In the format provided by the authors and unedited.

Convergence and divergence of bitternessbiosynthesis and regulation in CucurbitaceaeYuan Zhou1,2,3†, Yongshuo Ma1†, Jianguo Zeng3†, Lixin Duan4†, Xiaofeng Xue5, Huaisong Wang1,Tao Lin1,2, Zhiqiang Liu6, Kewu Zeng7, Yang Zhong1, Shu Zhang1, Qun Hu8, Min Liu1, Huimin Zhang1,James Reed9, Tessa Moses9, Xinyan Liu1, Peng Huang3, Zhixing Qing10, Xiubin Liu3, Pengfei Tu7,Hanhui Kuang8, Zhonghua Zhang1, Anne Osbourn9, Dae-Kyun Ro11, Yi Shang1,2*and Sanwen Huang1,2*

Differentiation of secondary metabolite profiles in closely related plant species provides clues for unravelling biosyntheticpathways and regulatory circuits, an area that is still underinvestigated. Cucurbitacins, a group of bitter and highlyoxygenated tetracyclic triterpenes, are mainly produced by the plant family Cucurbitaceae. These compounds have similarstructures, but differ in their antitumour activities and ecophysiological roles. By comparative analyses of the genomes ofcucumber, melon and watermelon, we uncovered conserved syntenic loci encoding metabolic genes for distinctcucurbitacins. Characterization of the cytochrome P450s (CYPs) identified from these loci enabled us to unveil a novelmulti-oxidation CYP for the tailoring of the cucurbitacin core skeleton as well as two other CYPs responsible for the keystructural variations among cucurbitacins C, B and E. We also discovered a syntenic gene cluster of transcription factorsthat regulate the tissue-specific biosynthesis of cucurbitacins and that may confer the loss of bitterness phenotypesassociated with convergent domestication of wild cucurbits. This study illustrates the potential to exploit comparativegenomics to identify enzymes and transcription factors that control the biosynthesis of structurally related yet uniquenatural products.

The structural diversity of plant secondary metabolites in phy-logenetically related species is likely to be associated withadaptation to different ecological niches1. In the plant family

Cucurbitaceae, bitter compounds known as cucurbitacins canserve as protectants against generalists, and also as feeding attrac-tants for specialists2,3, in mediating the co-evolutionary associationbetween herbivores and cucurbits including cucumber (Cucumissativus L.), melon (Cucumis melo L.) and watermelon (Citrulluslanatus (Thunb.) Matsum. and Nakai)3. To date, 12 categories ofcucurbitacins have been discovered, most of these from cucurbitplants4. Although a cucurbitacin C (CuC) biosynthetic module con-sisting of nine genes has been identified in cucumber and four of thepathway enzymes have been characterized5, the genetic basis ofcucurbitacin diversity among the wider cucurbits is unknown.Cucurbitacins also exhibit wide-ranging pharmacological potential,including anti-inflammatory, purgative and antitumour activities4,6,7.It is of particular interest that cucurbitacins B (CuB) and E (CuE)(Fig. 1a), the major bitter compounds isolated from melon8 andwatermelon9 respectively, display much stronger antitumouractivities than the structurally similar CuC9–11. Understandingthe evolutionary and genetic basis of this metabolic diversityand specificity could provide crucial clues for chemical ecologyand tools for crop breeding, as well as opening up opportunities

for production and engineering of these plant-derived chemicals forpharmaceutical applications.

The non-bitter cucurbit cultivars that are used for production ofvegetables and fruits for human consumption have been domesti-cated from their extremely bitter progenitors. In cucumber,expression of the CuC biosynthetic genes is controlled by twotissue-specific basic helix–loop–helix (bHLH) transcription factors(TFs) in the leaves (Bl, bitter leaf ) and fruit (Bt, bitter fruit)5.Mutations within the promoter of Bt effectively remove fruit bitter-ness, and this trait of non-bitterness has been selected and fixedduring the domestication process5. However, it is unclear whetherthe mechanisms underlying CuC regulation and domestication incucumber also prevail in other cucurbits. Since genome mininghas become a powerful strategy for metabolic studies in bothmicrobes and plants12–20, we envisioned that investigation of thegenome sequences of three cucurbit plants (cucumber21, melon22

and watermelon23) would provide a unique opportunity to under-stand the regulatory and biochemical principles dictating cucurbitacindiversity in cucurbits.

Here, by applying a comparative genomic study, we report thatindependent mutations within syntenic transcription factor genesin the three cucurbits may result in marked decreases in fruit bitter-ness, in turn giving rise to converged domestication of bitter wild

1Institute of Vegetables and Flowers, Chinese Academy of Agricultural Sciences, Key Laboratory of Biology and Genetic Improvement of Horticultural Cropsof the Ministry of Agriculture, Sino-Dutch Joint Laboratory of Horticultural Genomics, Beijing 100081, China. 2Agricultural Genome Institute at Shenzhen,Chinese Academy of Agricultural Science, Shenzhen 518124, China. 3Horticulture and Landscape College, Hunan Agricultural University, National ChineseMedicinal Herbs (Hunan) Technology Centre, Changsha 410000, China. 4Institute of Botany, Chinese Academy of Science, Beijing 100093, China.5Institute of Apiculture Research, Chinese Academy of Agricultural Sciences, Beijing 100093, China. 6School of Life Science, Shanxi University, Taiyuan030006, China. 7School of Pharmaceutical Sciences, Peking University Health Science Centre, Beijing 100191, China. 8College of Horticulture and ForestrySciences, Huazhong Agricultural University, Wuhan 430070, China. 9John Innes Centre, Norwich Research Park, Norwich NR4 7UH, UK. 10School ofPharmacy, Hunan University of Chinese Medicine, Changsha 410208, China. 11Department of Biological Sciences, University of Calgary, Calgary T2N 1N4,Canada. †These authors contributed equally to this work. *e-mail: huangsanwen@caas.cn; shangyi@caas.cn

ARTICLESPUBLISHED: XX XX 2016 | ARTICLE NUMBER: 16183 | DOI: 10.1038/NPLANTS.2016.183

NATURE PLANTS | www.nature.com/natureplants 1© 2016 Macmillan Publishers Limited, part of Springer Nature. All rights reserved.

SUPPLEMENTARY INFORMATIONARTICLE NUMBER: 16183 | DOI: 10.1038/NPLANTS.2016.183

NATURE PLANTS | www.nature.com/natureplants

1

Supplementary Methods, Figures, Tables

Supplementary Methods

Quantitative RT-PCR analysis

Total RNA was isolated using an RNA extraction Kit (Qiagen). First-strand cDNA

was synthesized from 1 μg total RNA using the iScript cDNA synthase kit (Bio-Rad).

Primer specificity was checked by sequencing and blast analysis (listed in

Supplementary Table 2). PCRs were performed on an ABI 7900 using SYBR Premix

(Roche) according to the manufacturer’s instruction. Three technical replicates and

three independent biological experiments were performed in all cases. Relative gene

expression was performed using the comparative 2-∆∆Ct or 2-∆Ct method.

Phylogenetic analysis

Sequence alignments and phylogenetic analyses, based on a neighbor-joining method,

were carried out with the Clustal X program. A phylogenetic tree was created with the

MEGA5.0 program; the number of bootstrap replications was 1,000. Sequence

exchanges were detected using Geneconv and visual inspection35, 36.

Yeast one-hybrid assay

The primers used for cloning are listed in Supplementary Table 2. Yeast cells were

co-transformed with pHIS2 bait vector harboring the promoter of target-gene and

pGADT7 prey vector harboring the ORF of transcription factor. Yeast cells

transformed with the empty pGADT7 vector and pHIS2 harboring the corresponding

promoter was used as negative control. Transformed yeast cells were grown on

SD-Leu-Trp and SD-Leu-Trp-His plates supplemented with proper concentration of

3-AT (Sigma). These plates were then incubated for five days at 30°C.

2

Trans-activation ability assay in tobacco leaves

For effector construct, transcription factor was fused to the binary vector

(pCAMBIA1300) downstream of the 35S promoter using the primers listed in

Supplementary Table 2. Reporter construct was composed of the promoter linked to

the reporter gene Luciferase. Each promoter of the potential CuB or CuE biosynthetic

enzymes were isolated using the primers listed in Supplementary Table 2. The Luc

cDNA was PCR-amplified using the forward primer

5’-TCCCCCGGGATGGAAGACGCCAAAAAC-3’ and reverse primer

5’-CGGGATCCTT-ACACGGCGATCTTTCCGC-3’ from pGL3-Basic vector

harboring the Luc cDNA. Each promoter was fused into the pCAMBIA1300 vector

by replacing the 35S promoter via ligation or using the In-Fusion Cloning Kit

(Clontech). The Luc cDNA was then fused to the SmaI/BamHI sites downstream of

the promoter. The construct was transformed into Agrobacterium tumefaciens strain

GV3101. After transformation, strain was grown at 28°C for 24 h in LB media

supplied with kanamycin (50 mg/L) and rifampicillin (50 mg/L). Cells were harvested

by centrifugation and resuspended in 10 mM MES buffer containing 10 mM MgCl2

and 200 μM acetosyringone (Sigma) to a final OD600 of 0.6. For co-infiltration, equal

volumes of the strains were mixed and infiltrated into fully expanded leaves of

four-week-old tobacco plants using a needleless syringe. After infiltration, plants were

grown first under dark for 12 h and then with 16 h light/8 h dark for 48 h under

normal condition. Luc activity was observed with a CCD imaging apparatus

(AndoriXon, Andor, UK). Experiments were repeated independently for at least three

times with similar results.

3

EMSA assay

EMSA was performed using recombinant CmBr-His, CmBt-His, ClBr-His or

ClBt-His protein purified from E. coli. Protein purification was performed according

to the manufacturer’s instruction (Novagen). The promoter fragment used for the

EMSA assay was amplified by PCR using the primers listed in Supplementary Table 2.

Each promoter fragment was labeled with digoxigenin-dUTP (Roche) according to

the manufacture’s instruction. Binding reaction was performed in 20 μL of binding

buffer composed of 10 mM Tris-HCl pH 8.0, 20 mM NaCl, 0.4 mM MgCl2, 0.25 mM

EDTA, 0.25 mM DTT and 10% glycerol and in the presence of 0.04 mg/mL of poly

(deoxyinosinic-deoxycytidylic) sodium salt [poly (dI-dC)] (Sigma). Binding reaction

was carried out using 20 ng of fusion protein and 30 ng for each digoxigenin-labeled

promoter fragment at 25oC for 30 min. Samples were separated on 5%

polyacrylamide gels (19:1 acryl:bisacrylamide) in 0.25 mM Tris-borate-EDTA at 4°C.

After the sample transfer process, the nylon filter was exposed under ultraviolet light

and then washed twice with standard saline citrate buffer, each for 15 min. The filter

was blocked with blocking reagent (provided in the 10-block kit, Pierce) for 2 h, and

then incubated with anti-digoxigenin-alkaline phosphatase (AP) antibody for 1 h. The

AP substrate was added into the mixture and subsequently analyzed by

autoradiography.

UPLC analysis of cucurbitacins

Plant sample was frozen in liquid N2 and ground in a mortar and pestle. The resultant

powder (0.5 g) was added to methanol (2 mL) and homogenized for 15 min, followed

by a centrifugation at 10,000 g at 4°C for 10 min. The solution was filtered through a

0.22 μm membrane and analyzed with the UPLC system (Waters

4

ACQUITY) equipped with a BEH RP-C18 column (1.7 μm, 100 × 2.1 mm). The

mobile phase consisted of water (solvent A) and acetonitrile (solvent B). The UV

wavelength was 230 nm. The flow rate was 0.25 mL/min, and the injection volume

was 3 μL. A linear gradient with the following proportion of phase B (tmin, B%) was

used: (0, 10), (5, 25), (10, 100).

GC-MS analysis of yeast extract

Yeast extract was derivatized using MSTFA at 60°C for 30 min and analyzed on a

GC-MS system (Agilent 7000B) equipped with an Agilent HP-5MS column (5%

phenyl methyl silox, 30 m × 250 μm internal diameter, 0.25 μm film). The front inlet,

transfer line, and ion source temperatures were set at 280°C, 250°C and 230°C,

respectively. The oven temperature program used was as follows: 70°C for 2 min,

then 20°C/min to 260°C, final 10°C/min to 300°C for 10 min. The flow rate of the

carriage gas (He) was 1 mL/min. Split injection (split ratio 5:1). The mass spectral

data between m/z 50-800 were recorded.

HPLC-APCI-qTOF analysis of yeast extract

The yeast extracts were dissolved in methanol and the solution was filtered through

0.22 μm membrane prior to injection. Chromatography was performed on an Agilent

1200 HPLC system using a ZORBAX SB-C18 column (5 μm, 4.6 × 100 mm, Agilent).

The column temperature was maintained at 35°C. The mobile phase consisted of 0.1%

formic acid aqueous solution (v/v, solvent A) and methanol/0.1% formic acid (v/v,

solvent B). The flow was 0.4 mL/min, and the injection volume was 15 μL. A linear

gradient with the following proportion of phase B (tmin, B%) was used: (0, 30), (2, 30),

(10, 60), (15, 95), (30, 95). MS was performed in an Agilent 6510 Q-TOF equipped

5

with an atmospheric pressure chemical ionization (APCI) in positive ion model. The

optimum MS conditions consisted of the corona current of 4 μA, a capillary voltage

of 3.0 kV, a skimmer voltage of 65 V, and a segmented fragmentor at 135 V. The gas

temperature, vaporizer temperature, drying gas flow rate and nebulizer pressure were

300°C, 350°C, 4.5 L/min, and 40 psi, respectively. Nitrogen was used as the collision,

drying, and nebulizer gas. MS spectra were acquired at 150-1000 m/z using an

extended dynamic range and a scan rate of 1.5 spectra/s by varying collision energy

with mass.

Compound purification from yeast

Yeast harboring CmBi, CPR and Cm890 was cultured (120 L) and then centrifuged at

6,000 rpm for 10 min, the filtrate (110 L) was twice extracted with the same volume

of PE (60~90°C), and the PE was gathered. The pellet (220 g) resuspended in 1 L 20%

KOH/50% EtOH (1:1) and treated with alkali lysis method for 1 h at 95°C. Lysate

was twice extracted with the same volume of PE. After retrieval of the PE, the residue

(3 g) was subjected to column chromatograph (CC) on a silica gel (200-300 mesh, 3 ×

10 cm) with a gradient of PE-ethyl acetate (15:1–1:1) to yield six fractions (Fr.1–Fr.6).

Fr.5 was further purified by semi-preparative HPLC with the mobile phase of

methanol (flow rate 2.0 mL/min, retention time 17.5 min) to yield compound 1. Fr.3

was purified by semi-preparative HPLC (methanol/water, 90:10, 2.0 mL/min,

retention time 21.2 min) to yield compound 2.

Yeast harboring CmBi, CPR, Cm890 and Cm180 was cultured (250 L) and

treated using the same approach as used for compounds 1 and 2. After retrieval of the

PE, the mixed residue (7 g) was subjected to CC on a silica gel (200-300 mesh, 5 × 23

cm) with a gradient of PE-ethyl acetate (2:1-1:2) to give seven fractions (Fr.1-Fr.7).

6

Fr.6 was purified by semi-preparative HPLC (methanol/water, 85:15, 2.0 mL/min,

retention time 15 min) to yield compound 3.

Compound structure elucidation by MS/MS and NMR data analysis

NMR analyses of compounds 1-3 were performed on a BrukerAvance III 500 system

with CDCl3.

Compound 1 was assigned a molecular formula of C30H48O2 determined on the basis

of its positive HRAPCIMS [M+H]+ ion peak at m/z 441.3745 (calculated value,

441.3728). In positive ion mode with the CE of 20 V, MS/MS of compound 1 gave

clear information of fragment ions (Supplementary Fig. 5a). As demonstrated in the

proposed fragmentation pathway of compound 1 (Supplementary Fig. 5b), the ion

peaks at m/z 423, 405 were observed to correspond to a loss of one and two H2O

molecules, respectively. The fragment ion peak at m/z 289 could be explained by the

loss of a neutral fragment C10H16O produced by Retro-Diels-Adler (RDA) reaction.

The crucial fragment ion at m/z 179 produced by a cleavage of its side chain with the

loss of molecule of 110 (C8H14) suggested the presence of an oxygen on the C or D

ring. The 1H and 13C NMR data (Supplementary Fig. 6 and Supplementary Tables 3, 4)

assigned by HSQC and HMBC experiments (Supplementary Fig. 7), revealed seven

singlets for tertiary methyl at δH 0.94, 1.01, 1.02, 1.11, 1.16, 1.59, 1.68, one doublet at

δH 0.89 (d, J= 6.4 Hz), two olefinic linkages (δH 5.66, δC 120.8, C-6; δH 5.08, δC 125.0,

C-24), one hydroxylated carbon (δC 76.3, C-3) and one carbonyl group (δC 214.9,

C-11) in compound 1. Analysis of the above data suggested compound 1 to be a

carbonylated cucurbitane triterpene. Detailed HMBC correlations, including from

H-19 (δH 1.11) to C-8 (δC 43.9), C-9 (δC 48.8), C-10 (δC 35.5) and C-11, from H-12

(δH 2.43, 2.92, both double peak, J= 14.5 Hz) to C-11, C-13 (δC 48.9) and C-18 (δC

7

16.9) supported this assignment, and this analysis further indicated the structure of

compound 1 closely resembling cucurbitadienol37, except for the main difference that

C-11 was carbonylated in 1. Based on the above analysis, compound 1 was elucidated

to be 11-carbonylcucurbitadienol.

Compound 2 was assigned a molecular formula of C30H48O3 determined on the basis

of its positive HRESIMS [M+H]+ ion peak at m/z 457.3683 (calculated value,

457.3692). Comparison of the MS of compound 2 and 1 suggested one more oxygen

atom in the formula of 2. The positive ion mode MS/MS with the CE of 20V

(Supplementary Fig. 8a) and the proposed fragmentation pathway of compound 2

(Supplementary Fig. 8b) showed ion peaks at m/z 439, 421, 403 which were

corresponding to the loss of one, two and three H2O molecules, respectively. The

fragment ion peak at m/z 305 could be deduced from a RDA reaction with the loss of

a neutral fragment C10H16O. The important fragment ion at m/z 179 produced by a

cleavage of the side chain with the loss of molecule of 126 (C8H14O) revealed the

additional oxygen on the side chain. The 1H and 13C NMR data (Supplementary Fig. 9

and Supplementary Tables 3, 4) assigned by HSQC and HMBC experiments

(Supplementary Fig. 10), revealed eight singlets for tertiary methyl at δH 0.91, 1.02,

1.03, 1.11, 1.16, 1.27, 1.60, 1.67, two olefinic linkages (δH 5.65, δC 120.7, C-6; δH

5.08, δC 124.3, C-24), two hydroxylated carbons (δC 76.2, C-3; δC 75.0, C-20), one

carbonyl group (δC 214.7, C-11). By comparing the NMR data with those of

compound 1, the structure of 2 showed one more hydroxyl on its skeleton, and was

eventually identified as the 20-hydroxyl derivative of compound 1 as the CH3-21 of

compound 2 possessed a single peak in the 1H NMR spectrum, of which the C-20

chemical shift transferred to low field at δC 75.0 (ΔδC 39.3 ppm) correspondingly.

This assignment was further confirmed by the major HMBC cross-peaks of H-17 (δH

8

2.03)/C-20, H-21 (δH 1.27)/C-17 (δC 50.9), H-21/C-20. Correlation between H-21 and

H-18 (δH 0.91) was observed in the NOESY spectrum (Supplementary Fig. 11) of 2

suggested 20-OH to be β-orientated. Based on the above analysis, compound 2 was

identified as 11-carbonyl-20β-hydroxycucurbitadienol.

Compound 3 was determined as C30H48O4 on the basis of its positive HRESIMS

[M+H]+ ion peak at m/z 473.3643 (calculated value, 473.3625). Comparison of the

MS data of compound 3 and 2 suggested one more oxygen atom in 3. As deduced by

the positive ion mode MS/MS (Supplementary Fig. 13a) analysis, the proposed

fragmentation pathway of compound 3 (Supplementary Fig. 13b) showed ion peaks at

m/z 455, 437, 419 which were corresponding to the loss of one, two and three H2O

molecules successively. The fragment ion peak at m/z 389 might be the product of a

loss of CH2O from m/z 419. The ion m/z 329 could be produced by the cleavage of the

side chain of C30H46O3 (m/z 455). The 1H and 13C NMR data (Supplementary Fig. 14

and Supplementary Tables 3, 4) assigned by HSQC and HMBC experiments

(Supplementary Fig. 15), revealed eight singlets for tertiary methyl at δH 0.92, 1.00,

1.01, 1.12, 1.22, 1.27, 1.61, 1.67, two olefinic linkages (δH 5.71, δC 121.3, C-6; δH

5.08, δC 124.3, C-24), three hydroxylated carbons (δC 68.5, C-2; δC 78.7, C-3; δC 75.0,

C-20), one carbonyl group (δC 214.3, C-11). Comparison of the NMR data between

compound 3 and 2 revealed the only difference that there was one more hydroxyl

group on the position of C-2 in compound 3, with the result that the chemical shift of

C-3 moved to relative low field at δC 78.7 (ΔδC 2.5 ppm). This variation was further

confirmed by the HMBC correlations between H-2 (δH 3.90) and C-3; H-3 (δH 3.45)

and C-2. The β-orientation of 2-OH was determined by the NOESY correlations

(Supplementary Fig. 16) between H-2 and H-10 (δH 2.30), H-2 and H-3, H-2 and

H-28 (δH 1.22). Based on the above analysis, compound 3 was unambiguously

9

identified as 11-carbonyl-2β, 20β-dihydroxycucurbitadienol.

Supplementary References

35. Sawyer, S. Statistical tests for detecting gene conversion. Mol. Biol. Evol. 6,

526-538 (1989).

36. Kuang, H., Woo, S. S., Meyers, B. C., Nevo, E., & Michelmore, R. W. Multiple

genetic processes result in heterogeneous rates of evolution within the major cluster

disease resistance genes in lettuce. The Plant Cell, 16, 2870-2894 (2004).

37. Tan, J.W., Liu, J.K., Dong, Z.J., Liu, P.G. & Ji, D.G. Lepida acid A from

basidiomycetes Russula lepida. Chin. Chem. Lett. 10, 297-299 (1999).

10

Supplementary Figures

Supplementary Figure 1. Colinearity analysis among cucumber, melon and

watermelon. (a-e) Comparative genomics of the three sequenced cucurbits at the

collinear regions where the cucumber CsBi (a), Cs890 (b), Cs490 (c), Cs540 and Cs550

(d), CsBl and CsBt (e) genes are located. The genes encoding different types of

enzymes are indicated by different colors. Orthologous relations are depicted by solid

lines between the different syntenic loci. Boxes represent partial genes and triangles

represent frame-shift mutations within genes. The silent CYP gene Csa180 in the

cucumber cluster is indicated by the dark blue hatched coloring. (f and g) Unrooted

phylogenetic tree of CYPs at the collinear region where cucumber Cs540 and Cs550

genes are located (f), and bHLH TFs at the collinear region where CsBl and CsBt are

located (g). 1,000 bootstrap replicates were performed with neighbor-joining method

and the bootstrap values are indicated at the nodes. The bar represents a 5% difference

in nucleotide sequences. Gene accessions are listed in the Table 1.

11

Supplementary Figure 2. Elucidation of first committed step for CuB and CuE

biosynthesis. (a) Phylogenetic tree of the oxidosqualene cyclases (OSCs) from melon,

watermelon, cucumber and squash (Cucurbita pepo). CmBi and ClBi are clustered

with C. sativus CsBi and a cucurbitadienol synthase OSC previously characterized

from C. pepo (CPQ) in the same sub-clade (indicated in red). (b) Mass spectra for a

cucurbitadienol standard and the products generated by CmBi and ClBi .

12

Supplementary Figure 3. UPLC-qTOF-MS analysis of acetyltransferase catalytic

reaction products. (a to b) CmACT, ClACT and CsACT are able to acetylate CuD

and CuI in vitro to give CuB (a) and CuE (b), respectively. (c) CmACT and ClACT

are able to acetylate deacetyl-CuC in vitro to give CuC. Specific products are

indicated by red arrows.

13

Supplementary Figure 4. Elucidation of catalytic steps in the CuB and CuE biosynthetic pathways. (a) UPLC-qTOF-MS analysis of extracts prepared from yeast expressing OSC, Cytochrome P450 reductase (Melo3C017461 in melon and Cla018830 in watermelon), and each candidate CYP with APCI in positive ion mode. One expected product (indicated by red arrows) is generated by the CYP87D20 enzymes Cm890, Cl890A, Cl890B, and Cs890. EIC 441.3727, extracted ion chromatogram of the accurate parent ion at m/z of 441.3727 [M+H]+. (b) UPLC-qTOF-MS analysis of the extracts used in (a) with ESI in positive ion mode. One expected product (indicated by red arrows) is generated by the CYP87D20 enzymes Cm890, Cl890A, Cl890B, or Cs890. EIC 457.3676, extracted ion chromatogram of the accurate parent ion at m/z of 457.3676 [M+H]+. (c)

14

UPLC-qTOF-MS analysis of the extracts prepared from yeast accumulating 11-carbonyl-20β-hydroxycucurbitadienol and expressing each candidate CYP gene. One expected product peak (indicated with red arrows) is generated by the CYP81Q59 enzymes Cm180 and Cl180. EIC 473.3625, extracted ion chromatogram of the accurate parent ion at m/z of 473.3625 [M+H] +. (d and e) Comparison of the catalytic efficiency of Cl890A and Cl890B in generating 11-carbonxylcucurbitadienol (d) and 11-carbonyl-20β-hydroxycucurbitadienol (e) (mean ± SEM, n=3 biological replicates).

15

Supplementary Figure 5. MS/MS analysis of the intermediate product

(11-carbonylcucurbitadienol, compound 1). (a) Detailed APCI-MS/MS in positive

ion mode with the collision energy (CE) of 20V. (b) The proposed fragmentation

pathway of compound 1. The crucial fragment ion at m/z 179 produced by a cleavage

of its side chain with the loss of molecular weight of 110 (C8H14) suggested the

presence of an oxygen on the C or D ring.

16

Supplementary Figure 6. 1H NMR (a) and 13C NMR (b) results of

11-carbonylcucurbitadienol. Chemical shift at δC 214.9 observed in 13C NMR

spectra indicated the presence of a carbonyl group.

17

Supplementary Figure 7. 2D NMR spectra of 11-carbonylcucurbitadienol. (a)

HSQC spectra of compound 1. (b) HMBC spectra and key correlations of compound

1. The assignment of 11-carbonyl was supported by the correlations from H-12 (δH 2.43 and 2.92) and H-19 (δH 1.11) to C-11 (δC 214.9) (crossing lines).

18

Supplementary Figure 8. MS/MS analysis of the intermediate product

(11-carbonyl-20β-hydroxycucurbitadienol, compound 2). (a) Detailed ESI-MS/MS

in positive ion mode with the CE of 20V. (b) The proposed fragmentation pathway of

compound 2. The important fragment ion at m/z 179 produced by a cleavage of its

side chain with the loss of molecular weight of 126 (C8H14O) revealed additional

oxygen on the side chain.

19

Supplementary Figure 9. 1H NMR (a) and 13C NMR (b) results of

11-carbonyl-20β-hydroxycucurbitadienol. By comparison to compound 1, the

chemical shift at δC 75.0 observed in 13C NMR spectra (b) indicated the presence of

an additional hydroxyl group in compound 2. In addition, the doublet of the proton

peak of CH3-21 changing to a singlet suggests the substituting of hydroxyl on C-20

(a).

20

Supplementary Figure 10. 2D NMR spectra of

11-carbonyl-20β-hydroxycucurbitadienol. (a) HSQC spectra of compound 2. (b)

21

HMBC spectra and key correlations of compound 2. The assignment of 20-OH was

supported by the correlations from H-17 and H-21 to C-20 (crossing lines).

22

Supplementary Figure 11. NOESY spectra and key correlations of

11-carbonyl-20β-hydroxycucurbitadienol. The β-orientation of 20-OH was

determined by the correlations between H-21 and H-18 (crossing lines).

23

Supplementary Figure 12. Comparison of the catalytic products of Cs890 and

Cs540. (a) UPLC-qTOF-MS analysis of the extracts prepared from yeast expressing

CsBi, CPR (Csa1G423150), Cs890 and Cs540 with APCI in positive ion mode.

19-hydroxycucurbitadienol (C30H50O2) is showed by EIC at m/z 443.3884 [M+H] +,

and 11-carbonxylcucurbitadienol (C30H48O2) is showed by EIC 441.3727 [M+H]+. (b)

The ion abundance of 11-carbonxylcucurbitadienol is about 4 fold that of

19-hydroxycucurbitadienol (mean ± SEM, n=3 biological replicates).

24

Supplementary Figure 13. MS/MS analysis of the intermediate product

(11-carbonyl-2β, 20β-dihydroxycucurbitadienol, compound 3). (a) Detailed

ESI-MS/MS on positive ion mode with the CE of 20V. (b) Proposed fragmentation

pathways of compound 3.

25

Supplementary Figure 14. 1H NMR (a) and 13C NMR (attached proton test, APT)

(b) results of 11-carbonyl-2β,20β-dihydroxycucurbitadienol. Comparing to those

of compound 2, the chemical shift at δC 68.5 and a lower field movement of C-3 to δC 78.7 (ΔδC 2.5 ppm) observed in the 13C NMR spectra (b) indicated an additional

hydroxyl group on the position of C-2 in compound 3.

26

Supplementary Figure 15. 2D NMR spectra of 11-carbonyl-2β,

20β-dihydroxycucurbitadienol. (a) HSQC spectra of compound 3. (b) HMBC

spectra and key HMBC correlations of compound 3. The presence of 2-OH was

confirmed by the correlations from H-3 to C-2 and H-2 to C-3 (crossing lines).

27

Supplementary Figure 16. NOESY spectra and key correlations of

11-carbonyl-2β, 20β-dihydroxycucurbitadienol. The β-orientation of 2-OH was

determined by the correlations from H-2 to H-10, H-3 and H-28 (crossing lines).

28

Supplementary Figure 17. Direct interaction between the promoters of the genes

for CuB biosynthetic candidate enzymes and the potential regulators from melon.

(a) Yeast one-hybrid analysis of the interaction between the bitterness regulators

CmBr and CmBt and promoters of CuB biosynthetic enzymes. P, promoter; AD,

activation domain; 3-AT, 3-amino-1,2,4-triazole; SD-L-T, synthetic dropout media

w/o leucine and tryptophan; SD-L-T-H, synthetic dropout media w/o leucine,

tryptophan, and histidine. (b) CmBr and CmBt activate luciferase reporter expression

by binding to promoters of CuB biosynthetic candidate genes in a tobacco leaf assay

system. Images show activated (right, reporter + effector) and non-activated (left,

reporter only) status of infiltrated tobacco leaves, respectively.

29

Supplementary Figure 18. Recruitment of CmBr and CmBt to the promoters of

CuB biosynthetic candidate genes. EMSA assays showed that CmBr or CmBt is

recruited to the promoter of Cm160 (a), Cm170 (b), Cm180 (c), CmBi (d), CmACT (e),

Cm710 (f), Cm490 (g), Cm890 (h), but not to the promoter of Cm609 (i). The

promoter regions from -2,000 upstream from the start codon are represented

schematically. The black vertical lines indicate the localization of E-boxes and red

horizontal lines indicate the regions used in EMSA. His, His tag; +/-,

presence/absence of protein.

30

Supplementary Figure 19. Direct interactions between the promoters of the CuE

biosynthetic candidate genes and the potential regulators from watermelon. (a)

Yeast one-hybrid analysis of the interaction between the regulators ClBr and ClBt and

promoters of CuE biosynthetic candidates. P, promoter; AD, activation domain; 3-AT,

3-amino-1,2,4-triazole; SD-L-T, synthetic dropout media w/o leucine and tryptophan;

SD-L-T-H, synthetic dropout media w/o leucine, tryptophan, and histidine. (b) ClBr

and ClBt activate luciferase reporter expression by binding to promoters of CuE

biosynthetic genes in a tobacco leaf assay system. Images show activated (right,

reporter + effector) and non-activated (left, reporter only) status of infiltrated tobacco

leaves, respectively.

31

Supplementary Figure 20. Recruitment of ClBr and ClBt to the promoters of

CuE biosynthetic candidate genes. EMSA assays showed that ClBr or ClBt is

recruited to the promoter of Cl160 (a), Cl170 (b), Cl180 (c), ClBi (d), ClACT (e),

Cl710 (f), Cl490 (g), Cl890A (h), Cl890B (i), Cl510 (j), but not to the promoter of

Cl560B (k). The promoter regions from -2,000 upstream from the start codon are

represented schematically. The black vertical lines indicate the localization of

E-boxes and the red horizontal lines indicate the regions used in EMSA. His, His tag;

+/-, presence/absence of protein.

32

Supplementary Figure 21. No interaction between the promoters of the CuE or

CuB biosynthetic genes and other bHLH TFs. (a) Yeast one-hybrid analysis of the

interaction between Cm609 and the promoter of CmBi; Cl509 and the promoter of

ClBi. P, promoter; AD, activation domain; 3-AT, 3-amino-1,2,4-triazole; SD-L-T,

synthetic dropout media w/o leucine and tryptophan; SD-L-T-H, synthetic dropout

media w/o leucine, tryptophan, and histidine. (b) Cm609 and Cl509 failed to activate

the luciferase reporter by binding to promoters of CmBi and ClBi, respectively, in a

tobacco leaf assay system. Images show activated (right, reporter + effector) and

non-activated (left, reporter only) status of infiltrated tobacco leaves, respectively. E,

effector. (c) EMSA analysis showed that Cm609 and Cl509 are not recruited to the

promoters of the CmBi and ClBi genes, respectively. His, His tag; +/-,

presence/absence of protein.

33

Supplementary Figure 22. Identification and characterization of the genetic

variant associated with fruit bitterness of watermelon. (a) The premature

translational termination of the fruit-specific bitterness regulator ClBt caused by a

frameshift mutation is associated with fruit bitterness in watermelon. (b) Yeast

one-hybrid analysis of the interaction between ClBt or ClBtdel and the promoter of

the CuE biosynthetic gene ClBi. P, promoter; AD, activation domain; 3-AT,

3-amino-1,2,4-triazole; SD-L-T, synthetic dropout media w/o leucine and tryptophan;

SD-L-T-H, synthetic dropout media w/o leucine, tryptophan, and histidine. ClBtdel

represents the truncated ClBt. (c) ClBtdel failed to activate expression of the

luciferase reporter by binding to the promoter of the CuE biosynthetic gene ClBi in a

tobacco leaf assay system. Images show activated (right, reporter + effector) and

non-activated (left, reporter only) status of infiltrated tobacco leaves, respectively. E,

effector. (d) EMSA assay showed that ClBtdel cannot recruited to the promoter of the

ClBi gene. His, His tag; +/-, presence/absence of protein.

34

Supplementary Table 3. 1H NMR (500 MHz) data of compounds 1-3. J in Hz and δ

in ppm.

No. 1 2 3

1 1.46, m; 1.49, m 1.40, m; 1.46, m 1.44, m; 1.58. m

2 1.65, m; 1.78, m 1.64, m; 1.76, m 3.90, m

3 3.47, m 3.46, m 3.45, d, J=1.9 Hz

6 5.66, d, J=5.9 Hz 5.65, d, J=6.0 Hz 5.71, d, J=5.7 Hz

7 1.90, m; 2.37, m 1.91, m; 2.37, m 1.92, m; 2.38, m

8 1.94, d, J=7.5 Hz 1.94, d, J=8.0 Hz 1.97, d, J=8.0 Hz

10 2.29, d, J=12.0 Hz 2.26, d, J=12.0 Hz 2.30, d, J=12.5 Hz

11

12 2.43,d, J=14.5 Hz

2.92,d, J=14.5 Hz

2.52, d, J=13.4 Hz

2.93, d, J=13.4 Hz

2.54, d, J=14.5 Hz

2.91, d, J=14.5 Hz

15 1.26, m; 1.35, m 1.29, m; 1.37, m 1.33, m; 1.37, m

16 1.85, m; 2.02, m 1.90, m; 1.99, m 1.92, m; 1.96, m

17 1.70, m 2.03, t, J=9.3 Hz 2.03, t, J=9.5 Hz

18 0.94, s 0.91, s 0.92, s

19 1.11, s 1.11, s 1.12, s

20 1.40, m

21 0.89, d, J=6.4 Hz 1.27, s 1.27, s

22 1.05, m; 1.44, m 1.39, m; 1.46, m 1.39, m; 1.44, m

23 2.00, m 1.83, m 1.84, m

24 5.08, t, J=6.5 Hz 5.08, t, J=7.0 Hz 5.08, t, J=7.0 Hz

26 1.68, s 1.67, s 1.67, s

27 1.59, s 1.60, s 1.61, s

28 1.16, s 1.16, s 1.22, s

29 1.02, s 1.02, s 1.00, s

30 1.01, s 1.03, s 1.01, s

35

Supplementary Table 4. 13C NMR (125 MHz) data of compounds 1-3. δ in ppm.

No. 1 2 3

1 20.7 20.6 29.4

2 28.6 28.6 68.5

3 76.3 76.2 78.7

4 41.7 41.7 41.7

5 139.6 139.6 137.7

6 120.8 120.7 121.3

7 24.1 23.9 24.0

8 43.9 43.1 43.0

9 48.8 49.0 48.8

10 35.5 35.5 33.8

11 214.9 214.7 214.3

12 48.5 48.7 48.3

13 48.9 48.9 48.9

14 49.5 49.9 49.9

15 34.5 33.9 33.9

16 24.9 23.1 23.1

17 49.6 50.9 50.9

18 16.9 18.8 18.9

19 20.2 20.2 20.2

20 35.7 75.0 75.0

21 18.5 26.2 26.2

22 36.3 44.2 44.3

23 27.9 21.9 21.9

24 125.0 124.3 124.3

25 131.3 132.0 132.1

26 25.9 25.8 25.9

27 17.8 17.8 17.8

28 25.5 25.5 25.5

29 27.4 27.4 26.7

30 18.3 18.5 18.4