Total RNA was isolated from Cucurbita pepo cultivar golden glory and Cucumis sativus var. beit alpha ten-day-old roots using Spectrum Plant Total RNA Kit (Sigma-Aldrich). First-strand cDNA was synthesized from total RNA of roots using the GoScriptTM Reverse transcriptase kit (Promega) following the manufacture’s instructions.

Squash orthologs of genes in the cucurbitacin cluster and pathway were searched for in the Cucurbit Genomics Database (http://cucurbitgenomics.org/) by BLASTP analysis using cucumber sequences as query. Annotations of the genomic region were corrected using FgenesH gene-finder (Solovyev et al., 2006).

Primers designed for cloning candidate genes are given in Table S1 . Amplification was done by polymerase chain reaction (PCR) in a SensoQuest Labcycler using homemade Phusion X7 polymerase. The following program was applied: 95°C for 1 min; 30 cycles of 95°C for 20 s, 50°C for 30 s, 72°C for 1 min and 30 s; 72°C for 5 min (elongation time for CpCPQ was increased to 2 min). Gel purified amplicons were ligated to the respective cloning system: i) Gateway cloned into pDONR 207 (Invitrogen) and subsequently to the pJCV51 binary vector (VIB-UGent Center for Plant Systems Biology, Vector ID: 4_60) ii) the USER modified binary vector pEAQ : HT-DEST (ref) by USER™ Enzyme (New England Biolabs) iii) pJET 1.2/blunt vector from CloneJet PCR cloning kit (Thermo Fisher Scientific). Ligation reactions were transformed into E. coli competent cells using the heat shock method and successful constructs were retrieved 16 hours after. Sequences of constructs were confirmed by Sanger sequencing service (Macrogen, Europe). PJCV51 empty vector from Dong et al. (2018) was used in this study.

Two uracil-specific excision-reagent (USER) vectors with uracil, tryptophan, leucine, or histidine as auxotrophic markers were used for genomic integration. The expression plasmids pESC-URA-USER (pCfB132), pESC-LEU-USER (pCfB220), and pESC-HIS-USER (pCfB291) were gift from Prof. Irina Borodina (Technical University of Denmark, Kongens Lyngby, Denmark) and pESC-TRP-USER (pWUS) was constructed by replacement of the URA3 gene with TRP1 on pCfB132. In addition, the synthetic fragments bear flanking regions containing specific restriction sites and a generic sequence compatible for USER cloning. By applying the USER cloning technique, one or two genes of interest in a uni- or bidirectional promoter of choice were simultaneously integrated into the backbone vectors by specific annealing of the overhang created between the PCR parts and digested USER cassette (Geu-Flores et al., 2007). The CsCPR and Cp P450 genes were amplified and cloned by USER cloning into the yeast expression vectors using primers listed in Table S1 to introduce compatible USER overhangs with the corresponding promoter part ( Table S2 ) and the digested AsiSI/Nb.BsmI USER cassette of the pESC-URA/LEU/HIS/TRP-USER backbone.

Yeast integrative vectors were constructed in Escherichia coli cells (E. cloni^®; Lucigen) utilizing USER cloning technique. Isolated ORFs of squash genes in the cucurbitacin cluster and pathway and S. cerevisiae tHMG1, ERG1, ERG12 and ERG20 were amplified by PCR using primers given in Table S1 . Three integrative vectors (pAS1_X-3, pAS2, pAS3_X-3) ( Table S2 ) were used to insert Oxidosqualene cyclases (OSCs) and terpene pathway boosting genes on site 3 of chromosome X, a region known for high gene expression with low to no impact on growth rate (Mikkelsen et al., 2012). Empty integration vectors were kindly provided by Victor Forman, University of Copenhagen, Denmark. pAS1_X-3 contains an ampicillin resistance gene (AmpR), a Kluyveromyces lactis URA3 gene and a region homologous to the insertion site, X3-DOWN, and a homologous region to pAS2. pAS2 contains homologous regions to pAS1_X-3 and pAS3_X-3. pAS3_X-3 contains homologous regions to pAS2 and the integration site X3-UP. PCR fragments were USER cloned into AsiSI/Nb.BsmI digested empty vectors with PCR amplified GAL1/GAL10 promoters from S. cerevisiae, S. mikae and S. arboricola, to form pInt1-X ( Table S2 ). All generated vectors were confirmed by sequencing.

Prior to yeast transformation, integrative vectors were linearized by digestion with NotI according to manufacturer’s protocol.

A list of the plant species used and their providers can be found in Table S3 .

Seeds for Cucurbita pepo cultivar golden glory and Iberis umbellata were first sterilized in a 50-ml falcon tube by washing in 30 mL of deionized water with 300 µL of Mild Cream Soap (ABENA, Denmark) and shaken for 20 min. Seeds were then rinsed in 70% ethanol for 1 min. Afterwards, seeds were submerged in 1.5% NaOCl and shaken for no more than 15 min. Subsequently, the seeds were rinsed in sterile water 6 times. The wet seeds were placed on a sterile petri dish in the sterile bench to surface dry. When the seeds were surface dried, seeds were placed on plates with ½ Murashige and Skoog media (MS) agar supplemented with 3% glucose sealed with parafilm (Merck KGaA, Darmstadt, Germany) and placed in a climate chamber. The climate chamber was from Fitotron type SGC120-H from 2016 (Weiss Technik UK Ltd., Epinal Way, Loughborough, United Kingdom) and the settings were: light at 79 uE, 25°C, 80% humidity, 16-hour photoperiod.

Rizhobium rhizogenes strain LBA9402 was transformed with pJCV51, pJCV51:CpCUCbH1 through electroporation and recovered on S.O.C medium for 2 hours at 28°C before being plated on YEB agar supplemented with spectinomycin (50 µg/ml) and rifampicin (50 µg/ml). A single colony was selected for each transformation and grown overnight in liquid YEB supplemented with spectinomycin and rifampicin at 28°C under constant agitation (180 rpm). Subsequently, 50 μL of the R. rhizogenes suspension was placed on solid YEB medium supplemented with spectinomycin and rifampicin, and grown for two days at 28°C and colonies resuspended into PS buffer [10 mM PIPES (Sigma-Aldrich P8203)/KOH pH 6.8, 200 mM Sorbitol] supplemented with acetosyringone to a concentration of 100 μM. Finally the OD₆₀₀ of the suspension was adjusted to 0.4.

For squash, cotyledons were transformed with R. rhizogenes ten days after germination. Cotyledons were inoculated by bruising with a sterile syringe needle dipped in the R. rhizogenes suspension. Four incisions were cut on the abaxial side of the cotyledon perpendicular to the central vein and where placed with the abaxial side down onto ½ MS (3% glucose) agar plates without antibiotics and incubated in the dark for two days. Afterwards, the inoculated cotyledons were transferred to ½ MS (3% glucose) agar plates supplemented with cefotaxim (500 µg/ml), carbenicillin (250 µg/ml) and kanamycin (50 µg/ml) and incubated under light for a week, cotyledons were subsequently transferred to new plates where cefotaxim and carbenicillin concentrations were halfed. Transformed hairy roots expressing mRFP would start to emerge after three weeks of tissue culture. These roots were excised from leaves and subcultured every two weeks with cefotaxim and carbenicillin and kanamycin. The concentrations of cefotaxim and carbenicillin were reduced 50% in each subculturing step until no antibiotic was used, nevertheless kanamycin concentration remained constant.

In the case of I. umbellata optimal age for transformation was optimized to be 5 weeks after germination for leaves and stems. For the leaves a similar procedure to that of squash was used, but stems were inoculated by cutting them transversally with a sterile razor blade dipped in the R. rhizogenes suspension. Subsequent cultivation were performed the same way as described above for squash cotyledons. All incubation periods were done in the climate chamber.

Squash HRs were grown 2 weeks in MS agar plates before being transferred to temporary immersion bioreactors (CIRAD Ltd., France) which were previoiusly sterilized by autoclaving at 121°C for 15 min. The TIR were filled with 400 ml of ½ MS media, HRs were flooded for intervals of 30 min every 8 hours for a period of 14 days. The flow rate of inlet air was 60 l/h.

For MeJ treatment, a 100mM stock solution of MeJ of 95% purity (Sigma Aldrich) was prepared in 96% ethanol and filter sterilized. After incubating HRs for 14 days in TIR, 400 µl of the MeJA stock solution was added to make a ∼100 µM MeJA treatment. In the same manner, 30 µl of MeJA sterilized stock solution were pipetted in the ½ MS agar near 14-day-old HR (square plates typically had 30 ml of agar). Samples were collected after 24 hours of adding MeJA, frozen in liquid nitrogen and stored at -70°C until further processing. All experiments in TIRs were carried out in triplicates and statistical significance was tested using one-way ANOVA (p < 0.05).

pEAQ-HT-DEST expression vector (Sainsbury et al., 2009) constructs harboring eGFP and CpCUCbH1 (described above) were transformed into Rhizobium (Agrobacterium) tumefaciens strain AGL1. Colonies of A. tumefaciens were picked in the morning and precultured in 5 ml of Luria-Bertani (LB) media supplemented with kanamycin (50 µg/ml). Afterward, 12 ml of LB media containing kanamycin was inoculated with 50 µL of the preculture and incubated at 28°C overnight without allowing OD₆₀₀ to exceed 1.5. The cultures were subsequently centrifuged at 3000g, the resulting cell pellet resuspended in infiltration buffer (10 mM MgCl2, 10 mM MES, pH 5.6, and 100 µM acetosyringone), and the OD₆₀₀ was adjusted to 0.4. The suspension was incubated for 1 h on a shaker at room temperature. Age of infiltration for tissues of each species was optimized: for cotyledons of squash and cucumber were infiltrated 5 days after germination and examined 5 days after infiltration, whereas cotyledons from Luffa were infiltrated 7 days after germination and examined 7 days after infiltration. In the case of E. elaterium leaves of 2-month-old plants were infiltrated and examined 5 days after infiltration, while for Nicotiana benthamiana (tabacco) leaves were infiltrated when plants were 5-weeks old and harverted 5 days after infiltration. For infiltration experiments plants were grown in soil (Pindstrup substrate no. 2) in a glasshouse with a 16-h day at 28°C and an 8-h night at 28°C. Infiltrated tissue for confocal microscopy was harvested the same day while for metabolite analysis tissues were frozen in liquid N₂ and stored at −70°C until further processing. All experiments were carried out in triplicates and statistical significance was tested using one-way ANOVA (p < 0.05).

Gel electrophoresis confirmed that the primers designed for qRT-PCR ( Table S4 ) amplify only one amplicon. Each tissue was analyzed in triplicate, synthesizing cDNA from 1 µg of total RNA using the GoScript™ Reverse transcriptase kit (Promega), and final volume was adjusted to 10 ng/µl. All quantitative PCRs for the three primer sets were performed in the CFX384 real-time system (Bio-Rad) under the following conditions: 4 µL of PowerUp SYBR Green Master Mix (Thermo Fisher Scientific), 1 µL of forward primer (2 µm), 1 µL of reverse primer (2 µm), 1 µL of cDNA (10 ng/µl), and 1 µl of MilliQ water. The PCR conditions were initial 95°C for 5 min, followed by 40 cycles of 95°C for 15 s, 50°C for 15 s, and 72°C for 30 s.

For visualization of in planta transient expression, confocal Laser Scanning Microscopy (CLSM) was done essentially as previously described (Laursen et al., 2016). Briefly, leaf discs from infiltrated tissue described above were excised, mounted in water, and observed by CLSM. Cell imaging was performed using an SP5x laser scanning confocal microscope equipped with a DM6000 microscope (Leica, Germany). Images were recorded using a 63x water immersion objective lens. Excitation/emission wavelengths were 488/500–550 nm for eGFP. The images were sequentially acquired and processed using the LAS X software (Leica).

The yeast strain EGY48 (Thomas and Rothstein, 1989; Ellerström et al., 1992; Ignea et al., 2011) was transformed with yeast integrative vectors using the lithium acetate method (Gietz and Schiestl, 2007), generating strains EGY48-EV, EGY48-CpCPQ. Subsequently, EGY48-CpCPQ was co-transformed with plasmids pCfB220 containing the Cucumis sativus Cytochrome P450 Reductase (pCfB220:CsCPR) and pCfB291 containing one of the following P450s: CpCYP88L4, CYP88L2, CYP88L7, CpCYP87D19, CpCYP89A140; a full list of the strains generated is presented in Table S4 and the description of the plasmids used is in Table S2 . Colonies following transformation were genotyped for correct integration using PCR. For triterpenoid production, pre-cultures of the strains were grown overnight until saturation in 10 ml synthetic defined media containing 2% glucose. Then, the cells were washed twice with water and heterologous genes induced by inoculation into 10 mL synthetic defined media with 4% galactose and 2% raffinose with or without 10 mM β-methyl cyclodextrins and grown for 96 h at 30°C and 150 rpm.

Protein-coding sequences were deduced using the Expasy translate tool (http://web.expasy.org/translate/) and manually curated. Models for the evolutionary histories for the bHLH TFs, CYP88s and ACT protein sequences were inferred separately by using the maximum-likelihood method based on the JTT matrix-based model (Jones et al., 1992). The tree with the highest log likelihood was selected (−4579.11 for the bHLH, - 6,515.95 for the CYP88 and −8,676.26 for the ACT trees, respectively). Initial trees for the heuristic search were obtained by applying the neighbor-joining method to a matrix of pairwise distances estimated using a JTT model. The trees are drawn to scale, with branch lengths measured in the number of substitutions per site. The analysis involved 21, 11 and 14 amino acid sequences for the bHLH TFs, CYP88s and ACT trees, respectively. All positions containing gaps and missing data were eliminated. In total, there were 187, 434 and 366 positions in the final data sets for the bHLH, CYP88 and ACT trees, respectively. The statistical significance of each node was tested by the bootstrap method using 1,000 iterations. The evolutionary analyses were conducted in MEGA7 (Kumar et al., 2016). Accession numbers for sequences used in phylogenetic trees are given in Table S5 .

Strains EGY48-CpCPQ+CsCPR+CYP88L2 and EGY48-CpCPQ+CsCPR+CYP88L2 were grown overnight until saturation in 10 ml synthetic defined media containing 2% glucose. Then, the cells were washed twice with water and heterologous genes induced by inoculation into 10 mL synthetic defined media with 4% galactose and 2% raffinose with 10 mM β-methyl cyclodextrins. In addition, 100 µl of either 1mM Cucurbitacin B, D, E or I were added to the cultures and grown for 96 h at 30°C and 150 rpm. Additionally, production of either Cuc B or E by feeding Cuc D or I to a cultures of strain EGY48-CpACT1 grown in a similar fashion as described above was used as controls, to make sure that feeding of yeast with cucurbitacins was indeed possible. All reactions were carried out in triplicates.

WinRHIZO, a root image analysis software (Arsenault et al., 1995) was used for determination of primary root length, number of root tips and total root length phenotypes of HR lines. Pictures for root phenotyping were taken from HRs in agar plates 5 days after sub-culturing using a Nikon D5600 DSLR camara with an AF-P DX 18-55 VR lens (Nikon, Japan).

For GC-MS analysis, 50 µL of extract was aliquoted into a glass insert and evaporated under vacuum. The glass inserts were sealed with air-tight magnetic lids into GC-MS vials and derivatized by the addition of 30 μL of trimethylsilyl cyanide as described before (Khakimov et al., 2013). All steps involving sample derivatization and injection were automated using a MultiPurpose Sampler (MPS; Gerstel). After reagent addition, the sample was transferred into the agitator of the MPS and incubated at 40°C for 40 min at 750 rpm. Immediately after derivatization, 1 μL of the derivatized sample was injected in splitless mode. The spilt/splitless injector port was operated at 320°C. The septum purge flow and purge flow to split vent at 2.1 min after injection were set to 3 and 15 mL min−1, respectively. The GC-MS system consisted of an Agilent 7890A GC device and an Agilent 5975C series MSD (Agilent Technologies). GC separation was performed on an Agilent HP-5MS column (30 m × 250 μm × 0.25 μm) by using hydrogen carrier gas at a constant flow rate of 1.2 ml/min. The GC oven temperature program was as follows: initial temperature, 40°C; equilibration time, 2 min; heat up to 270°C at the rate of 12°C/min; heat at the rate of 6°C/min until 310°C; and hold for 10 min. Mass spectra were recorded in the range of 50 to 700 mass-to-charge ratio with a scanning frequency of 2.2 scans/s, and the MS detector was switched off during the first 20 min of the run, since all targeted molecules eluted after this retention time. The transfer line, ion source, and quadrupole temperatures were set to 290°C, 230°C, and 150°C, respectively. The mass spectrometer was tuned according to the manufacturer’s recommendation by using perfluorotributylamine. The MPS and GC-MS devices were controlled using vendor software Maestro (Gerstel).

First, 100 mg of frozen powdered material were extracted with 600 µL of methanol (containing 50 µM of protopanaxatriol as internal standard) in a 1.5 mL vials with screw cap. Vials were ultrasonicated for 20 min at room temperature, the vials were then centrifuged for 10 min at 13,000 rpm and the supernatant was filtered through 0.22 µm centrifugal filters (UFC30GV, Merck Millipore). All treatments consisted of at least three replicates (specific replicate numbers are specified in figure legends). Statistical significance was tested usin one-way ANOVA (p < 0.05).

For extraction of triterpenoids from yeast cultures incubated with β-methyl cyclodextrin, 10 mL of culture was extracted using 10 mL ethyl acetate. The mixture was vortexed for 2 minutes, then centrifuged at 3,000g for 10 minutes and the organic phase collected, evaporated under vacuum and the residues were resuspended in 200µL of methanol. For extraction of triterpenoids from yeast cultures not containing β-methyl cyclodextrin, the cell pellet from 10 ml culture was resuspended in 500µL 10% KOH, 80% EtOH and saponified for 2 hours at 70°C. The samples were then extracted three times with 500µL hexane, which was pooled, evaporated under vacuum and the residues resuspended in 200µL of Methanol.

A Dionex UltiMate™ 3000 RS UPLC system from Thermo Scientific™ (Waltham, MA USA) was used and equipped with a KINETEX^® XB-C18 column (2.1 mm × 100 mm, 1.7 µm, Phenomenex). Mobile phase A was prepared using MilliQ water and 0.05% formic acid, whereas mobile phase B consisted of acetonitrile and 0.05% formic acid. The gradient was t = 0 min, 5% B; t = 40 min, 100% B; t = 45 min, 100% B; t = 46 min, 5% B; t = 50 min, 5% B. The chromatographic run lasted 50 min with a flow rate of 0.3 ml/min. For detection, a Bruker compact™ QTOF mass spectrometer (Bremen, Germany) was used and operated in negative mode. The mass spectrometer’s settings were: dry gas flow rate 8 L min−1 at 220°C, capillary 4500 V, collision energy 7 eV and collision RF 500Vpp, transfer energy 100 µs, pre-Pulse storage 5µs. The QTOF operated with the mass range set from 50 to 1200 m/z and was calibrated with sodium formate clusters at the beginning of every injection. The injection volume was 5 μL per sample. Acquisition of LC-MS data was performed under Bruker DataAnalysis 4.3. Quantification of cucurbitacins was performed in more than three lines for each HR construct/treatment. Calibration curves with cucurbitacins B,D,E&I (Extrasynthese, Lyon, France) were made for quantification of cucurbitacin production in hairy roots lines.

Suggested chemical structures for the putative cucurbitacins in Figures 7 ; S7 . have a metabolite identification confidence level of 4, according to the Metabolomics Standards Initiatives (Sumner et al., 2007; Schrimpe-Rutledge et al., 2016) and was achieved by molecular formula generated from accurate mass spectrometry data (± 5 ppm) and MS/MS fragmentation pattern.

Our previous work established a HR transformation protocol for Squash (Dong et al., 2018) based on their growth on agar plates. Recently we showed that I. amara produce Cucs B,D,E&I in all organs (Dong et al., 2021), suggesting that Iberis species could be an alternative to squash HR for metabolic engineering of cucurbitacins. Hence, in this study a HR transformation protocol for Iberis umbellata was developed, to enable a comparison between I. umbellate and squash as HR platforms for metabolic engineering of cucurbitacins ( Figures 3A-C ). Due to the ∼7-fold higher transformation efficiency and shorter time taken to generate analyzable HRs, further work was carried out in squash HRs as opposed to I. umbellata HRs (table in Figure S1 ).

To establish a reproducible and scalable production platform, growth of squash HRs on agar plates, in liquid culture and in temporary immersion reactors (TIRs) was compared. In terms of cucurbitacin production, metabolite profiling by LC-MS showed that accumulation of Cuc E significantly increased ( ∼2-fold) in empty vector (EV) HR lines grown in TIRs as compared to those in plates but not for Cucs B&I ( Figure 3D ). Subsequent attempts to increase cucurbitacin yields of HR grown on plates and TIRs involved inducing their production with Methyl-Jasmonate (MeJA). Although the mean values appear higher for HRs grown on plates treated with MeJA (5.78 ± 1.96, 9.16 ± 2.91 and 51.98 ± 12.87 ng/mg DW for Cuc I, Cuc B and Cuc E, respectively) when compared to untreated control (5.63 ± 1.14, 6.82 ± 0.74 and 31.93 ± 2.68 ng/mg DW for Cuc I, Cuc B and Cuc E, respectively), there was no significant difference in cucurbitacin production ( Figure 3D ). Similarly, addition of MeJA did not significantly elevate cucurbitacin levels in TIRs. In terms of biomass production, HRs on plates grew slowly and were limited in space, whilst growth in liquid media inside Erlen Meyer flasks resulted in a drastic variation of biomass production within lines ( Figure S2A-C ). In summary, upscaling and reproducible production of biomass as well as higher cucurbitacin yields is best in TIRs ( Figure S2D ), probably due to facilitated aeriation, and is not affected by addition of MeJA.

Our previous work showed that overexpression of squash squalene epoxidases can be used to increase cucurbitacin levels in both HRs and the tabacco transient-expression systems by increasing the triterpenoid precursor 2,3-oxidosqualene (Dong et al., 2018). In this paper the approach of increasing cucurbitacins through overexpression of transcription factors was attempted.

Previously, a TF named CsBL, belonging to the clade II of the bHLH superfamily was reported to increase cucurbitacin C levels in cucumber leaves. Hence, an orthologous bHLH TF that would induce cucurbitacin accumulation in squash HRs and could be used as a tool in metabolic engineering of cucurbitacins was searched for in squash. BLASTp was used to searched the squash genome for putative orthologous sequences of the CsBL TF. This identified two sequences in the loci Cp4.1LG05g03810.1 and Cp4.1LG09g01220.1 which were 57% and 36% identical at aa level, respectively. These genes were termed Cucurbitacin inducing bHLH transcription factor 1 (CpCUCbH1) and CpCUCbH1-like.

A phylogenetic analysis showed that CpCUCbH1 was the most likely putative ortholog of CsBL while CpCUCbH1-like formed a separate clade from these sequences ( Figure 3E ). Overexpression of CpCUCbH1 in squash HR grown on plates increased ∼5-fold the production of Cuc I&B, and ∼3-fold of Cuc E ( Figure 3D ). To confirm that CpCUCbH1 was the functional ortholog of CsBl, the expression levels of known genes in the cucurbitacin pathway were quantified using RT-qPCR. With the exception of CpCYP87D19, there was a significant increase in the expression of all tested genes in the CpCUCbH1 overexpressing lines as compared to the empty vector (EV) lines ( Figure 4 ). These results suggest that CpCUCbH1 is indeed an ortholog of CsBl, and that the evolution of specific TFs regulating the production of cucurbitacins occurred at least before the split of the Cucurbitae and Benincasea tribes in the Cucurbitacea family. Interestingly, although growing HR lines in TIRs increased expression of genes in the cucurbitacin cluster and pathway as compared to HRs in agar plates, growth in TIR did not increase expression of CpCUCbH1 ( Figure 4A ). This suggests there are other mechanisms besides the clade II TFs of the bHLH superfamily which may be involved in the regulation of cucurbitacin biosynthesis which respond to different environmental stimuli.

It is noteworthy that the increase of cucurbitacins in CpCUCbH1-overexpressing HR lines was followed by a distinctive HR phenotype of reduced lateral roots, resulting in lines without the typical fish-bone phenotype of HRs, in the end this affected the biomass produced ( Figure S3 ). Hence, propagation efforts to produce enough biomass to explore the cucurbitacin production of CpCUCbH1-overexpressing HR lines in TIRs were not attempted due to the poor growth reported by these lines.

An orthologous bHLH TF sequence of Momordica grouping with CsBl and CpCUCbH1 in the phylogetic tree ( Figure 3E ) made us suspect that the bHLH induction of cucurbitacins could be widespread in the Cucurbitaceae. To determine if CpCUCbH1 could be used as a biotechnological tool to induce cucurbitacins in different species of the Cucurbitaceae, cotyledons or leaves of several Cucurbitaceae species were transiently infiltrated, and cucurbitacin levels, and expression levels of selected cucurbitacin biosynthetic genes were measured. To test if the squash bHLH could induce cucurbitacin biosynthesis in other Cucurbitaceae species, CpCUCbH1 was first transiently expressed in cucumber cotyledons. Since genes in the cucurbitacin biosynthesis pathway of cucumber have been reported, their expression could be monitored by qPCR compared to cucurbits in deeper Cucurbitaceae lineages that exhibit a lack of genetic resources. overexpression of CpCUCbH1 resulted in ~8 fold higher accumulation of Cuc C ( Figure 3F ) and higher expression of genes in the Cuc C pathway ( Figure 4B ). Having confirmed this, we next tested induction of cucurbitacins in deeper Cucurbitaceae lineages. Luffa cylindrica (Luffa) is a member of the Sicyeae tribe which split from the Cucurbitae ( Figure 1 ) 39 ± 3 Million Years ago (MY) (Schaefer et al., 2009) and produces Cucs B,D,E&I in leaves (Rehm et al., 1957); cotyledons of Luffa were successfully infiltrated and constructs were overexpressed, as observed from fluorescence of the GFP control constructs ( Figure S4A-D ). However, cotyledons of Luffa did not accumulate Cuc B,D,E&I, and overexpression of CpCUCbH1 did not induce production of cucurbitacins ( Figure S4E ). This incapacity of CpCUCbH1 to induce cucurbitacin biosynthesis was also observed for squash cotyledons ( Figure S4F ).

Finally, transient-expression experiments were performed on Ecballium elaterium to test if CpCUCbH1 would also increase cucurbitacin production in species of deeper phylogenetic clades in the Cucurbitaceae. Ecballium elaterium is a cucurbit which accumulates Cucs B,D,E&I (David and Vallance, 1955) in its leaves and belongs in the Bryonieae tribe, a tribe that split from the Cucurbitae 41 ± 3 Million Years ago (MY) (Schaefer et al., 2009). The transient-infiltration of E. elaterium leaves experiments for overexpression of CpCUCbH1 significantly increased levels of Cuc D ∼4-fold but not of Cucs B,E&I ( Figure 3G ). Induction of cucurbitacins in the deeper lineage of the Momordiceae was also attempted, but unfortunately leaves of Momordica proved difficult to infiltrate.

These results suggest that induction of cucurbitacin biosynthesis by clade II bHLH TF may require additional factors and is partially conserved in the Cucurbitaceae. Nevertheless, here it is shown that CpCUCbH1 overexpression can be used as a tool to induce cucurbitacins in selected clades of this family.

The results in this paper show that cucurbitacin biosynthesis in Cucurbitaceae species can be increased by overexpression of CpCUCbH1 in squash HRs and by environmental conditions such as growth in TIRs. Previously the class II bHLH TFs CsBl and CsBt in cucumber and ClBr in Citrullus lanatus were reported to regulate cucurbitacin biosynthesis in leaf, fruit and roots, respectively (Zhou et al., 2016). In this study it was observed that the bHLH class II in the Cucurbitaceae splits into two clades in which one contains functional homologs of CsBl. In particular, CpCUCbH1 can induce biosynthesis of Cucs B,D,E&I in squash HRs, Cuc C in cucumber cotyledons but only Cuc D in E. elaterium leaves. Furthermore, transient overexpression of CpCUCbH1 alone did not induce cucurbitacin biosynthesis in Luffa and squash cotyledons. Consistent with the latter observation, Brzozowski et al. (Brzozowski et al., 2020) reported that cucurbitacin biosynthetic genes are not expressed in squash cotyledons and thus accumulation of cucurbitacins in cotyledons depends on transport of these metabolites from the roots; in addition, cucurbitacin accumulation trait in squash appears to be cultivar dependent.

It is known that bHLH TFs may make heterocomplexes with MYB TFs to be functional, and that the function of these heterocomplexes can be regulated by protein interactions such as by WD40 repeat proteins (Li, 2014; Pireyre and Burow, 2015). The absence of a MYB TF or presence of a WD40 protein could be a reason why cotyledons of squash and Luffa transiently overexpressing CpCUCbH1 did not produce cucurbitacins. When Shang et al. (Shang et al., 2014) restored Cuc C production in cotyledons of the non-bitter cucumber line XY-3 by transiently expressing CsBl, they were complementing a mutation in the CsBl TF, and since cucumber cotyledons typically produce Cuc C it is probable that the cofactors to induce Cuc biosynthesis were also present. In conclusion, further work is needed to understand the intricacies of cucurbitacin biosynthesis regulation throughout the Cucurbitaceae species and plant tissues.

When HRs were grown in different conditions, it was found that growth in TIR boosted Cuc production and induced the known biosynthetic genes ( Figures 3 , 4 ). The work described here was consistent with other papers where culture of HRs in TIR had been shown to increase yields of plant specialized metabolites (Thakore et al., 2017; Kochan et al., 2018); but perhaps, more interesting in this study was that this increase in expression of cucurbitacin biosynthetic genes and cucurbitacin production was not related to CpCUCbH1 expression ( Figures 3 , 4 ). In addition, class IV bHLH TFs inducing terpenoid production in Medicago truncatula and Cataranthus roseus are known to be induced by jasmonate signaling pathway, but this was apparently not the case for the expression of CpCUCbH1 since expression did not increase after MeJA treatment.

The cucurbitacin pathway continues to evolve in the Cucurbitaceae family, and thus not all Cucurbitaceae species have the same cucurbitacin profiles ( Figure 1 ). Cucs B,D,E&I are the most widespread cucurbitacins, their presence can be observed starting in the Telfairieae tribe after the split from the Momodiceae (Rehm et al., 1957; Schaefer et al., 2009); nevertheless, cucumber in the Benincaseae tribe has lost the ability to produce Cucs B,D,E&I and instead biosynthesizes Cuc C. Cucurbitacins in the Momordiceae species in general lack the C16 hydroxylation (Okabe et al., 1982). In the case of cucumber, the absence of Cucs B,D,E&I is due to the lack of function of the CYP81Q59 catalyzing the C2 hydroxylation ( Figure 2 ) (Shang et al., 2014). We identified an ACT, IaCUCA1, from I. amara that acetylates at the C3 postion. A phylogentic analysis showed that it evolved independently of the Cucurbitaceae ACTs that acetylate the C16 and C25 hydroxyl groups (Thakore et al., 2017), but evolved out of a clade in the Brassicaceae which has a predisposition for acetylating triterpenoids. To our knowledge, the acetylation of the hydroxyl group in C3 of cucurbitadienol has not been reported before. The fact that no 3-acetyl-16-hydroxy-cucurbitadienol was found in crude extracts of I. amara, most likely reflects the promiscuity of IaCUCA1; indeed, ACTs have been shown to be promiscuous acting in several substrates like the AtTHAA2 (Huang et al., 2019) from which IaCUCA1 diverged from. Nevertheless, the ability to create novel cucurbitacin modifications, reflects that I. amara represents an additional potential reservoir for cucurbitacin-modifying genes as the pathway has evolved independently in the two lineages.

Cucumber and Momordica diverged 46 ± 3 MY (Schaefer et al., 2009), but they both report the ability to hydroxylate at the C19 position. In contrast, there are no reports of cucurbitacins with C19 hydroxylation in squash despite of it diverging from cucumber only 30 ± 4 MY. The data presented in this paper shows that CsCYP88L2 and McCYP88L7 enzymes are responsible for the C19-hydroxylation in cucumber and Momordica, respectively, and suggests that CsCYP88L2 and McCYP88L7 must act early on the cucurbitadienol backbone as they are unable to hydroxylate highly oxygenated Cucs B,D,E&I ( Figures 6 ; S8 ) possibly due to steric hindrance in the active site. Taking into consideration the suggested early action of these P450s and the lack of ions with m/z values for a deacetylated cucurbitacin in the MS2 spectra ( Figure 7B ), then a likely structure for the new observed peak could be one of those with molecular formula of C₃₀H₄₂O₇ in Figure S7B ; especially the middle structure as this molecule would not be acetylated by the promiscuous CpACT1.