Cytological and molecular characterization of carotenoid accumulation in normal and high-lycopene mutant oranges

Ripe Cara Cara sweet orange contains 25 times as much carotenoids in flesh as Newhall sweet orange, due to high accumulation of carotenes, mainly phytoene, lycopene and phytofluene. Only yellow globular chromoplasts were observed in Newhall flesh. Distinct yellow globular and red elongated crystalline chromoplasts were found in Cara Cara but only one type of chromoplast was present in each cell. The red crystalline chromoplasts contained lycopene as a dominant carotenoid and were associated with characteristic carotenoid sequestering structures. The increased accumulation of linear carotenes in Cara Cara is not explained by differences in expression of all 18 carotenogenic genes or gene family members examined, or sequence or abundance of mRNAs from phytoene synthase (PSY) and chromoplast-specific lycopene β-cyclase (CYCB) alleles. 2-(4-Chlorophenylthio)-triethylamine hydrochloride (CPTA) enhanced lycopene accumulation and induced occurrence of red crystalline chromoplasts in cultured Newhall juice vesicles, indicating that carotenoid synthesis and accumulation can directly affect chromoplast differentiation and structure. Norflurazon (NFZ) treatment resulted in high accumulation of phytoene and phytofluene in both oranges, and the biosynthetic activity upstream of phytoene desaturase was similar in Newhall and Cara Cara. Possible mechanisms for high carotene accumulation and unique development of red crystalline chromoplasts in Cara Cara are discussed.

Pigment changes in flesh of Newhall and Cara Cara oranges during development. The flesh colour pattern was ascribed to differences in content and composition of chlorophylls and carotenoids. Both cultivars accumulated a small quantity of chlorophylls at the early stages, which decreased gradually until S5 (Fig. 1B). In Newhall, only trace amount of carotenoids were present until S5 and then increased steadily during fruit maturation. However, in Cara Cara, large amount of carotenoids accumulated in flesh tissues even at S1, and then increased to around three fold during fruit maturation (Fig. 1B). During the whole development and maturation stages, carotenoids in flesh of Cara Cara were between 19 and 56 times higher than in Newhall (Fig. 1B).
Scientific RepoRts | 7: 761 | DOI: 10.1038/s41598-017-00898-y This increment resulted mainly from enhanced accumulation of carotenes, but, taking S8 fruit as an example, the amount of total xanthophylls in the two sweet orange cultivars was similar ( Fig. 2A).
There were major differences in the composition of carotenoids in flesh in the two sweet orange cultivars. In Newhall ripe fruit, the three most abundant carotenoids were all xanthophylls, 9-cis-violaxanthin, violaxanthin, and luteoxanthin, while in Cara Cara the linear carotenes, phytoene, lycopene, and phytofluene were the most abundant ( Fig. 2A). These differences between two cultivars were also observed at other developmental stages (Fig. 2B).

Light microscopy of chromoplasts in flesh cells of Newhall and Cara Cara oranges during fruit development.
A difference in chromoplast development was observed for the two cultivars. Only yellow globular chromoplasts were observed in all flesh cells of mature Newhall fruit, while two distinct types, yellow globular ones and red crystalline ones, were found in different flesh cells of Cara Cara (Fig. 3). Crucially, for any individual cell, only one type of chromoplast was present and cells contained exclusively either red chromoplasts or yellow ones, often with cells containing each type found adjacent to each other (Fig. 3). A single cell with both types of chromoplasts was not observed, and this precluded the possibility of transition between two types of chromoplasts.
During fruit development, red crystalline chromoplasts appeared before the yellow globular ones. As shown in Fig. 3, in flesh tissue of Cara Cara fruit at S3, some cells harbored red crystalline chromoplasts while the others were devoid of chromoplasts, while at this stage no chromoplast could be observed in flesh cells of Newhall. At Figure 1. Fruit shape, size, color, carotenoids and chlorophylls in Newhall and Cara Cara oranges. (A) External and internal colors of Newhall and Cara Cara fruits during development and ripening at different stages (S1-S8). Bar, 2 cm. (B) Changes in peel CCI (citrus colour index), fruit size (transverse diameter), content of flesh total carotenoids and flesh chlorophylls during fruit development and ripening in Newhall and Cara Cara oranges. The bars represent SE from three biological replicates. later stages, individual flesh cells contained substantial numbers of either yellow or red chromoplasts for Cara Cara and only yellow chromoplasts for Newhall (Fig. 3).
Despite the fact that Cara Cara flesh appeared to have a uniform red color in cross section (Fig. 1), a close examination showed that juice vesicles of different colours, from yellow to red, could be found under the microscope, even from neighboring sites (Fig. 4A). However, inside each vesicle, which contains multiple cells, two types of cells with distinct chromoplasts could still be found, no matter what the colour of the vesicle was (Fig. 4B). The color difference between these three groups of vesicles resulted from the different percentage, around 60%, 30% and 10% respectively, of cells with red chromoplasts. In situ analysis of carotenoids in flesh chromoplasts of Newhall and Cara Cara oranges by Raman microspectroscopy. It was technically not feasible to collect sufficient protoplasts harboring either type of chromoplasts for carotenoid analysis by HPLC. However, by in situ Raman confocal microscopy, we were able to establish that the red crystalline chromoplasts contain lycopene as a dominant carotenoid, as manifested by the presence of three peaks around 1510 cm −1 , 1156 cm −1 and 1004 cm −1 , consistent with the spectrum produced by tomato lycopene 37 , while the globular ones, either from Cara Cara or Newhall, did not (Fig. 5). The spectrums for globular chromoplasts from Cara Cara and Newhall were similar (Fig. 5), probably indicating other similarities in carotenoid composition.
Ultrastructure of plastids in Newhall and Cara Cara flesh cells during fruit development. Differences in ultrastructure of plastids in flesh cells of the two cultivars were observed by TEM. During early stages (S2 and S4), the plastids contain starch granules and thylakoid structures (Fig. 6). At later stages (S6 and S8), consistent with the data from microscopy observation, only globular chromoplasts were found in Newhall flesh cells while both globular and crystalline ones were found in Cara Cara but in different flesh cells. The globular ones in Newhall and Cara Cara shared similar ultrastructure, containing thylakoid membranes and plastoglobules, while crystalline chromoplasts, present in Cara Cara only, had a distinct ultrastructure, characterized by the presence of large plastoglobules, crystals, crystal remnants and undulating membranes (Fig. 6).
Sequences of PSY1, LCYB and CYCB in Newhall and Cara Cara. Since Cara Cara possesses high amount of carotenes including lycopene, here the sequences of PSY1, LCYB and CYCB were checked for possible mutation, since it has been established in other studies that these are potential explanations for differences in carotenoid content. For each gene, two highly homologous sequences were obtained. This is consistent with the fact that orange is a diploid plant with heterogeneous genetic background.   Sequencing PSY1 cDNA revealed the existence of two alleles, named PSY1-a and PSY1-b, in both oranges ( Supplementary Fig. S2). Twelve SNPs and an indel of 6 bp, resulting from of gain or loss of two repeats of TAA, were found when comparing two alleles ( Supplementary Fig. S2). However, no sequence difference was observed between two orange cultivars, i.e., the PSY1-a sequences for both cultivars are identical and this is also true for the PSY1-b sequences ( Supplementary Fig. S2). Sequence BLASTing of both alleles in NCBI (http://blast.ncbi. nlm.nih.gov/Blast.cgi) revealed that PSY1-a was identical to PSY in C. unshiu, while PSY1-b was identical to C. maxima. This is consistent with the result of Velasco and Licciardello (2014) 38 who suggested that sweet orange arose from natural crosses between mandarin and pummelo.
Similarly, two alleles for LCYB and CYCB were identified from Newhall and Cara Cara. While 16 SNPs were found between LCYB alleles and 26 SNPs in CYCB alleles, no sequence difference was found between two cultivars ( Supplementary Figs S3 and S4). The sequence of the CYCB-a allele was identical with β-LCY2a, i.e., allele a of β-LCY2, a synonym for CYCB, in Star Ruby grapefruit 32 while CYCB-b had a synonymous SNP difference when compared with β-LCY2b 32 .
Carotenogenic gene mRNA levels in flesh of Newhall and Cara Cara oranges. The mRNA levels for carotenogenic genes in flesh of Newhall and Cara Cara oranges were compared (Fig. 7). Two members of DXS were expressed and the DXS1 transcript accounted for over 95% of total DXS transcripts in both orange cultivars, and this was also true for PSY (Fig. 7). Transcript levels for DXS1 and PSY1 increased during fruit development, but no significant difference in transcript abundance of DXS1 and PSY1 was observed between two cultivars. Expression of genes involved in serial conversion of phytoene into lycopene was similar for most samples between the two cultivars, except for higher transcript level of carotene isomerase (CRTISO) and ζ-carotene isomerase (ZISO) in Newhall at the last two stages of fruit development. However, during the early stages, despite the existence of differences in carotenoid accumulation, the transcript abundance of CRTISO and ZISO was similar in the two cultivars. Expression of both LCYB and lycopene ε-cyclase (LCYE) was over 20 fold lower than that of CYCB, and the transcript level of CYCB was similar in both oranges. Transcript levels of carotene β-ring hydroxylase (BCH) were lower and that of ZEP higher in Cara Cara at some stages but not others. Furthermore, the transcript level of three carotenoid cleavage genes, carotenoid cleavage dioxygenase gene family member 1 (CCD1), nine-cis-epoxycarotenoid dioxygenase gene family members 2 and 3 (NCED2, NCED3), were similar overall in the two cultivars.
The transcript levels for each of the PSY1 and CYCB allele in yellow and red juice vesicles separated from the same fruit of Cara Cara (Fig. 8A) were analysed by quantitative real-time PCR using allele-specific PCR primers (Supplementary Table S1), and it was found that the expression of both PSY1 and CYCB alleles was slightly higher in Cara Cara red juice vesicles than in Cara Cara yellow vesicles and Newhall vesicles, but no obvious bias in expression between alleles was observed (Fig. 8B).

Effects of NFZ and CPTA treatments on in vitro cultured flesh tissues. As expected, NFZ treatment
induced an albino phenotype in both Newhall and Cara Cara flesh tissues ( Fig. 9) and, although the amount of total carotenoids increased, colorless carotenoids phytoene and phytofluene were the predominant ones, accounting for over 90% of total carotenoids for both oranges at four weeks following treatment (Fig. 10). The increment  Table S2).
For both oranges, flesh tissues became deeper red following CPTA treatment during 4-weeks in culture (Fig. 9). The red pigment in CPTA-treated flesh tissue was mainly lycopene. In Newhall, lycopene accumulated up to 79% and 78% of total carotenoids following 2-and 4-week CPTA treatment, respectively, and accumulation of small amounts of phytoene and phytofluene was also observed (Fig. 10). In contrast, control Newhall flesh tissues did not contain phytofluene and lycopene, and only trace amounts of phytoene. In Cara Cara, CPTA treatment also raised the percentage of lycopene among total carotenoids; however, phytoene was still the most predominant carotenoid (Fig. 10). At four weeks following CPTA treatment, the increment in total carotenoids between treated flesh tissues and control was 21.72 μg g −1 FW in Newhall and 10.31 μg g −1 FW in Cara Cara (Supplementary Table S2).
The Newhall juice vesicles following 4-week CPTA treatment were taken for microscopic analysis, and it was found that red crystalline chromoplasts occurred, especially in the deeper red part (Fig. 11). In addition, the CPTA treated Newhall albedo tissue was also checked and red crystalline chromoplasts were found in all cells ( Supplementary Fig. S5). However, no colored plastids were observed in control Newhall albedo layers. In Cara Cara control albedo layers (without CPTA), red crystalline chromoplasts could also be observed in some cells ( Supplementary Fig. S5).

Discussion
Differences in carotenoid accumulation, varying in some cases up to hundreds of fold in total carotenoid content, have been reported between wild type and mutants, as well as among cultivars, in various plants 11,18,19,[39][40][41] . The tremendous differences often resulted from no or trace accumulation of carotenoids in a mutant or cultivar, such as the r tomato and white-fleshed loquats caused by the loss of function mutation of SlPSY1 and EjPSY2A 14,19 , and Early Moonbeam watermelon 39 . Here the difference in carotenoid accumulation between two Figure 7. Expression of carotenogenic genes in flesh of Newhall and Cara Cara oranges during fruit development and ripening. The bars represent SE from three biological replicates . Asterisks indicate significant differences, P < 0.05 and P < 0.01 are designated by * and **, respectively. The abbreviation for the genes is the same as listed in Fig. 1. sweet orange cultivars is a different case, since the ordinary orange Newhall contains normal level of carotenoids while the mutant Cara Cara contains over 20 times more in flesh during fruit development. Higher-than-normal accumulation of carotenoids in mutant fruits has also been reported in hp1, hp2 and hp3 tomatoes as well as a red-fleshed papaya and Hong Anliu sweet orange, however, the increment in fruits of these mutants was moderate, not exceeding six fold 12,29,42 .
Tremendously enhanced accumulation of carotenoids was reported in the curd of a cauliflower mutant Or, but the carotenoid accumulated was mainly β-carotene 43 , quite different from the linear carotenes in Cara Cara. Accumulation of lycopene in fruit of red-fleshed mutants rather than wild-type or ordinary cultivars has been reported for several plants, including papaya and citrus 12,28,29,32 . However, the carotenoid composition in Cara Cara fruit is unique among these lycopene-accumulating mutants, characterized by accumulation of higher amounts of another two other linear carotenes, phytoene, which was twice the amount of lycopene, and phytofluene, as well as normal amounts of xanthophylls ( Fig. 2A). This contrasts with the results for ordinary tomato and red-fleshed watermelon, where the amount of phytoene is normally less than 10% of lycopene 44,45 , for red-fleshed grapefruits, which contains high amount of phytoene and phytofluene but the xanthophyll amount is quite low 28 , and for red-fleshed Hong Anliu sweet orange, which does not contain phytoene in flesh tissues 29 . Thus the unique carotenoid accumulation in Cara Cara may indicate involvement of an unusual mutation.
The possible cytological mechanisms for the unique carotenoid accumulation in Cara Cara were investigated. Carotenoids are synthesized and accumulated in plastids. There are several types of plastids in plants, and chromoplasts are the main plastid type for carotenoid accumulation in ripe fruit. Chromoplasts vary in their  Table S1). The bars represent SE from three biological replicates. A significance test was conducted and values with the same letter indicates no significant difference at P < 0.05.
Scientific RepoRts | 7: 761 | DOI:10.1038/s41598-017-00898-y Figure 9. In vitro cultured flesh tissues of Newhall and Cara Cara following CPTA and norflurazon treatments. The 0 week tissues were separated from fruit at the S5 stage. CPTA, 2-(4-chlorophenylthio)-triethylamine hydrochloride. morphology and types of carotenoid-accumulating substructures and can be classified as globular, crystalline, membranous, fibrillar, tubular and vesicle types 46 . Here yellow and globular chromoplasts were found in both oranges but red and crystalline ones were only observed in Cara Cara (Fig. 3).
Chromoplast development may involve either transition from other plastids such as chloroplasts, amyloplasts, etc., or direct biogenesis from proplastids, and particular patterns of development are important for carotenoid accumulation in different wild type and mutant plants. In hp1, hp2 and hp3 tomatoes, the enhanced accumulation of carotenoids was related to increased chromoplast number and size 42 . The Or cauliflower mutant possesses a much higher ability for conversion of proplastids and other colorless plastids to yellow chromoplasts and accumulates high amounts of β-carotene 24 . Overexpression of the cauliflower Or in potato resulted in enhanced carotenoid accumulation in tubers and promoted the development of chromoplasts with similar ultrastructure to those in Or cauliflower 47 . Through Raman microscopy, we showed that the red crystalline chromoplasts in Cara Cara contain lycopene, while globular chromoplasts do not (Fig. 5). Therefore, it is clear that occurrence of red crystalline chromoplasts is closely related to the accumulation of lycopene, and probably other linear carotenes as well. Association of the occurrence of crystalline chromoplasts with lycopene accumulation has also been observed in red-fleshed papaya 12 . In peppers, it has also been shown that variations in shape, size and ultrastructure of chromoplasts are associated with differences in carotenoid amount and composition in different cultivars 48 .
Because of the close association between occurrence of red crystalline chromoplasts and the accumulation of carotenes in Cara Cara, it might be tempting to conclude that the development of this type of chromoplast determines the unique carotenoid accumulation in citrus. However, through CPTA treatment, it was shown that enhanced linear carotene accumulation can result in development of crystalline chromoplasts in ordinary orange flesh (Fig. 11), which suggests that chromoplast sink capacity is not the factor limiting the accumulation of carotenoids in ordinary citrus, and development of chromoplasts, even of a specific type, can be stimulated to adapt for the enhanced biosynthesis of carotenoids, and, in this case lycopene accumulation caused crystalline chromoplasts to develop where they normally do not occur. Perturbation of plastid development by modified carotenoid  synthesis has also been reported in other plants. In tomato overexpressing SlPSY1, chromoplast-like structures developed prematurely during fruit development 49 , and in maize, it was found that enhanced carotenoid biosynthesis through modification of natural PSY1 to increase enzyme activity resulted in distorted plastid shape and formation of fibrillar plastoglobulin 50 .
In summary, the causative relationship between chromoplast development and carotenoid accumulation is complicated, and may not be the same in different cases. In ordinary orange, accumulation of carotenoids might be limited by metabolism rather than chromoplast development, and in Cara Cara, occurrence of red crystalline chromoplasts might be not the reason for, but the result of enhanced carotenoid accumulation.
Several possible metabolic control points for enhanced carotenoid accumulation in Cara Cara were evaluated. PSY is the enzyme which catalyses production of the first carotenoid molecule phytoene, and is implicated as a major metabolic control point for carotenoid synthesis 49 . Therefore, it is important to evaluate whether PSY is involved in higher carotenoid accumulation in Cara Cara. Three PSY members are present in the citrus genome, however, in flesh tissues, only PSY1 is strongly expressed, accounting for over 95% of total PSY transcripts in both orange cultivars, while the expression of PSY2 is quite weak (Fig. 7 of this study; Peng et al. 51 ) and PSY3 transcripts are undetectable (data not shown; Peng et al. 51 ). Evolutionarily, oranges are offsprings from several natural crossing between pummelo and mandarin 52 . Therefore, oranges are heterozygous genetically. Here we observed two alleles for PSY1, with over 98% nucleotide homology between alleles, and there was no difference in sequence or allele expression between the two sweet orange cultivars in this study ( Fig. 8; Supplementary Fig. S2). These data suggested that PSY1 is not the gene responsible for high carotene accumulation in Cara Cara, in accordance with the suggestions made previously by Alquézar et al. 30 . Recently, posttranscriptional regulation of PSY activities was reported in Arabidopsis 53 , however, this may not occur in Cara Cara, since the increment in phytoene accumulation during a 4-week incubation of juice vesicle tissues on NFZ medium was slightly lower (49.32 μg g −1 FW) in Cara Cara than in Newhall (60.06 μg g −1 FW) (Supplementary Table S2), indicating that PSY activity is not higher in Cara Cara.
Lycopene cyclase genes include LCYB and LCYE, predominantly expressed in green tissues, as well as CYCB, expressed specifically in chromoplast-containing tissues 16 . CYCB is the gene with strongest transcript abundance in flesh tissues of oranges, being around 100 times higher in amount than LCYB and LCYE mRNA (Fig. 7). Similar to the results for PSY1, two alleles of CYCB were found, but no difference was observed in either allele between the two cultivars, and no obvious bias in expression of alleles was found (Fig. 8), which is quite similar to the data obtained by Alquézar et al. 35 for comparison between a white-fleshed and a red-fleshed grapefruit. Therefore, CYCB is unlikely to be the gene responsible for high carotene accumulation in Cara Cara. This conclusion was also supported by the data from CPTA treatment of cultured flesh tissues, where CPTA-treated Newhall flesh tissues contained much higher amounts of lycopene than phytoene, but did not have a carotenoid composition resembling the Cara Cara control ( Fig. 10; Supplementary Table S2).
Though the carotenoid amount is about 25 times less in flesh of ripe Newhall fruit compared to Cara Cara, the Newhall flesh tissues can be induced to accumulate considerable amount of carotenoids by CPTA and NFZ (Fig. 10, Supplementary Table S2), implying that the upstream biosynthesis ability is not lower in Newhall. Therefore, it would be valuable to further evaluate whether the carotenoid degradation capacity is different between these two cultivars and if it is, to identify the gene responsible for the high carotenoid amount in Cara Cara. This is an important area for future study since in some other plant tissues, e.g., white-fleshed peach fruit and white-coloured chrysanthemum petals, enhanced carotenoid degradation activity has been reported to be a factor resulting in absence of carotenoid accumulation 40,54 . Allelic expression analysis of key carotenoid catabolic genes should be an important focus for future study of the possible mechanism(s) and nature of the mutation responsible for the high carotene content of Cara Cara fruit.
Another interesting aspect of this study is observation of the occurrence of two different types of cells with distinct types of chromoplasts in Cara Cara flesh. To our knowledge, the presence of adjacent cells in the same tissue, each with an exclusive type of chromoplast, either globular or crystalline ones, but not both, has not been reported in plants previously.
Cara Cara is a mutant of ordinary orange 36 . The possibility that the mutation is in the plastid genome is not likely because in that case co-existence of both types of chromoplasts in a single Cara Cara cell would be expected. A remaining question would be how a mutation can result in the occurrence of two different types of cells with distinct type of chromoplasts. It would be reasonable to suggest a chimeric effect. However, although the two types of cells can be observed under the microscope, no distribution of sectors could be found with the naked eye (Fig. 1A), and these two types of cells can even be found distributed in a single juice vesicle (Fig. 4). These observations indicated that the fruit is not likely to be a chimera. Therefore the genetic background for each flesh cell should be same. Thus, comparison of the yellow and the red juice vesicles separated from Cara Cara flesh with modern -omics techniques can serve as an ideal starting point for further exploration of the mechanisms for high carotene accumulation.

Methods
Plant Materials. Fruits of Newhall (yellow-fleshed) and Cara Cara (red-fleshed) oranges (Citrus sinensis L. Osbeck) were sampled from an orchard in Linhai, Zhejiang, China. Fruits at eight different developmental stages, i.e., S1 to S8, from fruitlet to full mature (Fig. 1A) were collected from ten trees, with at least one representative fruit from each tree. The fruit were randomly separated into three biological replicates, with at least three fruit for each replicate. After color and size measurements of the fruit, the flesh tissue was separated and immediately frozen in liquid nitrogen, then stored at −70 °C for further analyses.
For allele expression analysis, juice vesicles from a single Cara Cara fruit at S6 were separated, with the most yellow ones (20% of the total) saved as yellow juice vesicles, and the 20% most red ones as red juice vesicles. Three fruit were involved with each as a biological replicate.
For NFZ and CPTA treatments, fruit were picked at S5 and were surface-sterilized as described 55 . Juice segments were excised from the equatorial region of the fruit, placed on Murashige and Skoog (MS) medium supplemented with 10% (w/v) sucrose and 1% (w/v) agar, pH = 5.8. The segments were placed with the endocarp side uppermost in the dark at 25 °C. CPTA and NFZ were added to the culture mediums to final concentrations of 0.2% and 0.1 mM, respectively. At 0, 2, 4 weeks after treatments, the segments were sampled, juice vesicles and albedo layer tissues separated, and stored at −70 °C for further analysis. For each sampling, nine segments were taken and separated into three biological replicates with three segments in each replicate.
Color and fruit size measurements. Peel color was measured using a Hunter Lab Mini Scan XE Plus colorimeter (Hunter Associates Laboratory, Inc., USA). The CIE L*a*b*color scale was adopted, and the data were expressed as L*, a*, b*, C*, H°, and the citrus color index (CCI) using the formula CCI = 1000 × a*/(L* × b*). Four random measurements per fruit were made. Fruit size at different stages was measured with Vernier calipers, using the transverse diameter as the fruit size.
Carotenoid extraction, quantification, and HPLC analysis. Carotenoids were extracted from flesh tissues, analysed and quantified by HPLC, according to a method previously described 28 . Briefly, around 0.5 g of fresh tissues were extracted with chloroform/methanol/Tris, and after centrifugation, the chloroform phase from two extractions were combined and dried under nitrogen gas. The residue was dissolved in diethyl ether and then saponified with 6% KOH. Water and chloroform were added to the saponified mixtures and after separation the chloroform phases were collected and dried under nitrogen gas. A Waters Alliance 2695 system (Waters Corporation, USA) equipped with a YMC reverse-phase C 30 column was used for HPLC analysis. Cell squashing and differential interference contrast (DIC) microscopy. The flesh tissue was cut into small pieces with a sterile scalpel blade and fixed overnight in 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.0). Then the fixing solution was removed and 0.1 M Na 2 EDTA was added to the samples, and then incubated at 60 °C for 2.5 h 56 . The cells were separated by squashing and observed and photographed with a Zeiss microscope (Germany).

Hyperspectral confocal Raman microscopy measurements. The Raman microscopy measurements
were performed on a Renishaw inVia laser-scanning confocal Raman microscope. The laser emitted light at 532 nm and the laser beam power was 50 mW. The signals were collected with the following parameters: spectral resolution less than 1 cm −1 , exposure time 1 s, confocality high. A single juice vesicle was used to make one freehand section. The testing laser beam was pointed precisely at a single chromoplast and each measurement determined separately using Raman microscopy 37 .
Transmission electron microscopy (TEM). Juice vesicles were separated and fixed overnight at 4 °C in 2.5% glutaraldehyde in 0.1 M phosphate buffer (pH 7.0). Samples were washed three times, 15 min for each, with phosphate buffer, then post-fixed with 1% OsO 4 in 0.1 M phosphate buffer (pH 7.0) for 1-2 h and washed three times in phosphate buffer. Samples were dehydrated by a linear gradient ethanol series (50%, 70%, 80%, 90%, 95%, and 100%), 15 min for each, and infiltrated by absolute acetone for 20 min. Samples were infiltrated with resin gradually (1:1 mixture of absolute acetone and the final Spurr resin mixture for 1 h, a 1:3 mixture of absolute acetone and the final resin mixture for 3 h, and finally the Spurr resin mixture overnight). Embedded samples were placed in capsules contained embedding medium and heated at 70 °C for 9 h. The samples were stained with uranyl acetate and alkaline lead citrate for 15 min each and observed in TEM using a Hitachi JEM-1230 (Japan).
Isolation of RNA and synthesis of cDNA. Total RNA was extracted from fruit frozen powder as previously described 22 . RNA integrity was electrophoretically verified and genomic DNA was eliminate by DNase I (RNase-free) (Fermentas MBI). The cDNA was synthesized using an iScript TM cDNA Synthesis Kit (Bio-Rad, USA). All procedures were according to the instruction manual.
Full-length gene cloning and sequencing. Primers for full-length cloning of PSY1, LCYB and CYCB were designed according to consensus sequences flanking coding sequences of each gene from C. unshiu, C. paradisi and C. sinensis deposited in GenBank and the sequences of primers are listed in Supplementary Table S3. PCR was performed using High Fidelity PCR Master kit (Roche, Switzerland) and general cloning procedures were followed. Sequences for both alleles were obtained from sequencing of at least ten recombinant plasmids for each gene.
Quantitative real-time PCR analysis. The quantitative real-time reactions were performed with a CFX96 instrument (Bio-Rad, USA) in a 20 μl system including 1 μl of each primer (Supplementary Table S1), 2 μl cDNA, 6 μl PCR-grade water, and 10 μl SsoFast TM EvaGreen Supermix (Bio-Rad, USA). The PCR programme was initiated with a preliminary step of 5 min at 95 °C, followed by 45 cycles at 95 °C for 5 s, 58 °C for 15 s, and 72 °C for 10 s. A melting curve was generated for each sample at the end of each run to assess the purity of the amplified products. The expression level of actin was used to normalize the mRNA levels for each sample, with abundance expressed as a multiple of actin. The expression level of each of the PSY1 and CYCB allele was measured by quantitative real-time PCR using allele-specific PCR primers (Supplementary Table S1). All PCR products were cloned and sequenced to verify the authenticity of the amplification.