Seeds of C. intybus L. var. Orchies were provided by Florimond-Desprez (Cappelle en pévèle, France). Following germination, seedlings were cultivated hydroponically on 0.5 X Murashige and Skoog (MS) medium (Murashige and Skoog, 1962) in a greenhouse for 6 weeks (16/8-h light/dark cycle; 24/18°C light/dark phase). Samples for cDNA synthesis and qRT-PCR experiments were collected in biological triplicates on a weekly basis, starting from week 3. Each plant was divided into three parts: root, young leaves and old leaves. All samples were immediately snap frozen in liquid nitrogen and stored at -80°C until needed. N. benthamiana were grown on soil under identical greenhouse conditions (16/8 h light/dark cycle; 24/18°C light/dark phase).

Methanolic extractions were performed identically for C. intybus and N. benthamiana. Lyophilized plant material was powdered using a ball mill and 20 mg were resuspended in 1 mL of a methanol/water/acetic acid mixture (75/23/2, v/v/v). The mixtures were then incubated for 1 h at 4°C under agitation in the dark. Homogenates were then clarified by centrifugation at 14,000 g for 15 min at 4°C, and the supernatants were filtered through a 0.45 µm filter. The extracts were analysed by HPLC-DAD and UPLC-MS-MS as previously described (Bernard et al., 2022). For the calculation of the kinetic parameters, another gradient was developed in HPLC-DAD to achieve a better separation of 5-CQA and CA. The mobile phase was composed of water (as solvent A) and acetonitrile (as solvent B) both acidified with 0.1% ortho-phosphoric acid. The solvents were delivered at a flow rate of 1.1 mL min ^-1. Oven temperature was set to 45°C. The following gradient of solvent B was used: 5-13% (0-1 min), 13-15% (1-11 min), 15-70% (11-12 min), 70% (12-13 min), 70-5% (13-14 min), 5% (14-21 min). For visualization and quantification, wavelength was set at 320 nm. All compounds were identified by comparison of their retention times, UV and mass spectra with those of commercial standards. Quantification was performed by reference to external standard calibration curves for each compound.

UHPLC-ESI-HRMS analysis was carried out on an ACQUITY UPLC I-class chain coupled to the Vion IMS Quadrupole time-of-flight high-resolution mass spectrometer (HRMS), equipped with an electrospray ionization (ESI) source (Z-spray) and an additional spray for the reference compound (Waters, Manchester, UK). A double detection was performed using a photodiode array (PDA) detector (UV detection between 210 and 400 nm) and by ESI-HRMS. One µL of the sample was injected into a Kinetex Biphenyl column (2.1 x 100 mm; 1.7 μm) (Phenomenex, Torrance, California, USA) heated to 55°C. Elution was performed with a mobile phase flow rate of 0.5 mL min^-1 consisting of a mixture of water (A) and methanol (B) containing 0.1% of formic acid. The gradient starts at t = 0, ratio (A/B) 80:20; at t = 0.5 min, 80:20; at t = 5 min, 40:60; at t = 6 min, 10:90; at t = 7 min, 10:90; at t = 7.5 min, 80:20; and at t = 10 min, 80:20. ESI-HRMS analysis was performed in positive ionization mode over the mass range 50–1000 Da with ESI source parameters as follows: capillary voltage 3 kV, sampling cone voltage 40 V. Data were acquired in high-definition MS^E (HDMS^E), a data independent analysis (DIA) mode using drift-time filtering, which consists of alternating a Low-Energy function (CE 6 eV) as well as a second High-Energy function (CE 20 to 50 eV) to obtain information on fragment ions. Data acquisition and processing was performed with UNIFI software (version 1.9.4, Waters).

Homologous searches in chicory were conducted using the Basic Local Alignment Search Tool (BLAST). The IbICS protein sequence (GenBank accession QOH99319) was used as a query in a tBlastn search against the published industrial chicory genome (Waegneer et al., 2023). Signal peptide prediction was carried out using TargetP 2.0 (https://services.healthtech.dtu.dk/services/TargetP-2.0/), SignalP 5.0 (https://services.healthtech.dtu.dk/services/SignalP-5.0/) and Deeploc 2.0 (https://services.healthtech.dtu.dk/services/DeepLoc-2.0/). The three programs gave the same prediction. The molecular weight and isoelectric point of the predicted proteins were obtained from the ExPASy compute pI/MW tool (https://web.expasy.org/compute_pi/). The nucleotide sequences of candidate genes were translated, and multiple sequence alignment was performed using Muscle algorithm (Edgar, 2004) in MEGA 11 software (Tamura et al., 2021) to identify conserved domains and pairwise similarity. Amino acid conservation and alignment were visualized using Jalview 2.11.2.6 (Waterhouse et al., 2009). Phylogenetic tree construction was based on the multiple sequence alignment using the neighbour-joining method (Saitou and Nei, 1987) with 1000 bootstrap replicates in MEGA 11 software.

Plant samples stored at -80°C were ground in liquid nitrogen using a mortar and pestle. Approximately 100 mg of powdered tissue was used for extraction. Total RNA was isolated using the NucleoSpin RNA Plus Kit (Macherey-Nagel). Prior to cDNA synthesis, complete removal of contaminating genomic DNA was performed using the Turbo DNA-free™ kit (Invitrogen) by doing a routine DNAse treatment on 10µg of purified RNA samples. RNA quality and concentration were assessed on an Agilent 2100 Bioanalyzer using the 2100 Expert software. For candidate gene amplification and subsequent cloning, RNA samples from different tissues (roots, young/old leaves) and developmental stages (3 to 6 weeks) were pooled. For qRT-PCR experiments, RNA was extracted from different parts of 3-week-old chicory plants. cDNA synthesis was performed using 1 µg total RNA with the SuperScript IV Reverse Transcriptase Kit (Invitrogen) and oligo (dT)₂₀.

The TIP41 gene was used as a reference for normalisation, based on previous experiments and assessment of its expression stability among samples used in this study (Delporte et al., 2015). CiCQEs and TIP41 specific primers were designed using Primer 3 software (https://primer3.ut.ee/) with default parameters (Tm: 60 ± 1°C; 18-25 nucleotides; 40-60% GC; amplicon length from 60 to 150 bp) ( Supplementary Table S1 ). The amplification efficiency of each primer set was determined to be 109.2, 110.3, 104.6 and 103.2% for CiCQE1, CiCQE2, CiCQE3 and TIP41, respectively. qRT-PCR was performed as described in Delporte et al. using the iQ SYBR Green Supermix (Bio-Rad) in a final volume of 20 µL, containing 10 µL of iQ™ SYBR^® Green supermix (2x), 100 ng of DNA-free RNA and 5 µM of each primer (Delporte et al., 2015). The obtained cycle thresholds were normalized to TIP41 values. The following program was used: initial denaturation at 95°C for 3 min followed by 40 cycles including 95°C for 10 s and 60°C for 30 s. Melting curves were generated for each sample to verify the amplification specificity. To control for genomic DNA contamination, reactions were also performed with non-retro-transcribed RNA. Relative gene expression levels were calculated according to the Pfaffl equation (Pfaffl, 2001).

Full-length sequences of CiCQE1 (Genbank accession PX093951), CiCQE2 (Genbank accession PX093952) and CiCQE3 (Genbank accession PX093953) without stop were amplified from a pool of cDNA (see above) with a proofreading polymerase (PrimeSTAR HS DNA polymerase, Takara). To minimise allelic variation, each amplification was performed in triplicate. The primers used are listed in Supplementary Table S1 . For CiCQE2, cloning from gDNA or by removing the signal peptide was also performed. Using Gateway cloning™, entry vectors pDONR-Zeo_CiCQE1, pDONR-Zeo_CiCQE2, pDONR-Zeo_CiCQE3 were obtained and used to generate the expression vectors pEAQ-HT-DEST3_CiCQE1, pEAQ-HT-DEST3_CiCQE2, pEAQ-HT-DEST3_CiCQE3, respectively. All constructs were verified by Sanger sequencing. The pEAQ-HT-DEST3 vector adds a C-terminus 6x His tag to the protein and co-expresses the silencing inhibitor P19. The constructs pEAQ-HT-DEST3_CiCQEs were introduced into Rhizobium radiobacter strain LBA4404 (Invitrogen) by electroporation (V=2,500, µF=25, Resistance=400 Ω) and selected on YEB plates containing 25 µg mL^-1 of rifampicin and 50 µg mL^-1 of kanamycin. Four independent clones per construct were used to inoculate 20 mL of liquid YEB medium containing 50 µg mL^-1 of kanamycin and grown overnight at 28°C with shaking at 210 rpm to an OD₆₀₀ of 0.8-1. The bacterial cultures were then pelleted by centrifugation for 10 min at 3,000 g at 20°C and resuspended in infiltration buffer (10 mM MgCl₂, 10 mM MES, 100 µM acetosyringone, pH 5.7) to a final OD₆₀₀ of 1. After a 3-hour incubation at room temperature, the abaxial epidermis of 6-week-old N. benthamiana leaves was infiltrated with the bacterial suspension using a needless syringe. Plants infiltrated with agrobacteria carrying the empty pEAQ-HT-DEST3 vector served as negative controls.

Transformed N. benthamiana leaves were harvested 3 days post-infiltration, immediately frozen in liquid nitrogen, and stored at -80°C until protein extraction. Frozen leaf tissue expressing the pEAQ-HT-DEST3_CQEs constructs was ground in liquid nitrogen using mortar and pestle. An appropriate volume of binding buffer (20 mM sodium phosphate, 0.5 mM NaCl, 20 mM imidazole, pH 7.4) was immediately added to the resulting powders (4 mL g^-1 FW). Following a 15-minutes incubation on ice, the crude extracts were clarified by centrifugation for 10 min at 4,000 g at 4°C and passed through a 0.22 µm filter (Sarstedt). Flow-throughs were used immediately for protein purification.

His-tagged proteins were purified on 1 mL HisTrap HP columns (GE Healthcare) loaded with Ni²⁺, following supplier protocols. Briefly, the column was first equilibrated with 10 mL of binding buffer (20 mM sodium phosphate, 0.5 mM NaCl, 20 mM imidazole, pH 7.4) before loading with filtered protein extracts. Unbounded proteins were then washed out with 15 mL of binding buffer. Finally, bound His-tagged proteins were eluted in five 1 mL fractions with elution buffer (20 mM sodium phosphate, 0.5 mM NaCl, 500 mM imidazole, pH 7.4). The enriched protein fractions were pulled and desalted on a pD-10 desalting column (GE Healthcare). Proteins were eluted in storage buffer (10 mM phosphate buffer pH 7.2, 20% glycerol) and concentrated ten-fold on Vivaspin 6 columns (Sartorius) with a 30 kDa mass cut-off. Concentrated His-tagged proteins were stored at -20°C until further use.

Protein concentration was estimated using the Bradford reagent (Bio-Rad). A standard calibration curve was generated using BSA ranging from 0 to 1 mg mL^-1. SDS-PAGE and immunoblot analyses were performed as described elsewhere. For Coomassie blue staining and Western blot analysis, 5 µg of purified proteins were loaded per lane. Proteins were transferred to a PVDF membrane (Trans-Blot Turbo Transfer Pack). Immunodetection was carried out using a mouse anti-penta-His antibody (1:1,200, Qiagen) as the primary antibody and a mouse anti-IgG antibody (1:2,500, Promega) as the secondary antibody. Revelation was performed using the NBT/BCIP Reagent (Bio-Rad).

Enzymatic assays were performed in a total volume of 15 µL of tri-buffer (0.1 M acetic acid, 0.1 M Tris, 20 mM MES, pH 6), containing 0.4 µg of purified recombinant his-tagged protein. For the evaluation of CiCQEs activity on chicory crude methanolic extracts, root and leaf extracts prepared as described were first dried using a SpeedVac concentrator and then resuspended in the same volume of tri-buffer. Standard assays were performed with 1 mM of substrates in the reaction mix and incubated for 1 h at 30°C. For kinetic analysis, varying concentrations of 5-CQA (10 µM to 10 mM) were used with 0.2 µg of CiCQEs, incubated for 15 min at 30°C based on preliminary experiments. The optimal pH was determined using 500 µM of 5-CQA in 15 µL of various buffers (0.1 M citrate, 0.1 M acetate, 0.1 M phosphate and 0.1 M tris) ranging from pH 3 to 9 containing 0.2 µg of purified CiCQEs. Reactions were run for 30 min at 30°C. All reactions were stopped by the addition of 45 µL of methanol, filtered through 0.45 µm filter plates (AcroprepTM PALL) and immediately analyzed by HPLC-DAD (Shimadzu). All reactions were performed in triplicate. Kinetic parameters were determined by fitting Michaelis-Menten curves using GraphPad Prism Software. Optimal pH was calculated using the same software by fitting a Gaussian curve to the data.

Esterase activity was assayed in a 100 µL reaction volume containing tri-buffer (0.1 M acetic acid, 0.1 M Tris, 20 mM MES, pH 6), 1 mM of p-nitrophenyl-acetate and 0.2 µg of purified CiCQEs. In control reactions, purified proteins were replaced with storage buffer (10 mM phosphate buffer pH 7.2, 20% glycerol). Enzymatic reactions were incubated at 30°C for 15 min. Absorbance at 347 nm was immediately measured using a Thermo Scientific Multiskan Spectrum. The concentration of p-nitrophenol was determined from an external calibration curve generated with a commercial standard of p-nitrophenol.

Statistical analysis was made on R 4.1.2. Depending on the data analyzed, ANOVA or Student’s t-test were performed.

A GDSL enzyme that catalyzes the formation of 3,5-DiCQA from two molecules of 5-CQA was recently identified in sweet potato and named isochlorogenate synthase (IbICS) (Miguel et al., 2020). To identify homologous enzymes involved in 5-CQA metabolism in chicory, a tBLASTn search was conducted against the published industrial chicory genome (Waegneer et al., 2023) using the IbICS protein sequence as query. Among the multiple retrieved GDSL homologs, the three best hits were selected and hereafter named caffeoyl-quinate esterases i.e. CiCQE1, CiCQE2, CiCQE3 for further analysis. Genomic sequence analysis revealed that CiCQE1 and CiCQE2 each consist of five exons and four introns, whereas CiCQE3 contains an additional exon ( Figure 2A ). Their coding sequences are 1182, 1173 and 1164 bp long, respectively. In particular, CiCQE3 has a large intron of 9 kb between exons 2 and 3. CiCQE2 and CiCQE3 are clustered on chromosome 6 together with several other predicted GDSLs, while CiCQE1 is located on chromosome 5. The predicted lengths of the encoded proteins are 394, 391, and 388 amino acids (AA), respectively. Signal peptide prediction using SignalP, TargetP and DeepLoc softwares suggested N-terminal signal peptides in all three proteins, addressing them to the secretory pathway ( Supplementary Figure S1 ). After removal of the signal peptides, the predicted protein molecular weights were 39.72 (CiCQE1), 38.91 (CiCQE2) and 39.19 (CiCQE3) kDa, with a isoelectric points (pI) of 5.15, 5.26 and 5.05, respectively. Pairwise sequence identity analysis revealed 64.14% similarity between CiCQE1 and CiCQE2, 63.71% between CiCQE2 and CiCQE3, and 69.29% between CiCQE1 and CiCQE3. All three candidate proteins contain the characteristic GXSXXDXG motif, which is conserved among members of the GDSL lipase/esterase family, as well as the four conserved blocks of the SGNH hydrolase family ( Figure 2B ). In addition, the catalytic triad Ser-Asp-His, which is essential for the esterase/transferase activity, is present in all three proteins, supporting their classification within the GDSL family. Phylogenetic analysis including several characterized GDSL proteins revealed that the three candidates cluster with IbICS and SlCGT (Teutschbein et al., 2010; Miguel et al., 2020), which are transferases using 5-CQA as a substrate as well as several GDSLs involved in the metabolism of cell wall components (Clauß et al., 2011; Zhang et al., 2017; Ursache et al., 2021) ( Figure 2C ). These results suggest a potential role for these chicory GDSLs in 5-CQA metabolism or related pathways.

To gain further insight into the potential roles of the candidates, their expression profiles were examined by quantitative real-time PCR (qRT-PCR). Given the differential accumulation of caffeoyl-quinic acids (5-CQA and 3,5-DiCQA) and caffeoyl-tartaric acids (CTA and DiCTA) between roots and leaves in chicory, expression analyses were conducted on both tissues, including leaves at different developmental stages. Transcript levels were normalized using TIP41, which was identified as the most stable reference gene among the samples tested. All three CiCQEs displayed higher expression levels in roots compared to leaves, although this difference was not statistically significant for CiCQE1 ( Figure 3 ). In particular, CiCQE3 showed a tenfold higher expression in roots than in leaves. As 5-CQA and 3,5-DiCQA are the only caffeic acid esters accumulated in roots, these expression patterns suggest that the CiCQEs candidates are more likely to be involved in the metabolism of these compounds rather than caffeoyl-tartaric acid esters (Legrand et al., 2016). However, except for CiCQE3, the high expression of CiCQE1 and CiCQE3 in leaves did not support their role in 3,5-DiCQA synthesis as this compound is almost completely absent in leaves.

To evaluate the enzymatic functions of CiCQEs, their corresponding genes were cloned and expressed in the plant heterologous system N. benthamiana. This system was chosen for its ability to perform appropriate post-translational modifications that are often lacking in Escherichia coli, resulting in inactive GDSL proteins (Teutschbein et al., 2010; Tan et al., 2014; Miguel et al., 2020). The CiCQE1, CiCQE2 and CiCQE3 coding sequences were amplified from chicory cDNA and fused to a C-terminal 6xHis tag for purification. N-terminal His-tag constructs were not generated, as each gene contains a predicted signal peptide at the N-terminus. Following transient expression in N. benthamiana leaves via Rhizobium-mediated transformation, proteins were extracted and purified on His-trap columns, based on histidine affinity to nickel. The presence and identity of CiCQE1 (39.72 kDa) and CiCQE3 (39.19 kDa) were confirmed by Coomassie blue staining and immunoblot analysis ( Figures 4A , B ). CiCQE2 did not give bands at the expected size despite multiple attempts of production and purification. To overcome potential problems of transcript processing or protein targeting, cloning from gDNA and of the truncated cDNA (without signal peptide sequence) were done and the resulting expression vectors were used to infiltrate tobacco. Unfortunately, no protein or activity could be detected in these conditions. Modeling and docking experiments performed with CiCQEs revealed that, despite sharing a similar 3D structure, 5-CQA might be bind in reverse position within the active site of CiCQE2, blocking the catalytic access of serine to the ester bonds ( Supplementary Figure S2 ). This candidate was therefore not considered for further analysis. To assess enzymatic activity in vitro, purified CiCQE1-6xHis and CiCQE3-6xHis proteins were incubated with methanolic extracts of chicory leaves or roots ( Figures 4C - E ). Incubation with both enzymes did not result in the formation of 3,5-DiCQA, as was initially expected on the basis of sequence homology with IbICS. Instead, both enzymes surprisingly led to the accumulation of (CA) together with a strong reduction in the levels of 5-CQA and 3,5-DiCQA. Similarly, the incubation of purified CQEs with standards of 5-CQA and 3,5-DiCQA resulted in substrate degradation and accumulation of CA, as verified by mass spectrometry (m/z of 181 in positive mode) ( Figures 4F , G ) ( Supplementary Figures S3 , S4 ). Both enzymes were unable to metabolize caffeoyl-tartaric acids but showed hydrolytic activities toward caffeoyl-shikimate (data not shown). Incubations made with other isomers of 5-CQA and 3,5-DiCQA showed that CiCQE1 acts specifically on 5-CQA and 3,5-DiCQA, while CiCQE3 uses 5-CQA, 3,5-DiCQA and 1,5-DiCQA ( Supplementary Figure S5 ). Collectively, these results suggest that CiCQE1 and CiCQE3 function as esterases rather than transferases, efficiently hydrolyzing 5-CQA and 3,5-DiCQA to release CA in vitro.

Previous data have shown that the two enzymes under investigation may most probably act as esterases. To test this, we monitored the esterase activity of both enzymes using p-nitrophenyl-acetate, a substrate commonly used to assess hydrolytic activities (Peng et al., 2016b). Upon hydrolysis, p-nitrophenyl-acetate releases p-nitrophenol, which has a stable absorbance at 347 nm, independent of pH, allowing for accurate quantification ( Figure 5A ). Purified CiCQE1 and CiCQE3 were then incubated with p-nitrophenyl-acetate after which the absorbance at 347 nm was measured. Both enzymes showed a significant accumulation of p-nitrophenol compared to the negative control ( Figure 5B ). CiCQE1 and CiCQE3 exhibit esterase activity, producing 1.38 and 1.12 µmol min^-1 mg^-1 enzyme of p-nitrophenol using p-nitrophenyl-acetate as substrate according to a calibration curve of p-nitrophenol. This analysis confirms that the candidate CiCQEs possess esterase activity, confirming their role as chlorogenate esterases in vitro.

To further investigate the enzymatic activity of CiCQE1 and CiCQE3, we determined their kinetic parameters in vitro using 5-CQA as substrate ( Figure 6A ). The Michaelis-Menten Km values for CiCQE1 were determined to be 0.684 mM, while the turnover Kcat was estimated to be 0.256 s^-1. In contrast, CiCQE3 exhibited a higher Km value of 1.740 mM, indicating a lower substrate affinity for 5-CQA. However, its Kcat was found to be 0.466 s^-1, suggesting a higher catalytic rate. Despite this, both enzymes display relatively high Km values, indicating a low affinity to 5-CQA under in vitro conditions. Nevertheless, given the relatively high endogenous concentration of 5-CQA, around 4 µmol g^-1 DW in chicory roots (Legrand et al., 2016), these compounds might be used in vivo as well. The Vmax values for CiCQE1 and CiCQE3 were comparable to those reported for AtCSE, when tested with caffeoyl-shikimate, which have significative activity in planta (Vanholme et al., 2013). The optimal pHs of CiCQE1 and CiCQE3 were then determined using different buffers depending on the pH range ( Figure 6B ). Both enzymes showed peak activity around pH 6. However, an increase in activity was observed at pH 4 in acetate buffer. These pH optima are consistent with the signal peptides of CiCQEs, which are predicted to direct proteins through the secretory pathway, either the vacuole or the apoplast. Both of this compartment are acidic cellular environments.

In addition to their activity on 5-CQA, both CiCQE1 and CiCQE3 were found to have hydrolytic activity on 3,5-DiCQA. However, the enzymatic hydrolysis of 3,5-DiCQA yields 5-CQA, which is also a substrate for CiCQEs, further hydrolyzed to CA ( Figure 4G ; Supplementary Figure S5 ). This sequential reaction hampers kinetic analysis, as it introduced overlapping substrate-product dynamics and prevented accurate determination of kinetic parameters for the initial hydrolysis of 3,5-DiCQA. Consequently, although both enzymes clearly hydrolyze 3,5-DiCQA, the catalytic efficiency toward this substrate could not be determined.

Tobacco leaves are known to accumulate high levels of 5-CQA. To investigate the functional role of CiCQE1 and CiCQE3 in planta, we analyzed the impact of their overexpression on the phenolic contents of N. benthamiana leaves. Methanolic extracts were prepared from leaves six days after agroinfiltration with constructs harboring either CiCQE1 or CiCQE3. Leaves infiltrated with an empty vector served as a negative control. Infiltration with the two constructs resulted in significant accumulation of two compounds not detected in the negative control ( Figure 7A ). One of these was identified as caffeic acid (CA), in agreement with the in vitro characterization. The second compound was purified and analyzed by UPLC/ESI- HDMS^E. A peak absent in the control was detected at retention time 2.89 min. It gave a [M+H]⁺ ion at mass-to-charge ration (m/z) of 251.1375 ( Figure 7B ). The elemental composition was determined to be C₁₃H₁₉N₂O₃. The HDMS^E spectra revealed fragment ions at m/z 234.1112, 163.0378 and 89.1064 which are specific of the fragmentation pattern of caffeoyl-putrescine ( Figure 7C ) (Baumert et al., 2001; Zhong et al., 2024). It should be noted that many fragment ions are present in Low Energy spectrum due to spontaneous fragmentation in the ESI source and most of them are found in the High Energy spectrum, differing only in their intensity. Based on the fragmentation pattern ( Figure 7C ), the fragment ion at m/z 234.1112 corresponds to the loss of NH₃ (17 Da) from the precursor ion 251.1375, i.e. a NH₃ loss from the terminal amine of putrescine part of caffeoyl-putrescine. The fragment at m/z 163.0378 corresponds to the caffeoyl and can be explained by loss of putrescine (88 Da) from caffeoyl-putrescine [M+H]⁺ ion at m/z 251.1375. The very low intensity fragment at 89.1064 observed in the Low Energy spectrum ( Figure 7B ) is diagnostic and corresponds to [M+H]⁺ of putrescine. The other fragments at m/z 145.0272, 135.0430, 117.0328 and 89.0379 came from the caffeoyl corresponding ion at m/z 163.0378. To conclude, HDMS^E fragmentation data are consistent with caffeoyl-putrescine as the second compound accumulated in N. benthamiana leaves. Caffeoyl-putrescine is a phenolamide synthesized from caffeoyl-CoA suggesting the formation of this intermediate in response to the overaccumulation of CA in transformed tobacco (Peng et al., 2016a; Cao et al., 2024). The accumulation of CA and caffeoyl-putrescine was consistent across biological replicates and correlated with a significant reduction in 5-CQA levels ( Figure 7D ). Although neochlorogenic acid (4-CQA), which is present in substantial amount in tobacco leaves, coelutes with 5-CQA in our experiments. Neither CiCQE1 nor CiCQE3 can use this isomer in vitro, as demonstrated in Supplementary Figure S5 . The observed reduction in chlorogenic acids level in tobacco is therefore only due to the consumption of 5-CQA. Regarding neochlorogenic acid (3-CQA), tobacco leaves also accumulated this compound, however, no decrease in its accumulation was observed following the expression of CiCQEs in accordance with in vitro experiments. These results in tobacco support a role for CiCQE1 and CiCQE3 in 5-CQA remobilization in planta.