Published BAHD acyltransferase sequences (D’Auria, 2006; Tuominen et al., 2011) were used in preliminary BlastP searches against the globe artichoke predicted proteome (Scaglione et al., 2016). The putative BAHD sequences found were aligned with previously characterized BAHD proteins using MUSCLE¹ and a manual inspection was conducted to exclude loci lacking the conserved motifs (HXXXD or DFGWG), with filtering for redundancy. Sequences which exhibited no HXXXD motif were removed. Target P (Emanuelsson et al., 2007) and Predotar (Small et al., 2004) software were used to predict in silico the occurrence of mitochondrial, plastid, and ER targeting sequences.

Globe artichoke putative BAHD sequences together with other characterized protein members (Data Sheet 1) belonging to the 8 clade-based classification reported in Tuominen et al. (2011) were aligned using the MAFFT v6.717 online server²; the FFT-NS-i iterative refinement method was run with default settings using the Blosum62 substitution matrix, leaving gappy regions. An UPGMA based phylogenetic tree was constructed and visualized with the FigTree graphical viewer³.

Paralogous genes are typically generated by a whole genome duplication (WGD) event (Ohno et al., 1968). The CoGe platform⁴ for comparative genomics was used to detect CQA-related paralogous BAHD genes within the globe artichoke genome. To compute chains of syntenic genes found within the complete genome sequence, DAGchainer software (with the ‘Relative gene order’ option activated and the ‘Maximum distance between two matches’ parameter set to 20) was used together with Quota-Align algorithm (with maximum distance between two blocks set to 20 genes), both implemented to the SynMap function within CoGe. The chromosomal locations of the ohnolog BAHD genes were visualized using CIRCOS ideograms generated by the software package from http://circos.ca.

Globe artichoke plants (F1 hybrid ‘Concerto,’ Nunhems) were grown up to the production of commercial immature inflorescences (heads) in an experimental field at Carmagnola (Torino). The following plant materials were harvested and stored at –80°C until required: (i) leaves from 6 weeks- and 1 year-old plants; (ii) external bracts of the inflorescence at the commercial stage; (iii) stems of the primary head at the commercial stage of the inflorescence.

RNA was isolated from 100 mg of globe artichoke tissues using Trizol reagent (Invitrogen) according to the manufacturer’s instructions. The total RNA was quantified and controlled for purity using a spectrophotometer and agarose gel electrophoresis. cDNAs were synthesised from 1.0 μg total plant tissue RNA using a High Capacity RNA-to-cDNA Kit (Thermofisher) according to the manufacturer’s instructions. For quantification of the levels of HQT, Acyltransf_1, and Acyltransf_2 (hereafter named HQT1, HQT2, and HQT3, respectively,) gene-expression in different globe artichoke tissues, a qPCR analysis was performed, using the primers reported in Supplementary Table 1. As a housekeeping gene for globe artichoke, actin (amplified with the primer combination ACT-Rt-For and ACT-Rt-Rev, Supplementary Table 1) was chosen for its stability and level of expression, comparable to those of the genes of interest, and whose expression remains stable in all tissues (Menin et al., 2012). 20 μL qPCRs were performed in three biological replicates for each tested tissue in the presence of fluorescent dye (GoTaq^® qPCR Master Mix, Promega). PCR reactions were carried out in 48-well optical plates using the iCycler Real-time PCR Detection System (Bio-Rad Laboratories, USA) as described in Menin et al. (2010).

Ground tissue (50 mg) of each plant biological replicate was extracted with 1 mL of 75% methanol containing 0.1% formic acid and sonicated (125 W, 20 kHz) for 15 min. Extracts were then centrifuged at 20,000 g and 22°C for 5 min, filtered through a 0.2 μm inorganic membrane filter (RC4, Sartorius, Germany) attached to a disposable syringe, and transferred to a glass vial. The LC-QTOF-MS platform consisted of a Waters Alliance 2795 HT HPLC system equipped with a Luna C18(2) pre-column (2.0 × 4 mm) and an analytical column (2.0 × 150 mm, pore size 100 Å, particle size 3 μm; Phenomenex), connected to an Ultima V4.00.00 QTOF mass spectrometer (Waters, MS Technologies). Degassed eluent A, ultra-pure water: formic acid (1000:1, v/v), and eluent B, acetonitrile: formic acid (1000:1, v/v) were used at a total flow rate of 0.19 mL min^-1. The gradient started at 5% B and increased linearly to 75% B over 45 min; afterward the column was washed with 100% B and equilibrated at 5% A for 15 min before the next injection. The injection volume was 5 μL, ionization was performed using an electrospray source, with detection in the positive mode. The identification of CQAs was carried out by comparing retention times and masses with those reported in Menin et al. (2010), using standard cynarin (1,3-dicaffeoylquinic acid) from Carl Roth (Karlsruhe) and chlorogenic acid from Sigma–Aldrich. Mean comparison was conducted using Tukey’s test. All the data were statistically analyzed using SPSS statistical software.

Seeds of the globe artichoke hybrid ‘Concerto’ were germinated for 2 weeks between two layers of wet filter paper; plantlets were then transplanted into pots in a greenhouse and grown in a climate room at 25°C with 60% relative humidity and a 16 h light: 8 h dark photoperiod cycle with light intensity of 300 μmol m^-2s^-1. The pTRV1 and pTRV2 vectors described by Liu et al. (2002) were used in this study. Two pTRV2 based constructs were employed: pTRV2-PDS [phytoene desaturase (PDS)] and pTRV2-PDS-HQT1. In order to identify the cDNA sequence of globe artichoke PDS, the cDNA sequence of tomato PDS (Liu et al., 2002) was used as query for blast searches in C. cardunculus EST database (Scaglione et al., 2012). 423 bp PDS fragment was PCR amplified from globe artichoke cDNA using primers with EcoRI and XhoI restriction sites (PDS-EcoF and PDS-XhoR, Supplementary Table 1). The resulting product was cloned into pTRV2 to form pTRV2-PDS. 400 bp fragment of HQT1 was PCR amplified from globe artichoke cDNA using primers with XhoI and SmaI restriction sites (HQT1-XhoF and HQT1-SmaR, Supplementary Table 1) and cloned into pTRV2-PDS vector.

The pTRV2 constructs were transformed into Agrobacterium tumefaciens strain C5801. The obtained recombinant A. tumefaciens strains were grown at 28°C and 80 rpm for 24 h in 5 mL of LB media containing kanamycin (50 mg L^-1) and tetracycline (10 mg L^-1). After an overnight incubation, 500 μL of the cultured cells were added to 25 mL of LB broth containing 10 mM 2-[N-morpholino] ethanesulfonic acid (MES) and 20 μM acetosyringone (4′-hydroxy-3′,5′-dimethoxyacetophenone, Sigma) and grown overnight at 28°C and 80 rpm. After the overnight incubation, the bacterial cultures were centrifuged for 20 min at 4,000 g and 4°C, resuspended in 10 mM MES buffer containing 10 mM MgCl₂ and 200 μM acetosyringone to a final OD₆₀₀ of 1–1.5, and incubated at room temperature under gentle shaking at 50 rpm for 3 h. The bacteria containing pTRV1 and the bacteria containing pTRV2 or its derivatives were then mixed together in 1:1 ratio. The cotyledons of globe artichoke were infiltrated with the mixed bacteria cultures using a 1 mL disposable syringe without a needle. The agroinfiltrated plants were then transferred to a climate room at 25°C with 60% relative humidity and a 16 h light/8 h dark photoperiod cycle with light intensity ranging from 300 to 400 μM m^-2s^-1. After 4 weeks the VIGS-silenced plant material (3 biological replicates) was collected and used for qPCR (as described in Quantitative PCR and LC-QTOF-MS Analysis in Globe Artichoke Tissues) and LC-PDA analyses. For the quantification of PDS gene-expression levels in silenced material we used primers reported in Supplementary Table 1.

For transient expression, the pEAQ-HT vector (extremely high-level expression, GATEWAY-compatible plasmid, Sainsbury et al., 2009) was kindly provided by Prof. Lomonossoff (JIC, Norwich UK). For the construction of the expression vectors containing HQT1, HQT2, and HQT3, sets of primers with attB1 and attB2 sites (Supplementary Table 1) were designed. The amplified fragments were first cloned by Gateway Recombinant Technology in pDONOR 207 vector through a BP recombination and subsequently transferred by LR recombination into the pEAQ-HT destination vector, originating the expression vectors pEAQ/HQT1, pEAQ/HQT2, and pEAQ/HQT3. These vectors and the empty vector pEAQ-HT, as a negative control, were introduced into Agrobacterium tumefaciens strain C5801 by the freeze-thaw method. Bacteria containing a single construct or the control vector were grown overnight at 28°C in 5 mL of L medium (10 g L^-1 bactotryptone, 5 g L^-1 Yeast extract, 5 g L^-1 NaCl, 1 g L^-1 D-glucose) with kanamycin (50 mg L^-1). The overnight cultures (2 mL) were then transferred into 20 mL of induction medium (L broth containing 10 mM MES and 20 μM acetosyringone) with kanamycin (50 mg L^-1), and grown as above. The cells were collected by centrifugation for 10 min at 4,000 g and resuspended in 50 mL of infiltration medium (10 mM MgCl₂, 10 mM MES, 200 μM acetosyringone) to an OD₆₀₀ of 1.0 and kept at room temperature for 3 h before being infiltrated into the abaxial air spaces of 2–4-week-old N. benthamiana plants. After 4 days, the infiltrated leaf material was collected and used for semi-qPCR and LC-QTOF-MS analyses.

For the quantification of transgene expression in transformed N. benthamiana plants a semi-qPCR analysis was performed. cDNAs were synthesized from 1.0 μg total RNA from leaves of HQT1, HQT2, HQT3 and control transformants using a High Capacity RNA-to-cDNA Kit (Thermofisher) according to the manufacturer’s instructions. Semi-qPCR amplifications were performed by using specific primers designed by Primer 3 software⁵ for globe artichoke HQT1, HQT2, and HQT3 and tobacco elongation (EF) factor as a housekeeping gene (Nt-EF-For and Nt-EF-Rev, Supplemental Table 1). The thermal cycling program included one step at 95°C for 5 min, followed by 25 cycles of three steps (94°C for 30 s, 57°C for 30 s and 72°C for 30 s). Amplified products were visualized on a 1.5% agarose gel.

Transiently transformed N. benthamiana grinded tissues (100 mg) were suspended in 300 μl of 70% (v/v) methanol and sonicated for 20 min in a water bath. After centrifugation (10,000 g for 10 min), supernatants were filtered with a 13 mm diameter, 0.22 μm pore diameter PTFE syringe filter and analyzed on the LC-PDA-MS/MS analytical platform. Analyses were carried out on a Shimadzu Nexera X2 system equipped with a photodiode detector SPD-M20A in series to a triple quadrupole Shimadzu LCMS-8040 system provided with electrospray ionization (ESI) source (Shimadzu, Dusseldorf Germany). An Ascentis^® Express RP-Amide column (100 mm × 2.1 mm i.d., 2.7 μm particle size, Supelco, Bellefonte, PA) was used. The analysis conditions were: mobile phase: eluent A: 0.1% formic acid in water; eluent B: 0.1% formic acid in acetonitrile; mobile phase gradient was as follows: 5–25% B in 20 min, 25–100% B in 10 min, and 100% B for 1 min. Injection volume: 5 μL; the flow rate of the mobile phase was 0.4 mL min^-1 and the column was maintained at 30°C. UV spectra were acquired in the 210–450 nm wavelength range.

The identification of the components was based on their UV spectra and mass spectral information in Multiple Reaction Monitoring (MRM) mode in both positive and negative ionization mode (respectively, ESI+ and ESI-). MS operative conditions: heat block temperature: 400°C; nebulizing gas (nitrogen) flow rate: 3 L min^-1; drying gas (nitrogen) flow rate: 15 L min^-1; desolvation line (DL) temperature: 250°C. Collision gas: argon (230 kPa). Transitions monitored: ESI⁺: m/z 355.00 →163.00 for CGAs and m/z 517.00 →163.00, m/z 517.00 →145.00, m/z 517.00 →135.00 for diCQAs (dwell time: 20 ms, collision energy -35 V, event time: 0.096 s); ESI^-: m/z 513.00 →179.00 for CGAs and m/z 515.00 →179.00, m/z 515.00 →191.00, m/z 515.00 →135.00 for diCQAs (dwell time: 20 ms, collision energy: 35 V, event time: 0.096 sec). The MRM transitions were selected on the basis of the fragments obtained by analyzing the CGAs and diCQAs standards in full-scan mode in both ESI⁺ and ESI^- in the range of 300–1200 m/z, with a scan speed of 1000 μ sec^-1 and then in product ion scan mode in both ESI+ and ESI - in the range of 100–550 m/z, with a scan speed of 1000 μ sec^-1 and using as precursor ions: 355.00 m/z [M+H]⁺ for ESI⁺ and 353.00 m/z [M-H]^- for ESI^- for CGAs and 517.00 m/z [M+H]⁺ for ESI⁺ and 515.00 m/z [M-H]^- for ESI^- for diCQAs.

For the quantification of CGAs and diCQAs the external calibration method based on the following transitions: ESI+: m/z 355.00 →163.00 for CGAs and m/z 517.00 →163.00 for diCQAs was adopted. A five points calibration curve was built for CGAs analyzing in triplicate the pure standards in the range of 5–500 μg mL^-1 while a four points calibration curve was built for diCQAs analyzing in triplicate the pure standards in the range of 5–100 ng mL^-1. The determination coefficient (R²) was in all cases higher than 0.992.

Chlorogenic acid (CGA), neochlorogenic acid (neoCGA), cryptochlorogenic acid (cryptoCGA), and CoA were purchased from Sigma–Aldrich, while the necessary diCQAs (1,3; 1,5; 3,5; 3,4; 4,5 isomers) were provided from TransMIT (Marburg, Germany).

VIGS silenced artichoke grinded tissues (100 mg) was suspended in 300 μl of 70% (v/v) methanol and sonicated for 20 min in a water bath. After centrifugation (10,000 g for 10 min), the supernatant was filtered with a 13 mm diameter, 0.22 μm pore diameter PTFE syringe filter and analyzed on a Shimadzu XR system equipped with a photodiode detector SPD-M20A (Shimadzu, Dusseldorf Germany). An Ascentis^® Express C18 column (150 mm × 2.1 mm i.d., 2.7 μm particle size, Supelco, Bellefonte, PA) was used and the analysis conditions were: mobile phase: eluent A: 0.1% formic acid in water; eluent B: 0.1% formic acid in acetonitrile; mobile phase gradient was as follows: 5–25% B in 10 min, 25–40% B in 5 min, 40–100% B in 5 min and 100% B for 1 min. Injection volume: 5 μL; the flow rate of the mobile phase was 0.4 mL min^-1 and the column was maintained at 30°C. UV spectra were acquired in the 210–450 nm wavelength range. The identification of the CQAs was carried out by comparing retention times and UV spectra with those of the commercially available standards. For the quantification of CGA and 3,5-diCQA the external calibration method based on the LC-PDA profiles acquired at 325 nm was adopted. A four points calibration curve was built for both compounds analyzing in triplicate the pure standards in the range of 1–100 μg/ml for CGA and in the range of 0.5–10 μg/ml for 3,5-diCQA. The determination coefficient (R²) were 0.998 for CGA and 0.999 for 3,5-diCQA.

Mean comparison was conducted using Tukey’s test. All the data were statistically analyzed using SPSS statistical software.

The full length sequences of the HQT1, HQT2, HQT3 genes were amplified from globe artichoke cDNA using attB specific primers (Supplementary Table 1) and recombined into the pDONR207 Entry vector through a Gateway strategy. The amplicons were cloned into pK7WGF2 (Karimi et al., 2002) producing pK7-35S:GFP:HQT1, pK7-35S:GFP:HQT2, and pK7-35S:GFP:HQT3, respectively. As a control, the unmodified vector for expression of GFP, pK7WGF2 (under the control of the 35S promoter), and an endoplasmic reticulum-targeted pBIN-GFP-KDEL construct were also agro-infiltrated into Nicotiana benthamiana leaves. The expression constructs pK7-35S:GFP:HQT1, pK7-35S:GFP:HQT2, and pK7-35S:GFP:HQT3, and the pK7WGF2 and pBIN-GFP-KDEL vectors alone (controls) were transformed into Agrobacterium tumefaciens strain C5801. The obtained recombinant A. tumefaciens strains were grown at 28°C and 220 rpm for 24 h in 5 mL of L media containing spectinomycin (100 mg L^-1) and tetracycline (10 mg L^-1). The overnight cultures (2 mL) were then transferred into 20 mL of induction medium [L broth containing 10 mM MES and 20 μM acetosyringone with spectinomycin (100 mg L^-1)], and grown as above. The cells were collected by centrifugation at 4,000 g and resuspended in 50 mL of infiltration medium (10 mM MgCl₂, 10 mM MES, 200 μM acetosyringone) to an OD600 of 1.0 and kept at room temperature for 3 h before being infiltrated into the abaxial air spaces of 5-week-old N. benthamiana plants. N. benthamiana plants were grown from seeds on soil in a climate chamber at 25°C (16 h light)/25°C (8 h dark). The localization of fluorescent proteins was analyzed 4 days post-agroinfiltration in small leaf samples (1 cm² leaf explant from at least three independent agro-infiltrated plants) by confocal laser scanning microscopy. All images were acquired and processed using a Leica TCS SP2 confocal microscope and software (Leica Microsystems GmbH, Wetzlar, Germany) as described in Eljounaidi et al. (2015). GFP and plastid fluorescence were both excited at 488 nm with emission recorded at 500–525 nm and 600–640 nm, respectively. A scanning resolution of 1024 × 1024 pixels was chosen and serial optical sections were acquired with either 1 or 2 μm resolution along the z-axis. Quantification of transgene expression in transformed N. benthamiana plants was performed by a semi-qPCR analysis (as described in Transient Heterologous Expression in N. benthamiana).

A survey of the globe artichoke genome (release v1.0, Scaglione et al., 2016), showed the existence of 74 genes with high similarity to previously characterized BAHD acyltransferases (Menin et al., 2010; Tuominen et al., 2011). By considering only those carrying the catalytic site HXXXD, the number was reduced to 69. A phylogenetic analysis of these 69 globe artichoke BAHD proteins together with 43 representative proteins of the BAHD family from 21 species highlighted seven major clades (Figure 2). Four globe artichoke proteins clustered into Clade I, corresponding to the group classified as Clade II by Tuominen et al. (2011), defined by the characterized Glossy2 and CER2 homologs in Zea mays (Tacke et al., 1995) and A. thaliana (Negruk et al., 1996; Xia et al., 1996), involved in the extension of long chain epicuticular waxes, which are important both for restricting water loss and for defence against pathogens. Clade II contained 15 globe artichoke proteins and is sister to the Clade IIIa reported by Tuominen et al. (2011), along with acyltransferases which utilize a range of alcohol substrates to produce volatile esters (D’Auria, 2006). Nine globe artichoke proteins were grouped in Clade III, corresponding to the Clade IIIb described in Tuominen et al. (2011), which lacked any functionally defined homolog. No globe artichoke BAHD proteins occurred in Clade IV which contained sequences related to barley agmatine coumaroyl transferase (ACT), an enzyme involved in the biosynthesis of anti-fungal hydroxycinnamoyl agmatine derivatives (Burhenne et al., 2003). Clade V can be subdivided further into several subgroups, as reported in D’Auria (2006), three of which contained characterized enzymes: the first clustered three globe artichoke proteins along with enzymes that are involved in biosynthesis of volatile esters; the second grouped one globe artichoke protein with enzymes, from Taxus species, involved in the production of the compound paclitaxel; the third clustered nine globe artichoke sequences with enzymes that use hydroxycinnamoyl/benzoyl CoA as acyl donor. Clade VI is sister to the group classified as Clade Ib by Tuominen et al. (2011) and includes ten globe artichoke members which lack any functionally defined homolog. Finally, Clade VII, corresponding to the group classified as Clade Ib by Tuominen et al. (2011), grouped 13 globe artichoke paralogous proteins along with the characterized enzymes involved in modification of phenolic glycosides, predominantly anthocyanins (Suzuki et al., 2002).

We focused our attention on Clade V, in particular on the sub-group containing the biochemically characterized hydroxycinnamoyltransferases (HCT/HQT) involved in biosynthesis of lignin and chlorogenic acid. In this sub-group the globe artichoke sequences included are: HCT (AAZ80046, Comino et al., 2007), Acyltransf_1 and Acyltransf_2 (ADL62854, ADL62855.1, Menin et al., 2010; renamed in this paper as HQT2 and HQT3, respectively) and HQT (ABK79689, Comino et al., 2009; renamed in this paper as HQT1) plus an extra-HCT (KVH99042) functionally predicted in silico. The remaining four predicted sequences are related to previously characterized enzymes: KVI01309.1 (locus Ccrd_020450) and KVI06149.1 (locus Ccrd_015498) appeared similar to an ω-hydroxypalmitate O-feruloyl transferase (HHT1, Lotfy et al., 1995, 1996), while KVI11866.1 (locus Ccrd_009712) and KVH57169.1 (locus Ccrd_025673) are similar to a spermidine hydroxycinnamoyl transferase (SHT, Grienenberger et al., 2009).

Many BAHD multiple gene copies were observed in the globe artichoke genome, and the five CQA-related genes appeared pairwise highly similar (Figure 3A), exhibiting the functional HXXXD or DFGWG domains (Figure 3B); they appeared as duplicated genes, but resident in different chromosomes in syntenic segments, likely fruit of a WGD event. In particular, HQT2 (in Chr8), HQT3 (in Chr2), and HQT1 (unplaced scaffold 311, formerly mapped on Lg5/Chr5, Comino et al., 2009) appeared as ohnologous genes (Ohno et al., 1968), in the same way as for the HCT (Chr3) and extra-HCT (Chr6) genes, as depicted in Figure 3C. Only HQT-like enzymes involved in CGA synthesis were selected for in vivo functional investigation.

Expression of the genes encoding for HQT1, HQT2, and HQT3 was analyzed by qPCR and compared to the levels of CQAs at the same stage of plant development. The gene expression profiles are shown in Figure 4A. The expression profile for HQT1 was higher in bract, with a level of expression 4.3-fold higher than 1-year leaf tissue. Expression levels of HQT3 were notably high in stem tissue, where transcripts were 17-fold more abundant than 1-year leaf tissue. Expression of HQT2 was greatest in both stem and 6-week leaf tissues.

To evaluate any relationship between gene expression and metabolite content, CGA and 1,3-diCQA were quantified in the same globe artichoke tissues by liquid chromatography coupled to mass spectrometry (LC-MS) of methanol extracts of freeze dried plant material, and compared to two original standards. Significant differences in CGA contents were observed (p ≤ 0.05) among leaf and other tissues (Figure 4B), but statistical differences in 1,3-diCQA contents were not found for any of the analyzed tissues.

Four weeks after infiltration, plants inoculated with pTRV1 and pTRV2 vector showed no obvious differences compared with the control in overall shoot and leaf morphology. The virus was detected in plants agro-infiltrated with pTRV2 vector, while no virus was detected by PCR in control plants (data not shown).

To determine if endogenous gene silencing can also be elicited by TRV-mediated VIGS, we inserted PDS marker into a pTRV2 VIGS vector. 4 weeks after infiltration, a photobleached phenotype was observed, mainly localized in proximity of the main veins of young leaves of ‘Concerto’ globe artichoke seedlings (Figure 5A). No photobleaching phenotype was observed in plants infected with pTRV2 empty vector. To monitor the silencing level of PDS, a qPCR analysis was performed. The results revealed that PDS transcript levels in photobleached leaves were reduced by more than 50% compared to the controls (Figure 5B). The VIGS approach was applied for silencing the HQT-like acyltransferases involved in the final steps of the caffeoylquinic acid pathway. To achieve single gene silencing (i.e., to avoid post-transcriptional silencing of closely related gene sequences), sequence identity of more than 22 nt with other genes has to be avoided (Gaquerel et al., 2013). This pre-requisite was achieved for HQT1, while due to the high level of identity between HQT2 and HQT3 it was not possible to perform single gene silencing on these genes. ‘Concerto’ globe artichoke seedlings were infiltrated with a mixture of Agrobacteria transformed with pTRV2-PDS-HQT1. The photobleached leaf phenotype correlated with a marked reduction in the expression level of the genes introduced into the silencing vector. PDS transcript levels in photobleached leaves were 70% compared to the control. The HQT1 transcript levels in leaves were reduced to 50% of those found in the control (Figure 5B). No cross-silencing of HQT2 and HQT3 was detected in HQT1 silenced leaves (Figure 5B).

The VIGS-silenced leaves were analyzed by (LC)-PDA, comparing their UV profiles to those obtained from a CQA standards mixture. The silencing of HQT1 (Figure 6) resulted in a significant reduction in content of both chlorogenic and 3,5-dicaffeoylquinic acids (33,6 ± 31,3 vs 184,85 ± 59,65 μg/g FW and 2,35 ± 0,95 vs 18,8 ± 2,1 ng/g FW, respectively).

Agrobacteria transformed with a pEAQ expression vector containing HQT1, HQT2, and HQT3 were infiltrated into N. benthamiana leaves. Plants agro-infiltrated only with the empty vector were used as negative controls. 4 days after infiltration, transformed leaves were assayed for monoCQA and diCQAs content by LC-MS analysis.

N. benthamiana leaves transiently expressing HQT1, HQT2, and HQT3 were characterized by a significant increase in mono and diCQAs content (Figure 7, Supplementary Image 1). In particular the overexpression of HQT1, HQT2, and HQT3 determined a 2–5 fold increase on mono CQAs and a 1–4 fold increase for diCQAs. The diCQA content in tissue extracts of transiently transformed leaves is between 6.4 and 64.9 ng/mL while the mono CQAs content is 2,000–4,000 higher ranging between 8.7 and 326.8 μg/mL (Figure 7). Transient transformants and controls (plants transformed with empty vector) were also tested for expression of the HQT1, HQT2, and HQT3 transgenes by semi-qPCR using specific primers for globe artichoke HQT1, HQT2, and HQT3 sequences. Expression of the transgene was demonstrated in all HQT1, HQT2, and HQT3 transformants, while no amplification was detected in control pEAQ-HT transformed plants (Supplementary Image 2).

No putative targeting sequences predicting mitochondrial, plastid and ER localization were found for HQT1, HQT2, and HQT3 proteins. For the in vivo assessment of subcellular localization of the enzymes, N. benthamiana plants were infiltrated with Agrobacteria suspension harboring the expression constructs pK7-35S:GFP:HQT1, pK7-35S:GFP:HQT2, pK7-35S:GFP:HQT3 as well as the pK7WGF2 and the endoplasmic reticulum-targeted pBIN-GFP-KDEL vector as controls. Expression of the fusion genes was confirmed using semi qPCR with gene specific primers (Supplementary Image 3).

The subcellular localization of each protein was analyzed by confocal laser scanning microscopy (Figure 8). All three GFP-tagged proteins accumulated at the periphery of the cells and in cytoplasmic strands (Figures 8A–C). Significant differences in protein distribution can anyway be highlighted: GFP:HQT1 (Figure 8A) appeared to be excluded from the nucleoplasm (Figure 8A); GFP:HQT2 (Figure 8B), by contrast, also diffused in the nucleus, generating a very similar pattern to that observed for free cytosolic GFP (Figure 8D) but different from the localization of GFP-KDEL in the endoplasmic reticulum (Figure 8E); lastly, GFP:HQT3 localization (Figure 8C) can be described as cytosolic, even if the transient expression of this construct resulted in a weaker accumulation of fluorescent signal compared to the previous two fusion proteins.

Confocal microscopy imaging of GFP fusion constructs is compatible with the presence of the chimeric proteins in the cytoplasm rather than in other organelles, such as the vacuole or the endoplasmic reticulum, in line with with the in silico predicted localization of HQT1, HQT2, and HQT3 proteins in the cytoplasm.