Decreased Protein Abundance of Lycopene β-Cyclase Contributes to Red Flesh in Domesticated Watermelon

The gene ClLCYB contributes to the red flesh color of domesticated watermelon through decreased protein level to convert lycopene to β-carotene in the carotenogenesis pathway.

Abstract

Red-fleshed watermelons (Citrullus lanatus) that accumulate lycopene in their flesh cells have been selected and domesticated from their pale-fleshed ancestors. However, the molecular basis of this trait remains poorly understood. Using map-based cloning and transgenic analysis, we identified a lycopene β-cyclase (ClLCYB) gene that controls the flesh color of watermelon. Down-regulation of ClLCYB caused the flesh color to change from pale yellow to red, and ClLCYB overexpression in the red-fleshed line caused the flesh color to change to orange. Analysis of ClLCYB single-nucleotide polymorphisms using 211 watermelon accessions with different flesh colors revealed that two missense mutations between three haplotypes (ClLCYB^red, ClLCYB^white, and ClLCYB^yellow) were selected and largely fixed in domesticated watermelon. Proteins derived from these three ClLCYB haplotypes were localized in plastids to catalyze the conversion of lycopene to β-carotene and showed similar catalytic abilities. We revealed that ClLCYB protein abundance, instead of ClLCYB transcript level, was negatively correlated with lycopene accumulation. Different amounts of ClLCYB protein degradation among the ClLCYB haplotypes were found in ClLCYB transgenic Arabidopsis (Arabidopsis thaliana) lines. After treatment with the proteasome inhibitor MG132, the concentration of ClLCYB^red increased noticeably compared with other ClLCYB proteins. These results indicate that natural missense mutations within ClLCYB influence ClLCYB protein abundance and have contributed to the development of red flesh color in domesticated watermelon.

The flesh color of watermelon (Citrullus lanatus) is an important trait predominantly determined by the composition and levels of different carotenoids that accumulate within flesh cells. Cultivated watermelons exhibit a wide range of flesh colors (Holden et al., 1999; Tadmor et al., 2005). Red-fleshed watermelon accumulates lycopene, a naturally occurring red-colored carotenoid, in flesh cells (Collins et al., 2006). Henderson et al. (1998) described five different colors of watermelon flesh: white, salmon yellow, canary yellow, orange, and red. The flesh of wild and semi-wild watermelon species is white, with little carotenoid accumulation (Holden et al., 1999; Tadmor et al., 2005). Cultivated canary yellow and yellow fruits accumulate violaxanthin and neochrome (Bang et al., 2010). Orange varieties show increased levels of β-carotene at the expense of lycopene (Tadmor et al., 2005). Previous works have reported that red-fleshed watermelons contain more lycopene per unit of fresh weight than tomato (Solanum lycopersicum) and that this lycopene is easier for the body to absorb (Edwards et al., 2003; Naz et al., 2014). Considering the nutritional and sensory functions of lycopene and other carotenoids, increasing the carotenoid content is the main objective in watermelon fruit quality improvement.

Modern, intensely colored cultivated watermelons have been selected from their pale-colored ancestors through the long history of watermelon domestication (Chomicki and Renner, 2015; Paris, 2015; Guo et al., 2019). The wild watermelon species Citrullus colocynthis and Citrullus amarus, found in Africa, have pale-colored and bland- or bitter-tasting flesh (Guo et al., 2013; Paris, 2015). The closest wild relative of sweet dessert qwatermelon is Citrullus mucosospermus (also called C. lanatus ssp. mucosospermus), which exhibits a wide range of fruit flavors and color patterns (Chomicki and Renner, 2015; Guo et al., 2015, 2019). Modern cultivated watermelons are characterized as sweet tasting and tender, with intensely colored flesh that ranges in hue from yellow to deep red (Watanabe et al., 1987; Tadmor et al., 2005). The red-colored flesh trait was selected early and continually during the domestication process (Paris, 2015). This overall process followed the general pattern of crop domestication in which domesticated plants were derived from small samples of wild source populations (Ladizinsky, 1985; Paris, 2015). It is currently unclear whether watermelons with red flesh and other flesh colors, such as yellow and orange, were domesticated independently or whether one was selected from another. The domestication footprints and molecular mechanism of red flesh color selection are still unknown.

Information regarding the inheritance of watermelon flesh color has been previously reported (Poole, 1944; Robinson et al., 1976; Henderson, 1989, 1991). For example, Shimotsuma (1963) reported that the flesh color in an F2 population derived from a cross between a white-fleshed wild line and a red-fleshed cultivar displayed a ratio of 12 white (Wf_ _ _):3 yellow (wfwfB_):1 red (wfwfbb). Wf is considered a major flesh color-controlling locus with epistatic effects, and B is thought to control the yellow flesh color. We previously localized Wf on chromosome 2 and pinpointed the B gene in a small region on chromosome 4 that contained a gene orthologous to lycopene β-cyclase (LCYB; Zhang et al., 2014). The canary yellow color (C gene) is dominant to the pink (c) color but epistatic to the red (Y) color (Poole, 1944; Henderson et al., 1998). Bang et al. (2007) reported that the watermelon C locus, associated with the dominance of canary yellow-colored flesh over red-colored flesh (c), contained the LCYB gene. The LCYB gene was also located in the quantitative trait locus responsible for lycopene content in watermelon flesh (Liu et al., 2015). The locus controlling red flesh has been named b, c, or Y based on different crossing combinations (Porter, 1937; Poole, 1944; Henderson et al., 1998). However, the candidate red-flesh (rf) gene remains unidentified. LCYB catalyzes the conversion of lycopene to β-carotene and represents a key point in the carotenoid biosynthetic pathway (Cunningham, 2002; Li and Yuan, 2013; Sun et al., 2018). In tomato, many color mutants result from mutations in the chromoplast-specific lycopene β-cyclase (CYCB) gene (Ronen et al., 1999, 2000). However, the function and mechanism of LCYB during watermelon flesh color domestication remain unknown.

Here, we describe ClLCYB as the rf gene in watermelon. The accumulation of lycopene in watermelon flesh is correlated with changes in ClLCYB protein levels instead of transcript levels. Natural missense mutations in ClLCYB influence ClLCYB protein abundance and contribute to the red flesh color in domesticated watermelon.

RESULTS

Identification of ClLCYB as the Candidate rf Gene

To identify the rf gene in watermelon, we first developed an F2 mapping population by crossing the red-fleshed cv 97103 and the pale yellow-fleshed cv Cream of Saskatchewan (CS). The carotenoid composition and content in these two parents and wild white-fleshed watermelon ‘PI296341-FR’ are listed in Table 1. The F2 population was evaluated to determine flesh color inheritance (Fig. 1A). In the F1 generation, all fruits had pale yellow flesh. The F2 plants segregated at a ratio of 784 (pale yellow):254 (red). The result of the χ² goodness-of-fit test fit a 3:1 segregation ratio (Table 2), indicating that the difference in flesh color between the red-fleshed cv 97103 and cv CS is controlled by a single gene. Using map-based cloning, we mapped the gene to an ∼129-kb region on chromosome 4 (between 15,370,022 and 15,499,154) of the watermelon genome (v2; Guo et al., 2019). The region contained five predicted protein-coding genes (Fig. 1B; Supplemental Table S1). Among these genes, Cla97C04G070940, which was annotated as LCYB, was the only one predicted to be related to carotenoid biosynthesis (Fig. 1B). Cla97C04G070940 shared 96% nucleotide identity with melon (Cucumis melo) LCYB and 95.8% and 71% identity with the characterized cucumber (Cucumis sativus) and tomato LCYBs, respectively. We therefore considered Cla97C04G070940 (named ClLCYB) as the candidate rf gene.

Table 1. Carotenoid composition and content in the flesh of cv 97103, cv CS, and wild white-fleshed PI296341-FR watermelons.

Values are given in mg kg^–1 fresh weight. ND, Not detected.

Carotenoids Detected	Watermelon Fruits
	97103	CS	PI296341-FR
Phytoene/phytofluene	Trace	Trace	Trace
ζ-Carotene	Trace	ND	Trace
Lycopene	45.361	ND	ND
β-Carotene	4.356	0.215	0.005
α-Carotene	ND	ND	0.125
Lutein	0.013	0.017	0.052
Zeaxanthin	ND	ND	Trace
Violaxanthin	ND	0.245	ND
Neoxanthin	ND	0.328	Trace

Figure 1.

Map-based cloning of the rf-controlling gene in cv 97103. A, Fruit phenotypes of red-fleshed cv 97103, pale yellow-fleshed cv CS, and their F1 and F2 populations. Bars = 5 cm. B, The candidate locus was mapped to chromosome 4 between markers BVMS00661 and BVWS00807. The numerals indicate the number of recombinants identified among 254 F2 red-fleshed plants. The candidate gene was narrowed to a 129-kb genomic DNA region between left and right flanking markers and cosegregated with marker P. A detailed view of the ORFs within the 129-kb mapping interval shows the five candidate ORFs. LYCB-like Cla97C04G070940 (ORF4, ClLCYB) is a candidate rf gene. C, Variation in ClLCYB protein levels in cv 97103 and cv CS. aa, Amino acids.

Table 2. Segregation of pale yellow and red flesh colors in the F1 and F2 progeny derived from a cross between '97103' and 'CS' watermelon plants.

PA, Parent A; PB, parent B.

Generation	Genotype	Observed, Pale Yellow:Red	Expected, Pale Yellow:Red	χ2	P
97103 (PA)	rfrf	0:10	0:1	0	1
CS (PB)	RFRF	10:0	1:0	0	1
F1	RFrf	12:0	1:0	0	1
F2	Segregating	784:254	3:1	0.0790	0.7786

RACE was first used to obtain the full-length cDNA sequence of ClLCYB in both cv 97103 and cv CS (Supplemental Fig. S1). Based on our results, the 3′-untranslated region (UTR) of ClLCYB is 370 bp, and we predicted an upstream open reading frame (uORF) in the 5′-UTR in the two parental lines (Supplemental Fig. S1). The ClLCYB gene consisted of one exon in both parents and encoded proteins with 504 amino acids. In the coding sequence (CDS) region, there were two single-nucleotide polymorphisms (SNPs), namely G676T and G1305C, resulting in a Val (V)-to-Phe (F) substitution at amino acid 226 (V226F) and a Lys (K)-to-Asn (N) substitution at amino acid 435 (K435N), respectively (Fig. 1C). Reverse transcription quantitative PCR (RT-qPCR) revealed that ClLCYB was expressed ubiquitously in the examined watermelon organs, including roots, stems, leaves, flowers, and fruits, but ClLCYB expression was highest in leaves, flowers, and fruits (Supplemental Fig. S2).

Watermelon Flesh Color Is Altered by Transgenic Manipulation of ClLCYB

To verify that ClLCYB is the corresponding rf gene, we transformed a ClLCYB antisense construct into the pale yellow-fleshed cv CS and overexpressed ClLCYB in the red-fleshed cv ZZJM line. The resulting antisense transgenic population contained 26 independent T0 plants. Eleven transgenic T0 plants with reduced ClLCYB mRNA levels were self-pollinated to generate T1 and T2 plants; three of these lines (cllcybR1, cllcybR2, and cllcybR3) that displayed obvious changes in flesh color are shown in Figure 2A. Line cllcybR12, which did not show any reduction in ClLCYB transcript levels, was used as the negative control (Fig. 2, A and C). Compared with that in the pale yellow-fleshed control line, the lycopene accumulation in the three down-regulated transgenic lines increased significantly, changing the pale yellow flesh color to red (Fig. 2D). Other altered phenotypes, such as bleached leaves and stems, were observed in the transgenic lines (Fig. 2B). These results indicate that down-regulation of ClLCYB can lead to the generation of red-fleshed watermelons in pale-fleshed lines.

Figure 2.

Phenotypes of ClLCYB antisense transgenic lines. A, Longitudinal sections of the fruits of ClLCYB knockdown lines (cllcybR1–cllcybR3) and the control line (CK; line cllcybR12) at 34 d after pollination (DAP). Bars = 1 cm. B, Leaf and stem phenotypes of the ClLCYB knockdown lines presented in A. Bars = 10 cm. C, ClLCYB transcript levels in the ClLCYB knockdown lines presented in A. ClLCYB expression levels were measured by RT-qPCR. Watermelon ACTIN was used as an internal control. D, Lycopene content measured by HPLC in the transgenic fruits presented in A. The error bars indicate sd (n = 3). A significant difference (*P < 0.05 and **P < 0.01) was found in comparison with control plants. FW, Fresh weight.

Overexpression of ClLCYB in the red-fleshed line cv ZZJM resulted in an orange flesh color. Three T1 plants of the 19 identified independent T0 plants (cllcybOE1, cllcybOE2, and cllcybOE3) that displayed obvious changes in flesh color are shown in Figure 3A. Line cllcybOE18, which did not show any increase in ClLCYB transcript level, was used as the negative control (Fig. 3, A and C). Compared with the red-fleshed control line, the three ClLCYB overexpression lines displayed β-carotene accumulation, which changed the flesh color from red to orange (Fig. 3D). Other altered phenotypes in overexpression lines included yellow stems (Fig. 3B). These results show that up-regulation of ClLCYB can produce orange-fleshed fruits in red-fleshed cultivars.

Figure 3.

Phenotypes of ClLCYB-overexpressing transgenic lines. A, Longitudinal sections of fruits of ClLCYB overexpression lines (cllcybOE1–cllcybOE3) and the control line (CK; line cllcybOE18) at 34 DAP. Bars = 1 cm. B, Leaf and stem phenotypes of the ClLCYB overexpression lines presented in A. Bars = 5 cm. C, ClLCYB transcript levels in the ClLCYB overexpression lines presented in A. ClLCYB expression levels were measured by RT-qPCR. Watermelon ACTIN was used as an internal control. D, Lycopene and β-carotene contents were measured by HPLC in the transgenic fruits presented in A. The error bars indicate sd (n = 3). A significant difference (**P < 0.01) was found in comparison with control plants.

Selection Pressure on ClLCYB Underlies the Domestication of Red-Fleshed Watermelon

Our results above clarify the important role of ClLCYB in controlling red flesh formation in red-fleshed lines, and these results prompted us to hypothesize that ClLCYB plays an important role during watermelon domestication based on flesh color. We analyzed the sequence variation of the ClLCYB gene in a collection of 211 watermelon accessions belonging to four Citrullus species, C. colocynthis, C. amarus, C. mucosospermus, and C. lanatus (Zhang et al., 2016). C. lanatus was divided into the C.L. cultivar and C.L. landraces (Kim et al., 2013; Zhang et al., 2016; Zhao et al., 2016). These accessions were collected from various regions worldwide, and their fruits displayed various flesh colors (Supplemental Table S2). We sequenced the ClLCYB gene (including 2 kb upstream and downstream of the gene) and identified 189 SNPs in all 211 accessions (Supplemental Table S2). Phylogenetic reconstruction in which 211 × 189 SNPs were used revealed several interesting patterns of relatedness among the entries (Fig. 4A). As shown in Figure 4A, different base colors show the five species and subspecies, and the font colors mimic the flesh colors. We found that the red- and pink-fleshed watermelons formed a distinct clade, which was different from the subspecies classification.

Figure 4.

Phylogenetic analysis of ClLCYB. A, Phylogenetic tree of the ClLCYB alleles in 211 watermelon accessions. The phylogram shows ClLCYB generated from 39,879 SNPs from 211 watermelon accessions (four main watermelon subspecies: C. colocynthis [CC, denoted with a blue background], C. amarus [CA, denoted with a green background], C. mucosospermus [CM, denoted with a dark gray background], C.L. landrace [LR, denoted with a gray background], and C.L. cultivar [CL, denoted with a white background]). Flesh color is denoted by letter color: white (white-fleshed lines), gray (pale yellow-fleshed lines), yellow (yellow-fleshed lines), orange (orange-fleshed lines), pink (pink-fleshed lines), and red (red-fleshed lines). B, Graphs of π estimates for SNPs distributed within the ClLCYB locus. π estimates are shown for white and pale yellow, yellow and orange, and pink and red accessions. C, Representative genotypes of the SNPs and the different flesh color phenotypes of the five subspecies.

The genetic diversity analysis revealed little genetic diversity within the ClLCYB allele in the red-fleshed cultivars (mean nucleotide diversity [π] sil = 0.000216). By contrast, the diversity was greater at most loci of the ClLCYB allele in the yellow-fleshed cultivars and white-fleshed individuals (mean π sil = 0.094272 in yellow-fleshed individuals, mean π sil = 0.258859 in pale-colored individuals, averaged across all loci; Fig. 4B). This phenomenon was evidence of a selective sweep in ClLCYB during red-fleshed watermelon domestication. The results of association tests between flesh color and sequence variation showed that six SNPs were correlated with the red flesh phenotype (P < 0.001), three of which were located in the CDS region (Fig. 4C; Supplemental Table S2). The most correlated position was a missense amino acid position, V226F at G676T (P = 6.95E⁻¹⁹), followed by the 981-bp downstream position (P = 1.31E⁻¹⁵), the synonymous SNP L4L at A12G (P = 1.47E⁻¹⁵), the missense amino acid position K435N at G1305C (P = 1.68E⁻⁰⁶), the −1,836-bp upstream position (P = 1.54E⁻⁴), and the 818-bp downstream position (P = 4.92E⁻⁴). As shown in Figure 4C, all non-red-fleshed accessions had nucleotide T at the V226 sites, but the red-fleshed C. lanatus had G at the V226 site. Moreover, the SNPs were either G or A at the L4L site and G or C at the K435N site in the non-red-fleshed C. lanatus. The red-fleshed lines of C. lanatus were domesticated and had homozygous nucleotides at these correlated SNP sites. This finding indicates strong selection pressure on ClLCYB in the red-fleshed cultivars.

Three Domesticated Haplotypes of ClLCYB Are Present in the Same Subcellular Localization and Share Similar Catalytic Functions

Based on the two nonsynonymous SNPs, we revealed three haplotypes of ClLCYB among accessions with different flesh colors. Type I consisted of ClLCYB^red with Val (V) at residue 226 and Lys (K) at residue 435 in all the red-fleshed lines, including cv 97103. Type II consisted of ClLCYB^yellow with Phe (F) at residue 226 and Asn (N) at residue 435 in the cultivated lines that did not accumulate lycopene, such as cv CS and cv JLM. Type III consisted of ClLCYB^white with Phe (F) at residue 226 and Lys (K) at residue 435 in wild and semiwild white-fleshed watermelon lines. We aligned these three ClLCYB protein sequences, as shown in Figure 5A.

Figure 5.

Catalytic function and localization of three ClLCYB proteins. A, Three ClLCYB haplotypes found in watermelons with different flesh colors: ClLCYB^red with Val (V) at residue 226 and Lys (K) at residue 435 in all the red-fleshed lines; ClLCYB^yellow with Phe (F) at residue 226 and Asn (N) at residue 435 in the cultivated lines that did not accumulate lycopene, such as cv CS and cv JLM; and ClLCYB^white with Phe (F) at residue 226 and Lys (K) at residue 435 in wild and semiwild white-fleshed watermelons. B, Subcellular localization of three ClLCYBs. ClLCYB-mCherry (ClLCYB^white, ClLCYB^yellow, and ClLCYB^red) and mCherry vector (vector control) were transiently produced in watermelon fruit protoplasts incubated in the dark for 12 h. The protoplasts were observed under chromoplast autofluorescence (column 1), under bright field (column 2), under mCherry fluorescence (column 3), and in merged images (column 4) with a confocal microscope. Bars = 20 µm. C, HPLC analysis of the color complementation of a pAC-LYC bacterial line with three ClLCYB constructs. The pAC-LYC plasmid (Cunningham and Gantt, 2007) enables lycopene production and accumulation in E. coli. pET28a expression vectors (Promega) carrying the three ClLCYB constructs were transformed into pAC-LYC cells, with empty pET28a as a control. Carotenoid pigment composition was examined in cultures of E. coli containing the plasmid indicated above. HPLC detection was performed at 440 nm. D, Michaelis-Menten profiles for the activities of three ClLCYBs and the concentration of the substrate lycopene in 5 mm NADPH, pH 6.5, at 25°C.

We first examined the subcellular localization of these three types of ClLCYB proteins in watermelon fruit protoplasts. ClLCYB-mCherry fusion proteins and free mCherry vectors (controls) were transiently produced in flesh protoplasts that were isolated from fruits of cv 97103 at 14 DAP. The fluorescence signals of all of the ClLCYB-mCherry fusion proteins precisely coincided with the chromoplast autofluorescence (green), in contrast to the ubiquitous distribution of free mCherry (Fig. 5B). These results imply that all three ClLCYB types are localized within chromoplasts. These natural missense mutations in the three haplotypes of ClLCYB did not influence protein localization.

To determine whether the amino acid substitution in three ClLCYB proteins disrupted the enzymatic activity of ClLCYB, we cloned these alleles into an expression vector to express these ClLCYB proteins in a lycopene-producing Escherichia coli strain expressing the pAC-LYC plastid (Cunningham and Gantt, 2007). HPLC was used to analyze carotenoids extracted from the bacterial cells. As shown in Figure 5C, the strains containing vector control\pAC-LYC exhibited a main single peak, whose retention time and absorbance spectrum corresponded to those of lycopene, and the cultures were red. By contrast, extracts from pET28a-ClLCYB\pAC-LYC cells showed mainly β-carotene accumulation, and the cultures turned orange. Color complementation with these ClLCYB constructs demonstrated that these gene products function as lycopene β-cyclase. Enzyme kinetic studies were performed with the three purified ClLCYBs to determine the K_m values for the substrate lycopene. The steady-state kinetics were derived using the Michaelis-Menten equation, as shown in Figure 5D. The observed K_m and V_max values were 8.59 μm and 0.06 μmol min⁻¹ for ClLCYB^red, 9.15 μm and 0.055 μmol min⁻¹ for ClLCYB^yellow, and 9.26 μm and 0.057 μmol min⁻¹ for ClLCYB^white, respectively (Table 3). The kinetic parameters suggested that the ClLCYB^red protein has fairly good catalytic attributes. These results show that the enzymatic activity of ClLCYB^red in red-fleshed watermelon is not responsible for lycopene accumulation.

Table 3. K_m and V_max values of three ClLCYB proteins in watermelons with different flesh colors.

Enzyme	Km	Vmax
	μm	μmol min–1
ClLCYBred	8.59 ± 0.619	0.060 ± 0.017
ClLCYByellow	9.15 ± 2.710	0.055 ± 0.006
ClLCYBwhite	9.26 ± 2.519	0.057 ± 0.006

Differential Protein Abundance and Degradation of ClLCYBs Contribute to Lycopene Accumulation in Fruits

Our results indicate that different ClLCYB haplotypes determine the formation of red- or other-colored flesh. However, the enzymatic function and protein localization of these alleles does not account for the flesh color variation. These findings prompted us to infer that lycopene accumulation in red-fleshed watermelons may be caused by a decrease in ClLCYB^red transcript or protein abundance. To clarify the relationship between ClLCYB gene expression and fruit flesh color development, we first examined the transcriptional expression of ClLCYB in lines with different flesh colors. The fruit flesh color phenotypes of 20 representative watermelon accessions spanning the genetic diversity of genus Citrullus are shown in Figure 6A and Supplemental Table S3 (Guo et al., 2013). The lycopene content, sugar content, and ClLCYB gene expression level in the flesh were measured at 34 DAP. The results showed that the variation in ClLCYB transcript level among these 20 lines was not correlated with flesh color (Fig. 6A). Pearson’s correlation analysis revealed that lycopene content and ClLCYB gene expression level were not significantly correlated (r = 0.17918, P = 0.4497; Supplemental Table S3). For example, in the white-fleshed line PI595203 (no. 5), the ClLCYB transcript levels were approximately the same as those in the red-fleshed lines. Our results indicate that the endogenous expression level of ClLCYB is not the reason for watermelon flesh color differentiation.

Figure 6.

Decreased ClLCYB protein levels lead to red-fleshed watermelon. A, ClLCYB protein level was negatively correlated with the lycopene contents in the chromoplasts of the fruits of the 20 watermelon accessions. The fruit morphology of 20 selected watermelon lines is shown at the bottom. Lycopene content (red line) was measured by HPLC as described in “Materials and Methods.” Relative protein levels (blue line) were measured as described in “Materials and Methods.” Relative gene expression levels (black line) were measured by RT-qPCR. Watermelon ACTIN was used as an internal control. The error bars indicate sd (n = 3). B, Representative western blot showing ClLCYB protein accumulation in the chromoplasts of the fruits of the 20 watermelon accessions. Western blotting was performed with anti-ClLCYB and anti-ACTIN antibodies. C, Representative western blot showing the amounts of ClLCYB:MYC fusion proteins in the six differently treated seedlings with anti-MYC antibodies (top) and Rubisco stained with Ponceau S (bottom). Arabidopsis transgenic lines expressing ClLCYB^red (with V226 and K435), ClLCYB^yellow (with F226 and N435), chimeric ClLCYB^VN, and ClLCYB^white (with F226 and K435) were grown on MS medium for 2 weeks and then transferred to liquid medium that contained 30 μm CHX, 50 μm MG132, or a mixture of protease inhibitors for 1 h. D, Graphs showing the quantification of protein levels relative to total ClLCYB levels in C. The graphs were constructed by normalizing the band intensity values to those of the corresponding bands in the first treatment. The error bars indicate sd (n = 3). The bars marked with different letters are significantly different (ANOVA, P < 0.05).

To determine the amount of ClLCYB protein in the lines with different flesh colors, we extracted intact chromoplasts from the 20 watermelon accessions mentioned above and prepared a specific antibody against ClLCYB (Supplemental Fig. S3). Western blotting was used to reveal the ClLCYB protein content within total chromoplast protein of these selected lines (Fig. 6B). As shown in Figure 6, A and B, compared with the other lines, the red-fleshed lines had significantly lower levels of the ClLCYB protein. A significant negative correlation between ClLCYB protein level and lycopene accumulation was detected in these watermelon lines (r = −0.83452, P < 0.001; Supplemental Table S3). Overall, ClLCYB protein accumulation was significantly suppressed in the cells of red-fleshed lines, regardless of the ClLCYB mRNA expression level. This finding indicated that additional posttranslational modifications may be one of the regulatory mechanisms reducing ClLCYB abundance in red-fleshed lines.

To determine whether the natural amino acid changes in ClLCYB are responsible for the decreased protein abundance in the red-fleshed lines, we constructed four 35S-ClLCYB-MYC plasmids that expressed ClLCYB^red (with V226 and K435), ClLCYB^yellow (with F226 and N435), ClLCYB^white (with F226 and K435), or chimeric ClLCYB^VN (with V226 and N435) and transformed these plasmids into Arabidopsis (Arabidopsis thaliana). Four transgenic lines were chosen based on their high expression levels (Supplemental Fig. S4). These four transgenic lines were grown on Murashige and Skoog (MS) medium and then transferred to liquid medium that contained 30 μm cycloheximide (CHX), 50 μm proteasome inhibitor MG132, or a mixture of protease inhibitors for 1 h. Western-blot analyses were used to measure MYC fusion protein accumulation in the transgenic lines (Fig. 6C). The levels of these four fusion proteins in transgenic seedlings under six treatments were measured as shown in Figure 6D. The results showed that both the MG132 and protease inhibitor treatments could block the degradation of ClLCYB proteins (Fig. 6, C and D). This trend was obvious after the application of CHX, which blocks protein synthesis. As shown in Figure 6D, clear accumulation of the ClLCYB^red-MYC fusion protein was observed after MG132 treatment compared with the protease inhibitor treatments, whereas ClLCYB^yellow-MYC was more sensitive to protease inhibitor treatment than to MG132 treatment. The function of MG132 and the protease inhibitor showed no obvious differences between ClLCYB^white and chimeric ClLCYB^VN. These results suggest that, compared with other haplotypes, ClLCYB^red is more sensitive to the posttranslational ubiquitin-proteasome system, which could explain the low ClLCYB^red protein abundance detected in flesh cells.

DISCUSSION

Watermelon fruit flesh color is an indicator of the nutritional benefits provided by carotenoids and is a phenotypic trait important for consumer preference. Distinctly different from its wild pale-colored ancestors, red-fleshed watermelon accumulates substantial amounts of lycopene in its fruit. Carotenoids exhibit many health-promoting activities, including provitamin A activity (Grune, et al., 2010), enhanced immune system function, and skin photoprotectant activity (Cooper, 2004; Bonet et al., 2015; Milani et al., 2017). Lycopene is believed to be associated with reduced prostate cancer risk and inhibition of obesity (Fenni et al., 2017; Rowles et al., 2017). Cloning of the key gene controlling red flesh formation in watermelons has long been pursued. In our research, we found that the ClLCYB gene (the rf gene) controls red flesh color in watermelon, and we present biochemical evidence that decreased ClLCYB protein abundance in red-fleshed lines is the reason for lycopene accumulation in watermelon.

LCYB is an important enzyme in the carotenogenesis pathway and catalyzes the conversion of lycopene to β-carotene. The carotenoid biosynthesis pathway is conserved in plants, but lycopene overaccumulation in fruits is not a common phenotype. Tomato and watermelon are two popular fruits that accumulate lycopene as the major fruit carotenoid, which gives them their typical red color (Tadmor et al., 2005). In tomato, several attempts have been made to engineer high fruit lycopene levels, but few have been successful. Up-regulation of the carotenogenesis pathway via increased expression of PSY, bacterial PDS (CrtI), and LCYB in tomato leads to an increase in the β-carotene content (Römer et al., 2000; Fraser et al., 2002). By contrast, down-regulation of LCYB in tomato leads to a slight increase in lycopene content (Rosati et al., 2000; Wan et al., 2007). In our research, we provide evidence that down-regulation of LCYB expression leads to the accumulation of lycopene in pale-colored watermelon. A better transgenic analysis method for clarifying the function of the missense V226F and K435N point mutations is CRISPR-mediated base editing (Komor et al., 2016; Shimatani et al., 2017). However, at present, the reported C-to-T and A-to-G base-editing systems unfortunately do not work for ClLCYB (Komor et al., 2016; Gaudelli et al., 2017). We believe that, in the near future, an increase in lycopene content by CRISPR base editing of the LCYB gene will facilitate the generation of red-fleshed Cucurbitaceae species.

In general, investigations have revealed the plausible effects of missense mutations on protein stability (Zhang et al., 2012; Socha and Tokuriki, 2013; Wang et al., 2013), protein-protein interactions (Liu et al., 2010), protein localization and function (Faso et al., 2009; Marti et al., 2010; Ji et al., 2016), and many other aspects. In our research, the two mutations in ClLCYB^red were correlated with protein accumulation and degradation by various posttranslational modifications. Ubiquitination is one of the central posttranslational modifications affecting the fate of proteins and regulates a wide range of plant developmental processes (Moon et al., 2004). Moreover, proteolysis involves the degradation of specific proteins in different compartments of cellular organelles (Adam and Clarke, 2002). Our data provide an explanation for the lower ClLCYB protein abundance in red-fleshed watermelon. Treatment with MG132, a well-known proteasome inhibitor, effectively blocked the decrease in ClLCYB^red in red-fleshed fruit. Ubiquitin was reported previously to bind to its target via a Lys (K) residue, followed by ubiquitination and clearance of the target via the 26S proteasomal complex (Pickart, 2001; Smalle and Vierstra, 2004; Mattiroli and Sixma, 2014), which suggests that the Lys (K) residue at position 435 in ClLCYB^red plays a critical role in regulating ClLCYB degradation. Another possible posttranslational modification of the ClLCYB^yellow protein in yellow-fleshed fruits is Asn (N)-linked glycosylation at position N435, which forms an N435-G436-T437 cluster. Kornfeld and Kornfeld (1985) noted that the Asn (N) glycosylation sites are Asn-X-Ser/Thr tripeptides, where X can be any amino acid except Pro (P) or Asp (D). ClLCYB^yellow could be a glycoprotein in watermelon fruit. The influence of different 5′-UTRs on the translation efficiency of ClLCYB mRNA may be another possible explanation (Satoh et al., 2004). In our study, no correlated SNPs were found located in or near the 5′-UTR, but we detected a possible uORF coding 24 amino acids located at −275 to −200 bp before the main ORF in all the tested accessions (Supplemental Fig. S1). These explanations of the difference in ClLCYB protein abundance caused by translational and posttranslational modification represent a valuable addition to the limited information available on the regulation of the carotenoid biosynthesis pathway.

Domestication is the process by which humans take wild species and acclimatize them for breeding and survival outside of natural conditions. It is believed that the domestication of red-fleshed fruits was a conscious effort undertaken by humans. The appearance of red-fleshed watermelon may be a natural mutation, but the phenotype spread and was strengthened through breeding efforts by humans. Notably, all the red-fleshed lines among the 211 investigated accessions expressed the ClLCYB^red protein, suggesting that ClLCYB^red alleles are widely distributed in the watermelon population. Compared with the other alleles, the ClLCYB^yellow alleles were more common in cultivated species with fruits that presented other flesh colors, such as orange, canary yellow, yellow, and white (red was excluded). This finding raises the question of whether yellow-fleshed watermelon is the ancestor of red-fleshed cultivars. In our phylogenetic tree, SNP classification allowed reconstruction of the genealogical tree of flesh color formation with great accuracy to detect patterns in the distribution of shared historical lineages. Our findings indicate that red-fleshed watermelon was domesticated from a pale-colored semiwild ancestor, whereas watermelons with yellow flesh and other flesh colors, which do not accumulate lycopene, were domesticated as another branch. Taken together, these results support flesh-color selection pressure around the ClLCYB region and that this ClLCYB^red haplotype has been selected for and largely fixed in domesticated watermelon. We speculate that the selection of red-fleshed watermelons benefits the consumption habits of humans and other animals and improves fruit quality and flavor, leading to increased consumption and thereby promoting seed dispersal.

MATERIALS AND METHODS

Plant Materials

Red-fleshed watermelon (Citrullus lanatus ‘97103’) and pale yellow-fleshed watermelon (‘Cream of Saskatchewan’) plants were crossed. F1 and F2 populations were generated by controlled pollinations in the greenhouse. Briefly, 12 F1 plants were self-pollinated to obtain F2 seeds. One thousand thirty-eight F2 individuals were subsequently genotyped, and 209 wild and cultivated watermelon accessions were grown in our laboratory (Zhang et al., 2011). The seeds of each watermelon variety were sown at the same time, and the plants displayed uniform fruit development. All fruits for evaluation were harvested at full maturity, cut longitudinally, imaged, and categorized into different flesh-color groups.

Carotenoid Extraction and Analysis

Carotenoids from mature watermelon flesh (0.5 g fresh weight) and Escherichia coli were extracted and analyzed using a Nexera HPLC system (Shimadzu) in accordance with the method described by Bang et al. (2010). Quantification of the relative concentrations of individual carotenoids was achieved by comparing the individual peak areas with a calibration curve constructed by using commercial lycopene and β-carotene standards (Sigma-Aldrich). All samples were analyzed in triplicate.

Map-Based Cloning and Sequencing of ClLCYB

An F2 mapping population resulting from a cross between cv 97103 and cv CS was used to map the red flesh-controlling gene. Sequence-tagged site markers were designed based on the insertions/deletions between cv 97103 and cv CS resequencing data. The BVWS markers are listed by Ren et al. (2012). The watermelon reference v2 genome and gene annotations and expressions were downloaded from the Cucurbit Genomics Database (http://cucurbitgenomics.org/organism/21; Zheng et al., 2019). The newly designed markers were named according to their physical positions. The cetyl-trimethyl-ammonium bromide method was used to extract genomic DNA from watermelon. The molecular lesion responsible for ClLCYB (gene identifier Cla97C04G070940) was identified by PCR amplification and sequencing data of the Cla97C04G070940 genomic region from cv CS, cv 97103, and cv 209 other watermelon accessions (Guo et al., 2019); the sequences were subsequently compared using DNAMAN (www.lynnon.com). The primer sequences are listed in Supplemental Table S4.

RACE and RT-qPCR Analysis

Total RNA was extracted using a Quick RNA Isolation Kit (Huayueyang Biotechnologies). cDNA was synthesized from 1 mg of total RNA using SuperScript III transcriptase (Invitrogen). The 5′ and 3′ ends of the transcripts were determined using a SMARTer RACE 5′/3′ Kit (Clontech Laboratories). PCR was then carried out to amplify the cDNA fragments of ClLCYB using degenerate primers. RT-qPCR analyses were performed with three independently generated samples from each watermelon line at the same developmental stage. SYBR Green I was added to the reaction system, and the reaction was conducted on a Roche 480 system in accordance with the manufacturer’s instructions. The relative expression levels of ClLCYB were normalized against those of the watermelon ACTIN gene (gene identifier Cla97C02G026960) transcript. Each RT-qPCR experiment was repeated three times. The relative expression of the genes was calculated with the sd for three biological replicates. The primer sequences used for the above studies are listed in Supplemental Table S4.

Phylogenetic Analysis of the ClLCYB Gene in Watermelon Accessions

We used the resequencing data (Guo et al., 2019) together with the sequenced ClLCYB gene to identify the SNPs in 211 watermelon accessions (Supplemental Table S2). The watermelon reference v2 genome was downloaded from the Cucurbit Genomics Database. Trait Analysis by Association, Evolution, and Linkage (TASSEL) software (v5.2.44; Bradbury et al., 2007) was used to calculate kinship and π. The resultant kinship matrix was used in conjunction with the maximum likelihood method to perform association analyses.

Construction of ClLCYB Transgenic Watermelon

ClLCYB transgenic analysis of watermelon was completed using pYBA1302 vectors (Yan et al., 2012). For the antisense RNA analysis in pale yellow-fleshed cv CS, the full-length ClLCYB CDS was amplified from cv CS by PCR primers with different adaptors. The fragment was then inserted into the EcoRI/BamHI sites of pYBA1302, which was subsequently transformed into Agrobacterium tumefaciens strain C58/ATCC 33970. ClLCYB overexpression in red-fleshed cv ZZJM was also performed using the pYBA1302 vector. The full-length ClLCYB CDS was amplified from cv CS and then inserted into the EcoRI/XhoI sites of pYBA1302. The A. tumefaciens-mediated transformation protocol was described by Tian et al. (2017). PCR amplification with specific primers (an upstream primer of the 35S promoter and a downstream primer of the ClLCYB gene) and an AS013 PAT/Bar Kit (Envirologix) were used to confirm the transgene insertion in the transformed watermelon plants. The primer sequences used are listed in Supplemental Table S4.

Functional Analysis of ClLCYBs in E. coli

The full-length cDNAs of ClLCYB^red (red-fleshed lines with V226 and K435), ClLCYB^white (wild white-fleshed lines with F226 and K435), and ClLCYB^yellow (non-red-fleshed lines with F226 and N435) were isolated from fruit cDNA libraries using the LCYB-F and LCYB-R primers. The verified PCR products were subsequently cloned into pET28a expression vectors, resulting in the pET28a-LCYB^red, pET28a-LCYB^white, and pET28a-LCYB^yellow plasmids. The pAC-LYC plasmid (Cunningham et al., 1996) was kindly provided by Dr. Li Li and contains all the genes for lycopene formation. pAC-LYC plasmids were transformed or cotransformed with pET28a-LCYB plasmids into E. coli BL21. The double transformants were plated onto Luria-Bertani agar supplemented with ampicillin (100 μg mL⁻¹) and chloramphenicol (50 μg mL⁻¹). A 50-mL culture of E. coli double transformants was grown in Luria-Bertani medium at 37°C to an OD₆₀₀ of 0.8. Protein expression was induced with the addition of isopropyl-β-d-thiogalactopyranoside to the growth medium at a final concentration of 0.1 mm. The bacterial culture was grown at 30°C for an additional 24 h. The E. coli cells were harvested by centrifugation, and their color development was monitored.

Enzyme Assay and Kinetic Characterization of ClLCYBs

Using recombinant plasmid pGEX-ClLCYB BL21(DE3) expression systems, three ClLCYB-GST proteins were produced. The purity of the recombinant ClLCYB-GST protein reached greater than 95% following purification with Pierce glutathione agarose (Thermo Scientific, no. 16101). The kinetic parameters were examined by performing enzymatic assays under the optimal conditions according to the method described by Schnurr et al. (1996). The assay mixture contained six different concentrations of lycopene (Sigma-Aldrich), 5 mm NADPH, and the purified enzyme (50 μg mL⁻¹ protein) in 500 μL of extraction buffer that contained 50 mm Tris/malate (pH 6.5), 4 mm MgCl₂, and 6 mm MnSO₄. Incubations were performed in the dark at 30°C for 2 min, 5 min, 15 min, 30 min, or 1 h. The assay was terminated by adding methanol (1.5 mL) containing 6% (w/v) KOH and heating at 60°C for 20 min. The remaining substrate and the products formed were extracted from the aqueous incubation mixture with diethyl ether:petrol (1:9, v/v). Analysis of the reaction products was performed using a Nexera HPLC system (Shimadzu) in accordance with the method described by Bang et al. (2010). The kinetic parameters (K_m and V_max) of these ClLCYBs were measured using the Michaelis-Menten equation.

Subcellular Localization Analysis

To determine the subcellular localization of the three types of ClLCYB proteins, the proteins were inserted into pYBA1138-mCherry vectors in which mCherry was at the C terminus, driven by the constitutive 35S promoter. The plasmids were purified and transformed into watermelon fruit protoplasts according to a previously described method (Zhang et al., 2017), with slight modifications. The mCherry fluorescence and chromoplast autofluorescence were analyzed using an LSM700 microscope at wavelengths of 587 and 500 nm for excitation and 610 and 509 nm for emission, respectively.

Chromoplast Protein Extraction

Crude chromoplasts from watermelon flesh were isolated and purified with a discontinuous Suc gradient according to the methods of Zhang et al. (2017). The fruit samples were carefully selected based on the DAP and their flesh color. Three biological replicates were collected at each sampling. Briefly, approximately 250 g of fresh material was ground in 600 mL of ice-cold extraction buffer that contained 50 mm HEPES (pH 7.5), 2 mm EDTA, 330 mm sorbitol, and 5 mm 2-mercaptoethanol. The suspensions were subsequently filtered with four and eight layers of cheese cloth and then centrifuged at 5,000g for 5 min at 4°C. The crude chromoplast pellets were gently suspended in 5 mL of extraction buffer that contained 50% (w/v) Suc, overlaid with a discontinuous Suc gradient (50%, 30%, and 17% [w/v] in extraction buffer), and centrifuged at 62,000g for 45 min at 4°C. Intact chromoplasts between the 50%/30% and 30%/17% layers were carefully collected and washed. The total intact chromoplasts were then ground in liquid nitrogen to extract the proteins according to the method of Zhang et al. (2017). The total protein levels extracted from the plants were first measured using a Qubit protein assay kit and a Qubit 2.0 fluorometer (Invitrogen).

Production of Mouse Monoclonal Anti-ClLCYB Antibodies

Specific monoclonal antibodies against ClLCYB were produced using truncated ClLCYB protein (481 amino acids) produced in mice by Abmart. ClLCYB proteins were produced by using the recombinant plasmid pET28a (+)-ClLCYB BL21(DE3) expression system. Six BALB/c mice were immunized subcutaneously with the proteins. Spleen cells obtained from the immunized mice were fused with SP2/0 myeloma cells according to standard procedures. Positive hybridomas were cloned, and IgG was purified by protein G affinity chromatography from ascitic fluid. Western blotting was performed to evaluate the specificity and titer of the anti-ClLCYB antibodies using bacterially expressed recombinant ClLCYB.

Western-Blot Analysis

For western-blot analysis, protein samples were added to the loading buffer (62.5 mm Tris [pH 6.8], 2% [w/v] SDS, 5% [v/v] mercaptoethanol, 10% [v/v] glycerol, and 0.02% [w/v] bromophenol blue) and boiled for 3 min. Then, the proteins were centrifuged for 3 min at 12,000g. The extracts were separated by 12% (w/v) SDS-PAGE with anti-ClLCYB. Anti-ACTIN (Enzo BML-PW8370, 1:1,000) was used as a control. An Azure Biosystem imaging system and AzureSpot analysis software were used for densitometric analyses of western blots. Values were normalized to those for ACTIN, which served as a loading control. The quantification results were calculated at least three times for three biological replicates and normalized to those of the control.

Construction of Transgenic ClLCYB-MYC-Overexpressing Arabidopsis

Sequences encoding the full-length ClLCYB gene from red-fleshed cv 97103 (ClLCYB^red) and pale yellow-fleshed cv CS (ClLCYB^yellow) as well as the full-length sequence of the wild ClLCYB^white were amplified by PCR with different adaptors. To obtain the chimeric ClLCYB^VN encoding V226 and N435, the PCR products of ClLCYB^yellow and ClLCYB^red were digested by SapI and then linked together. These four ClLCYB fragments were subsequently inserted into the pSuper1300-MYC vector. The resulting constructs were then transformed into Arabidopsis (Arabidopsis thaliana; Columbia-0 background) by vacuum infiltration (Bechtold and Pelletier, 1998).

Chemical Treatment and Quantification of the ClLCYB-MYC Protein Levels in Transgenic Arabidopsis

Two-week-old (grown under a 16-h-light/8-h-dark photoperiod on MS medium) Arabidopsis seedlings carrying 35S-ClLCYB-MYC were grown in agar in petri dishes and treated with different chemicals. The seedlings were then transferred to synthetic liquid medium (one-half-strength MS medium consisting of 1% [w/v] Suc) that contained 30 μm CHX (Sigma-Aldrich), 50 μm MG132 (dissolved in 1% [v/v] dimethyl sulfoxide; Sigma-Aldrich), or a mixture of protease inhibitors (Roche Complete Protease Inhibitor Cocktail) for 1 h at room temperature. After treatment, the total protein from the seedlings was extracted with protein extraction buffer (50 mm Tris-HCl buffer [pH 7.5], 150 mm NaCl₂, 10 mm MgCl₂, 1 mm phenylmethylsulfonyl fluoride, 0.1% [v/v] Nonidet P-40, and Roche Complete Protease Inhibitor Cocktail), transferred to 1.5-mL tubes, and then centrifuged at 12,000g for 10 min at 4°C. Protein samples from transgenic Arabidopsis were separated by 12% (w/v) SDS-PAGE with anti-MYC (Abmart M20002, 1:5,000) antibodies to detect the protein levels as described previously. To confirm equal efficiencies of protein extraction and loading, an identical parallel gel was stained with Ponceau S.

Accession Numbers

Sequence data from this article can be found in the DDBJ/ENA/GenBank data libraries under accession number AGCB02000000 and in the Cucurbit Genomics Database (http://cucurbitgenomics.org/organism/21).

Supplemental Data

The following supplemental materials are available.

• Supplemental Figure S1. ClLCYB (Cla97C04G070940) mRNA sequences in cv 97103 and CS.

• Supplemental Figure S2. ClLCYB (Cla97C04G070940) expression patterns in different parts of the red-fleshed cv 97103.

• Supplemental Figure S3. Preparation and testing of the monoclonal antibody for ClLCYB (Cla97C04G070940).

• Supplemental Figure S4. Transcriptional and translational expression of ClLCYB in the four Arabidopsis transgenic lines expressing ClLCYB^red, ClLCYB^yellow, chimeric ClLCYB^VN, and ClLCYB^white.

• Supplemental Table S1. List of candidate genes located at the rf locus.

• Supplemental Table S2. Flesh color, classification, and SNPs of the ClLCYB gene of all 211 watermelon accessions.

• Supplemental Table S3. Lycopene content, °Brix values, and ClLCYB transcript and protein levels of 20 naturally occurring germplasms.

• Supplemental Table S4. List of primers used for genotyping, cloning, and expression analysis.