Largely different carotenogenesis in two pummelo fruits with different flesh colors

Carotenoids in citrus fruits have health benefits and make the fruits visually attractive. Red-fleshed ‘Chuhong’ (‘CH’) and pale green-fleshed ‘Feicui’ (‘FC’) pummelo (Citrus maxima (Burm) Merr.) fruits are interesting materials for studying the mechanisms of carotenoid accumulation. In this study, particularly high contents of linear carotenes were observed in the albedo tissue, segment membranes and juice sacs of ‘CH’. However, carotenoids, especially β-carotene and xanthophylls, accumulated more in the flavedo tissue of ‘FC’ than in that of ‘CH’. Additionally, the contents of other terpenoids such as limonin, nomilin and abscisic acid significantly differed in the juice sacs at 150 days postanthesis. A dramatic increase in carotenoid production was observed at 45 to 75 days postanthesis in the segment membranes and juice sacs of ‘CH’. Different expression levels of carotenogenesis genes, especially the ζ-carotene desaturase (CmZDS), β-carotenoid hydroxylase (CmBCH) and zeaxanthin epoxidase (CmZEP) genes, in combination are directly responsible for the largely different carotenoid profiles between these two pummelo fruits. The sequences of eleven genes involved in carotenoid synthesis were investigated; different alleles of seven of eleven genes might also explain the largely different carotenogenesis observed between ‘CH’ and ‘FC’. These results enhance our understanding of carotenogenesis in pummelo fruits.


Introduction
Carotenoids, especially lycopene, contribute to the red-fleshed color of many plants, such as tomato (Solanum lycopersicum) [1], papaya (Carica papaya) [2,3], red carrot (Daucus carota) [4], watermelon (Citrullus lanatus) [5], pink guava (Psidium guajava) [6], gac (Momordica cochinchinensis) [7], and red-fleshed citrus (Citrus) [8]. These red-colored fruit fleshes have health benefits and are visually attractive. Recently, the mechanism of red-fleshed pigmentation due to carotenoids has attracted increasing amounts of research attention, and relevant candidate genes have been verified in papaya [2], carrot [9] and tomato [10]. Recently, the composition and content of carotenoids in red-fleshed citrus were reported to largely differ a1111111111 a1111111111 a1111111111 a1111111111 a1111111111 whole developmental stages and carotenogenesis based on the expression levels of carotenogenesis genes have not been elucidated. In this study, the carotenoid contents in the fruit tissues of 'CH' and 'FC' at five developmental stages were investigated. Furthermore, the expression levels of 13 carotenogenesis genes were analyzed, and 11 of those genes were cloned and sequenced. High expression levels of CmZDS, CmBCH and CmZEP in the segment membranes (SMs) and JS of 'CH' and different alleles of 7 genes in the JS might be directly responsible for the largely different carotenogenesis between the 'CH' and 'FC' pummelo fruits.

Materials and methods
All protocols in this manuscript share the same protocols, 10.17504/protocols.io.qsjdwcn.

Materials and reagents
'CH' and 'FC' pummelo (Citrus maxima) fruits exhibiting typical characters at each corresponding fruit development stage were collected from the scientific research orchard of the Forestry Science Academy of Lishui, Zhejiang Province, in 2011. The fruits were collected at 45, 75, 100, 125 and 150 days postanthesis (DPA) (Fig 2). The fruits of 'CH' and 'FC' were commercially mature between 125 DPA and 150 DPA. The flavedo, albedo, SMs and JS of the fruits were carefully dissected [22]. and Acidless (WS, pale green-fleshed) pummelo fruits at full maturity were collected from the National Citrus Breeding Center at Huazhong Agricultural University. Three biological replicates for each sample were prepared, with each biological replicate consisting of 6-8 fruits.
All the samples were treated with liquid nitrogen and then stored at -80˚C. Some of them were lyophilized in a Heto Lyolab 3000 (Heto-Holten A/S, Allerød, Denmark) and homogenized to a powder in liquid nitrogen for the extraction of metabolites.

Extraction and determination of carotenoids
Carotenoids were extracted from 1.0 g of lyophilized samples and then measured in accordance with the methods of Liu et al. [22] and Lee [34]. The carotenoids were identified by comparing the retention times and the absorption spectra with those of authentic standards and were quantified by peak areas. The peak areas were converted to concentrations by comparisons with the authentic standards of known concentrations measured by high-performance liquid chromatography (HPLC). Authentic carotenoid standards of antheraxanthin, α-carotene, β-cryptoxanthin, all-trans-lutein, all-trans-violaxanthin and phytoene were purchased from CaroteNature (Lupsingen, Switzerland), while those of β-carotene and all-translycopene together with standards of limonin and nomilin were purchased from Sigma Co. Ltd. (St. Louis, MO, USA).

Extraction and determination of limonin and nomilin in limonoid aglycones
In accordance with the methods of Li et al. [35], 3.0 g of lyophilized powder was extracted with 50 mL of dichloromethane via a FexIKAvarioControl system (IKA-Werke GmbH & Co. KG, Staufen, Germany). The extraction solution was collected and then dried under vacuum in a 5301 concentrator (Eppendorf, Hamburg, Germany) after 15 cycles of Soxhlet extraction (approximately 3 h), after which 1 mL of acetonitrile was ultimately added. Prior to the HPLC analysis, 1 mL of sample tissue was filtered through a micropore film filter (0.22 μm).
With respect to HPLC analyses, the same instrument was used to analyze the carotenoids, while a C 18 HPLC column (4.6×150 mm, 5 μm, Agilent, USA) was used for limonin and nomilin separation. For quantification, an isocratic elution model was applied with acetonitrile:10% methanol (40:60, v/v) at an elution speed of 1 mL/min and an injection volume of 20 μL (Manners, 2007). Both limonoids were detected at a wavelength of 210 nm under ambient temperature.

Extraction and determination of phytohormones
In accordance with the methods of Pan et al. [36] and Ding et al. [37], phytohormone extractions were carried out with 50 mg of lyophilized powder in 500 μL of chilled buffer (isopropanol:water:concentrated hydrochloric acid = 100:50:0.1 v/v/v). After the mixture was incubated at 0˚C for more than 12 h and after another 500 μL of chilled buffer was added, the mixture was shaken at a speed of 100 g for 1 h at 4˚C. Afterward, 2 mL of chilled dichloromethane was added to the sample, after which the sample was shaken for 1 h at 4˚C. The samples were subsequently centrifuged at 12,000 g for 10 min at 4˚C to form two phases. The lower phase was collected and concentrated using a nitrogen evaporator. The samples were redissolved in 0.15 mL of methanol and then centrifuged at 12,000 g for 15 min at 4˚C, after which 20 μL of the supernatant of each sample was filtered through 0.22-μm micropore filters for injection into a reverse-phase C18 Gemini HPLC column for HPLC-electrospray ionization-tandem mass spectrometry (HPLC-ESI-MS/MS) analysis.
In accordance with the methods of Ma et al.
[38], the phytohormone extracts were separated with HPLC (Agilent 1100, Agilent Technologies, Palo Alto, CA, USA) and measured via HPLC-ESI-MS/MS (API 3000 mass spectrometer, Applied Biosystems, Foster City, CA, USA), and the MS/MS conditions of each analysis were set in accordance with the methods of Pan et al. [36].
Authentic standards of IAA, ABA, SA and JA were purchased from OlChemIm (OlChemIm, Olomouc, Czech Republic).
Reverse transcription was performed as described by Liu et al. [22], while qRT-PCR was performed using an ABI 7500 Real Time System (PE Applied Biosystems, Foster City, CA, USA) in accordance with the methods of Liu et al. [22]. For each sample, the expression level of each biological replicate was determined by the mean value of at least three technical replicates. The mean threshold cycle (Ct) and the standard deviation (SD) for each sample were obtained in accordance with the methods of Gao et al. [39].

Amplification and sequencing of gene alleles
To amplify the whole coding sequences of the eleven genes identified mainly from the JS of both pummelo fruits, primers were designed in accordance with the methods of Liu et al. [22] and Gao et al. [39] or via Primer Premier 5; the primers are listed in S2 Table. The whole coding sequences of the CmDXS alleles could not be amplified, possibly due to their low expression levels in all fruit tissues. The whole coding sequences of at least 10 clones for each carotenogenesis gene were sequenced from the cDNA mainly from the JS in each pummelo fruit. In particular, with respect to CmPSY, CmLCYb1, CmLCYb2, CmLYCe and CmNSY, nearly 50 clones of each gene in each pummelo fruit were obtained. However, due to the difficulties in obtaining new clones, only 15 clones of CmZEP were obtained for 'FC'.
A Phanta 1 Max Super-Fidelity DNA Polymerase Kit (Vazyme Biotech Co., Ltd) was used for allele amplification. The PCR products were electrophoresed and then screened on 1.5% agarose gels, after which they were recovered using an EZNA 1 Gel Extraction Kit (Omega, USA) in accordance with the manufacturer's instructions.
A vector was connected to the PCR amplification products using a Zero Background pTO-PO-Blunt Simple Cloning Kit (Aidlab Biotechnologies Co., Ltd) in accordance with the manufacturer's instructions. For the selection of positive clones, 100 ng/μL ampicillin and universal M13 primers were used, and the positive clones were sent to Tshingke Biological Technology Co. (Wuhan, China) for sequencing.
Phylogenetic analysis. Homologous gene sequences of each carotenogenesis gene were obtained from the NCBI database (https://blast.ncbi.nlm.nih.gov/Blast.cgi). Their phylogenetic relationships were analyzed using MEGA 7.0.14. The maximum likelihood method was used to construct phylogenetic trees with 1000 replicate bootstrap supports, and values greater than 50% are shown at nodes.

Data analysis
The developmental stages of 'CH' and 'FC' (Fig 2) were processed via Photoshop software. The metabolite and gene expression profiles were processed by OriginPro 8. Average significant difference analyses were examined at a threshold of P<0.05 and were conducted with ANOVA in SAS (SAS Institute, Inc., USA).

Carotenoid production in the tissues of 'CH' and 'FC' pummelo fruits during development
'CH' and 'FC' pummelo fruits at five developmental stages were selected for analyses. The red-colored pigmentation in 'CH' started from the albedo at 75 DPA, followed by the SMs and JS at 100 DPA; this pigmentation ultimately peaked at 150 DPA in the SMs and JS. No red color was observed in any tissue of 'FC' at any stage (Fig 2). Combined with the content and composition of individual carotenoids detected by HPLC, lycopene contributed mainly to the red color in the albedo, SMs and JS of 'CH'.
In the flavedo, lutein, violaxanthin, α-carotene and β-carotene were found in both pummelo fruits during development. Except at the first stage, the total carotenoid content in 'FC' was significantly higher than that in 'CH', which is in accordance with our previous report concerning fruits at the commercially mature stage [11]. During fruit development, the total carotenoid content in 'CH' and 'FC' sharp decreased from 304.90 μg/g (dry weight) to 5.12 μg/g (dry weight). The dominant carotenoid in the flavedo was β-carotene, followed by lutein, before 125 DPA (Fig 3A).
In the albedo tissue, after the first stage (45 DPA), lycopene was the dominant carotenoid in 'CH', followed by phytoene and β-carotene. A dramatic accumulation of lycopene content in 'CH' was observed: it increased from trace levels at 45 DPA to the highest level at 125 DPA, after which it decreased beginning at 150 DPA. In contrast, a small amount of carotenoids were synthesized in 'FC'. Only low levels of lutein at 75 DPA and α-carotene at both 75 DPA and 100 DPA were observed; all other carotenoids were at trace or nondetectable levels throughout the fruit development of 'FC' (Fig 3B).
In the SMs, the production of total carotenoids rapidly increased in 'CH' after 75 DPA. A large amount of lycopene accumulated, followed by β-carotene and phytoene. A dramatic accumulation of lycopene in 'CH' was observed from 45 DPA to 75 DPA. Afterward, the content increased by nearly 4-fold at 100 DPA and peaked at 125 DPA (accounting for 90.2% of the total carotenoids). However, lycopene and the total carotenoids both decreased dramatically at full maturity (150 DPA) (accounting for 93.9% of the total carotenoids). The contents of β-carotene and phytoene increased, exhibiting a trend similar to that of lycopene. In addition, the two carotenoids accounted for 6.2% of the total carotenoids after 45 DPA. However, low carotenoid biosynthesis was observed in 'FC', with only small amounts of violaxanthin and α-carotene detected at 75 DPA and 100 DPA, respectively ( Fig 3C).
Likewise, in the JS of the 'CH' pummelo fruits, a large amount of lycopene was observed, followed by β-carotene and phytoene. A dramatic accumulation of lycopene in 'CH' was observed from 45 DPA to 75 DPA. Afterward, the level increased by nearly 3-fold at 100 DPA and peaked at 150 DPA, accounting for 89.5% of the total carotenoids. The contents of β-carotene and phytoene increased and exhibited a continuously increasing trend until the last development stage as did lycopene. On average, the two carotenoids accounted for 6.5% of the total carotenoids after 45 DPA, which is quite similar to the results in the SMs. As in the SMs, small amounts of carotenoids accumulated in the JS of 'FC'; the highest levels were detected for violaxanthin at 125 DPA. Moreover, less contents of lutein, α-carotene and β-carotene were observed before 100 DPA, after which point the contents decreased to nondetectable or trace levels. In addition, undetectable or trace levels of phytoene and lycopene were observed at all stages of 'FC' (Fig 3D).

Contents of limonin and nomilin in 'CH' and 'FC' pummelo fruits
The contents of limonin and nomilin in the flavedo of 'CH' fruits were significantly lower than those in the 'FC' fruits from 100 DPA to 150 DPA. However, in the albedo, the contents of both limonoid aglycones in 'CH' were significantly lower than those in 'FC' at the two early stages, after which the contents dramatically increased at the last stage. However, both limonin and nomilin in the albedo of 'FC' continued to decrease to the lowest observed levels at 150 DPA. Interestingly, the highest limonin and nomilin production was observed in the SMs of both pummelo fruits; the limonin and nomilin contents in the SMs were 25.0-, 7.5-, 9.5-fold and 29.5-, 12.5-, 8.5-fold greater than those in the flavedo, albedo and JS of 'CH', respectively, and 5.5-, 3.0-, 7.0-fold and 2.0-, 2.5-, 6.5-fold greater than those in the flavedo, albedo and JS of 'FC', respectively. Notably, in SMs and JS, the limonin content was significantly higher in 'CH' than in 'FC' at both 100 DPA and 150 DPA. Additionally, at 150 DPA, the nomilin content in 'CH' was markedly higher than that in 'FC'. In contrast, both the limonin and nomilin contents were significantly lower in 'CH' than in 'FC' at 125 DPA (Fig 4A).

Phytohormone contents in 'CH' and 'FC' pummelo fruits
Owing to the large differences in the contents of carotenoids and limonoid aglycones in the JS at the last developmental stage, we measured the concentrations of phytohormones in the JS to determine whether such differences were caused by the regulation of phytohormones. ABA was lower in 'CH' than in 'FC', which was consistent with the results for Star Ruby (redfleshed) and Marsh (white-fleshed) grapefruits [21], while IAA was higher in 'CH' than in 'FC'. In addition, the contents of JA and SA were at the same levels in both pummelo fruits (Fig 4B).
Expression levels of carotenogenesis genes during fruit development. In the flavedo, the expression levels of CmDXS, CmDXR and CmZISO were significantly higher in 'FC' than in 'CH' before 125 DPA, while the expression levels of CmPSY, lycopene β-cyclase gene 1 (CmLCYb1), CmBCH and CmZEP were constantly higher in 'CH' throughout all developmental stages or during most developmental stages. At 150 DPA, the expression levels of only CmZISO and CmLCYe were downregulated in 'CH', and the expression level of only CmZDS was the same in both 'CH' and 'FC', whereas the expression levels of the other 8 genes were significantly upregulated in 'CH'. Throughout all the stages, the largest fold change in gene expression was observed for CmBCH, followed by CmPSY, CmPDS and CmZDS, while the expression of CmLCYb1 remained relatively stable (Fig 5).
In the albedo, the expression levels of CmDXS, CmDXR, CmPDS, CmZDS, CmZISO, CmCRTISO and CmZEP were significantly higher in 'FC' than in 'CH' (if not at the same level). However, compared with those in 'CH', the expression levels of CmPSY, CmLCYb1 and CmLCYe in 'FC' were higher or the same at three stages. Notably, a generally increasing trend was observed in the expression levels of CmPDS, CmZDS, CmZISO, CmLCYe and CmZEP in both pummelo fruits, although larger fold changes occurred in 'FC'. Compared with that in 'FC', the expression level of CmLCYb1 in 'CH' was quite stable, but it drastically increased after 125 DPA (Fig 5).
The expression levels of 13 genes in the SMs were also evaluated. Higher expression levels of CmZDS, CmLCYb1, chromoplast-specific lycopene β-cyclase gene 2 (CmLCYB2), CmLCYe, CmBCH and CmZEP were detected in 'CH' than in 'FC' (if not at the same level). The expression levels of CmDXS, CmDXR, CmPSY, CmPDS, CmZISO and CmCRTISO were higher in 'CH' than in 'FC' in at least three stages. Notably, throughout all stages, the expression levels of CmLCYe, CmBCH and CmZEP were significantly higher in 'CH' than in 'FC'. Interestingly, with the exceptions of the upregulation of CmDXR, CmPSY and CmLCYb1, the expression levels of the other nine genes were downregulated at 150 DPA in 'FC'. Compared with that of the other genes, the expression of CmLCYb1 in 'CH' was relatively steady. The expression levels of CmLCYb2 and CmNSY were lower in 'FC' than in 'CH' at four of the five stages, while in 'CH' alone, the levels remained comparatively stable but were elevated at 125 DPA or 150 DPA (Fig 5).
In the JS, the expression levels of 13 genes were also determined. Notably, at 45 DPA, the expression levels of all the genes except CmCRTISO were higher in 'FC' than in 'CH'. After 45 DPA, the expression levels of CmZDS, CmBCH and CmZEP were all significantly higher in 'CH' than in 'FC' (if not at the same level); the same results were obtained for CmDXS,  CmDXR, CmZDS, CmZISO and CmLCYb1. Upregulated expression levels of CmPSY, CmPDS and CmCRTISO were observed at the last stage in 'FC'. However, relatively steady expression levels of CmDXS, CmLCYb1 and CmLCYe were observed in 'CH' throughout the fruit developmental stages (Fig 5).

Alleles of carotenogenesis genes in red-and pale-/pale green-fleshed pummelo fruits
Considering the large amounts of carotenoids (especially lycopene) in the albedo, SMs and JS of 'CH', the sequences of 11 carotenogenesis genes were therefore checked for possible mutations. CmPDS, CmZISO, CmBCH and CmNSY shared one common allele in the JS of both pummelo fruits, while for the other 7 genes, two alleles were obtained for each gene ( Table 1). The differences in nucleotide sequences of each carotenogenesis gene are shown in Table 1. Sites of various InDels and single amino acid polymorphisms are also indicated (S1-S11 Figs). Regarding CmPSY, which encodes the rate-limiting enzyme involved in carotenogenesis, CmPSYa and CmPSYb, which have one nucleotide InDel (site 183-185 causing CmPSYa N and CmPSYbamino acid residues, respectively) and one different nonsynonymous site (site 215 causing CmPSYa R and CmPSYb Q amino acid residues, respectively), are present in 'CH' and 'FC', respectively (Table 1). However, in the other eight investigated pummelo fruits, the same differences at both sites of CmPSY were observed regardless of flesh color, and six alleles (referred to as CmPSYa to CmPSYf) were investigated (Table 1; S2 Fig). Thus, the difference in both sites might not be the factor that contributes to the large accumulation of carotenoids. However, among the four genes encoding enzymes that have cyclization functions in 'CH' and 'FC', the CmNSY alleles were identical in both pummelo fruits (Table 1), whereas for CmLCYb1, CmLCYb2 and CmLCYe, two alleles were obtained for each gene (Table 1; S6-S8  Figs).
Moreover, a phylogenetic tree consisting of 11 genes was generated and analyzed via the full-length region of the studied genes and homologous genes obtained from the NCBI database (https://blast.ncbi.nlm.nih.gov/Blast.cgi); we named them the same as the closest previously reported genes in phylogenetic trees. The CmDXR and CmZISO genes were first cloned in citrus regardless of genome sequencing (S1-S11 Figs).

Tissue-specific carotenogenesis in 'CH' and 'FC' pummelo fruits
Carotenoids are extremely important for plant growth and human dietary needs. Understanding the mechanism of carotenoid biosynthesis and accumulation is of great significance. In this study, we investigated the carotenoid profiles of two pummelo fruits that have different flesh colors. In 'FC', trace or small amounts of carotenoids were synthesized in the albedo, SMs and JS, while the accumulation of carotenoids was high in the flavedo. In accordance with the results of our previous report [11], high amounts of β-carotene, lutein and violaxanthin were detected in the flavedo of both pummelo fruits, and no lycopene was detected. Additionally, in the flavedo, except at 45 DPA, the production of total carotenoids was higher in 'FC' than in 'CH' at the corresponding stages.
However, in the albedo, SMs and JS of 'CH', the increased carotenogenesis could be largely attributed to the accumulation of lycopene, followed by that of β-carotene and phytoene. Thus, the data imply that the production of phytoene increased and that lycopene cyclization and βcarotene hydroxylation are blocked in the 'CH' fruit tissues. Moreover, the results of this study indicated that carotenoid synthesis is tissue specific, which is clearly demonstrated by the Table 1

CmDXR
CmDXRa CH 1422(473) T(F) 20 (7   carotenoid synthesis in the flavedo and JS (Fig 3A and 3B). Clearly, different carotenoid profiles in the tissues of the two pummelo fruits indicate that the carotenoid biosynthesis pathway is relatively triggered in the JS of 'CH', while the pathway is suppressed in the flavedo during developmental stages. Moreover, the SMs of 'CH' also accumulated large amount of lycopene and thus more total carotenoids as observed in the JS; however, the reduction in these compounds in the SMs and the increase in the JS at the last stage again imply that carotenoid biosynthesis or lycopene accumulation is tissue specific, which is in good agreement with other reports in citrus fruit [15,40] and carrot [41,42].

Changes in other metabolites in 'CH' and 'FC' pummelo fruits
In addition to the different carotenoid profiles in both pummelo fruits, differences in the production of other terpenoids such as volatile terpenoids have also been previously reported [5], and different contents of limonoid aglycones and phytohormones were also observed in the present study (Fig 4). Carotenoids, volatile terpenoids, limonoids and ABA are all terpenoids and share the same precursor pyruvate [43] (Fig 1), and the relationships among terpenoid metabolic pathways are complicated [27,30,44]. Moreover, ABA is a derivative of carotenoids and regulates carotenoid biosynthesis via feedback inhibition [17]. In the present study, lower ABA contents were detected in 'CH' than in 'FC', which was similar to results reported for Star Ruby (red-fleshed) and Marsh (white-fleshed) grapefruit [21], suggesting that carotenoid metabolism was blocked in 'CH'. Ethylene is highly important for fruit ripening and influences the regulation of carotenoid production throughout fruit development. As a closely related phytohormone of ethylene, IAA also indirectly restricts the accumulation of carotenoids and thus plays a role in controlling the onset of fruit ripening [18]. In contrast, the endogenous IAA content was reduced in carotenoid-deficient tomato lines [45]. Compared with 'FC', 'CH' accumulated a higher IAA content at 150 DPA (Fig 4B), which also indicates an identical accumulation trend between carotenoids and IAA at the last stage. JA has been proposed to enhance lycopene accumulation in tomato fruits [46], and SA is a relatively far downstream product of β-carotene [47]. Thus, from the perspective of metabolomic networks, in redfleshed 'CH', other terpenoids are involved in the accumulation mechanism of carotenes, including phytoene, lycopene and β-carotene. However, it cannot be presumed that the largest difference in metabolism between 'CH' and 'FC' lies in the carotenoid biosynthesis pathway until their entire metabolic profiles are revealed.

Critical genes involved in the carotenoid biosynthesis pathway
Based on our results of carotenogenesis, it might not be appropriate to simply attribute the accumulation of lycopene, phytoene or β-carotene in 'CH' to the upregulated expression of upstream genes and the downregulated expression of downstream genes, as Liu et al. [22] reported for sweet oranges. It would be very helpful if the functions of the enzymes coded by various alleles of all related genes were clarified, which will require additional reliable and rigid verification due to the large numbers of alleles of all the investigated genes. In this study, the expression of two genes in the MEP pathway upstream of the carotenoid biosynthetic pathway and eleven genes in the latter pathway, as well as the alleles of the eleven genes, was evaluated ( Fig 5; Table 1). Overall, the higher expression of CmZDS, CmBCH and CmZEP might contribute to the higher lycopene content in the SMs and JS of 'CH'. Except at the first stage, the expression levels of genes involved in the MEP pathway (CmDXR and CmDXS) were significantly higher in the JS of 'CH' than in those of 'FC', implying that the biosynthesis of the initial carotenoid metabolites in the carotenogenesis pathway was largely induced in 'CH'. With respect to the rate-limiting gene CmPSY in carotenogenesis, we observed differences in two sites between CmPSYa and CmPSYb ( Table 1). Considering that 'AAT' repeats are correlated with high carotenoid (especially lycopene and β-carotene) accumulations in citrus [48], we rechecked for 'AAT' repeats ('N' repeats at the amino acid level) and carotenoid accumulation in the 10 pummelo fruits investigated (Table 1). Both CmPSY alleles are present in pummelo fruits that present flesh of various colors ( Table 1), suggesting that the differences in CmPSY alleles are not associated with flesh color. Posttranslational regulation is important with respect to supporting the function of carotenogenesis enzymes. Two alternative splice variants with different 5' untranslated regions (UTRs) were identified in the Arabidopsis PSY gene; these variants resulted in different translation rates for regulating AtPSY enzyme activity [49]. The LCYb1 [31,32,44,50], LCYb2 [41,42,44,50,51], LCYe [50] and NSY [52] genes function in catalyzing the cyclization of lycopene. Compared with the nonfunctional β-LCY2b, functional β-LCY2a was shown to have higher expression levels in navel oranges and whitefleshed marsh grapefruits, whereas nonfunctional β-LCY2b was expressed at much higher levels in red-fleshed Star Ruby fruits [21,31]. Since β-carotene (the direct product of β-LCY) also significantly accumulates in Star Ruby fruit [19,21,31], it cannot be excluded that functional LCYb and other cyclases such as LCYe and NSY are responsible for the accumulation of β-carotene [32,52]. In addition, the higher expression levels of CmLCYb1 and CmLCYb2 in the JS of 'CH' might explain the higher β-carotene content throughout the fruit development stages.
Notably, the expression of the downstream genes CmBCH and CmZEP was constantly significantly higher in the JS of 'CH' than in those of 'FC', which is similar to results reported in Star Ruby grapefruit [31] but opposite those in 'Cara cara' navel orange [13]. Yan et al. reported that maize crtRB1 alleles that encode an enzyme that exhibits BCH function are associated with reduced transcript expression and higher β-carotene concentrations [53]. However, in the present study, both the expression levels of CmBCH and the β-carotene accumulation were higher in the albedo, SMs and JS of 'CH' than in those of 'FC' at most development stages (Figs 3 and 5). Interestingly, compared with wild-type Arabidopsis plants, Arabidopsis mutants lacking ZEP activity exhibited a marked sixfold increase in total seed carotenoid contents, indicating that ZEP is a major contributor to carotenoid contents [54]. Furthermore, fruit-specific RNAi-mediated suppression of the 9-cis-epoxycarotenoid dioxygenase (SlNCED1) gene was accompanied by increased contents of lycopene and β-carotene in tomato fruits [55]. The carotenoid biosynthesis, maintenance and degradation genes also turned out to be regulatory factors involved in Actinidia fruit development [56].

Other factors that might regulate carotenoid biosynthesis
Transcription factors also play important roles in carotenoid accumulation. Overexpression of CsMADS6 resulted in higher lycopene contents and higher β-carotene contents in the sepals and fruits of tomato, respectively, by binding the promoters of LCYb1, PSY, PDS and CCD1 and upregulating their expression [57]. In a stay-green Ougan mutant (Citrus reticulata), CrMYB68 downregulated the expression of two carotenoid pathway genes, CrBCH and CrNCED5, which resulted in total flavedo carotenoid contents that were lower than those in the wild type [17]. The Mimulus lewisii R2R3 MYB transcription factor mutant rcp1 contains low amounts of carotenoids exhibits low expression levels of almost all carotenoid biosynthesis pathway genes [58]. Luo et al. reported that a stay-green protein, SlSGR1, dramatically reduced lycopene biosynthesis via the direct inhibition of SlPSY1 activity [59]. However, only one of these factors, ripening inhibitor (RIN), has been confirmed to regulate carotenoid accumulation in tomato fruits via interaction with the promoter of SlPSY1 [60].
Phytochrome-interacting factors (PIFs) bind to the PSY promoter, which suppresses its expression and reduces carotenoid accumulation in Arabidopsis [61]. Overall, since we did not find critical variants in carotenogenesis genes, the upstream regulators deserve extensive study in future research.

Conclusion
The development stage in which dramatic carotenoid production in the albedo, SMs and JS of 'CH' occurred was from 45 DPA to 75 DPA, and the total carotenoids increased by more than 39-, 40-and 14-fold, respectively. Different expression levels of carotenogenesis genes, especially CmZDS, CmBCH and CmZEP, in combination might be directly responsible for the largely different carotenoid profiles between these two pummelo fruits. The lycopene accumulation in specific 'CH' fruit tissues might strongly be caused by the different alleles of the CmDXR, CmPSY, CmCRTISO, CmLCYb1, CmLCYb2, CmLCYe, and CmZEP genes. The different numbers of 'AAT' microsatellite repeat units in PSY transcripts were not related to lycopene accumulation in red-flesh pummelo fruits.