Transcriptome analysis in petals and leaves of chrysanthemums with different chlorophyll levels

Chlorophylls (Chls) are magnesium-containing tetrapyrrole macromolecules responsible for the green color in plants. The Chl metabolic pathway has been intensively studied and nearly all the enzymes involved in the pathway have been identified and characterized. Synthesis and activity of these enzymes are tightly regulated in tissue- and developmental stage–specific manners. Leaves contain substantial amounts of Chls because Chls are indispensable for photosynthesis. In contrast, petals generally contain only trace amounts of Chls, which if present would mask the bright petal color. Limited information is available about the mechanisms that control such tissue-specific accumulation of Chls. To identify the regulatory steps that control Chl accumulation, we compared gene expression in petals and leaves of chrysanthemum cultivars with different Chl levels. Microarray and quantitative real-time PCR analyses showed that the expression levels of Chl biosynthesis genes encoding glutamyl-tRNA reductase, Mg-protoporphyrin IX chelatase, Mg-protoporphyrin IX monomethylester cyclase, and protochlorophyllide oxidoreductase were well associated with Chl content: their expression levels were lower in white petals than in green petals, and were highest in leaves. Among Chl catabolic genes, expression of STAY-GREEN, encoding Mg-dechelatase, which is a key enzyme controlling Chl degradation, was considerably higher in white and green petals than in leaves. We searched for transcription factor genes whose expression was well related to Chl level in petals and leaves and found three such genes encoding MYB113, CONSTANS-like 16, and DREB and EAR motif protein. From our transcriptome analysis, we assume that a low rate of Chl biosynthesis and a high rate of Chl degradation lead to the absence of Chls in white chrysanthemum petals. We identified several candidate transcription factors that might affect Chl accumulation in chrysanthemum petals. Functional analysis of these transcription factors will provide a basis for future molecular studies of tissue-specific Chl accumulation.


Background
Chlorophylls (Chls) are magnesium-containing tetrapyrrole macromolecules responsible for the green color in plants. Because Chls play a central role in light harvesting and energy transduction in photosynthesis, mature leaves contain a substantial amount of Chls [1]. Petals of many flowering plants contain Chls at early developmental stages [2]. As petals develop, Chl content decreases, and petals of fully opened flowers contain only trace amounts of Chls. The absence of Chls in petals is an important trait that enables flowers to be visually distinguished by pollinators against a background of leaves when the flowers are ready to be pollinated.
The Chl metabolic pathway can be divided into three distinct phases (Fig. 1) [3][4][5]: (1) biosynthesis of Chl a from glutamate; (2) interconversion between Chl a and b (Chl cycle); and (3) degradation of Chl a into a nonfluorescent Chl catabolite. Nearly all enzymes involved in the pathway are identified and characterized. In higher plants, synthesis and activity of these enzymes are tightly regulated in tissue-and developmental stage-specific manners.
There is increasing evidence that the Chl biosynthesis pathway is transcriptionally regulated in a tissue-specific manner [4,6]. Several transcription factors have been identified as negative or positive regulators of the Chl biosynthesis pathway. GOLDEN2-LIKE (GLK), LONG H YPOCOTYL5 (HY5), and GATA, NITRATE-INDUCIBLE, CARBON-METABOLISM INVOLVED (GNC) regulate Chl biosynthesis in leaves [7][8][9], whereas GNC-LIKE (GNL)/CYTOKININ RESPONSIVE GATA FACTOR1 (CGA1) enhances Chl biosynthesis in petals and stamens [9]. PHYTOCHROME-INTERACTING FACTOR 1 (PIF1) and PIF3 inhibit the biosynthesis of protochlorophyllide by interacting with photoreceptors in leaves in the dark [10]. Chl degradation is also regulated at the transcriptional level. Recently, ORE1 and ANAC046 have been shown to increase the expression of Chl catabolic genes [11,12]. Changes in the expression of genes for these transcription factors influence Chl content in Arabidopsis leaves, but limited information is available about the regulation of Chl metabolism in non-photosynthetic tissues.
As mentioned above, it is a distinct disadvantage for insect-pollinated flowers to accumulate Chls in petals. Therefore, even if a mutation that results in green flowers occurs, such plants may be eliminated from their natural habitat. In contrast, in ornamental plant breeding, greenflowered mutants are preferably selected and developed as cultivars. In this study, we used white-and green-flowered cultivars of chrysanthemum (Chrysanthemum morifolium Ramat.) to investigate the mechanisms that regulate Chl content in petals. We compared the expression profiles of genes related to Chl metabolism between these cultivars and searched for genes whose expression levels are positively or negatively associated with Chl content. We also searched for genes encoding transcription factors whose expression is regulated in coordination with Chl content in petals. From these data, we identified candidate transcription factors controlling Chl accumulation in chrysanthemum petals.

Plant materials
The chrysanthemum (Chrysanthemum morifolium Ramat.) cultivars Feeling White (FW), Feeling Green (FG), and Feeling Green Dark (FGD) (Japan Agribio Co., Shizuoka, Japan) were grown under natural daylight in a greenhouse at our institute (Tsukuba, Ibaraki, Japan). FGD and FW were bud sports arising from FG. Other white-and greenflowered chrysanthemum cultivars were purchased from the local market in Tsukuba. Petals of ray florets and the 3rd visible leaves from the top were harvested, immediately frozen in liquid nitrogen, and stored at −80°C until use.

Microarray analysis
We previously constructed a chrysanthemum expressed sequence tag (EST) database with 213,204 contigs [13]. Short contigs (<100 bp) and no-hit contigs (<197 bp) were omitted, and 176,026 contigs were chosen for custom oligonucleotide array design. Oligonucleotides of about 60 bases representing each EST were designed by Takara Bio (Ohtsu, Shiga, Japan) as described by Ohba et al. [14]. Oligonucleotides were synthesized on a glass surface (4 × 180 k) with SurePrint technology (Agilent Technologies, Palo Alto, CA, USA).
For microarray analysis, we chose petals at S2 because the difference in chlorophyll content among FW, FG, and FGD became evident at this stage. Total RNA was extracted from ray floret petals of FW, FG, and FGD at S2, and mature leaves of FG using Trizol reagent (Thermo Fisher Scientific, Waltham, MA, USA) and an RNeasy Mini Kit (Qiagen, Hilden, Germany). RNA integrity was evaluated using an Agilent 2100 Bioanalyzer (Agilent Technologies). Total RNA samples (100 ng) were used as starting materials. Three biological replicates were examined by the one-color method according to Nashima et al. [15]. Spot signal values were calculated with Agilent Feature Extraction version 9.1 software. Hierarchical clustering using Pearson correlation was performed with the tree-clustering tool of the Subio platform (Subio Inc., Kagoshima, Japan). The datasets for hierarchical clustering were normalized using a low signal cutoff (mean raw signal <100), log 2 -transformation, global normalization, and centering.
Gene ontology (GO) categorization was performed according to the to the annotation of the Arabidopsis gene corresponding to each probe [13]. GO terms were obtained from The Arabidopsis Information Resource (TAIR, http://www.arabidopsis.org/) (ATH GO GOSLIM , updated 2013 May 15).

Quantitative real-time PCR analysis
Quantitative real-time PCR (RT-qPCR) was performed as described previously [16]. cDNA for each gene was amplified by RT-PCR, cloned into the pGEM-T Easy Vector (Promega, Madison, WI, USA), and sequenced. Primers for RT-qPCR were designed from the sequences (Additional file 1 Table S1). Each plasmid was serially diluted 10-fold and used for a standard curve assay. The transcript copy number was determined by relating the RT-qPCR signal for each gene to a standard curve. The mRNA levels were calculated relative to that of actin (IABW01167629), which was constitutively expressed in leaves and petals as indicated by our microarray data. Analysis was performed in biological triplicate, and statistical significance was analyzed by Tukey-Kramer multiple-comparison test (P < 0.05).

Chlorophyll analysis
Tissues were ground into powder in liquid nitrogen and extracted with acetone. The samples were centrifuged at 10,000×g for 10 min, and the supernatants (80 μl) were mixed with 20 μl of water. Pigments were analyzed by high-performance liquid chromatography (HPLC; X-LC, Jasco, Tokyo, Japan) using a reversed-phase column (Symmetry C8, 150 × 4.6 mm; Waters, Milford, MA, USA) according to Zapata et al. [17]. The analysis was performed in biological triplicate.

Transmission electron microscopy
Tissues were cut into small pieces (approximately 1 mm 3 ), fixed, dehydrated, and embedded in Quetol 651 (Nisshin EM Co., Tokyo, Japan) as described previously [12]. Ultrathin sections were cut with a diamond knife on an ultramicrotome (Ultracut UCT, Leica Microsystems, Wetzlar, Germany). Sections were picked up on copper grids, stained with uranyl acetate and lead citrate, and observed under a transmission electron microscope (TEM) (JEM-1200EX; JEOL Ltd., Tokyo, Japan) at an acceleration voltage of 80 kV. Digital images were taken with a CCD camera (Veleta; Olympus Soft Imaging Solutions GmbH, Münster, Germany).

Chlorophyll content in petals and leaves
To identify genes coordinately expressed with Chl content, we used ray floret petals and leaves containing different Chl levels. In the white-flowered cultivar FW, small amounts of Chls accumulated in petals at the early developmental stage (S1) (Fig. 2). At the late stage (S3), Chl content decreased to extremely low levels. Larger quantities of Chls were detected in petals of the greenflowered cultivars FG and FGD than in those of FW especially at S3 (Additional file 4 Figure S2). At S3, FGD petals contained more Chls than FG and FW petals. At this stage, Chl content in FG leaves was 12.85 times that in FG petals and 6.56 times that in FGD petals. Because there was no significant difference in leaf Chl content among FW, FG, and FGD (Additional file 2 Figure S1), leaves of FG were used for further study.

Gene ontology classification of differentially expressed genes
Using microarray data, we extracted genes with the expression levels in FW or FGD petals >5 times greater or lower than in FG petals and examined the distribution of gene ontology terms (Fig. 3). In the cellular component category, genes associated with "plastid", "chloroplast", "other cytoplasmic component", "other intracellular component", and "other membrane" were highly enriched among the downregulated genes in FW petals. In the biological process category, genes associated with "electron transport or energy pathways", which include photosynthesis-related genes, were also highly represented among the downregulated genes in FW petals. No marked differences were found in the molecular function category.

Microarray and RT-qPCR analyses of chlorophyll-related gene expression
The chrysanthemum custom oligonucleotide array constructed in this study covered most genes involved in Chl metabolism (Additional file 3 Table S2). Microarray analysis was performed using this array and the expression profiles of selected genes of interest were further analyzed using RT-qPCR for a wider range of developmental stages of each tissue.

Chl biosynthesis
Many of the genes involved in Chl biosynthesis showed higher expression in FG and FGD petals than in FW petals (Fig. 4a). In particular, genes in group 1, CHLH (encoding Mg-protoporphyrin IX chelatase [Mg-chelatase] H subunit), CRD (encoding Mg-protoporphyrin IX monomethylester cyclase), HEMA1 (encoding glutamyl-tRNA reductase), and PORC (encoding protochlorophyllide oxidoreductase C) showed significantly higher expression in FG and FGD petals than in FW petals; the highest expression of these genes was found in FG leaves (P < 0.01). There were no significant differences in the expression levels of HEMD (encoding uroporphyrinogen III synthase), CHLG (encoding Chl synthase), CHLM (encoding Mg-protoporphyrin IX methyltransferase), HEMG2 (encoding protoporphyrinogen III oxidase), or CHLI (encoding Mg-chelatase I subunit).
Because the expression patterns of group 1 genes were positively matched with Chl content, we hypothesized that these genes play an important role in determining Chl content in petals and further analyzed their expression by RT-qPCR. As in microarray analysis, the levels of CHLH, CRD, HEMA1, and PORC transcripts were highest in leaves and were extremely low in FW petals (Fig. 5). In FGD, the expression of all four genes tended to increase as petals matured.

Chl cycle
Among genes related to the Chl cycle, the expression of two group 2 genes, NOL (Non-yellow coloring onelike, encoding Chl b reductase subunit) and HCAR (encoding hydroxymethyl Chl a reductase), was well correlated with Chl content: it was lower in FW petals than in FG and FGD petals and was remarkably high in FG leaves (Fig. 4b). RT-qPCR analysis of NOL and HCAR showed similar expression patterns to those observed in the microarray analysis (Fig. 5). In FG and FGD, the expression of group 2 genes tended to increase as petals matured.

Chl degradation
Among genes related to Chl degradation, the expression of only SGR (STAY-GREEN, encoding Mg-dechelatase) showed negative correlation with Chl content, with the highest expression in FW petals and extremely low in leaves (Fig. 4c). There was no clear relationship between Chl content and expression levels of RCCR (encoding red Chl catabolite reductase), PaO (encoding pheophorbide a oxygenase), and PPH (encoding pheophytinase). RT-qPCR analysis showed that the expression levels of SGR drastically increased as petals matured in FW, FG, and FGD, whereas that in leaves was very low (Fig. 5). Because an oligo probe specific to the SGR-like gene, a homologue of SGR, was absent from our custom oligonucleotide array, SGR-likespecific primers were designed based on its chrysanthemum EST sequence, and RT-qPCR was performed. Unlike SGR expression, that of SGR-like was lower in petals than in leaves during the course of development.

Comparison of gene expression between green and white petals
To confirm our findings, we compared the expression of group 1 and 2 genes and SGR (Fig. 4) in S3 petals of several white-and green-flowered cultivars (Additional file 4 Figure S2). Chl content in S3 petals was <1.4 nmol/g fresh weight (g FW) in white-flowered cultivars and 147-424 nmol/g FW in green-flowered cultivars (Fig. 6a). Among group 1 genes (CRD, CHLH, HEMA1, and PORC), only the expression of CRD was significantly higher in green petals than in white petals in all cultivars tested (Fig. 6b). Expression of the other three genes tended to be higher in green petals than in white petals in most cultivars compared. There was a considerable variability among cultivars in the expression levels of group 2 genes (NOL and HCAR) and SGR, with no significant correlation between transcript levels and Chl content.

Transcription factors differentially expressed in FW and FGD petals in comparison with FG petals
To identify transcription factors involved in the regulation of Chl accumulation in petals, we analyzed transcription factor genes whose expression was positively or negatively related to Chl content. Tables 1 and 2 list transcription factor genes differentially expressed between FG and FW petals and between FG and FGD petals, respectively, in our microarray data (P < 0.01 and fold change > ±15).
(See figure on previous page.) Fig. 3 Gene ontology classification of genes differentially expressed in FW, FG, and FGD petals. Expression of genes in stage 2 petals was compared between FW and FG and between FGD and FG. Genes were classified into categories (cellular component, molecular function, and biological process) after being mapped to the Gene Ontology database. The y-axis indicates the percentage of genes in each functional category, calculated from the number of genes whose expression in FW or FGD petals was >5 times greater or lower than that in FG petals We identified 11 transcription factor genes expressed significantly higher in FW petals than in FG petals ( Table 1, upper part). Five of them encoded MYB domain proteins and showed high fold change values (25.8-60.4). Four of these MYB sequences were most closely related to MYB113 in the Arabidopsis R2R3-Myb family and were designated MYB113-like. Arabidopsis MYB113 is involved in flavonoid biosynthesis [18,19]. However, the expression of one of the MYB113-like genes was not associated with that of the flavonoid biosynthesis genes dihydroflavonol 4-reductase, chalcone synthase, anthocyanidin synthase, and chalcone isomerase (Additional file 5 Figure S3). A gene encoding an ovate family protein showed the highest fold change value of 84.8. Expression of 27 transcription factor genes was down-regulated in FW petals, including 4 DREB and EAR motif protein (ERF) genes, 5 homeobox protein 40 (HD40) genes, 5 integrase-type DNA-binding superfamily protein genes, and 3 B-box type zinc finger protein with CCT domain (CONSTANS-like, COL) genes (Table 1, lower part). The COL gene sequences were most closely related to COL6 or COL16 in Arabidopsis and were designated COL16-like.
We identified 14 transcription factor genes expressed significantly higher in FGD petals than in FG petals ( Table 2, upper part). Among them, two genes encoding MYB3 proteins showed the highest fold change values (100.5 and 91.4). Three genes encoding MYB4 also showed high fold change values (19.3-42.8). Ten transcription factor genes were expressed at significantly lower levels in FGD petals than in FG petals ( Table 2, lower part). Six of them encoded mini zinc finger 2 proteins and showed extremely low relative expression in FGD petals (fold change values from −68.9 to −179.5).
To validate petal color-specific expression, we performed RT-qPCR analysis of 11 transcription factor genes in S3 petals of white-and green-flowered chrysanthemum cultivars. These genes were selected as representatives of different groups of transcription factors identified by the  (Tables 1 and 2). Among them, the expression of MYB113-like was restricted to white petals (Fig. 7a), although there was a variability in the expression levels among cultivars. In contrast, the expression of ERF and COL16-like tended to be higher or was significantly higher in green petals than in white petals (Fig. 7a). There was no correlation between the transcript levels of the other 8 genes and Chl content (Additional file 6 Figure S4).
We also analyzed the expression of MYB113-like, COL16-like, and ERF during development of FW, FG, and FGD petals, and in FG leaves. The expression of a MYB113-like gene in FW increased as petals matured, whereas it remained low in FG and FGD petals, and was also low in leaves (Fig. 7b). The expression of a COL16-like gene was extremely low in FW petals and high in FG leaves. FGD petals showed  * Genes whose expression levels in the white and green petals were analyzed by RT-qPCR (Figs.6, 7, and S3) higher COL16-like gene expression than did FW and FG petals, and the expression increased as FG and FGD petals matured. The expression of an ERF gene was highest in FG and FGD petals and increased as petals matured. We also examined the expression profiles of transcription factors previously identified as negative or positive regulators of Chl biosynthesis. GLK, HY5, and GNC enhance Chl biosynthesis [7][8][9]. ESTs encoding these transcription factors showed different expression patterns: GNC and HY5 expression in petals was positively and negatively correlated with Chl content, respectively. There was no marked difference in the expression level of GLK in petals of FW, FG and FGD. PIF1 and PIF3, have been shown to negatively regulate Chl biosynthesis [10]. However, in our microarray analysis, PIF1 and PIF3 expression levels were positively associated with Chl content in FW, FG, and FGD petals (Additional file 7 Figure S5). The results suggest that these transcription factors are not involved in the regulation of Chl accumulation in petals and that different transcription factors control Chl accumulation in leaves and petals.

Comparison of plastid ultrastructure in FW, FG, and FGD petals
We compared plastid ultrastructure among FW, FG, and FGD petals at different developmental stages and FG leaves by using TEM. At S1, epidermal cells of all petals contained electron-dense small plastids (Fig. 8a). Mesophyll cells of all petals also contained small plastids with lower electron density (Fig. 8b). At S2, the development of membrane structures was evident in the plastids of mesophyll cells of FG and FGD petals, whereas those of FW cells contained plastids with destroyed thylakoid membranes (Fig. 8b). At S3, we observed plastid debris in both epidermal and mesophyll cells of FW petals ( Fig. 8a and b; Additional file 8 Figure S6). In contrast, S3 FG and FGD petals still contained plastids with a welldeveloped thylakoid system. In leaves, plastids were larger than those in FG and FGD petals at S3 and contained large starch granules (Fig. 8c). Plastids were found only in mesophyll but not in epidermal cells (Additional file 8 Figure S6).

Discussion
To clarify the key factors that determine different levels of Chl accumulation in leaves and petals of chrysanthemums, we performed microarray analysis and searched for the differences in the expression levels of genes. GO classification showed that, among the downregulated genes in FW petals, genes associated with "plastids", "chloroplasts", and "electron transport or energy pathways" were highly represented. These results indicate that many of the genes related to photosynthesis whose products function in the chloroplast were transcriptionally inactive in FW petals. TEM observation of petals supports the hypothesis: at the late stage of development, green petals contain chloroplasts with well-developed thylakoid membranes, whereas white petals have lost chloroplasts. TEM observation also showed that, in green petals, both mesophyll and epidermal cells contain plastids, whereas only mesophyll cells contain plastids in leaves. The result indicates that leaves and petals have different patterns of Chl accumulation: Chls are present only in mesophyll cells in leaves, whereas they are present in both epidermal and mesophyll cells in green petals. Therefore, it is reasonable to assume that the regulatory mechanisms for Chl accumulation differ in leaves and petals. Microarray analysis showed that the expression of many Chl biosynthesis genes was lower in white petals than in green petals. In particular, the expression of genes encoding CHLH, CRD, HEMA1, and PORC was well associated with Chl content in petals and leaves of the FW, FG, and FGD cultivars. It is worth noting that the expression of the genes encoding these enzymes is induced by light and they are considered as key enzymes for Chl biosynthesis during photomorphogenesis [20][21][22]. We propose that Chl biosynthesis activity is lower in white petals than in leaves and green petals because of the remarkably low levels of expression of these genes.
There was no association between the expression levels of most Chl catabolic genes and Chl content in the petals of FW, FG, and FGD. However, SGR b a Fig. 7 RT-qPCR analysis of transcription factor genes differentially expressed in petals of white-and green-flowered chrysanthemum cultivars. Mean values (± SD) of three biological replicates are shown. Different letters indicate significant differences by Tukey-Kramer multiple-comparison test (P < 0.05). Designations of petal developmental stages and cultivars are as in Figs. 2 and 6 expression in mature petals of these cultivars was markedly higher than that in leaves. SGR is responsible for stay-green phenotypes of leaves in rice and Arabidopsis [23,24] and also non-photosynthetic tissues such as fruits of tomato, pepper, and kiwi, in which Chl degradation normally occurs at the onset of fruit ripening [25,26]. Several lines of evidence indicate that SGR plays a key role in the initiation of Chl degradation by destabilizing protein-pigment complexes in the thylakoid membranes [27,28]. Very recently, Shimoda et al. [29] revealed that SGR is a Mg-dechelatase, which catalyzes the conversion of Chl a to pheophytin a, which is the first step of Chl degradation. We then assume that SGR maintains high Chl catabolic activity and contributes to the absence or low level of Chls in chrysanthemum petals. Our results on the expression patterns of Chl metabolic genes in chrysanthemums are similar to our previous results on carnation and Arabidopsis [6].
Based on our results, we hypothesize that a low rate of Chl biosynthesis and a high rate of Chl degradation lead to the absence of Chls in white chrysanthemum petals. We assumed that factors that suppress