Transcriptome dynamics underlying flower color in marigold (Tagetes erecta L.)

Background: Marigold (Tagetes erecta L.) is an important ornamental plant with a wide variety of flower colors. Despite its economic value, few biochemical and molecular studies have explored the generation of flower color in this species. Results: To study the mechanism underlying marigold petal color, we performed a metabolomics analysis and de novo cDNA sequencing on the inbred line ‘V-01’ and its petal color mutant ‘V-01M’ at four flower developmental stages. A total of 49,217 unigenes were identified from 24 cDNA libraries. Based on our transcriptomic and metabolomic analyses, we present an overview of carotenoid biosynthesis, degradation, and accumulation in marigold flowers. The carotenoid content of the yellow mutant ‘V-01M’ was higher than that of the orange inbred line ‘V-01’, and the abundances of the yellow compounds lutein, neoxanthin, violaxanthin, zeaxanthin, and antheraxanthin were significantly higher in the mutant. During flower development, the carotenoid biosynthesis genes were upregulated in both ‘V-01’ and ‘V-01M’, with no significant differences between the two lines. By contrast, the carotenoid degradation genes were dramatically downregulated in the yellow mutant ‘V-01M’. Conclusions: We therefore speculate that the carotenoid degradation genes are the key factors regulating the carotenoid content of marigold flowers. Our research provides a large amount of transcriptomic data and insights into the marigold color metabolome.

its use as a potted plant, a landscaping plant, and a cut flower, marigold is an important source of lutein, an oxygenated carotenoid with nutritional, pharmacological, and health benefits. The consumption of lutein-rich foods can effectively reduce the chance of developing macular degeneration, cataracts, and atherosclerosis, as well as the development of certain cancers [2,3]; therefore, the international demand for lutein is constantly increasing. Lutein is the main pigment in marigold petals, accounting for up to 90% of the total carotenoids in these flowers [4]. Marigold is one of the main raw materials from which lutein is extracted, making it a promising cash crop.
The accumulation of lutein in plants is also determined by its degradation. The degradation of lutein and other carotenoids involves the carotenoid cleavage dioxygenase (CCD) and 9-cis epoxy carotenoid cleavage dioxygenase (NCED) enzymes. CCD performs a major role in the degradation of a series of xanthophylls, such as lutein, zeaxanthin, violaxanthin, neoxanthin, and antheraxanthin, whereas NCED specifically catalyzes the degradation of zeaxanthin.
In general, genes encoding carotenoid biosynthesis and degradation enzymes are expressed in specific patterns in the various organs and at different developmental stages. Their expression patterns are finely regulated by various transcription factors, but only a few studies have explored these regulatory pathways in detail [6]. Ralf et al. [7] found that an Arabidopsis thaliana APETALA2 (AP2)/ethyleneresponsive element-binding protein transcription factor (AP2/ERF), RAP2.2, binds to the PSY and PDS promoters. The rap2.2 knockdown mutant displayed a decreased level of PSY and PDS expression, in addition to a 30% reduction in its carotenoid content. A phytochrome-interacting factor (PIF) was also shown to bind to the PSY promoter and inhibit its expression [8]. Similarly, a MADS box transcription factor, RIN (ripening inhibitor), was found to interact with the PSY promoter and participate in carotenoid accumulation in tomato (Solanum lycopersicum) fruit [9]. These transcription factors are known to regulate carotenoid accumulation in general; however, no transcription factors have been reported to regulate lutein biosynthesis or degradation specifically.
The lutein contents of different marigold varieties can vary more than 100 fold, resulting in differing petal colors, which can be white, cream, yellow, and orangered. We previously developed the marigold inbred line 'V-01', which has orange petals, and recently identified a natural mutant derived from this population, 'V-01M', which displayed identical developmental and botanical characteristics to 'V-01', except that it produced yellow petals. Here, we used metabolome and transcriptome sequencing techniques combined with bioinformatics to analyze these two marigold genotypes and identify the genetic mechanisms underpinning their different flower colors. This work improves our understanding of the transcriptional mechanisms by which carotenoid accumulation and degradation are regulated.

Plant materials
The marigold (Tagetes erecta L.) inbred line 'V-01' (orange petals) and its natural mutant 'V-01M' (yellow petals) (Fig. 1A) were grown in a climate chamber at 22°C with 70% relative humidity and a 16-h/8-h day/night photoperiod. Flower development was divided into four stages: pre-flowering (I), unopened flower (II), semi-open flower (III), and full flowering (IV) (Fig. 1B). Flowers were harvested from both plants at these four developmental stages, immediately frozen in liquid nitrogen, and stored at -80°C until required for the transcriptomic and metabolomic analyses. Three replicates were prepared for each sample.

Pigment extraction
Petals (100 mg fresh weight) were frozen in liquid nitrogen, ground into powder, and extracted with a solution of n-hexane:acetone:ethanol (2:1:1, v/v/v). The extract was vortexed for 30s, then an ultrasound-assisted extraction was carried out for 20 min at room temperature. The extract was centrifuged at 12,000 rpm for 5 min, after which the supernatant was collected and evaporated under a nitrogen gas stream. The extract was then reconstituted in 75% (v/v) methanol and centrifuged, and the supernatant was collected for the liquid chromatography-mass spectrometry using Cufflinks. The cuffdiff command was used to filter out the differential genes with a mapping read sum greater than 10 in both samples, |log2 (fold change)| > 1, P-value ≤ 0.05, and Q-value ≤ 0.05. GO and KEGG significant enrichment analyses was performed, and a hypergeometric test (phyper) was used to identify any GO/KEGG terms that were significantly enriched in the differentially expressed genes (DEGs) compared with all of the expressed genes (P-value < 0.05).

mutant 'V-01M'
The marigold inbred line 'V-01' (orange flowers) and its natural mutant 'V-01M' (yellow flowers) had very similar botanical characteristics, except for their petal color (Fig. 1A). Marigold flower development can be divided into four stages ( To analyze the differences in color at the biochemical level, we assessed the accumulation of carotenoids in the Stage-IV ligulate flowers of 'V-01' and 'V-01M' using HPLC and mass spectrometry. A total of nine carotenoids were detected, which could be divided into two subgroups: carotenes (orange pigments) and xanthophylls (yellow pigments). The carotenes included α-carotene, β-carotene, lycopene, and capsanthin, while the xanthophylls included lutein, violaxanthin, zeaxanthin, neoxanthin, and antheraxanthin. The xanthophylls was significantly more abundant in the yellow mutant 'V-01M' than in the orange line 'V-01' (Fig. 1C and D). In 'V-01M', the contents of all five xanthophylls were significantly higher than in 'V-01', especially for lutein and zeaxanthin (Fig. 1D). In contrast, no significant difference of carotenes were detected in 'V-01' and 'V-01M' (Fig. 1C).
These results showed that the higher accumulation of yellow pigments  Table S1). The raw sequencing data were filtered to remove low-quality reads that could affect the data quality and subsequent analysis. We obtained 132.954 Gb of clean reads, which were used to  (Table 2). Among the 33,810 unigenes with a match in the NR database, 8.8% were most similar to sequences from grape (Vitis vinifera),, followed by sesame (Sesamum indicum; 7%), robusta coffee (Coffea canephora; 6.2%), and wild tobacco (Nicotiana tomentosiformis; 4.3%) (Fig. 3A). The predicted function and gene classification of the marigold unigenes were identified using the KOG and GO databases. A total of 1,779 unigenes were annotated as 'signal transduction mechanisms' based on the KOG database, and the most common category was 'general function prediction only' (3,200 unigenes) (Fig. 3B).
Furthermore, the unigenes were annotated with GO terms, with the most common biological process categories determined to be 'metabolic process' and 'cellular process' (Fig. 3C).
Expression dynamics of the DEGs during flower development Cufflinks software was used to identify DEGs between the four developmental stages in both of the marigold genotypes. The FPKM values were used to estimate the gene expression levels, and volcano plots were constructed to describe the distribution of all DEGs identified in the library comparisons (Fig. 4). These results indicated that, in both 'V-01' and 'V-01M', the most dramatic change in the expression of the genes occurred between developmental Stages I to III, as well as the comparison between Stages I and IV. This suggested that a large number of genes are significantly differentially expressed throughout flower development.
Furthermore, the fold changes in the expression of the DEGs between Stages II and III were greater than those of the DEGs from the comparison of Stages I and II in both 'V-01' and 'V-01M', suggesting that a more dramatic change in gene expression occurred between Stages II and III than between Stages I and II.
This suggested that Stages II and III are the key phases of flower development with the most dramatic changes in gene expression. This is consistent with the observed accumulation of carotenoids and the color changes in the marigold flowers beginning in Stage III (Fig. 1B). The coloring of the ligulate marigold flowers began at Stage III, and the 'V-01' (orange) and 'V-01M' (yellow) began to visibly differentiate during this stage.
The most dramatic changes in gene expression, both in fold change (Fig. 4)  The 'V-01M' flowers accumulated significantly more xanthophylls than 'V-01' (more than a ten-fold difference); therefore, we also investigated their expression of HYD-B (TR20167) and HYD-E (TR27505), which are involved in the biosynthesis of a series of xanthophylls. Surprisingly, no significant difference was observed in the expression of either HYD-B or HYD-E between 'V-01' and 'V-01M' in any of the four developmental stages, suggesting that this is not the reason for the color differences observed in these lines.
We next examined the expression of the genes responsible for xanthophyll degradation, the most important of which encode the CCD enzymes. Among the three CCD genes identified in our transcriptome data, TR9765 was noticeably downregulated (8.96-fold decrease) in 'V-01M' compared with 'V-01' at developmental Stage III. Similarly, another CCD gene, TR16287, was expressed to a level 4.30 times lower in 'V-01M' than in 'V-01' at developmental Stage III (Fig. 7).
In addition to the CCDs, the degradation of zeaxanthin also involves the NCEDs. At developmental Stage IV, two of the four NCED genes identified in the marigold transcriptome were expressed at a dramatically lower level in 'V-01M' than in 'V-01'; 22.62-fold and 12.35-fold decreases were observed in the expression of TR22914 and TR2330, respectively, in the mutant flowers (Fig. 7).
The CCDs and NCEDs are the enzymes responsible for the degradation of xanthophylls; therefore, the low expression of CCDs and NCEDs in the 'V-01M' mutant likely resulted in its accumulation of xanthophylls and consequently the yellow color of its flowers.

Discussion
Carotenoids play an important role in photosynthesis, and their degradation produces a series of plant volatiles, as well as strigolactones and abscisic acid phytohormones [10,11]. Moreover, carotenoids are widely used in the food and pharmaceutical industry; for example, lutein and similarly structured carotenoids can protect retinal cells in the eye against oxidative stress, and a number of studies have suggested that the supplementation of lutein can maintain eye health and lower the risk of various chronic eye diseases [12]. Lutein is the major pigment in marigold petals, making this plant one of the most important sources of this xanthophyll in the pharmaceutical industry. In some cultivars, lutein can account for approximately 90% of the total carotenoids in the marigold petals [4]; however, the lutein contents of the different varieties of marigold can vary substantially, with more than 100-fold differences detected between some lines [13]. Many studies have therefore examined the genetic regulation of carotenoid accumulation in marigolds, which are considered a model plant for analysis of this pathway.     Table S1. Quality analysis of all reads from 24 samples.
Supplemental Table S2. Length distribution of the assembly transcripts and unigenes. Venn diagram of the number of differentially expressed genes in the four stages of marigold Expression patterns of genes encoding enzymes putatively involved in the biosynthesis and d

Supplementary Files
This is a list of supplementary files associated with the primary manuscript. Click to download.