Transcriptome Analysis of Key Genes Involved in Color Variation between Blue and White Flowers of Iris bulleyana

Iris bulleyana Dykes (Southwest iris) is an extensively distributed Iridaceae species with blue or white flowers. Hereby, we performed a systematic study, employing metabolomics and transcriptomics to uncover the subtle color differentiation from blue to white in Southwest iris. Fresh flower buds from both cultivars were subjected to flavonoid/anthocyanin and carotenoid-targeted metabolomics along with transcriptomic sequencing. Among 297 flavonoids, 24 anthocyanins were identified, and 13 showed a strong down-accumulation pattern in the white flowers compared to the blue flowers. Significant downregulation of 3GT and 5GT genes involved in the glycosylation of anthocyanins was predicted to hinder the accumulation of anthocyanins, resulting in white coloration. Besides, no significant altered accumulation of carotenoids and expression of their biosynthetic genes was observed between the two cultivars. Our study systematically addressed the color differentiation in I. bulleyana flowers, which can aid future breeding programs.

sogdiana [9]. The subtle color variants of the Iris genus, ranging from dark purple, through blue, pink, and violet, to yellow and white flowers, have been the focus of scientists for many decades [10,11].
Iris plants are generally dominated by two types of pigments: flavonoids/anthocyanins and carotenoids. Bluepurple colors are mainly attributed to anthocyanin pigments, while orange, yellow, and pink colors are attributed to carotenoid synthesis. Various studies have identified multiple genes involved in the flavonoids/anthocyanins biological pathways for which alteration in gene expression induces color mutation. These genes include CHS (chalcone synthase) in parsley [39], petunia [40], tobacco [41], and safflower [42]; CHI (chalcone isomerase) in petunia [43], tobacco [44], and carnation [45]; F3H (flavanone-3-hydroxylase) in carnation [46], cineraria [47], saussurea [48], and peony [49]; DFR-dihydroflavonol 4-reductase in lily [50], gentian [51], peony [49], and saussurea [52]; ANS (anthocyanidin synthase) in gerbera [53] and peony [49]; glycosyltransferase (GT) in Veronica persica [54] and Bellis perennis [55]. Besides, some known transcription factors have also been reported to play a regulatory role in pigmentation, i.e., MYB, bHLH, and WD40 [56][57][58]. However, Iris bulleyana Dykes has not been characterized for its color formation. Due to its wide distribution in southwestern China and as a model species for studying the color formation, insight into the mechanisms underlying pigmentation will facilitate understanding the color formation and further breeding of colorful cultivars. Hereby, we have profiled the transcriptome and metabolome of Southwest iris (I. bulleyana Dykes) and its white variant (I. bulleyana Dykes f. alba YT Zhao) to pinpoint the genetic mechanism underlying flower color variation. Our study discussed the differential expression of key genes in carotenoids and anthocyanin biosynthesis pathways for their potential involvement in color formation in iris.

Results
Southwest iris (I. bulleyana Dykes) generally has blue petals; however, another variant with white petals (I. bulleyana Dykes f. alba YT Zhao) is also present (Figure 1). To understand the genetic variation underlying this variation, we performed transcriptomic and targeted metabolomics following sample collection from Southwest iris and its white variant.
2.1. The Differential Landscape of Metabolites between Blueand White-Colored Southwest Iris. Randomly selected fresh flower buds of Southwest iris and its white variant were collected and subjected to targeted metabolomics, revealing the differential landscape of metabolites, specifically anthocyanin and carotenoids.
Furthermore, carotenoid metabolites were also investigated in both groups. Eleven carotenoids were identified using targeted metabolomics (Additional Files 3, 4, and 5). There was no significant differential accumulation of carotenoids in blue and white flowers. However, α-carotene depicted higher accumulation in blue flowers compared to white, while zeaxanthin and xanthophyll (lutein) both up accumulated in white flowers. The changes in accumulation patterns of these three carotenoids were statistically nonsignificant, suggesting a neglected role of carotenoids in color differentiation from blue to white flowers. The accumulation pattern of carotenoids in purple and white flowers explained the conserved yellow stripes on both flowers.

Differential
Landscape of Expressed Genes between Blueand White-Colored Southwest Iris. In order to analyze the   Figure 2(a)). Principal component analysis (PCA) differentiated both color variants into two groups, and biological replicates were closely grouped ( Figure 2(b)). PCA results suggested high reliability of transcriptome data for further analysis.
Based on previously published reports suggesting the involvement of MYB and bHLH transcription factors as a key regulators in plant pigmentation [56][57][58], we identified 158 MYBs and 122 bHLHs. However, their expression was conserved between the Southwest iris and its white variant.

Proposed Mechanisms of Blue/White Color Formation in
Southwest Iris. In the two Southwest iris variants, we identified 13 anthocyanins differentially accumulated. The initial anthocyanins are very unstable and can easily degrade [59,60]; therefore, they need to be glycosylated and transferred into vacuoles for pigmentation. The 3GT and 5GT genes play this function [61], and because they were significantly downregulated in the white flower of Southwest iris, anthocyanin glucosides could hardly be produced, resulting in no blue coloration. In contrast, the high activity of 3GT and 5GT in the blue Southwest iris favored the formation and accumulation of anthocyanin glycosides, contributing to the blue color of the flowers (Figure 3). We did not observe any change in the carotenoid pathway, which explains the conserved yellow stripes in the flowers of both genotypes (Figure 3).
To further confirm the expression of identified genes in the development of flower color, we performed qRT-PCR for three groups of selected genes related to flower color regulation, viz., flavonoid biosynthesis, anthocyanin biosynthesis, and carotenoid biosynthesis. The qRT-PCR results have been presented in Figure 4. Interestingly, the genes 3GT and 5GT showed significantly lower expression patterns in white flowers compared to blue flowers (Figure 4(b)), which further confirms our hypothesis that downregulation of 3GT and 5GT genes resulted in white coloration. Besides, genes related to carotenoid synthesis did not show significant differential expression in both flowers (Figure 4(a)), supporting our transcriptome results.

Discussion
Flower colors, with their eye appeal and aesthetic value, have been the focus of many biological studies [12][13][14], and genetic pathways for color development have been well characterized. Carotenoids, flavonoids, and betalains are primary 7 BioMed Research International metabolites characterized for their role in pigmentation in flower and fruit color. However, certain species-specific variations due to mutation, activities of regulatory genes, and multigene influence have also been reported [12][13][14]35]. Therefore, this study was systematically designed utilizing metabolomics and transcriptomics to uncover flower color differentiation between Southwest iris (I. bulleyana Dykes) with blue flowers and its white variant (I. bulleyana Dykes f. alba YT Zhao).
Anthocyanins, a branch of flavonoids, have many biological functions in higher plants. Previously published literature suggested the essential role of anthocyanins in plant pigmentation. For instance, the red seed coat in peanuts has a strong association with anthocyanins [62]. A study by Qiu et al. demonstrated a significant increase in total anthocyanins in purple passion fruit compared to yellow [63]. White, yellow, blue, and pink Primula vulgaris [64] showed a gradual increase in total anthocyanin content as the color deepened. Moreover, anthocyanins play a critical role in plant defense responses against biotic and abiotic stress conditions [65,66]. In iris, the presence/absence of anthocyanins is a critical factor for color development [19]. Flavonoid-targeted metabolomics identified 13 anthocyanins showing significant down-accumulation in white flowers compared to the blue flowers, which are predicted to favor the blue coloration. Cyanidin 3-O-glucosyl-malonylglucoside [67,68], delphinidin O-malonyl-malonylhexoside [69], delphinidin 3-O-glucoside (mirtillin) [70][71][72], and delphinidin 3-O-rutinoside (tulipanin) have been previously reported for their active role in blue color pigmentation in perianths. Differential accumulation of anthocyanins pertaining to different flower colors and their corresponding shades has been reported in different iris species [73][74][75][76]. Further, anthocyanins, as biological/chemotaxonomic markers, have been used for the taxonomic classification ofspecies and cultivars [77,78].
Dp3pCRG5G (delphinidin-3-pcoumaroylrutinoside-5glucoside) is the most common anthocyanin in iris species and is generally responsible for blue-colored perianths is different iris species such as Dutch iris, Siberian iris, and I. germanica [19]. However, the precursor of DP3pCRG5G, delphinidin is very unstable [60], which requires further glycosylation for stabilization to the end product Dp3pCRG5G. Our transcriptome results suggest a downregulation of 3GTanthocyanidin 3-O-glucosyltransferase in the white flower [79]. The downregulation of the 3GT gene is predicted to inhibit the synthesis of delphinidin 3-glucoside [80,81]. Furthermore, a downregulation of another gene 5GT (anthocyanidin 5-O-glucosyltransferase) was also observed in the white flower, which may result in reduced levels of delphinidin 3-rutinoside [82]. Florio et al., characterized acyltransferase, complemented by 5GT, for differential accumulation of delphinidin-3-rutinoside and nasunin [82]. Contrary to our results, a study concerning gentian identified delphinidin 3,5,3 ′ -O-triglucosideas a stable blue pigmentregulated by the coexpression of 3GT and 5GT [83]. Another study concerning rose petal coloration identified 5,3GT as a contributor to petal coloration by catalyzing glycosylation at two different positions on anthocyanidin [84]. However, we observed a conserved expression of 5,3GT in blue and white flowers. Interestingly, targeted metabolomics suggested a significantly higher accumulation of cyanidin 3-Ogalactoside in blue flowers compared to white; however, we did not identify UDP-galactose: anthocyanidin 3-Ogalactosyltransferase from the transcriptome data. UDPgalactose has been reported previously to influence the accumulation patterns of cyanidin 3-O-galactoside [85]. The reason for the differential accumulation of cyanidin 3-O-galactoside in the blue and white iris is unclear and requires further study to understand the accumulation pattern. Further insights into substrate recognition, utility, and structure-activity of 3GT and 5GT could provide significant results for pigmentation in the iris.
Moreover, we identified yellow stripes on both flowers, which were explained by similar accumulation patterns of carotenoids in purple and white flowers. Carotenoid biosynthesis has been well-documented in many plant species [34,86,87]. Yellow, orange, and red colors in plants are mainly attributed to carotenoid accumulation patterns [88]. A study concerning Iris germanica L. demonstrated the role of the phytoene synthase gene (crtB) in managing yellow color by increasing metabolite flux into carotenoid biosynthesis pathways [2]. However, in this study, there were no significant differences in accumulation patterns of carotenoids in purple and white flowers, explaining the conserved yellow stripes on both flowers. Moreover, the gene identified in carotenoid biosynthesis pathways depicted nonsignificant differences in purple and white flowers.
In contrast to our results, a recent report by Wang et al. [89] suggested a shunted anthocyanin pathway due to the absence of naringenin, a key compound in the pathways, as a major constraint in color differentiation from blue to white in Iris laevigata Fisch. However, in our study, naringenin chalcone was detected with a similar expression pattern of the corresponding CHI gene in both blue and white flowers, which highlights that various mechanisms are involved in the color variation in different Iris species.
Altogether, the down-accumulation of various anthocyanins, probably due to the strong downregulation of 3GT and 5GT, plays a major role in color differentiation between blue and white flowers in the Southwest iris. Further functional verification of these genes can provide a valid reference for the differential pigmentation pattern in the Southwest iris.

Differential Expression
Analysis of Identified Genes. The read numbers mapped to each gene were counted using fea-tureCounts v1.5.0-p3 [55]. Then, calculating the expected number of FPKM (fragments per kilobase of exon model per million reads mapped) of each gene based on the length of each gene and reads count mapped to the gene. DEGs between blue and white groups of colored samples were identified using the DESeq R package (v1.18.0) [91] and edgeR package (v 3.24.3). The threshold p value in multiple tests to judge the significance of gene expression difference was based on the false discovery rate (FDR) method. When FDR ≤ 0:05 and FPKM values showed at least a 2-fold difference among samples, the gene was considered a significant DEG. DEGs commonly detected by both packages were used in this study.

Validation of Gene Expression
Using qRT-PCR. To verify the RNA-seq data, qRT-PCR was used following total RNA extraction from flower bud samples in three replicates, using the Tiangen RNAprep Pure Plant kit (Tiangen Biotech, Beijing, China), following the manufacturer's protocol. Twenty genes related to flavonoid/anthocyanin and carotenoid pathways of the transcriptome data were selected, and corresponding primers were designed for qRT-PCR using the Oligo-7 software (Additional File 8). The primers were synthesized by Sangon Biotech (Shanghai, China). Actin was used as an internal reference gene for qRT-PCR analysis of the target genes [92]. The cDNA was extracted from RNA and used as a template to make the reaction for qRT-PCR by using Takara qPCR kit SYBR Premix Ex TaqTM II (Tli RNa-seH Plus). Three biological repeats were used for each qRT-PCR reaction.

KEGG Enrichment Analysis of DEGs.
To test the statistical enrichment of Kyoto Encyclopedia of Genes and Genomes (KEGG) pathways, the GOseq R package was used. The KEGG pathways enriched with DEGs (FDR < 0:05) were detected using KOBAS 2.0 software [62] based on the method of overrepresentation analysis (ORA). The adjusted p value of significantly corroborated KEGG terms was less than 0.05.

Data Availability
RNA-seq data is available at the SRA database in National Center of Biotechnology Information with the accession number PRJNA676187 (https://www.ncbi.nlm.nih.gov/bioproject/ ?term=PRJNA676187).

Disclosure
The funder has no role in study design, data collection and analysis, decision to publish, or preparation of the manuscript.

Conflicts of Interest
The authors declare that they have no competing interests.

Authors' Contributions
Conceptualization was prepared by L M, Y Z, F C, XL, and J W; methodology was prepared by X L, L M, and Y Z; software analysis was carried out by X L, L M, and Y Z; validation was prepared by L M, Y Z, W J, Q D, W D, F X, G C, and X W; formal analysis was prepared by L M, Y Z, W J, Q D, W D, F X, G C, and X W; investigation was prepared by L M, Y Z, W J, Q D, W D, F X, G C, and X W; resources were prepared by L M and Y Z; data curation was prepared by L M, Y Z, W J, Q D, W D, F X, G C, and X W; writing of the original draft preparation was carried out by L M and Y Z; writing in review and editing was carried out by F C and J W; visualization was prepared by L M; supervision was prepared by F C and J W; project administration was carried out by F C and J W; and funding acquisition was carried out by F C and J W. All authors have read and approved the final version of the manuscript. The cofirst authors are Lulin Ma and Yiping Zhang.