Research Article

Biosynthesis, regulation, and domestication of bitterness in cucumber

See allHide authors and affiliations

Science  28 Nov 2014:
Vol. 346, Issue 6213, pp. 1084-1088
DOI: 10.1126/science.1259215


Cucurbitacins are triterpenoids that confer a bitter taste in cucurbits such as cucumber, melon, watermelon, squash, and pumpkin. These compounds discourage most pests on the plant and have also been shown to have antitumor properties. With genomics and biochemistry, we identified nine cucumber genes in the pathway for biosynthesis of cucurbitacin C and elucidated four catalytic steps. We discovered transcription factors Bl (Bitter leaf) and Bt (Bitter fruit) that regulate this pathway in leaves and fruits, respectively. Traces in genomic signatures indicated that selection imposed on Bt during domestication led to derivation of nonbitter cucurbits from their bitter ancestors.

Biosynthetic pathway holds roots of domestication

The wild cucumber is a spiky, bitter relative of what we now grow in our gardens. The bitterness comes from cucurbitacin, which helps the plant to ward off herbivores. Cucurbitacin is also useful to people for its anti-tumor properties. Shang et al. have now worked out the biosynthetic pathway of cucurbitacin. Along the way, they discovered genetic traces of the domestication process and unraveled the mystery of why some cucumbers, if grown in chilly conditions, become bitter.

Science, this issue p. 1084

Plant specialized metabolites play essential roles in mediating interactions between the plant and its environment and constitute a valuable resource in discovery of economically important molecules. In the plant family Cucurbitaceae, a group of highly oxygenated tetracyclic and bitter triterpenes, the cucurbitacins, mediated the coevolution between cucurbits and herbivores. They serve either as protectants against generalists or feeding attractants to specialists (13). Widely consumed as vegetables and fruits, cucurbits were domesticated from their wild ancestors that had extremely bitter fruits. Drought and temperature stress can increase the bitterness in certain domesticated cultivars, which can affect fruit quality and marketability. Molecular insights into the occurrence and domestication of bitterness in cucurbits remain largely unknown.

Despite their presence in fruits as a negative agricultural taste, cucurbitacins have for centuries been exploited for anti-inflammatory and hepatoprotective activities, in the form of traditional herbal medicines (4, 5). Bitter fruits and leaves of wild cucurbit plants have been used as a purgative and emetic in India (6). The bitter fruit stem of melon (in Chinese, “gua di”) is prescribed as a traditional hepatoprotective medicine whose effect and usage were well documented in Ben Cao Gang Mu, the Chinese Encyclopedia of Botany and Medicines composed by the Ming Dynasty physician Li Shi-Zhen in 1590 CE. Recent studies revealed that cucurbitacins can cause cell-cycle arrest, apoptosis, and growth suppression of cancer cells through the specific inhibition of the Janus kinase–signal transducers and activators of transcription (JAK-STAT) pathway (7, 8). At present, their low concentrations in plants and nonspecific cytotoxicity limit their pharmaceutical applications.

To date, plant metabolic diversification studies (9, 10), as well as recently reported gene clusters in plants [reviewed in (11)], indicate that clustering of functionally-related genes for the biosynthesis of secondary metabolites may well be a common feature of plant genomes. In cucumber, two interacting Mendelian loci were reported to control the bitterness, conferred predominantly by cucurbitacin C (CuC) (3, 12). The Bi gene (1) confers bitterness to the entire plant and is genetically associated with an operon-like gene cluster (13), similar to the gene cluster involved in thalianol biosynthesis in Arabidopsis (14). Fruit bitterness requires both Bi and the dominant Bt (Bitter fruit) gene. Nonbitterness of cultivated cucumber fruit is conferred by bt, an allele selected during domestication as indicated by population genomics (15). Exploiting these genetic clues, here we report the discovery of 11 genes involved in the biosynthesis, regulation, and domestication of cucumber bitterness.

First committed step in CuC biosynthesis

To identify genetic variants associated with Bi, a genome-wide association study was performed based on the variation map (15) of 115 diverse cucumber lines (Fig. 1A and table S1). The most significant single-nucleotide polymorphism (SNP) was located within the region where Bi had been mapped and resulted in a nonsynonymous change from cysteine (C) to tyrosine (Y) at residue 393 (C393Y) of the cucumber gene Csa6G088690 (Fig. 1B). In the 115 lines, this SNP explained the phenotype in all but one line, CG7744. In-depth analysis of the variation map identified a 1–base pair (bp) deletion at Csa6G088690 in CG7744 that resulted in a frameshift mutation at the 760th amino acid residue (FS760) (Fig. 1B). Genetic analysis pinpointed that Csa6G088690 defines the Mendelian Bi gene (fig. S1A).

Fig. 1 The Bi gene.

(A) Genome-wide association study for the bitter foliage trait. Red arrow, most significant association. Scale, –log10 of P value of SNPs. (B) Amino acid alignment between wild Csa6G088690 and two mutant alleles. (C) GC-MS analysis of extracts prepared from yeast INVSc1 that harbored Bi, two mutant alleles (C393Y and FS760), empty vector, or an authentic standard. TIC, total ion chromatograms; EIC 498, extracted ion chromatograms at a mass/charge ratio (m/z) of 498 [M+TMS (trimethysilyl)].

Bi is a member of the oxidosqualene cyclase (OSC) gene family. Phylogenetic analysis showed that Bi is the ortholog of cucurbitadienol synthase gene CPQ in squash (Cucurbita pepo) (16) (fig. S1B). We next used yeast to express Bi, as well as its two mutant alleles, C393Y and FS760, to test its biochemical function. As revealed by gas chromatography–mass spectrometry (GC-MS) analysis, formation of cucurbitadienol occurred only in the yeast strain expressing the wild-type gene (Fig. 1C and fig. S1, C and D). Thus, in cucumber, Bi encodes a cucurbitadienol synthase that catalyzes the cyclization of 2,3-oxidosqualene into the tetracyclic cucurbitane skeleton, the first committed step of CuC biosynthesis (fig. S1E).

A leaf-specific regulator of Bi

To investigate the molecular mechanism in regulating CuC biosynthesis, we searched for naturally occurring mutants and screened an ethylmethane sulfonate–induced cucumber mutant library and subsequently identified two nonbitter mutants (XY-3 and E3-231). The foliage expression level of Bi in the natural mutant XY-3 was reduced to less than 1% of that in the bitter isogenic line XY-2 (Fig. 2A), which indicated that cucurbitacin biosynthesis is disrupted in XY-3. Genomes of XY-2 and XY-3 were resequenced and compared to identify possible mutations. A SNP in the cucumber gene Csa5G156220 caught our attention, as it encodes a putative basic helix-loop-helix (bHLH) transcription factor (TF) expressed specifically in leaves (table S2). The mutation resides at the splicing site of the predicted intron that likely disrupts proper gene transcription (Fig. 2, A and B).

Fig. 2 The Bl gene.

(A) Expression of Bl (Csa5G156220) and Bi in nonbitter mutant XY-3 and bitter XY-2 cucumber lines (means ± SEM, n = 3). (B) Sequence alignment between wild Bl and two mutated alleles. (C and D) Transient expression of Bl in cotyledons complemented the nonbitter phenotype of XY-3. (C) Expression of Bl and Bi determined 7 days after agroinfiltration (means ± SEM, n = 6). Value obtained from control (CK) was set to 1 and used to obtain relative values for the test sample. INF, sample infiltrated with Bl; CK, sample infiltrated with empty vector. (D) Presence or absence of CuC detected by high-performance liquid chromatography (HPLC) analysis of extracts prepared from Bl or control infiltrated cotyledons. mAU, milli–arbitrary units. (E) Schematic of the Bi promoter region (2000 bp upstream of the start codon). Black vertical lines indicate locations of E-box motifs, and red horizontal lines indicate regions amplified in ChIP assays or used in EMSA. Localization of mutated E-box used in EMSA is indicated in red. (F) ChIP analysis of Bl recruitment to the Bi promoter region by PCR. ChIP assays conducted with or without (+/−) Myc antibody. INF, sample infiltrated with Bl; CK, sample infiltrated with empty vector. (G) qPCR analysis of Bl recruitment to the indicated Bi promoter region (means ± SEM, n = 3). (H) EMSA showing that Bl-His specifically binds, in vitro, to the E-box region within the Bi promoter. Lane identified by a red triangle indicates that the E-box element within the probe has been mutated from CANNTG to GANNTG. Comp, competitor (unlabeled probe); His, His-tag; +/−, presence or absence of protein or competitor; closed triangle, increasing amount of protein or competitor.

Resequencing of E3-231 revealed another SNP located within Csa5G156220 that caused a change from arginine (R) to lysine (K) at the 85th amino acid residue (R85K), which is located inside the basic domain (Fig. 2B). This mutation may affect regulatory ability, as the basic domain is essential for DNA binding ability for bHLH TFs (17). Genetic analyses showed that the mutations in XY-3 and E3-231 are actually two recessive alleles of the same gene (fig. S2A). Increased expression of both Bi and Csa5G156220 was also observed in cucumber plants either exposed to drought stress or treated with the phytohormone ABA (fig. S2, B and C), which indicated that abiotic stress may stimulate the bitterness biosynthesis in cucumber by up-regulation of Csa5G156220.

A cucumber cotyledon transient agro-infiltration expression system was developed to further confirm the in vivo function of Csa5G156220 (18). Increasing expression of Csa5G156220 in XY-3 cotyledons up-regulated expression of Bi, which in turn functionally complemented the nonbitter phenotype (Fig. 2, C and D, and fig. S3, A and B). Thus Csa5G156220 regulates the bitterness biosynthesis in cucumber leaves, and hence, this gene was named Bl (Bitter leaf).

Next, we investigated how Bl regulates Bi. Analysis of the Bi promoter region revealed the occurrence of seven E-box (CANNTG) sequences (Fig. 2E), a cis-element to which bHLH TFs could potentially bind (17). Yeast one-hybrid (Y1H) assay and a tobacco transient reporter (luciferase) activation system showed that Bl indeed could bind to this promoter (fig. S2, D and E). Chromatin immunoprecipitation (ChIP) assays were performed by using formaldehyde-fixed cotyledons of XY-3 that were transiently expressing a Bl-Myc fusion protein. As revealed by the polymerase chain reaction (PCR) products and quantitative real-time PCR (qPCR), Bl was selectively recruited to the Bi promoter region containing E-box elements (Fig. 2, F and G). Electrophoretic mobility-shift assays (EMSAs) also confirmed selective binding of Bl to the E-box elements within the Bi promoter (Fig. 2H). Thus, Bl regulates cucurbitacin biosynthesis by activating transcription of Bi in cucumber leaves.

A cucumber domestication gene

Bt was previously mapped to a 442-kilobase (kb) region on chromosome 5 that harbors 67 predicted genes (15). Bl and its two homologs (Csa5G157220 and Csa5G157230) are among these candidates and clustered in an 8.5-kb region (Fig. 3A). As Bl positively regulates Bi in cucumber leaves, we considered Csa5G157230 to be a candidate for Bt, given that it is specifically expressed in the fruit of the wild line, PI 183967, consistent with the distribution of bitterness in these plants (Fig. 3A and table S2). In addition, positive correlations were observed between expression levels of Csa5G157230 and Bi, and between fruit bitterness and gene expression in various cucumber lines, especially in those five extremely bitter wild lines (Fig. 3B). These studies established a correlation between Csa5G157230 expression and accumulation of bitterness in the fruit.

Fig. 3 The Bt gene.

(A) The Bt-mapped region on chromosome 5 overlaps with a large domestication sweep region showing almost zero nucleotide diversity in the domesticated population (top). Differential expression profiles of genes predicted within the Bt region illustrated by a gradient in red (bottom). Numeric expression values of predicted genes are shown in table S2. Candidate Bt gene is indicated in red. CuC content of wild and cultivated cucumber was compared (means ± SEM, n = 3). WF, wild fruit; CF, cultivated fruit; WL, wild leaf; CL, cultivated leaf. (B) High consistency observed between expression of Bt, Bi, and the CuC content in 21 cucumber lines, including five extremely bitter lines (means ± SEM, n = 3, indicated in red). Presence or absence of SV-2195 indicated by +/−. Genotype of SNP-1601 (Y: A or G, U: unknown). (C) High consistency among cold-stress treatment: expression of Bt, Bi, and CuC content in fruit of HAN, han, and F1 individual plants (means ± SEM, n = 3). (D and E) Transient expression of Bt in fruit complemented the nonbitter phenotype of cucumber line XinTaiMiCi-2. (D) Expression of Bt and Bi determined 15 days after agroinfiltration (means ± SEM, n = 3). Value obtained from control (CK) was set to 1 and used to obtain relative values for the test sample. INF, sample infiltrated with Bt; CK, sample infiltrated with empty vector. (E) Presence or absence of CuC detected by HPLC analysis of extracts prepared from Bt or control infiltrated fruits 15 days after agroinfiltration. mAU, milli–arbitrary units.

Next, we performed a local association analysis within the 442-kb region to further identify genetic variants associated with the extremely bitter phenotype. This led to finding 11 signals at the regulatory region of Csa5G157230, including 10 SNPs and one structural variant, a 699-bp insertion 2195 bp upstream of the Csa5G157230 start codon (SV-2195). Another variant was also identified at the regulatory region of Csa5G157230, a SNP at the 1601 bp upstream of the start codon (SNP-1601), which cosegregated with the Bt locus in a large F2 population (n = 1822). In the 115 lines, 22 carrying a homozygous “A” at SNP-1601 all bear nonbitter fruits (table S3). These analyses indicated that selection at the regulatory region of Csa5G157230 may down-regulate Csa5G157230 expression in cultivated lines, which results in reduced fruit bitterness.

In some cucumber lines, fruits become bitter under stress conditions. For instance, the fruits of the cucumber line HAN become bitter when plants were grown at a low temperature (18°C day, 12°C night), whereas, at a normal temperature (30°C day, 22°C night), the fruits are not bitter. We identified a natural HAN mutant (han), whose fruits were nonbitter even under such low temperature conditions. Resequencing both lines revealed a mutation corresponding to SNP-1601 (G in HAN and A in han). Genetic analysis showed that SNP-1601 cosegregates with the phenotype (fig. S4A). Our qPCR analysis indicated that SNP1601 is essential for regulating Bi expression in response to this environmental factor (Fig. 3C).

To confirm the in vivo function of Csa5G157230, a fruit transient gene expression system was developed (18). Expression of Csa5G157230 activated transcription of Bi and promoted biosynthesis of CuC in the fruit (Fig. 3, D and E). In parallel experiments, we expressed Csa5G157230 in XY-3 cotyledons, with the method described above. An increase in CuC content in the infiltrated XY-3 tissue was also observed (fig. S3C), which indicated that the TFs, Bl, and Csa5G157230 have a similar biochemical function and that they control CuC biosynthesis in different organs. Next, we tested whether Csa5G157230 could directly regulate the Bi gene. Here, we expressed the Myc-tagged protein in cotyledons of XY-3 to prepare sufficient material for ChIP assays. Similar to Bl, Csa5G157230 could bind to the E-box elements within the Bi promoter (fig. S4, B to F). Taken together, these studies provide strong support for the hypothesis that Csa5G157230 is the Bt gene, which activates Bi and regulates CuC biosynthesis in the fruit.

Nine genes in CuC biosynthetic pathway

To catalyze the formation of CuC, cucurbitadienol has to be further modified with a series of oxidation reactions and acetylation, likely catalyzed by cytochrome P-450 enzymes (P450s) and an acyltransferase (ACT). On chromosome 6, Bi colocalizes with four P450 genes (Csa6G088160, Csa6G088170, Csa6G088180, and Csa6G088710) and one ACT gene (Csa6G088700) within a 35-kb genomic region. Except for Csa6G088180, all other genes shared nearly identical expression patterns, with high expression occurring in leaves of line 9930 as well as in fruits of wild line PI 183967 (Fig. 4, A and B). In addition, these coexpressed genes were down-regulated in leaves of XY-3 as compared with XY-2 and in fruits of han as compared to HAN, and they were up-regulated in cucumber leaves under ABA treatment or drought stress (Fig. 4, C to F, and table S4). Furthermore, our studies showed that Bl and Bt could specifically bind to the promoters of these coexpressed genes and could activate their transcription (Fig. 4G, and figs. S5 and S6). Mutation (R85K) within the basic domain of Bl appeared to affect its binding ability to the CuC biosynthetic genes (fig. S5, C and D), which in turn is likely to result in the nonbitter phenotype of cucumber (E3-231). Although the Y1H assay showed that Bt could also interact with the promoter of Csa6G088180 (fig. S5A), Bt cannot activate Csa6G088180’s transcription (figs.S5B and S6C).

Fig. 4 Nine pathway genes that are coordinately regulated.

(A and B) Identification of coexpressed candidate enzymes by analyzing transcriptomic data acquired from cultivar 9930 (A) and wild line PI 183967 (B). Candidate enzymes are indicated with different colors according to their annotations. Low-expressed gene Csa6G088180 is indicated with hatched green and was used as a negative control in the following analyses. (C to F) Coregulation of candidate genes (means ± SEM, n = 3). Down-regulation of the nine genes in XY-3 as compared with XY-2 (C) and han as compared with HAN (D) (asterisk indicates samples prepared from plants grown under low temperature), and up-regulation of the nine genes in the presence of ABA treatment (E) or drought stress (F). (G) Summary of interaction between candidate gene promoter and Bl or Bt. Luc, luciferase trans-activation assay. (H) Function of enzymes elucidated by transient RNAi assays (means ± SEM, n = 6). RNAi sample in blue; CK in red. Value obtained from control (CK) was set to 1 and used to obtain relative values for the RNAi sample. CK, sample infiltrated with empty vector. More information is provided in figs. S5 to S7.

We failed in a search for the specific P450 within the cluster responsible for oxidizing cucurbitadienol, which suggests there should be other candidates located outside this 35-kb genomic region. We reasoned that other genes would be coexpressed with the Bi cluster and coregulated by Bl and Bt. Therefore, by applying the integrative bioinformatics and molecular biology approach described above, we identify four additional P450 genes (three on chromosome 3, Csa3G698490, Csa3G903540, and Csa3G903550, and one on chromosome 1, Csa1G044890) that are coexpressed with the Bi cluster and are activated by Bl and Bt in leaves and fruits, respectively (Fig. 4, A to F, and table S4).

The relation of CuC biosynthesis and these candidate tailoring enzymes was probed by using a transient RNA interference (RNAi) system acting on the bitter cotyledon of the cucumber line 9930 (18). RNAi-mediated target-specific down-regulation of transcripts for all these candidate genes resulted in a decrease in CuC content in the infiltrated cotyledons (Fig. 4H and fig. S7). Thus, Bl and Bt regulate bitterness formation in leaves and fruits, respectively, by direct transactivation of nine genes (one OSC, seven P450s, and one ACT) involved in the CuC biosynthetic pathway.

Three more steps in CuC biosynthesis

To characterize the biochemical function of these candidate P450s, we expressed each P450 in the engineered yeast (EY10) that accumulates 10 times as much cucurbitadienol as its original strain (18) (fig. S8). No expected product was detected from yeast extract at first (Fig. 5A). However, once an NADPH–cytochrome P450 oxidoreductase gene (CPR, Csa1G423150) was expressed with candidate P450 in the EY10, we detected a specific product catalyzed by Csa3G903540 (a member of CYP88 family, located outside the Bi cluster) (Fig. 5A). The structure of this purified product (compound 1) was interrogated by nuclear magnetic resonance (NMR) spectroscopy (figs. S9 and S10), which indicated that it was a derivative of cucurbitadienol in which the 19-CH3 was hydroxylated. The product of Csa3G903540 was named 19-hydroxy cucurbitadienol.

Fig. 5 Three more catalytic steps.

(A) GC-MS analysis of putative product (red arrow) generated by Csa3G903540 in the engineered yeast (EY10). Partial enlarged details are shown as insets. TIC, total ion chromatograms; EIC 586, extracted ion chromatograms at m/z of 586. The product structure (right) was elucidated by NMR (18). (B) Ultra-performance liquid chromatography coupled with quadropole time-of-flight mass spectrometry (UPLC-qTOF-MS) analysis of yeast extracts with electrospray ionization (ESI) on positive mode. EIC 459.3833, extracted ion chromatograms of the accurate parent ion at m/z of 459.3833. The product (indicated by red arrow) structure was elucidated by MS/MS and NMR (18). (C) UPLC-qTOF-MS analysis of the acetyltransferase catalytic reaction product. Deacetyl-CuC is acetylated by Csa6G088700-His in vitro (indicated by red arrow). A leaf-specific ACT (Csa5G639480) served as a negative control. Schematic of this biosynthetic pathway from deacetyl-CuC to CuC is shown at right.

We continued to search for downstream P450s using this same approach. As revealed by liquid chromatography–mass spectrometry (LC-MS) assays, we identified an expected peak in the yeast expressing Bi, CPR, Csa3G903540, and Csa6G088160 (a member of CYP81 family, located within the Bi cluster) (Fig. 5B). Tandem mass spectrometry (MS/MS) and NMR analysis revealed that a hydroxyl group was transferred to the C-25 position of 19-hydroxy cucurbitadienol and that the double bond between C-24,25 was shifted to the position of C-23,24 (figs. S11 and S12). The product (compound 2) of Csa6G088160 was named 19,25-dihydroxy cucurbitadienol.

From fresh bitter cucumber leaves, our NMR analysis identified a deacetyl CuC (figs. S13 and S14, compound 3). LC-MS analysis showed that the ACT enzyme (Csa6G088700) was able to acetylate this compound to yield CuC (Fig. 5C). These studies indicate that Csa6G088700 is the enzyme involved in the final step in the CuC biosynthetic pathway.

In summary, we discovered that two TFs regulate nine genes in the CuC biosynthetic pathway and propose a model as to how extremely bitter wild cucumber was domesticated into nonbitter cultivars (fig. S15). As revealed in this study, such regulators must contribute to the highly coordinated and efficient transcription of plant specialized metabolic pathways. The new knowledge on cucurbitacin biosynthesis will open a door for biological manufacturing and engineering of these triterpenoids as anti-tumor drugs, for example, in a manner similar to the biosynthesis of artemisinic acid, the antimalarial drug precursor (19, 20).


Materials and Methods

Figs. S1 to S15

Tables S1 and S8

References (2124)


  1. Materials and methods are available as supplementary material on Science Online.
  2. Acknowledgments: We thank J. Bohlmann and D.-K. Ro for critical comments on the manuscript and J.-J. Qi, X.-Z. Lin, T. Lin, X.-F. Xue, and X.-Y. Liu for bioinformatic and experimental assistance. The P450s were named according to the alignment made by D. Nelson ( This work was funded by the National Program on Key Basic Research Projects in China (the 973 Program; 2012CB113900), National Science Fund for Distinguished Young Scholars (31225025), National Natural Science Foundation of China (31272161, 31322047, and 31101550), Agricultural Science and Technology Innovation Program, and National Key Technology R&D Program (the 863 Program; 2012BAI29B04). This work was also supported by the Shenzhen Municipal and Dapeng District Governments. The Institute of Flowers and Vegetables has three pending patent applications relating the genes reported in this study. Supplementary materials contain additional data. This whole-genome shotgun project has been deposited at DNA Data Bank of Japan/European Molecular Biology Laboratory/GenBank under the accession ACHR00000000. The version described in this paper is version ACHR02000000. Genes reported in the study are deposited in the National Center for Biotechnology Information (NCBI), NIH, with accession numbers (KM655851–KM655862, KM677686–KM677688). RNA-seq data may be obtained from NCBI with the accession number SRA046916.
View Abstract

Navigate This Article