Field-grown chicories R01 to R13, furnished by Florimond-Desprez, are a breeding material chosen for their variability in the STL content based on the dosage of STLs conducted in 2018 and 2019 (unpublished data). They have been cultivated in 2020 (Coutiches, France) for their roots by Florimond-Desprez SA (Cappelle-en-Pévèle, France). As chicory is allogamous and not able to form stable lineages, each chicory R01 to R13 is a population sharing a common gene pool. For the sensory analysis, each field-grown chicory was furnished as a sample originating from a pool of all roots of three field plots. ChicBitter002, is an in vitro-propagated industrial chicory clone, which genome has been sequenced by the Gentiane platform (INRAE, Clermont-Fermond, France) for Florimond Desprez SA (Cappelle-en-Pévèle, France). The clone ChicBitter002 was used in a wild-type form (Ctrl) or transformed by R. rhizogenes (HR lines).

Several CiGAS gene candidates were identified in the industrial chicory genome ChicBitter002 with Blastp or tBlastn algorithms using the germacrene A synthase short form gene (Uniprot accession: Q8LSC2) and the germacrene A synthase long form gene (Uniprot accession: Q8LSC3) from Bouwmeester et al. [30] as queries and with 26,367 ‘high probability’ predictive gene models from the Chicbitter002 sequence (unpublished database, Florimond Desprez SA, Cappelle-en-Pévèle, France) and 31,631 predictive gene models from a witloof chicory genome (Cargese program) or their complete genome sequences as subjects. Nucleotide sequences of significant hits were aligned to the Chicbitter002 sequence using GMAP and their positions were compared to corresponding positions of predictive gene models of recently sequenced Asteraceae, such as Lactuca sativa, Cynara cardunculus var. scolymus, Artemisia annua and transcript sequences from unpublished RNA-seq data in industrial and witloof chicories [31, 32, 33, 34]. The sequence of the CiGAO gene (Uniprot accession: D5JBW8) was used to identify candidates using the same procedure as for CiGAS genes. The CiGAS gene candidates were used as queries in a tBlastn procedure by comparing them to the BAC clone sequences previously published [16]. Multiple protein sequence alignment from the whole set of CiGAS sequences was carried out using MAFFT (Supplementary Fig. 1) [35].

Guide RNAs (sgRNAs) for CiGAS-short form and CiGAO genes were designed using the software CRISPOR (crispor.tefor.net) based on their GC content and their Doench score [36]. Two sgRNAs were defined for each gene except the gene CiGAS-S5 for which only one target could be defined (Supplementary Table 1). Potential off-targets were checked with the tool Cas-OFFINDER and verified on the chicory ChicBitter002 genome (unpublished sequence, Florimond Desprez SA, France) to avoid the disruption of off-target gene coding region [37]. Two binary expression vectors for CRISPR/Cas9 were constructed as described previously [25, 38]. Briefly, each sgRNA was inserted by digestion/ligation into an intermediate plasmid pKanCiU6-1p-sgRNAscaffold, which includes C. intybus U6-1p promoter and a sgRNA scaffold. Four vectors containing the cassette “CiU6-1p-Guide-sgRNAscaffold” were obtained. Adaptors for Golden Gate restriction/ligation method were added to the end of each cassette by PCR using PrimeStar HS polymerase (Takara Bio Europe, St-Germain-en-Laye, France) using primers GG1-F & GG1-R (Supplementary Table 1) for plasmid containing CiGAS_T1 and CiGAO_T1 guide and primers GG2-F & GG2-R (Supplementary Table 1) for plasmid containing CiGAS_T2 and CiGAO_T2 guide. PCR products were purified with NucleoSpin Gel and PCR Clean-up kits (Macherey-Nagel, Düren, Germany) and were cloned in pYLCRISPR/Cas9P${}_{35s}$-B (Addgene plasmid #66190, Cambridge, MA, USA) by Golden Gate restriction/ligation method. Final plasmids pYLCRISPR-GASshort containing CiGAS_T1 and CiGAS_T2, and pYLCRISPR-GAO containing CiGAO_T1 and CiGAO_T2 were obtained.

In vitro plants of C. intybus L. ChicBitter002 clone were used in this study. Plants were maintained on H10 solid medium as previously described [25]. Wild-type R. rhizogenes 15834 strains (provided by Marc Buée, INRAE, Nancy, France) were transformed with pYLCRISPR-GASsh or pYLCRISPR-GAO by electroporation. Wild-type (WT) strains or transformed strains of R. rhizogenes were used to generate hairy root (HR) lines from ChicBitter002 chicory. HRs were obtained and maintained in hormone-free media as previously described and spontaneous shoots regenerated from HRs were transfered in H10 medium [25]. A selection of HR mutant lines and control plants were grown in pots for a period of 3 months in a S2 greenhouse. Genomic DNA from HRs was extracted using NucleoSpin Plant II kit (Macherey-Nagel, Düren, Germany) using the manufacturer’s protocol. RolB specific primers were used to confirm R. rhizogenes infection, and Cas9 presence was checked by PCR using primer pair C9-F and C9-R (Supplementary Table 1).

Genotyping of HRs was realized to select mutant lines. Primer pairs GAO and S1 to S5 (Supplementary Table 1) labelled with different fluorescent dyes (6-Fam or Hex) were designed to amplify the sequences of GAO gene or each potentially expressed copy of CiGAS-short form gene. The PCR assay was performed in a volume of 15 µL containing 1X PCR buffer with 2 mM MgCl${}_2$, 100 µM of each dNTP, 0.2 mg·mL${}^{-1}$ BSA, 133 nM of each primer, 0.3 U Taq polymerase (Applied Biosystems, Foster City, CA, USA) and 2 ng of template DNA. A touchdown (TD) procedure was applied: 5 min denaturation at 94 °C, followed by (a) 5 cycles of 30 sec at 94 °C, TD 30 sec at 60 °C (–1 °C per cycle), 30 sec at 72 °C and (b) 35 cycles of 30 sec at 94 °C, 30 sec at 55 °C, 30 sec at 72 °C, before 10 min of extension at 72 °C. The PCR products were analyzed on a 3100 Avant Genetic Analyzer capillary sequencer (Applied Biosystems). Genotyping data were automatically collected and analyzed (GeneMapper Software v3.5, Applied Biosystems, Foster City, CA, USA). Sequences of mutated genes were obtained by Sanger sequencing using S1 to S5 primers pairs for CiGAS-short form genes and one GAO primer pair for CiGAO gene (GATC service, Eurofins Genomics, Ebersberg, Germany).

Roots of edited chicories (n = 3 for each HR lines), WT 3-month-old chicories (n = 4) and field-grown chicories (R01 to R13) taproots were freeze-dried and powdered. Hundred mg of powder were extracted with 1 mL of water (LC-MS grade) under agitation (1400 rpm; Eppendorf ThermoMixer C, Hamburg, Germany) for 15 min at 35 °C. The supernatants were collected by centrifugation (15 min, 4 °C, 21,000 g) and boiled for 10 min at 100 °C under agitation (600 rpm). All aqueous extracts were finally filtered through a 0.45 µm PP Whatman UNIFILTER microplate (Cytiva, VWR, Rosny-sous-Bois, France) and transferred to vials for LC-UV analysis. STLs content was determined using an Ultimate 3000RS system equipped with an LPG-3400RS pump, a WPS-3000TRS autosampler and a DAD-3000RS (Thermo Fisher Scientific, Waltham, MA, USA). Chromatographic separation was achieved on an Uptisphere Strategy PHC4 column (3 µm, 150 $\times$ 3 mm: Interchim, Montluçon, France) at 50 °C with 0.1% (v/v) formic acid in acetonitrile (B) and 0.1% (v/v) formic acid in water (A) as mobile phases at flow rate of 0.6 mL·min${}^{-1}$. The gradient elution was as follows: isocratic at 10% B for 2.5 min; linear from 2 to 30% B in 40 min; linear from 30 to 85% B in 1 min; isocratic at 85% B for 2 min; linear from 85 to 10% B in 1 min; and isocratic at 10% B for 4 min. The injection volume was 5.0 µL. Fifteen STLs compounds were profiled based on this method and determined by LC-QTOF-HRMS analysis. Total STLs content, in all the analyses, was expressed as the sum of the peak areas of these 15 compounds, present in the UV chromatogram at 254 nm.

STL compounds were identified using an Ultimate 3000RS system equipped with a DAD-3000RS (Thermo Fisher Scientific, Waltham, MA, USA), interfaced with a high-resolution quadrupole time-of-flight mass spectrometer and equipped with an ESI source (Impact II, Bruker Daltonik GmbH, Bremen, Germany). One gram of freeze-dried powdered chicory root from the pool of field-grown chicories (R01-R13) was extracted with 10 mL of water (LC-MS grade) under agitation (1000 rpm) for 15 min at 35 °C. After centrifugation (15 min, 4 °C, 21,000 g) , the supernatant was immediately evaporated, and the residue was resuspended in 2% acetonitrile (v/v) containing 0.1% formic acid (v/v) and passed through a 0.2 µm hydrophilic PTFE syringe filter (Merck, Darmstadt, Germany) and transferred to a vial for LC-MS analysis. Chromatographic separation was achieved on an Uptisphere Strategy PHC4 column (2.2 µm, 150 $\times$ 3 mm; Interchim, Montluçon, France) at 30 °C with 0.1% (v/v) formic acid in acetonitrile (B) and 0.1% (v/v) formic acid in water (A) as mobile phases at flow rate of 0.5 mL·min${}^{-1}$. The gradient elution was as follows: isocratic at 2% B for 1 min; linear from 2 to 50% B in 32 min; linear from 50 to 100% B in 5 min; isocratic at 100% B for 7 min; linear from 100 to 2% B in 5 min; and isocratic at 2% B for 7 min. The injection volume was 2.0 µL. The mass spectrometer was operated in the positive electrospray ionization mode at 4.5 kV with the following parameters: nebulizer gas (N${}_2$), 2.4 bars; dry gas (N${}_2$), 10.0 mL·min${}^{-1}$; dry temperature, 250 °C; collision cell energy, 10.0 eV; end plate offset, 500 V. Sodium formate cluster ions were applied for instrument mass calibration and for re-calibration of individual raw data files. The acquisition was performed in full scan mode in the 90–1200 m/z range with an acquisition rate of 1 Hz and with mass accuracy 5 ppm. Software Compass otofControl 4.1 (Mount Vernon, NY, USA) was used for the operation of the mass spectrometer and for data acquisition, and data were processed by the software Compass DataAnalysis 4.4 (Bruker Daltonik GmbH, Bremen, Germany).

The sensory evaluation was performed at the agrifood platform of Polytech Lille (University of Lille, France) in laboratory rooms according to ISO 8589:2007. The panel was constituted from volunteers recruited among the personnel of the University of Lille and trained in accordance with ISO 8586:2012. Panelists were first selected based on their ability to recognize bitterness, sweetness, saltiness, and sourness. Four series of different concentrations of taste solutions in water were prepared by dilution of stock solutions as followed: bitter with caffeine (concentrations: 0.27, 0.35, 0.46, 0.59 g·L${}^{-1}$), sweet with sucrose (concentrations: 12, 24, 48, 96 g·L${}^{-1}$), salty with sodium chloride (concentrations: 2, 4, 8, 16 g·L${}^{-1}$) and sour with citric acid (concentrations: 0.6, 1.2, 2.4, 4.8 g·L${}^{-1}$); the lowest dilutions corresponding to the perception threshold for the substance evaluated. Participants were first asked to identify the four tastes for each concentration, as well as pure water blank. Subsequently, they ranked the different concentrations using a 10 cm horizontal line. Each test was repeated twice. Fourteen panelists were retained and to familiarize themselves with the bitterness of chicory, additional training with solutions of free STLs was performed. The six solutions were prepared according to the concentration range of free STLs from 0.62 mM to 0.019 mM, from a serial dilution of chicory root extract enriched in free STLs [39]. The solutions, including a pure water solution as a blank, were presented in plastic glasses accompanied with a plastic drop-pipette to pour solution into the mouth. Participants were invited to taste 7 mL of each dilution with two replications. They were asked to rate the different solutions according to the bitterness intensity using a 7-point ranking; the rank 1 corresponding to the least intense perceived concentration. At the end, twelve panelists were selected for the quantitative sensory evaluation. Finally, freeze-dried root powders from field-grown chicories (R01 to R13) were evaluated by the panel for their bitterness intensity according to ISO 6658:2017. The samples were served on plastic cups labeled with random three-digit numbers. The panelists were asked to rank the samples using a 10 cm horizontal line and evaluation was repeated twice in two sessions. The bitterness reference scale was constructed based on a simple linear regression model plotting bitter taste in function of relative total STL content. The bitterness degree of the samples from the roots of edited-chicory plants was then predicted by fitting their total STL content by means of the regression curve equation.

An ANOVA and Student’s t-test were performed to determined panelist performance (discrimination ability, consensus in notation, repeatability) and to test for bitterness differences between the chicory root samples from the field-grown chicories (R01-R13). Comparison of STL content of edited and WT 3-month-old chicories and assessment of the differences in panelist perception of bitter taster were tested by non-parametric ANOVA (Kruskal-Wallis) followed by post-hoc uncorrected Dunn’s test or post-hoc paired Wilcoxon, respectively. The assumptions of normality, homogeneity of variance, and independence were checked by Shapiro-Wilk test, Bartlett test, and Durbin-Watson test, respectively. The bitterness correlation with STL content was established by Pearson’s correlation test. The significance threshold was set to 0.05 for all statistical tests. All analyses were done with R Statistics 4.0.3 (r-project.org) and GraphPad Prism 9 software (Windows Inc., San Diego, CA, USA).

The genome of the industrial chicory ChicBitter002 was searched to identify CiGAS and CiGAO genes putatively involved in the STL pathway. A single copy of CiGAO gene was highlighted in the genome of ChicBitter002 (Supplementary Fig. 2). For CiGAS genes, quite similar data to Bogdanović et al. [16] were found with one gene for CiGAS-long form and five gene copies of CiGAS-short form. Sequence analysis confirmed the existence of GAS-S1, GAS-S2, GAS-S3 and CiGAS-S4b but CiGAS-S4a was not found. However a new CiGAS-short form candidate copy, named CiGAS-S5, was highlighted (Fig. 1). Sequence analysis confirmed that copies CiGAS-S1, CiGAS-S2, CiGAS-S3 and CiGAS-S5 could be expressed with a complete 7-exons structure and that copy CiGAS-S4b has a retrotransposon into the exon 3 that could hamper its expression. The five copies were identified at two different locations in the genome. On the one hand, CiGAS-S1, S2 and S3 were located on the same contig (contig_403_pilon) so they can be considered as three distinct genes but physically close on the chicory genome. Indeed, the CiGAS-S1 and the CiGAS-S3 are separated by ~85,000 bp, with CiGAS-S2 in between (Supplementary Fig. 3). It can also be noted that CiGAO gene is present on the same contig as the three CiGAS-short form genes, and ~85,000 bp away from the CiGAS-S3 copy (Supplementary Fig. 3). On the other hand, CiGAS-S4b and S5 also shared the same contig (contig_1740_pilon) and are ~76,000 bp apart. Based on these results, 4 copies of CiGAS-short can be considered as putatively expressed : copies S1, S2, S3 and S5 (Fig. 1). Since the copy S4b was presumably not expressed due to the presence of a retrotransposon, it was not targeted.

Fig. 1.

Schematic representation of the structure for CiGAS and CiGAO genes. GAS-S stands for the CiGAS-short form and GAS-LF stands for the CiGAS-long form. The exons and introns lengths are proportional to the nucleotide sequences of the genes. Colored box: exon, line: intron, red rectangle: retrotransposon.

The objective was to perform a KO of the 4 putatively expressed copies of CiGAS-short form and of the CiGAO gene. In order to design sgRNAs for the four putatively expressed copies of CiGAS-short form, the homology of these genes was investigated by comparing their exonic sequences using ClustalOmega tool [40]. All the putatively expressed copies (CiGAS-S1 to S5) are relatively close since they share between 86 to 97% identity. CiGAS-S1 and CiGAS-S2 sharing the highest identity (Supplementary Fig. 4). The newly identified CiGAS-S5 copy shares 86–93% identity with the other copies, confirming its link to the multiple copies of CiGAS-short form (Supplementary Fig. 4). The high level of homology between the potentially expressed copies of CiGAS-short form did not allow specific targeting of each copy, but it was possible to design sgRNAs in the third exon to target all copies simultaneously. Since a single copy of CiGAO was identified, two sgRNAs were designed at the beginning of the first exon in order to disrupt the gene function.

The use of the CRISPOR tool to determine the potential targets gave a list of sgRNAs that were then analyzed using the Cas-OFFINDER tool to search for potential off-targets. The results were checked on the ChicBitter002 genome, to avoid selecting sgRNAs with off-targets in gene coding regions. Finally, two sgRNAs were identified in the first exon of CiGAO and two sgRNAs were identified in the third exon of CiGAS-S1, CiGAS-S2 and CiGAS-S3. For the CiGAS-S5 copy, a single sgRNA common to the other copies was defined (Supplementary Table 1). The GAS-long gene was not targeted because it has been demonstrated to have minimal impact on STLs production in chicory roots [28].

Chicory leaves of 14-day-old vitroplants were infected with R. rhizogenes strain 15834 transformed with the binary vectors pYLCRISPR-GASshort or pYLCRISPR-GAO. Two weeks after R. rhizogenes transformation, selection of HRs based on their phenotype was carried out. Eighty-five HR lines were collected with 54 lines potentially transformed with the vector targeting the CiGAS-short form genes and 31 lines potentially transformed with the vector targeting the CiGAO gene. After verification of the presence of the T-DNA of R. rhizogenes and the integration of the binary vector, the number of transformed lines was reduced to 43 for CiGAS-short and 30 for CiGAO (Table 1). As integration of the binary vector is not always synonymous of mutation, a preliminary sorting of the mutants was performed by genotyping. Indeed, the action of CRISPR/Cas9 can generate a variety of mutated alleles and because multiple copies of CiGAS-short were highlighted, it is possible that all copies are not mutated at the same time. Then, a Sanger sequencing of the mutant HR lines identified by genotyping was performed to obtain the sequences of the mutants. Out of the 43 lines transformed with pYLCRISPR-GASshort, analysis of the mutations revealed 6 lines with one copy of CiGAS gene mutated, 4 lines with two mutated copies and 3 lines with three mutated copies, resulting in a mutation rate of 30.2% (Table 1). HRs transformed with pYLCRISPR-GAO showed a mutation rate of 20% with 6 mutated lines out of 30 transformed lines (Table 1).

As the chicory genome is diploid, the CRISPR/Cas9 system can induce two types of mutations. On the one hand, a monoallelic mutation could occur. In such case, one allele is mutated and the other remains wild-type (heterozygous) leading to an absence of full KO of the targeted gene. On the other hand, editing could generate a biallelic mutation where both alleles are mutated, either with the same mutation on each allele (homozygous), or with a different mutation (also heterozygous) that can lead to a KO of the gene (Table 2 and Fig. 2). Most of the identified mutations were small insertions of one nucleotide (HR1, HR2, HR4, HR5, HR6, HR8, HR9, HR10, HR12, HR13, HR14, HR16, or HR18) or small deletions of less than 10 nucleotides (HR1, HR2, HR5, HR6, HR7, HR9, HR10, HR11, HR13, HR15, HR17, HR18, or HR19), but in some case, larger deletions were found between the two targets T1 and T2 as shown for HR4, HR7, HR8, HR12, HR12*, HR15 and HR17 (Fig. 2). Some mutations cause a change in the coding frame leading to the KO of a gene or a gene copy and the potentially premature termination of protein translation. This type of event was observed for 5 HR lines mutated on CiGAS-short genes: 3 lines mutated on a single copy (HR1, HR2 and HR4), 1 line mutated on two copies (HR8) and 1 line mutated on 3 copies (HR12*); and in 2 HR lines mutated on CiGAO (HR18 and HR19) resulting in a gene KO frequency of 11.6% and 6.6% respectively (Tables 1,2). The R. rhizogenes-mediated transformation is a stable transformation, meaning that the T-DNA is inserted into the plant genome. As a result, the Cas9 gene is integrated into the genome of HR lines and can continue to exert its action and cause mutations. This case can be observed for the lines HR12 and HR12*. Originally, HR12 had a mutation on both CiGAS-S1 and CiGAS-S2copies but after a few months, an additional mutation appeared on the CiGAS-S5 copy.

Fig. 2.

Examples of sequences obtained by CRISPR/Cas9 editing. (A) Sequence analysis of the two alleles of hairy root lines transformed with pYLCRISPR-GASshort construct. (B) Sequence analysis of the two alleles of hairy root lines transformed with pYLCRISPR-GAO construct. The target sequences are in blue, and the PAM sequences are underlined. Mutations are highlighted in red and the changes in the nucleotide sequences are shown on the right of each allele. HR.x.x, Hairy Root x allele x; WT, Wild-Type; T1, Target 1; T2, Target 2; +, insertion; -, deletion.

For the rest of the paper, we will focus on only few mutants (HR2, HR3, HR9, HR12, HR12*, HR16 and HR18) to see which mutation event can affect the STL production.

Several methods to extract sesquiterpene lactones from chicory roots have been described using various solvents, but a method more representative of physiological conditions of human consumption was desired, since the objective of our project is to use industrial chicory root as functional eaten ingredient. Using water, an aqueous extract rich in sesquiterpene lactones was obtained with 15 compounds detected at a wavelength of 254 nm in the WT chicory ChicBitter002: lactucin, 8-deoxylactucin, lactucopicrin and their 11(S),13-dihydro derivatives, in addition of their oxalated forms and some glycosylated forms (Fig. 3). The same extraction method was used for the field-grown chicories used in sensory analysis and their STL composition was compared to that of the WT chicory ChicBitter002 “Ctrl” grown under controlled conditions (in vitro then in greenhouse). Fifteen identified compounds are present in all chicories, but a quantitative disparity was observed between the field-grown chicories (R01-R13) and the chicory “Ctrl” (Fig. 4A). As a general trend, the lactucin-15-oxalate and 8-deoxylactucin-15-glycoside content appeared to be higher in “Ctrl” plants compared with the field-grown chicories, whereas the different forms of 11(S),13-dihydrolactucin were in lower quantities. These differences could be due to the growing conditions or genotypic. However, if the distribution of STL content was examined according to the lactucin-like, 8-deoxylactucin-like, and lactucopicrin-like groups, the difference between chicories would fade, as would the total STL content (Fig. 4B). In the following analyses, the STL content of the chicories were therefore compared by considering their total STLs obtained by the sum of the 15 compounds detected in Fig. 3. This is based on the hypothesis that taking into account all STLs is a good indicator of bitterness. Indeed, the different studies on chicory have not clearly established a compound more involved than another in bitterness and when consuming chicory-based products, the consumer doest not taste the STLs separately in his mouth but perceives them as a whole and detects a bitter taste.

Fig. 3.

HPLC chromatogram at 254 nm of sesquiterpene lactones in root of the wild-type chicory ChicBitter002. (1) 11(S),13-dihydrolactucin-15-glycoside (DHLc-gly); (2) 11(S),13-dihydrolactucin-15-oxalate (DHLc-ox); (3) 11(S),13-dihydrolactucin (DHLc); (4) Lactucin-15-oxalate (Lc-ox); (5) Lactucin (Lc); (6) 8-deoxylactucin-15-glycoside (dLc-gly); (7) 11(S),13-dihydro-8-deoxylactucin-glycoside (DHdLc-gly); (8) 8-deoxylactucin-15-oxalate (dLc-ox); (9) 11(S),13-dihydro-8-deoxylactucin-15-oxalate (DHdLc-ox); (10) 8-deoxylactucin (dLc); (11) 11(S),13-dihydro-8-deoxylactucin (DHdLc); (12) 11(S),13-dihydrolactucopicrin-15-oxalate (DHLp-ox); (13) Lactucopicrin-15-oxalate (Lp-ox); (14) 11(S),13-dihydrolactucopicrin (DHLp); (15) Lactucopicrin (Lp). Identity of the molecules was confirmed by mass spectrometry.

Fig. 4.

STL content in roots of chicories according their peak area (mUA) obtained by HPLC analysis. (A) represents the distribution of each STL into the roots of chicories. (B) represents the STL levels of each chicories according to their structural group (Lc-like, dLc-like or Lp-like) and the total STL content of them. R01 to R13 are field-grown chicories used for sensorial analysis. Ctrl is wild-type ChicBitter002 grown under controlled condition. DHLc-gly, 11(S),13-dihydrolactucin-15-glycoside; DHLc-ox, 11(S),13-dihydrolactucin-15-oxalate; DHLc, 11(S),13-dihydrolactucin; Lc-ox, Lactucin-15-oxalate; Lc, Lactucin; dLc-gly, 8-deoxylactucin-15-glycoside; DHdLc-gly, 11(S),13-dihydro-8-deoxylactucin-glycoside; dLc-ox, 8-deoxylactucin-15-oxalate; DHdLc-ox, 11(S),13-dihydro-8-deoxylactucin-15-oxalate; dLc, 8-deoxylactucin; DHdLc, 11(S),13-dihydro-8-deoxylactucin; DHLp-ox, 11(S),13-dihydrolactucopicrin-15-oxalate; Lp-ox, Lactucopicrin-15-oxalate; DHLp, 11(S),13-dihydrolactucopicrin; Lp, Lactucopicrin. Lc-like, sum of all DHLc and Lc forms; dLc-like, sum of all DHdLc and dLc forms; Lp-like, sum of all DHLp and Lp forms; Total, sum of the 15 identified STLs.

A particularity of chicory hairy root lines is that they are able to regenerate spontaneous shoots. Thus, from the mutated hairy root lines we were able to regenerate whole plants that were grown in a greenhouse for 3 months. Applying the previously described extraction method, 3-month-old edited chicory roots (n = 3 for each HR lines selected) were analyzed to evaluate the impact of the CRISPR/Cas9 mutation on STL accumulation.

All the analyzed plants were derived from ChicBitter002 clone, with “Ctrl” corresponding to WT plants and “Ctrl_HR” corresponding to R. rhizogenes-infected chicory plants with no pYLCRISPR/Cas9P${}_{35s}$-B or binary vector pYLCRISPR-GASshort or pYLCRISPR-GAO. Comparison of “Ctrl” and “Ctrl_HR” showed an increase in STL content for R. rhizogenes-infected plants (Fig. 5). For the rest of the analysis, the genome-edited chicory lines were compared to “Ctrl_HR”. Seven edited-chicory lines were selected for their various mutation events (monoallelic or biallelic, one copy or several mutated copies). Determination of their STL content was carried out to represent the effect of each targeted gene on STL content.

Fig. 5.

Total sesquiterpene lactone content of roots of edited chicory regenerated from HR lines by HPLC. The peak areas (mUA) of STLs in the roots of 7 genome edited chicory regenerated from HR lines were analyzed by HPLC and compared to WT chicory lines (Ctrl and Ctrl_HR). Gray bars correspond to control chicory lines, blue bars to CiGAS-short edited lines and orange bars to CiGAO mutants. The histogram shows the modulation of total STL content (sum of all identified STLs) as a function of mutation on chicory lines. HR12 and HR12* are originally the same hairy root line except that HR12 is mutated for only CiGAS-S1 and CiGAS-S2 and HR12* is mutated for CiGAS-S1, CiGAS-S2 and CiGAS-S5. The letters indicate significantly different groups according to non-parametric one-way ANOVA and Dunn’s post-hoc test (p 0.05).

For the CiGAS-short form mutants, 5 lines have been selected: line HR3, a monoallelic mutant which carries a mutation in a single allele of CiGAS-S1; line HR2, a heterozygous biallelic mutant for CiGAS-S1 where editing events induce a change in the coding frame; line HR9, a monoallelic mutant for both CiGAS-S1 and CiGAS-S2, meaning that only one allele is mutated in each of the two genes; line HR12, a heterozygous biallelic mutant for CiGAS-S1 and a homozygous biallelic mutant for CiGAS-S2, meaning there is a change in the coding frame in both of these gene copies (Supplementary Figs. 5,6), and line HR12* which is the same as line HR12 except that there is an additional homozygous biallelic mutation for CiGAS-S5 (Supplementary Figs. 5–7). For the CiGAO mutants, the 2 selected lines were: line HR16, a monoallelic mutant and line HR18, a heterozygous biallelic mutant with a change of the coding frame of its gene (Supplementary Fig. 8). The nucleotide sequences of these lines are described in Fig. 2.

By comparing the total STL content of all these lines with the “Crtl_HR”, only three lines showed significative reduction in STL content: HR12, HR12* and HR18. All three share biallelic mutations. All other CiGAS-short and CiGAO edited lines showed no significant difference in total STL content (Fig. 5). Most are monoallelic mutants, and one is a biallelic mutant only for one copy of the GAS gene (CiGAS-S1). Analysis of the STL content of a chicory edited for both alleles of CiGAS-S2 with a change in the coding frame was also performed and showed no significant difference, but we did not have enough biological replicates to include this result in our data. Regarding the CiGAS-short form mutants, both the HR12 and HR12* edited line showed similar profiles where the total STL content was strongly reduced (Fig. 5). This trend was confirmed in total free forms and total oxalate forms of STLs (Supplementary Fig. 9).

All these data suggest that the two copies (S1 and S2) of CiGAS-short are important for STL production in our plants and must be simultaneously KO to significatively reduce the STL content. The CiGAO gene also plays an important role in STL production, but the impact was not as great as for the two copies of CiGAS-short.

Sensory analysis was conducted on the root powder of thirteen field-grown chicory plants (R01 to R13). Each chicory was assigned a bitterness score from 0 to 10, and panelists determined that R09 and R06 were the most bitter chicories (score of 7.83 and 6.13, respectively) and R04 and R13 were perceived as the least bitter chicories (score of 2.21 and 2.49, respectively) (Fig. 6, Supplementary Table 2). First, a correlation between bitterness and each STL of the field-grown chicories was sought to assess whether a single or multiple compounds are involved in bitterness. Of the fifteen STLs identified in this paper, six were not correlated with bitterness: DHLc-gly, DHLc-ox, DHLc, Lc-ox, dLc-ox and, DHLp-ox (Supplementary Table 3). The remaining nine were more or less correlated with an Pearson’s r ranging from 0.5587 to 0.9223 (Supplementary Table 3). Since bitterness cannot be attributed to a single compound, it is assumed that all STLs can be considered indicative of chicory bitterness. Next, the total STL content of each tasted chicory was assessed and a linear regression analysis was carried out to estimate whether the STL content can be correlated with the bitterness score (Fig. 6). A Pearson correlation allowed to establish a positive correlation (R${}^2$ = 0.46) between the perception of bitterness and the total STL content of chicories, with r = 0.68 (two-tailed p-value = 0.0103). Finally, using the resulting linear regression equation, the seven hairy roots edited by CRISPR/Cas9 for CiGAS-short and CiGAO were assigned theoretical bitterness scores based on their STL content and in ordert to project the hairy root mutants onto the linear regression curve, it was artificially elongated. These data were plotted in Fig. 6. Based on its STL content and its theoretical bitterness score (5.06), the “Ctrl” sample would appear to be more bitter than most field-grown chicories, except for R06 and R09 (Fig. 6). The projection on the correlation curve of the sample “Ctrl_HR”, which has a higher STL content than “Ctrl”, seems to indicate a high bitterness score (10.10). Therefore, the plant material generated by R. rhizogenes infection can initially be considered as very bitter. For the previously analyzed CRISPR/Cas9 edited-chicory lines, most of the theoretical bitterness scores obtained were in the range from 6.75 to 14.41, except the lines HR12, HR12* and HR18, which have theoretical bitterness scores of 1.29, 1.48 and 4.26, respectively (Fig. 6, Supplementary Table 2). The lines with the lowest theoretical scores are biallelic mutants for either CiGAS-S1 and CiGAS-S2, or CiGAS-S1, CiGAS-S2 and CiGAS-S5, or CiGAO, and had low STL contents compared to the “Ctrl_HR” line. The other lines correspond to mutant lines whose STL content did not differ significantly from the “Ctrl_HR” line.

Fig. 6.

Relationship between the total STL content and the bitterness score of chicories. Positive correlation (r = 0.68) between these two parameters was established using thirteen field-grown chicories (in black). CRISPR/Cas9 hairy root mutants (in red) were projected onto the linear regression curve using the equation Y = 0.04662 $\times$ X + 0.3942, where X was the total STL content of the samples. The significance of these results was confirmed by Pearson’s correlation test (two-tailed p-value = 0.0103).

It can be concluded that a complete KO of CiGAS-S1 and CiGAS-S2 genes could be sufficient to cause a drastic decrease in STL content that would be reflected in the bitterness score. HR12 and HR12* edited chicory lines almost lost their bitterness with a diminution of 86.35 $\pm$ 0.9% in their theoretical bitterness scores compared to “Ctrl_HR”. A full KO of the CiGAO gene is also responsible for a low bitterness score since it is 57.8% lower than the control “Ctrl_HR”.