To elucidate the composition of pigments in the floral tepals (petals and sepals), the contents of anthocyanins, carotenoids, and chlorophylls were measured in different Cymbidium species with various flower colors, including C. haematodes, C. kanran, C. tortisepalum, and C. sinense (Table 1 and Figure 1).

The results indicated that C. kanran-G exhibited the highest chlorophyll content (313.40 μg/g·FW), while the white-green (WG) of C. haematodes showed the lowest chlorophyll content of 30.29 μg/g·FW (Table 1). The highest carotenoid content was 67.52μg/g observed in C. sinense, while C. kanran-DR and C. kanran-G exhibited comparable levels. On the other hand, the lowest carotenoid content was found in C. haematodes-WG at7.70 μg/g·FW. Anthocyanins are critical coloration compounds within flavonoids that play important roles in the color formation as well as various biological processes. Notably high levels of anthocyanins were detected in C. kanran-DR, C. sinense, and C. haematodes-R, with the highest anthocyanin content of 207.47 µg/g·FW found in the tepals of C. kanran-DR. In contrast, C. tortisepalum-W exhibited the lowest anthocyanin content (2.63 µg/g·FW).

The results showed that the anthocyanin contents are significantly higher in red tepals of C. haematodes (DR, R, and LR) than those of YG and WG. The chlorophyll and carotenoid contents in C. haematodes are relatively low, with no significant difference among different colors. This indicated that chlorophyll and carotenoid have minimal influence on the flower color variations within C. haematodes. The anthocyanin contents in C. haematodes exhibit considerable variation across different colors, being highest in DR, followed by R and LR, while YG and WG display the lowest levels. The petals of C. sinense ‘Qihei’ is claret colored, which is darker than that of DR. However, it was found that anthocyanin content of DR reached 182.94 μg·g^-1·FW, whereas that in ‘Qihei’ was 161.84 μg·g^-1·FW. Thus, there might be other pigments except for anthocyanins, carotenoids and chlorophylls that determine the flower color differences between C. sinense and C. haematodes. These findings suggest that the anthocyanin concentrations are the main factors that contributed to different flower colors of C. haematodes. The wide spectrum of flower color ranges in C. haematodes makes it a good resource for studying flower color formation within Cymbidium species. Thereby, we performed further analysis to explore the mechanisms regulating anthocyanin biosynthesis in C. haematodes flowers.

Because no reference genome has been published for C. haematodes, and given its close relationship to C. sinense, we conducted transcriptome sequencing of the C. haematodes tepals using C. sinense genome as the reference genome. We also performed sequence comparisons between C. haematodes and C. sinense. Transcriptome sequencing was carried out on 15 samples representing five different color accessions of C. haematodes (DR, R, LR, YG, and WG). The comparison rates for 15 samples against the reference genome of C. sinense all exceeded 85% (Table 2), indicating that the genome met the requirements for gene analysis related to anthocyanin biosynthesis in C. haematodes.

According to the annotation information from the NR database of C. haematodes, a total of 20 structural genes related to anthocyanin biosynthesis pathway were identified (Figure 2). This includes three CsCHS genes, one CsCHI gene, four CsF3H genes, two CsFLS genes, three CsF3’H genes, two CsDFR genes, one CsANS gene and four CsUFGT genes (Table 3). These genes encode flavonoid biosynthesis enzymes belonging to different functional classes. For example, the CsCHS proteins belong to polyketide synthases, while CsDFR proteins are short-chain dehydrogenases/reductases. Additionally, CsF3H, CsANS and CsFLS proteins belong to 2-ODDs. The results of physical and chemical property analysis showed that nearly half of the proteins encoded by structural genes were unstable proteins with instability coefficient greater than 40. Moreover, most proteins were hydrophilic proteins with average hydrophilic index less than 0. The longest proteins were the three F3’Hs, with 515–531 aa in length. Subcellular localization of the proteins encoded by structural genes was predicted on cytoplasm, nucleus and chloroplast, among which over 65% proteins located on cytoplasm. Only two F3H proteins were predicted on nucleus, whereas all the F3’H proteins, one DFR, and one UFGT were located on chloroplast.

To investigate the evolution relationship of the structural genes in anthocyanin biosynthesis pathway, phylogenetic analysis was performed for amino acid sequences of the seven types of structural genes encoded proteins, including CHS, CHI, F3H, F3’H, DFR, ANS, and UFGT. A total of 66 protein sequences from 13 plant species were used to construct seven phylogenetic trees (one tree for each type of structural genes). In general, at least one structural gene from C. haematodes is most closely clustered with C. hybrid which is also from genus Cymbidium (Figure 3). For example, the CHS protein sequences from 10 species were used to construct a phylogenetic tree. The result shows that the CsCHS1 is closely related to CyCHS from C. hybrid, indicating high homology within the same genus. Both of them are clustered together with Oncidium (Figure 3A). However, the other two CsCHS proteins are distance from most Orchidaceae species. Similar patterns are also observed in the phylogenetic trees of CHI, F3H, F3H and DFR (Figures 3C–E). Nonetheless, CsANS was clustered together with Oncidium OhANS (Figure 3F), and CsUFGT4 showed high homology with Dendrobium (Figure 3G).

Moreover, a phylogenetic tree was also constructed to analyze the evolutionary relationships between C. haematodes 2-ODDs and 142 flavonoid biosynthesis-related 2-ODDs including monocotyledonous and dicotyledonous species (Figure 4). The proteins are mainly grouped into two major clades, corresponding to the DOXC28 and DOXC47 subgroups of the 2-ODD superfamily. In the F3H-containing DOXC28 group, the CsF3Hs are clustered with F3H proteins from Orchidaceae species Cymbidium, and those from other monocots. Similar patterns are also observed for the CsANSs and CsFLSs. Among the orchids, the C. haematodes 2-ODDs are most closely related to the 2-ODDs from Cymbidium species, indicating high homology and possible similar functional roles.

Multiple sequence comparison results showed that all six 2-ODD proteins contained H-x-D-xn-H and R-x-S conserved residues, which are the binding sites for Fe²⁺ ions and 2-oxoglutarate, respectively (Figure 5A). These two conserved residues are well known for the highly conserved sequences among most plant 2-ODD proteins. Conserved motif analysis of the 2-ODD proteins of three Cymbidium species, C. haematodes, C. goeringii, and C. ensifolium showed that most 2-ODD protein sequences had four motifs: motif1, motif2, motif6, and motif7, suggesting that these motifs might be the most conserved amino acid residues of the Cymbidium 2-ODDs (Figure 5B). All terminals of CsFLS, CsANS and CsF3H1 contain motif10, while the end motif of CsF3H3 is motif8, which might be a specific motif. Meanwhile, two F3H proteins (Mol019913/CsF3H2, Mol019912/CsF3H4) contain a specific motif (motif4). Among them, Mol019912 that located in the same evolutionary branch may have lost some sequences due to evolution or assemble mistakes. The above results indicate that there are diversity and differences in the sequences of the 2-ODD family protein among different species, which closely related to the functional diversity of those genes.

Analysis of the exon/intron structural patterns in homologous genes of three Cymbidium species (Figure 5C) revealed that the number of exons ranged from 2 to 6, suggesting exon loss or gain during gene evolution. The most common exon numbers were two or three, with 14 genes (53.8%) exhibiting 3 exons and 2 introns. Notably, JL001462 displayed a 5-exon and 4-intron structure, while CsF3H2 and JL003123 contained the highest number of exons and introns, with 6 exons and 5 introns. Furthermore, each gene possessed varying lengths of coding sequences (CDS). In contrast, untranslated regions (UTRs) were either short or entirely absent, with lower sequence similarity compared to CDS. This indicates that coding regions exhibit higher evolutionary conservation than UTRs.

To understand the potential roles of the structural genes involved in regulating anthocyanin biosynthesis, expression profiles of the structural genes in anthocyanin biosynthesis pathway were analyzed by using transcriptome sequencing data. Transcriptome sequencing was performed by using the tepals of three C. haematodes accessions with different colors (R, LR, and WG) and C. sinense ‘Qihei’. The results revealed distinct expression patterns of the structural genes in different colors of C. haematodes and C. sinense tepals (Figure 6). Four genes showed similar expression patterns. For example, CsF3’H3 and CsF3’H2 exhibit exclusive high expression levels in R tepals, whereas minimal expressions in other samples. CsUFGT4 and CsCHI also had relatively high expression levels in R, compared to LR, WG and C. sinense. Interestingly, the rest three CsUFGTs all highly expressed in the tepals of C. sinense, indicating functional divergence in the same gene subgroup. Except for the three CsUFGTs, CsDFR1 and CsDFR2 also had high expression levels in C. sinense, indicating there might be specific roles of these LBGs in anthocyanin biosynthesis in C. sinense. Moreover, CsDFR2 and CsUFGT2 also showed moderate to low expression in LR. Among three CsCHS genes, high expression levels of CsCHS2 and CsCHS3 in LR were detected, but CsCHS1 was slightly expressed in WG. These findings suggest that the structural genes have experienced functional divergence in regulating anthocyanin biosynthesis within the same subgroups and among different colors of C. haematodes and C. sinense. Despite that C. haematodes and C. sinense are closely related species, the molecular mechanisms in regulating flower color are obviously different. CsF3’H3 and CsF3’H2 are suggested as the potential key genes for anthocyanin biosynthesis of C. haematodes. However, further gene function verifications remain to be conducted.

To further verify the expression of structural genes in C. sinense and C. haematodes with different flower colors (DR, R, LR, YG, WG and W), qRT-PCR analysis was performed (Figure 7). The results showed that there were significant differences in the expression levels of each structural gene in tepals with different colors. Compared with C. sinense and the light colors of C. haematodes, CsF3’H2 genes showed significantly higher expression levels in both DR and R. Nonetheless, CsF3’H2 was lower expressed than CsF3H1 in C. sinense, indicating a more specific role in the flower color regulation in C. haematodes. The expression levels of CsF3H genes in C. sinense and R was higher than that of DR. Among them, the expression level of CsF3H1 was the highest in R, while the expression level of CsF3H2 was significantly higher in white WG. These results were largely consistent with transcriptome sequencing data. Nonetheless, the expression of CsANS (Mol002256) in C. sinense and DR were quite similar, both were significantly higher than that in other colors of C. haematodes. Thereby, CsANS may also play an important role in the determination of dark red flower colors in C. sinense and C. haematodes.

Three different structure gene-encoded proteins (CsANS, CsF3’H2, and CsF3’H3) were further selected for subcellular localization analysis. All the three cloned proteins were transient expressed in the epidermal cells of Nicotiana benthamiana leaves by using a 35S::GFP vector. The results revealed that CsF3’H3 is a cell membrane protein (Figure 8A), suggesting its potentially function involved in cell signaling process or other membrane-associated processes. On the other hand, CsF3’H2 is localized to the chloroplast (Figure 8B), which may have participated in plastid-related processes. Meanwhile, the diffuse CsANS-GFP fluorescence distributed throughout the cytoplasm, without overlapping with organelle-specific markers (Figure 8C). This suggests that ANS functions as a soluble cytoplasmic protein or may transiently interact with other cellular components under specific conditions. These findings provide crucial insights for further functional characterization of the structural genes in the tepals of C. haematodes.

In this study, the floral tepals of different Cymbidium species with various colors including C. haematodes, C. kanran, C. tortisepalum, and C. sinense were used as the experimental materials listed in Table 4. These Cymbidium species include five native C. haematodes accessions, two C. kanran accessions and two C. tortisepalum accessions with different flower colors and one C. sinense variety ‘Qihei’ (blackish red). The C. sinense variety ‘Qihei’ is native to Fujian province (China), whereas the C. haematodes accessions that collected from Yunnan province (China) have not been officially named yet, so they are numbered and represented by different flower colors: dark red (DR), red (R), light red (LR), yellow-green (YG), white-green (WG), and white (W). Likewise, the C. kanran and C. tortisepalum accessions were also presented by flower colors as C. kanran DR (dark red), C. kanran G (green), C. tortisepalum W (white) and C. tortisepalum R (red). The plant materials used in this study were all cultivated in green houses at the Orchid Germplasm Resources Paddy of Zhejiang Province (Hangzhou, China). Healthy and disease-free individuals from different flower color accessions of C. haematodes were selected. Due to the different flowering periods of Cymbidium species, the tepals were all sampled 10 days after flowering. Samples are taken from the perianth lobes on the same blooming condition, including sepals and petals. Three biological replicates were taken for each flower color. The samples of different flower color series of Cymbidium species were collected. The samples were quickly placed in liquid nitrogen for drying and freezing, and then stored at -80°C refrigerator for further analysis.

Each of 0.2 g samples of the tepals from different C. haematodes accessions (DR, R, LR, YG, and WG), C. kanran accessions (DR and G), C. tortisepalum accessions (R and W), and C. sinense “Qihei” were weighed, respectively. The samples were ground into powder quickly, with liquid nitrogen. Then, the anthocyanin contents in each sample were determined by colorimetry following the method described by Chen et al (Chen et al., 2016), and the total contents of chlorophyll and carotenoids were determined by spectrophotometry using modified protocols according to Wang (2006). Finally, the contents of main pigments in Cymbidium orchids were measured. The data were further analyzed using Microsoft Excel 2019 (Microsoft Corp., Redmond, WA) and GraphPad Prism software (vs. 8.4.3).

Based on the quantitative analysis of anthocyanin contents in C. haematodes, the structural genes in the anthocyanin biosynthesis pathway were obtained by searching for the gene annotation in the NR database of C. haematodes transcriptome data. The corresponding nucleotide sequences were obtained from C. sinense genome annotation file using TBtools software (Wang, 2006), and the conserved domains in amino acid sequences of these genes were predicted using NCBI CDD (https://www.ncbi.nlm.nih.gov/Structure/cdd/wrpsb.cgi). The basic physicochemical properties of the identified protein sequences, such as molecular weight (MW) and isoelectric point (pI), were analyzed using ExPASy proteomics website (https://web.expasy.org/). Subcellular localization was predicted using online website Cell-PLoc2.0 (http://www.csbio.sjtu.edu.cn/bioinf/Cell-PLoc-2/).

The amino acid sequences of reported structural genes from other plants including monocots and dicots were obtained from TAIR (http://www.arabidopsis.org/) and NCBI (https://www.ncbi.nlm.nih.gov/). CHS proteins from other species include Arabidopsis thaliana (AtCHS), Petunia hybrida, (PhCHS), Vitis vinifera (VvCHS), Antirrhinum majus (AmCHS), Oncidium var. Gower Ramsey (OgCHS), Paphiopedilum concolor (PcCHS), Dendrobium catenatum (DcCHS), Cymbidium hybrid (CyCHS), and Phalaenopsis equestris (PeCHS). CHI proteins from other species include AtCHI (A. thaliana), PhCHI (P. hybrida), DcCHI (D. catenatum), PeCHI (P. equestris), AsCHI (Apostasia shenzhenica), CyCHI (C. hybrid), and DhCHI (D. hybrid). F3H proteins from other species include AtF3H (A. thaliana), PhF3H (P. hybrida), VvF3H (V. vinifera), CyF3H (C. hybrid), DhF3H (D. hybrid), DoF3H (D. officinale), and AsF3H (A. shenzhenica). F3’H proteins from other species include AtF3’H (A. thaliana), PhF3’H (P. hybrida), VvF3’H (V. vinifera), LsF3’H (Lilium speciosum), CyF3’H (C. hybrid), DhF3H (D. hybrid), DcF3’H (D. catenatum), and AsF3’H (A. shenzhenica). DFR proteins from other species include AtDFR (A. thaliana), PhDFR (P. hybrida), AmDFR (A. majus), OhDFR (O.hybrid), CyDFR (C. hybrid), DcDFRA (D. catenatum), AsDFR (A. shenzhenica), and PeDFR (P. equestris). ANS proteins from other species include AtANS (A. thaliana), PhANS (P. hybrida), LsANS (L. speciosum), PeANS (P. equestris), CyANS (C. hybrid), DoANS (D. officinale), OhANS (O.hybrid), and DcANS (D. catenatum). UFGT proteins from other species include AtUG3GT (A. thaliana), NnUFGT (Nelumbo nucifera), VvUFGT (V. vinifera), and DcUFGT (D. catenatum).

The phylogenetic tree for plant 2-ODD gene family members was constructed with protein sequences from 43 species including four Cymbidium species (C. haematodes, C. sinense, C. ensifolium, and C. goeringri), and D. officinale (Do), P. equestris (Pe), A. shenzhenica (As), Sorghum bicolor (Sb), Setaria italica (Si), Ginkgo biloba (Gb), N. nucifera (Nn), Allium cepa (Ac), A. thaliana (At), Oryza sativa (Os), P. x hybrida (Ph), A. majus (Am), Zea mays (Zm), Rosa chinensis (Rc), V. vinifera (Vv), Fragaria x ananassa (Fa), Prunus cerasifera (Pc), Eustoma grandiflorum (Eg), Solanum lycopersicum (Sl), Camellia nitidissima (Cn), Gossypium hirsutum (Gh), Daucus carota (Dc), Vaccinium corymbosum (Vc), Paeonia lactiflora (Pl), Arachis hypogaea (Ah), Hibiscus sabdariffa (Hs), Salvia splendens (Ss), Lagerstroemia indica (Li), Ipomoea batatas (Ib), Strelitzia reginae (Sr), Brassica napus (Bna), Lilium regale (Lr), Glycine max (Gm), Punica granatum (Pg), Lycium chinense (Lc), Solanum tuberosum (St), Citrus sinensis (Cs), and Malus domestica (Md), by using a MEGAX software (V10.0, Tokyo Metropolitan University, Tokyo, Japan). MUSCLE program was used for multiple sequence alignment, and the Maximum Likelihood (ML) method was used with 1000 bootstrap replicates. The phylogenetic trees were visualized and modified using MEGAX and iTOL (https://itol.embl.de/index.shtml).

To obtain the conserved amino acid sequences of the DNA-binding domain, the ClustalX program was used to conduct multiple sequence alignment of the 2-ODD (ANS, F3H, and FLS) protein sequences. The sequences of the related structural genes of the homologous species C. goeringii and C. ensifolium were obtained using a TBtools software (Chen et al., 2023). The conserved motifs of the homologous sequences of genes in C. haematodes, C. goeringii and C. ensifolium were analyzed through MEME online website (http://meme-suite.org/tools/meme). The maximum value of the motifs was set to 10, and the rest were default parameters. The results were visualized by TBtools software (Chen et al., 2023).

Transcriptome sequencing was performed by using the tepals of C. sinense ‘Qihei’ and five different C. haematodes accessions (DR, R, LR, YG, and WG). The genome of C. sinense was used as the reference genome. Three biological replicates were performed for each sample. The total RNA of 18 samples was extracted according to Wang et al. (2022) and sequenced by Mateware Metabolic Biotechnology Co., LTD (Wuhan, China). Expression profiles of the 20 structural genes were obtained by FPKM (Fragments Per Kilobase of exon per million fragments Mapped) values and the Cufflinks software (http://cole-trapnell-lab.github.io/cufflinks, v2.2.1). The expression heatmap of structural genes was constructed using the TBtools software based on the FPKM values.

A total of 21 samples from plant materials with seven different flower colors, namely C. sinense (Mol), DR, R, LR, YG, WG, and W, were selected for qRT-PCR analysis of candidate structural genes in anthocyanin biosynthesis. There were three biological replicates for each flower color accession. Total RNA was extracted from each sample using the RNA extraction kit according to Wang et al (Wang et al., 2022), and the extracted RNA was subjected to electrophoresis for quality detection. The RNA with qualified quality was reverse transcribed into cDNA using a cDNA synthesis system (Invitrogen, Shanghai, China). Primer Premier 5 software was used to design qRT-PCR primers for the selected structural genes. The primers were synthesized and sequenced by Genepioneer Biotechnologies Co., Ltd (Nanjing, China) (Table 5). The 20 μL reaction system for qRT-PCR amplification was as follows: 1 μL template, 1 μL each of upstream and downstream primers, 10 μL of 2×SYBR Premix Ex Taq fluorescent dye, and 7 μL ddH₂O. The PCR amplification conditions were set by two-step method: 5 min of pre-denaturation at 95 °C, followed by 40 cycles of 10 s at 95 °C, 30 s at 60 °C, and 15 s at 95 °C, and then a melting curve program of 95 °C 15 s, 60 °C 60 s, and 95 °C 15 s. Mol017122 gene was chosen as the internal standard gene to analyze the relative fold differences. The reaction system and amplification program were the same as above. Each sample was repeated three times. The relative expression levels of the target genes were calculated using the 2^-ΔΔCt method, and the data were analyzed and plotted using the Graphpad Prism software (vs. 8.4.3).

Three key structural gene-encoded proteins (CsANS, CsF3’H2, and CsF3’H3) were selected for subcellular localization analysis. An artificial green fluorescent protein (GFP) was fused in-frame to the 3’-terminus of each target gene. The pCAMBIA1300-eGFP plasmid was digested with the BamHI restriction enzyme (TAKARA FastDigest™) in a single-enzyme reaction. Gene-specific homologous recombination primers were designed based on the target gene sequences using CE Design software. The target genes were subsequently amplified and cloned into the linearized pCAMBIA1300-eGFP vector via homologous recombination. The 35S promoter was employed to drive the expression of the three genes. The DsRed-RFP vector was used as nucleus marker. To validate subcellular localization, co-localization experiments were performed using either a nuclear/tonoplast/chloroplast specific marker or free GFP as a cytoplasmic reference control. The constructed vectors were transiently expressed in the epidermal cells of Nicotiana benthamiana through Agrobacterium-mediated transformation. Fluorescence of fusion protein constructs and the makers were detected under Olympus confocal laser microscope FV3000 (Hachioji City, Tokyo, Japan).

Statistics were performed with Graphpad Prism software (vs. 8.4.3) using one-way ANOVA (P < 0.05). Significant differences in values were indicated by * (P < 0.05), ** (P < 0.01), *** (P < 0.001), and **** (P < 0.0001). Expression analysis was performed using mean ± standard deviation (SD) values of three biological replicates.