A. arguta var. purpurea) and two progeny lines of a cross between A. macrosperma × A. melanandra (MaMe) were grown at the Plant and Food Research Orchard, Motueka, New Zealand as described previously 19 . Fruits were harvested at mature green stage and held at 20 °C for ripening. Skin peel and flesh tissues were separated and sampled into liquid nitrogen for mature green stage, colour change stage, and ripe stage.
RNA-sequencing and transcriptomic analysis. RNA

Transient over-expression of candidate genes in tobacco leaves. Over-expression vector pSAK277
with 35S promoter driving the expression of the candidate gene was transformed into Agrobacterium tumefaciens strain GV3101 by electroporation followed by incubation 19 . Nicotiana tabacum plants were grown under glasshouse conditions using natural daylight with extension to 16 h. Three leaves of the 6-week-old N. tabacum were infiltrated with Agrobacterium and kept under the same growth conditions. Leaves were photographed and harvested at 7 days after infiltration and stored at − 80 °C until analysis.
Anthocyanin quantification by high performance liquid chromatography (HPLC). Anthocyanin accumulation from transient over-expression of candidate genes in tobacco leaves was confirmed and measured by high performance liquid chromatography (HPLC-DAD). Approximately 300 mg of freeze-dried tissue powder were used to extract anthocyanin with acidified methanol (0.1% HCl) for two hours at room temperature. The supernatant was spin-dried and resuspended in 20% methanol followed by filtration by syringe filter and diluted with 20% methanol for analysis. Identification was achieved using an Acclaim PA2 C18 column (Dionex, ThermoFisher Scientific) maintained at 35 68,69 . Calli formed in the selection medium containing 150 mg/l of kanamycin were excised individually and transferred to fresh regeneration medium containing 150 mg/l of kanamycin for calli growth and bud induction. Approximately 12 weeks after transformation, half of the calli were powdered in liquid nitrogen and stored in − 80 °C until analysis. The rest of the calli were subcultured in fresh regeneration medium. Shoots generated from these calli were excised and subcultured in medium for root elongation before being potted and grown in the containment glasshouse.
Real time quantitative PCR expression analysis. RNA was isolated from the powdered calli using the Spectrum Plant Total RNA kit (Sigma-Aldrich, USA) and reverse transcribed into cDNA using the QuantiTect Reverse Transcription kit (Qiagen, USA) following manufacturer's protocols. Genes of the anthocyanin and proanthocyanin pathways were identified by BLAST with genes of known function on the Actinidia chinensis genome and the Actinidia arguta RNA-seq data 44,70 . Gene-specific oligonucleotide primers were designed using Geneious 10.0.8 and are summarised in Supplementary Table 9. RT-qPCR was carried out using the LightCycler 480 instrument with LightCycler 480 SYBR Green I Mastermix (Roche Diagnostics, USA). Each reaction volume was 5 µL and reactions were run in quadruplicate, and non-template control and water control were included in each run. The thermal cycling conditions were 95 °C for 5 min, followed by 50 cycles of 95 °C for 10 s, 60 °C for 10 s and 72 °C for 20 s, then a melting temperature cycle with continuous fluorescence data acquisition from 65 to 95 °C. The data output was analysed by the LightCycler480 software Version 1.5 (https ://lifes cienc e.roche .com/en_nz/produ cts/light cycle r1430 1-480-softw are-versi on-15.html) using the Target/Reference ratio to compare the expression level of the target genes normalised to the reference gene, elongation factor 1-α EF1α.
Proanthocyanin and anthocyanin quantification in A. arguta calli. The proanthocyanin and anthocyanin content of the calli was determined by liquid chromatography-mass spectrometry (LC-MS) using an LTQ linear ion trap mass spectrometer fitted with an ESI interface (ThermoFisher Scientific, San Jose, CA, USA) coupled to an Ultimate 3000 UHPLC and PDA detector (Dionex, Sunnyvale, CA, USA) as described previously 71 . For proanthocyanins, to each sample (~ 40 mg fresh weight) was added 1 ml of ethanol/water 95:5 (v:v) and 0.8 g stainless steel beads 0.9-2 mm (Next Advance Inc., NY, USA). Samples were bead beaten for 4 min (Bullet Blender 24 Gold, Next Advance Inc., NY, USA) and were extracted overnight. After centrifugation at 13,000 × g for 5 min, the supernatant was evaporated to dryness under a stream of nitrogen at 35 °C and reconstituted in 10% methanol (100 µl) for analysis. Compound separation was achieved using a Hypersil GOLD aQ 1.9µ C18 175 Å (Thermo Scientific, Waltham, MA, USA), 150 × 2.1 mm column maintained at 35 °C. The solvents were (A) water + 0.1% formic acid and (B) acetonitrile + 0.1% formic acid (flow rate, 200 µl/min). The initial mobile phase, 95% A/5% B, was held for 5 min, then ramped linearly to 90% A at 10 min, 83% A at 25 min, 77% A at 30 min, 70% A at 40 min, 3% A at 48 min and held for 5 min before resetting to the original conditions. The sample injection volume was 4 μl. The MS data were acquired in the negative mode using a data dependent LC-MS 4 method. This method isolates and fragments the most intense parent ion to give MS 2 data (daughter ions), then isolates and fragments the most intense daughter ion (MS 3 data), then granddaughter ion (MS 4 data).
The ESI voltage, capillary temperature, sheath gas pressure and sweep gas were set at − 10 V, 275 °C, 35 psi and 5 psi, respectively. Standards for catechin and epicatechin were used to quantitate proanthocyanin concentrations, which, with the exception of epicatechin which is reported as itself, are reported as catechin equivalents per mg of fresh weight (FW). For anthocyanins, to each sample (~ 10 mg fresh weight) was added 1 ml of methanol/formic acid 95:5 (v:v) and 0.8 g stainless steel beads 0.9-2 mm (Next Advance Inc., NY, USA). Samples were bead beaten for 4 min (Bullet Blender 24 Gold, Next Advance Inc., NY, USA) and were extracted overnight. After centrifugation at 13,000 × g for 5 min, the supernatant was evaporated to dryness under a stream of nitrogen at 35 °C and reconstituted in acetonitrile/formic acid/water (5:3:92 v:v:v; 100 µl) for analysis. Compound separation, identification and quantitation by LC-MS were as described 19 .
Statistical analysis. The statistical significance of the difference between the empty vector control and the treatment means in the dual luciferase transient assays was tested by two sample t-test. One-way ANOVA was used to determine the statistically significant differences in metabolite and gene expression analysis between the calli transformed with different candidate genes. Correlation matrix using Pearson's correlation with significance level was calculated and produced in R Studio version 1.2.5033.