Genome-wide identification and characterization of genes involved in carotenoid metabolic in three stages of grapevine fruit development

Carotenoids not only play indispensable roles in plant growth and development but also enhance nutritional value and health benefits for humans. In this study, total carotenoids progressively decreased during fruit ripening. Fifty-four genes involving in mevalonate (MVA), 2-C-methyl-D-erythritol 4-phosphate (MEP), carotenoid biosynthesis and catabolism pathway were identified. The expression levels of most of the carotenoid metabolism related genes kept changing during fruit ripening generating a metabolic flux toward carotenoid synthesis. Down regulation of VvDXS, VvDXR, VvGGPPS and VvPSY and a dramatic increase in the transcription levels of VvCCD might be responsible for the reduction of carotenoids content. The visible correlation between carotenoid content and gene expression profiles suggested that transcriptional regulation of carotenoid biosynthesis pathway genes is a key mechanism of carotenoid accumulation. In addition, the decline of carotenoids was also accompanied with the reduction of chlorophyll content. The reduction of chlorophyll content might be due to the obstruction in chlorophyll synthesis and acceleration of chlorophyll degradation. These results will be helpful for better understanding of carotenoid biosynthesis in grapevine fruit and contribute to the development of conventional and transgenic grapevine cultivars for further enrichment of carotenoid content.


Identification of 54 genes involving in grapevine carotenoid biosynthetic and catabolic pathway.
To determine the molecular basis and mechanism of the carotenoid biosynthetic/catabolic during grapevine fruit development, the genes involving in grapevine carotenoid biosynthetic/catabolic pathway were identified from the current genome. A total of fifty-four carotenoid biosynthesis-related genes were identified via BLAST-P search in NCBI using the Arabidopsis gene sequences as queries 1,15 . Subsequently, to verify the reliability of the initial results, a survey was conducted to confirm these genes using functional annotation of the grapevine transcriptome in three distinct stages of berry development (NCBI GEO Accession: GSE77218) 19 . As a result, 54 non-redundant carotenoid biosynthesis-related genes were identified and each carotenoid biosynthetic gene in grapevine was named based on the enzymatic reaction, similar to those given in the A. thaliana carotenoid biosynthetic pathway. Detailed information about each carotenoid biosynthetic and catabolic gene was showed in Table 1, including protein length, isoelectric points (pI), molecular weights, aliphatic index, grand average of hydropathicity (GRAVY) and subcellular localizations (Table 1).
Among the 54 carotenoid biosynthetic genes in grapevine, 10, 13, 24 and 7 genes were involved in MVA, MEP, carotenoid biosynthetic and catabolic pathways, respectively (Tables 1 and S1). These genes were members of 32 different gene families, and 22 genes were identified as single gene copy (Table 1). All these genes, only two (VvPSY3 and VvCCD8) were not expressed with a very low number of reads sequenced and an extremely low RPKM (reads per kilo base of exon model per million reads) value during three ripening stages. One transcript of VvMVK showed no change in expression level (|log2 fold-change (log2FC)| < 0.25). 20 transcripts were observed as up-regulated or down-regulated slightly (0.25 < |log2FC| < 1). Thirty-one transcripts were considered as significantly differentially-expressed genes (|log2FC|) ≥ 1 (Table S1). Fig. 1, 51 of 54 grapevine carotenoid biosynthetic genes were distributed unevenly throughout the 17 out of the 19 chromosomes, with eight genes located in Chromosome 4, while none of the carotenoid biosynthetic/catabolic genes mapped in Chromosomes 1 and 15 (Fig. 1). The remaining gene (VvDXS5, VvAACT1 and VvZEP3) had not yet been assembled to any chromosome according to the current genome.

Chromosomal distribution of grapevine biosynthetic genes. As shown in
Tandem and segmental duplications have been suggested to be two of the main causes for gene family expansion in plants 20 . The tandem duplication events were defined according to the methods of Holub 21 , where a chromosomal region within 200 kb containing two or more genes is defined as a tandem duplication event. There were five genes (VvDXS2/VvDXS4/VvDXS6 and VvCCD4a/VvCCD4b) clustered into two tandem duplication event regions on grape chromosome 2 and 11 (Table 1). Besides the tandem duplication events, 3 segregation duplication events (VvHMGR1/ VvHMGR3, VvGGPPS1/VvGGPPS2 and VvCHYB1/VvCHYB2) were also identified ( Fig. 2), indicating that some grapevine carotenoid biosynthetic genes were possibly generated by gene duplication. Moreover, the segregation duplication events can also provide a reference for the carotenoid biosynthetic gene evolutionary relationship and functional prediction.
Exon-intron organization of grapevine carotenoid biosynthetic gene. The exon-intron structure analysis was performed to gain more insight into the grapevine carotenoid biosynthetic genes. High variation was observed in numbers of exons and introns among grapevine carotenoid biosynthetic genes (Fig. 3 Variation of carotenoid and chlorophyll content during berry developing and ripening. In order to understand carotenoid and chlorophyll accumulation profiles in grapevine, the concentrations of carotenoid and chlorophyll in grape flesh were monitored at the three sampling time-points during fruit development  and ripening (Fig. 4). It was observed that the amounts of carotenoid and chlorophyll decreased significantly throughout fruit ripening stages. As expected, the total amount of carotenoids (expressed as μg/g fw) progressively decreased from 7.51 μg/g fw at the green stage to 2.19 μg/g fw at the ripe stage (Fig. 4). Similarly, the total contents of chlorophyll a and chlorophyll b decreased 3.7 and 3.2-fold during ripening, respectively (Fig. 4). The ratios of carotenoids/chlorophylls (~0.18) as well as the ratio of chlorophyll a/b (~2.0) remained constant throughout the sampling stages. Our results were consistent with the previous finding that carotenoid and chlorophyll contents of grape berries decreased with ripening especially from veraison to harvest 22-24 . Characterization and expression analysis of genes involving in the MVA pathway. Nine genes involved in six enzymatic steps were found in MVA pathway, which led to the formation of IPP and DMAPP. In addition, there was one FPS gene which catalyzed the synthesis of farnesyl diphosphate (FPP), the major substrate used by cytosolic and mitochondrial branches of the isoprenoid pathway 25,26 . Acetyl-CoA acetyltransferase (AACT) catalyses the condensation of two acetyl-CoA subunits to form acetoacetyl-CoA thus directing this central metabolite to the MVA pathway. In this study, VvAACT1 and VvAACT2 showed the highest expression levels at veraison and ripe stages, respectively. Interestingly, VvACAT2  mRNA expression (RPKM = 321.8 ± 71, the average value of the three ripening stages) had an approximate 5.8-fold higher than VvAACT1 (RPKM = 55.0 ± 13, also the average value of the three ripening stages) during berry developmental stages, suggesting that this enzyme may divert the metabolic flux of acetyl-CoA from the biosynthesis of fatty acids and amino acids toward the synthesis of isoprenoids ( Fig. 5 and Table S1). 3-Hydroxy-3-methylglutaryl-CoA reductase (HMGR), the first committed step of the MVA pathway for isoprenoid biosynthesis, catalyses the formation of MVA from 3-hydroxy-3-methyl-glutaryl-CoA (HMG-CoA) 27 . Three HMGR genes (VvHMGR1, VvHMGR2 and VvHMGR3) were identified in grapevine and two members (VvHMGR1 and VvHMGR3) were generated by gene duplication (Fig. 2). This was in agreement with previous reports that plant HMGR genes had arisen by gene duplication and subsequent sequence divergence 28 . Three members of VvHMGR exhibited significant differential expression patterns during fruit development. VvHMGR1 and VvHMGR2 showed the expression peak at green berry stage, whereas the highest expression of VvHMGR3 was detected at ripe stage ( Fig. 5 and Table S1), suggesting that VvHMGR might play a complicated role in the MVA pathway. Furthermore, the last three enzymes were encoded by single gene in grapevine. The expression of three genes exhibited no or slight changes throughout the berry developmental stages. These results indicated that the last three enzymatic steps of the MVA pathway might not be important control points in terpene biosynthesis.

Characterization and expression analysis of genes involving in the MEP pathway. In plants,
IPP and DMAPP entering in the biosynthesis of carotenoids are mainly synthesized by the MEP pathway in the plastids 12,29,30 . All these 13 grapevine genes were most possibly located on chloroplasts (Table 1) through protein subcellular localization prediction, which were consistent with the previous reports about the MEP pathway enzymes of Arabidopsis and Salvia 31, 32 .
The expressions of 11 genes involving in MEP pathway were declined during grapevine fruit development and ripening, only VvHDR and VvIDI showed an increasing expression during growth ( Fig. 6 and Table S1). 1-Deoxy-D-xylulose-5-phosphate synthase (DXS) catalyzes the first committed step of the MEP pathway and carotenoid biosynthesis. Six VvDXS members were identified in grapevine and displayed identical expression profiles, with the lowest and highest expression levels at veraison and green stage, respectively. Interestingly, the expression level of VvDXS1 was significantly higher than all other members (Table S1), suggesting that VvDXS1 might play predominant role in driving grapevine fruit carotenoid accumulation. 1-deoxy-D-xylulose-5-phosphate (DXP) is converted to MEP by the enzyme DXP reductoisomerase (DXR) encoded in grapevine by single gene. The expression profile of VvDXR was similar to VvDXS with its lowest and highest expression levels at veraison and green berry stages, respectively. MEP is subsequently converted into IPP and DMAPP by the consecutive action of five independent enzymes: 4-diphosphocytidyl-2-C-methyl-D-erythritol synthase (MCT), 4-diphosphocytidyl-2-C-methyl-D-erythritol kinase (CMK), 2-C-methyl-D-erythritol 2,4-cyclodiphosphate synthase (MDS), 1-hydroxy-2-methyl-2-(E)-butenyl 4-diphosphate synthase (HDS), and 1-hydroxy-2-methyl-2-(E)-butenyl 4-diphosphate reductase (HDR) 33 . VvMCT, VvCMK, VvMDS and VvHDS showed a slight decrease in expression levels during ripening, whereas VvHDR demonstrated increased mRNA expression during grapevine fruit ripening ( Fig. 6 and Table S1). Isomerization of IPP and DMAPP is catalyzed by isopentenyl diphosphate isomerase (IDI) encoded in grapevine by single gene whose mRNA expression profile remained strong and increased gradually throughout grapevine fruit ripening.
Characterization and expression analysis of carotenoid biosynthetic genes during grapevine berry development and ripening. A total of 24 genes were involved in carotenoid biosynthesis pathway, which led to the formation of carotenoids from the precursor geranylgeranyl diphosphate (GGPP). GGPP, the universal precursor for all plastid isoprenoids, is generated by geranylgeranyl diphosphate synthase (GGPPS) that catalyses the condensation of three IPP and one DMAPP units 34 . In our study, four VvGGPPS genes were identified and showed a steady decrease during grapevine fruit development ( Fig. 7 and Table S1). VvGGPPS2 was predominantly expressed in grapevine fruit ( Fig. 7 and Table S1), in agreement with the phenomenon that GGPPS2 was specific for chromoplasts in tomato flowers and/or fruits 35 . Phytoene synthase (PSY) the main bottleneck in the carotenoid pathway, catalyzes the condensation of two GGPP molecules to produce 15-cisphytoene 36,37 . Three VvPSY genes were identified and with different exprseeion patterns in grapevine. VvPSY1 was low at green fruit stage but started increasing at veraison, then decreased sharply at the ripening  stage. In contrast, VvPSY2 maintained low expression throughout fruit development, but increased sharply at ripening stage. VvPSY3 was no expressed in grapevine fruit ( Fig. 7 and Table S1). These results indicated that VvPSY genes showed a tissue-specific expression pattern.
The expression level of VvPDS depicted highest at green fruit stage, and then showed decreasing trend from veraison till to ripe stage. The expressions of VvZ-ISO, VvZDS and VvCRTISO kept increasing during fruit development and reached their picks at ripening stage. Additionally, VvCRTISO1 was much more highly expressed than VvCRTISO2 did with the progresses in growth season (Table S1), indicating VvCRTISO1 may have a more direct role in formation of lycopene. Furthermore, two different lycopene cyclases, lycopene β-cyclas (VvLCYB) and lycopene ε-cyclase (VvLCYE) were identified and showed an opposite expression pattern. Transcripts of VvLCYE decreased from veraison to harvest, whereas VvLCYB slightly increased through three periods of development (Table S1). The high expression levels of VvLCYB at harvest stage may help to maintain the amounts of γand β-carotene as intermediate metabolites for other compounds. Two VvCHYB were identified with a similar but quantitatively different expression pattern (Table S1). Neoxanthin synthase (VvNXS) and capsanthin-capsorubin synthase (VvCCS), two downstream genes of VvCHYB, also exhibited an opposite expression pattern. VvNXS were high expression at veraison and harvest stages, and VvCCS showed a very low expression. The expression pattern of two genes could be correlated to changes in ABA levels in grapevine berries that ABA levels typically peak at or around veraison and is thought to be responsible for the control of grape berry ripening 38, 39 . Characterization and expression analysis of carotenoid catabolic pathways during grapevine development and ripening. The steady-state level of carotenoids is dependent on the metabolic equilibrium between biosynthesis and degradation of carotenoids along with storage 40 . In our study, seven carotenoid cleavage dioxygenase (VvCCD) gene family members were identified and divided into two groups: four VvCCDs and three 9-cis-epoxycarotenoiddioxygenase (VvNCEDs). One member of VvCCD family (VvCCD1), predominantly expressed during fruit development, was found to increase and reached the highest value in the fruit ripen stage. Similar to VvCCD1, but to a lesser extent, VvCCD4b increased dramatically during grapevine development and ripening ( Fig. 7 and Table S1). The strong expression of two genes may help to control carotenoid turnover and contribute to the colors or aromas of grapevine fruit. Furthermore, three VvNCEDs showed different expression profiles in grapevine fruit. For example, VvNCED1 and VvNCED6 had the highest expression at green fruit stage, while VvNCED2 displayed the highest expression at ripe stage. The results agree with earlier studies that the expression pattern of VvNCEDs could not be correlated to changes in ABA levels in grapevine berries 40 . These results support the previous studies stating that different CCDs and NCEDs recognized different carotenoid substrates and cleaved at different sites, producing various apocarotenoids 41 .

Characterization and expression analysis of chlorophyll metabolism genes during grapevine fruit development and ripening. Color change often accompanies the onset of fruit ripening and it is
one of the most noticeable characteristics of grape fruit ripening. Color change in fruit typically involves chlorophyll loss and an increase in production of yellow, orange, red or purple pigments. The chlorophyll metabolic pathway can be divided into three distinct phases: (1) synthesis of chlorophyll a from glutamate; (2) interconversion between chlorophyll a and b (chlorophyll cycle); and (3) degradation of chlorophyll a into anon-fluorescent chlorophyll catabolite (Fig. S2) 42 . In this study, 29 genes involving in the three chlorophyll metabolic phases were identified (Table S2).
Twenty-one genes in chlorophyll biosynthesis, which catalyzed the formation of chlorophyll a from glutamate, exhibited gradual decrease during grapevine fruit ripening, except coproporphyrinogen III oxidase (VvCPOX) and protochlorophyllide reductase C (VvPORC) (Fig. 8 and Table S2). These results were consistent with a previous study indicating that chlorophyll degradation represented a dramatic change in metabolism at the onset of fruit ripening 43 . Further, four genes were identified in chlorophyll cycle pathway. With the exception of chlorophyll b reductase (VvNYC1), which was highly expressed at ripe stage, the other three genes also showed decreased expression patterns from the green fruit stage to the veraison/ripe stage. In chlorophyll degradation pathway, four genes were identified and the expression of three genes (chlorophyllase 2, VvCLH2; pheophytinase, VvPPH and pheophorbide a oxygenase, VvPaO) increased steadily during grapevine fruit ripening. VvPaO, a gene converting pheophorbide a into red chlorophyll catabolite, was highly expressed and showed a significant increase throughout the grapevine berry developmental stages, indicating that VvPaO may have an important role in the chlorophyll degradation. Transcriptomic datas demonstrated that the chlorophyll decline during grapevine fruit ripening was due to a acceleration in chlorophyll degradation and/or the blocking of chlorophyll synthesis.

Quantitative real-time PCR validation. Eight genes involved in MVA, MEP, carotenoid biosynthesis
and catabolism pathway were chosen for quantitative real-time PCR (qRT-PCR). Transcript abundance patterns were calculated over the entire course of berry development. The qRT-PCR results generally agreed (87.5%) with the changes in transcript abundance determined by RNA-seq (Fig. S3). Only VvDXS1, which showed lowest expression level in ripe stage, were different from that observed in RNA-seq. The qRT-PCR results suggested the reliability of the RNA-seq data.

Discussion
Carotenoid metabolism has received much attention due to the production of the phytohormones ABA and strigolactone; as well as the formation of norisoprenoids, impact flavour and aroma compounds in a number of commercially important fruits and flowers. In grapevine, carotenoids are not only responsible for the visual appeal of grape fruit, but also enhance nutritional value and health benefits for humans 44,45 . Thus, carotenoid content has become an important grapevine breeding objective and the study on grapevine carrotenoid metabolism has been focused. In this study, an integrative study combining carotenoid content and gene expression profiles was performed to gain insight into the synthesis and accumulation of carotenoids in grapevine during fruit development and ripening.
A toal of 54 carotenoid metabolism-related genes were identified and distributed on 17 of 19 chromosomes in grapevine, in agreement with the previous study in Brassica rapa 37 that the distribution of carotenoid metabolism genes was fractionation status at the whole-genome level. Furthermore, gene duplication and divergence events have been suggested to be the main contributors to evolutionary momentum 46 . In the current study, 11 of 54 carotenoid metabolism-related genes have arisen through either tandem or segmental duplication. Unlike Brassica rapa, two pair of tandem duplication events, VvDXS2/VvDXS4/VvDXS6 and VvCCD4a/VvCCD4b, were observed in the grapevine carotenoid genes (Fig. 2). Three pairs of segmental duplication (VvHMGR1/VvHMGR3, VvGGPPS1/VvGGPPS2 and VvCHYB1/VvCHYB2) were also identified (Fig. 2). Taken together, both tandem and segmental duplications existed in grapevine carotenoid metabolism-related genes, implying that gene duplications play a major role in genomic rearrangements and expansions of grapevine carotenoid metabolism. In addition, expression differentiation was also found between duplicated genes (Table S1). For example, in three pairs of segmental duplication, one copy was much more highly expressed than the other. In other genes, VvCCD4a and VvCCD4b showed the opposite expression (Table S1). The expression differentiation between duplicate genes has been reported and it increases the probability of the retention of duplicated genes in a genome 47,48 . These duplicated gene expression variations may be signs of subfunctionalization among different tissues and contribute to an increased complexity in the regulatory networks of the carotenoid pathway in grapevine.
Carotenoids are derived from the condensation of the 5-carbon precursors IPP and DMAPP, which are produced via the MEP pathway in plastids 49 . So far, both DXS and DXR are thought to be important rate-controlling steps of the MEP pathway and play important role in carotenoid flux regulation 6,50 . In grapevine, six VvDXS genes were divided into three distinct classes (Fig. S4), suggested that each VvDXS gene may play different role in terpenoid biosynthesis. VvDXS1, the member of the DXS1 clade, was highly and predominantly expressed compared with those of other VvDXS genes, indicating that VvDXS1 might play a predominant role in driving grapevine fruit carotenoid accumulation. The results are also in agreement with the previous study that Genes in the DXS1 clade played housekeeping roles and were probably involved in primary metabolism, such as the biosynthesis of carotenoids and the phytol chain of chlorophyll [51][52][53] . Futthermore, a positive correlation was found between the accumulation of DXR transcript and carotenoid production in Arabidopsis seedlings 50 or apocarotenoids in mycorrhizal roots from monocots 54 . Similar to Arabidopsis, the expression of VvDXR was also consistent with the situation of VvDXS1 with wide and high level expression during grapevine fruit development and ripening. In addition, the wide and high level expression of VvDXS1 and VvDXR may help to maintain the amounts of IPP and DMAPP as intermediate metabolites for carotenoid compounds and contribute to the colors or aromas of grapevine fruit.
The cytoplasmic MVA pathway also contribute to the synthesis of IPP and DMAPP (Fig. S1), which provide precursors for the biosynthesis of sesquiterpenes, polyterpenes, sterols, dolichol, and for ubiquinone formation in mitochondria. In our sthdy, the expression of VvAACT2 was significantly higher than that of VvAACT1 Scientific RepoRts | 7: 4216 | DOI:10.1038/s41598-017-04004-0 (5.8-fold), suggesting that VvAACT2 may play a more important role in cytosolic isoprenoid biosynthesis during grapevine berry development and ripening. Similar to grapevine, Arabidopsis AACT2 (AtAACT2) was six times more efficient than AtAACT1 did in complement the erg10 yeast mutant lacking AACT 55 . Three VvHMGRs, the rate-limiting step in MVA pathway, showed a completely differential expression pattern throughout the berry developmental stages, in agreement with previous studies that differential expressions of HMGR isozymes had a critical regulatory role for channeling and counter balancing carbon flux to the different downstream pathways of development or stress response 56 .
GGPPS is an important branch point enzyme in terpenoid biosynthesis since their enzymatic product GGPP is the direct precursor not only carotenoids, but also diterpenes, ABA and other compounds 57 . In grapevine, four divergent VvGGPPS isoforms showed different subcellular localization consistent with the subcellular compartmentalization of the diverse GGPP-dependent terpenoid pathways. VvGGPPS2 showed significantly higher expression level than any other member, indicating that VvGGPS2 had essential functions in carotenoid biosynthetic pathway during grapevine fruit development. A decreasing expression of VvGGPS2 from green to the red ripe stage had a significant positive correlation with a decrease of the total carotenoid content. PSY is a rate-limiting enzyme of carotenoid biosynthesis and is frequently regarded as the major bottleneck in carbon flux to carotenoids 15 . For example, over-expressing tomato PSY1 (SlPSY1) significantly increased the carotenoid content in tomato fruit 58 . In this study, three VvPSY exhibited different expression pattern. Unlike SlPSY1 with most highly expressed in ripe tomato fruits 59 , VvPSY1 was highly expressed in green and veraison stages, whereas VvPSY2 showed the highest expression in ripe grapevine fruits, confirming the complex regulation of the pathway. VvPSY3, a homologly with tomato SlPSY3 (Fig. S5), was not expressed in fruit, suggesting that VvPSY3 should be associated with root carotenogenesis, because SlPSY3 mainly expressed in root 60 . Furthermore, the cyclization of lycopene is a central branch point in the carotenoid biosynthetic pathway (Fig. S1). The expression of VvLCYB gradually increased and was significantly higher than that of VvLCYE during grapevine fruit ripening, especially in ripe stage of VvLCYB (Table S1). The induction of VvLCYB during grapevine fruit development and ripening might be responsible for maintain the amounts of β-carotene as intermediate metabolites for other compounds, such as ABA and C13-norisoprenoids. In grapevine, ABA concentrations increase and peak at the onset of ripening 61  The steady-state levels of carotenoid is dependent on the balance between biosynthesis and degradation along with storage 15 . Consistent with Arabidopsis CCD 62 , seven VvCCD family genes were divided into two groups, including four VvCCDs and three VvNCEDs (Table S1). VvCCD1 and VvCCD4b were found to increase stably and reached the highest values in the ripe stage. The significantly negative correlation between carotenoid content and gene expression profiles indicated that VvCCD1 and VvCCD4b preferentially involved in volatile compound production from different carotenoid substrates during grapevine fruit development and ripening. This negative correlation between the expression of CCD1 and CCD4 and carotenoid levels also exists in chrysanthemum flowers, potato, and strawberry [63][64][65] . Taken together, these results suggested that expression profile of genes coding for carotenoid biosynthetic pathway associated with the accumulation of carotenoid in grapevine fruit.

Conclusions
Fifty-four carotenoid biosynthetic genes were identified in grapevine genome. The composition of carotenoid biosynthetic genes could explain the metabolic profiles of carotenoid accumulation and help to elaborate the genetic mechanism of carotenoid biosynthesis in grapevine. The expression analysis of carotenoid biosynthetic genes showed that down regulation of VvDXS, VvDXR, VvGGPPS and VvPSY and a dramatic increase in the transcription levels of VvCCD were responsible for the reduction of carotenoids. The visible correlation between carotenoid content and gene expression profiles also suggested that transcriptional regulation of carotenoid biosynthesis pathway genes is a key mechanism of carotenoid accumulation. These results represent an important reference study for further characterisation of carotenoid biogenesis and accumulation in grapevine.

Materials and Methods
Plant material. Three-years-old 'Fujiminori' grapevine trees, grown in the standard field conditions at the Nanjing Agricultural University fruit farm, Nanjing, China, were chosen as the experimental material. Berries samples were collected at three time points: green stage (40 DAF), veraison (65 DAF) and ripe/harvest stages (90 DAF) throughout the growing season. Furthermore, in order to capture a representative biological selection of transcripts at each time-point, RNA for Illumina sequencing was purified from tissues of 40 berries sampled from 20 bunches.

Identification and analysis of the carotenoid biosynthesis-related genes in grapevine.
Carotenoid biosynthesis-related genes were identified via BLAST-P search in NCBI using the Arabidopsis gene sequences as queries. Gene sequences of Arabidopsis involved in the carotenoid biosynthetic pathway were acquired from the TAIR database (www.arabidopsis.org) and KEGG pathway database (http://www.genome.jp/ kegg/pathway.html). The grapevine genome and a set of annotated gene sequences from Grape genome database (http://genomes.cribi.unipd.it/grape/) were used to identify the carotenoid biosynthetic genes in grapevine. Subsequently, to verify the reliability of the initial results, a survey was conducted to confirm these genes using functional annotation of the grapevine transcriptome in three distinct stages of berry development (NCBI GEO Accession: GSE77218) 19 .
Scientific RepoRts | 7: 4216 | DOI:10.1038/s41598-017-04004-0 Determination of carotenoid and chlorophyll content. Accurately weighted 0.5 g of fresh grapevine sample was taken, and homogenized in tissue homogenizer with 10 ml of 95% ethanol. Homoginized sample mixture was centrifuge for 10,000 rpm for 15 min at 4 °C. The supernatant were separated and 0.5 ml of it is mixed with 4.5 ml of the 95% ethanol. The solution mixture was analyzed for Chlorophyll a, Chlorophyll b and carotenoids content in spectrophotometer (Parkin). The chlorophyll a, Chlorophyll b and carotenoid are respectively determined by spectrophotometric measurement at 664, 649and 470 nm as previous research 66 . Sequence feature analysis of grapevine carotenoid biosynthesis-related genes. The genomic sequence for each carotenoid metabolism gene was extracted from the whole genomic sequence according to gene location in the chromosome in the annotation file using a programmed Perlscript. Intron/exon structures were predicted using the Gene Structure Display Server (http://gsds.cbi.pku.edu.cn/chinese.php). The molecular weight and theoretical isoelectric point were predicted using the Prot-Param analyses (http://cn.expasy.org/ tools/protparam.html) 67 on the basis of their sequence. Conserved domains were searched using the Conserved Domain Database (http://www. ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi). The localizations of deduced proteins were predicted on the TargetP1.1 server (http://www.cbs.dtu.dk/services/TargetP/) 68 . Chromosomal locations, gene duplication and phylogenetic analysis of grapevine carotenoid biosynthesis-related genes. The information of chromosomal position of each carotenoid biosynthesis-related genes was obtained from the grapevine genome (http://www.genoscope.cns.fr/externe/ GenomeBrowser/Vitis/), and then the map was drafted using MapInspect software (http://www.plantbreeding. wur.nl/uk/software-mapinspect.html). MCScanX software (http://chibba.pgml.uga.edu/mcscan2/) was used to detect the gene duplication events, with the E-value set below 1 × 10 −5 following the description in a previous study 69 . The diagrams were generated by the program Circos version 0.63 (http://circos.ca/). Phylogenetic trees were constructed in MEGA7.0 software using the neighbor-joining (NJ) method and maximum likelihood (ML) method with 1000 bootstrap replications.
Expression analysis of grapevine carotenoid biosynthesis-related genes. The expression patterns of carotenoid biosynthetic genes in grapevine were acquired from gene expression omnibus (GEO) database of NCBI (GSE77218), which measured using RNA-Seq data 19 . Differential expression was analyzed and calculated based on the count values of each transcript between libraries using edgeR (the Empirical analysis of Digital Gene Expression in R) software 70 . The thresholds for judging significant differences in transcript expression were "FDR < 0.001" and "|log2 fold-change (log2FC)| ≥ 1". Transcripts with |log 2 FC| < 0.25 were assumed to no change in expression levels. Other transcripts (0.25 < |log2FC| < 1) were considered as "up-regulated slightly" or "down-regulated slightly".
qRT-PCR validation. qRT-PCR was performed to verify the expression patterns revealed by the RNA-seq study. Total RNA samples of three stages of grapevine fruit development were extracted using Trizol reagent (Invitrogen, Carlsbad, CA, USA). Purified RNA samples were reverse-transcribed using the PrimeScript RT Reagent Kit with gDNA Eraser (Takara, Dalian, China) as per the manufacturer's protocol. Eight transcripts were selected for the qRT-PCR assay. Gene specific qRT-PCR primers were designed using Primer3 software (http:// primer3.ut.ee/) (Table S3). qRT-PCR was carried out using an ABI PRISM 7500 real-time PCR system (Applied Biosystems, USA). Each reaction mix was composed of 10 μl 2 × SYBR Green Master Mix Reagent (Applied Biosystems, USA), 2.0 μl cDNA sample, and 400 nM of gene-specific primers in a final volume of 20 μl. PCR conditions were: 2 min at 95 °C, followed by 40 cycles of denaturation at 95 °C for 10 s and annealing at 60 °C for 40 s. The relative mRNA level for each gene was calculated using the 2 −ΔΔCT formula 71 . A primer pair was also designed for TC81781 (The Institute for Genomic Research, Release 6.0), encoding an actin protein (housekeeping gene). At least three replicates of each cDNA sample were performed for qRT-PCR analysis.