Introduction

Loropetalum chinense var. rubrum belongs to the Hamamelidaceae (witch-hazel family) [1] and is mainly distributed in the belt south of the middle Yangtze River to the north of the Tropic of Cancer in China. It originated in the Hunan Province and played an important role in the landscape [2]. L. chinense var. rubrum is an evergreen plant with an elegant tree shape and brightly coloured foliage. Its leaves, flowers, and roots were used in traditional Chinese medicine for treating cough, burns, abdominal pain, etc. [3]. Recently, the plant has gained a lot of interest and has been widely cultivated for its ornamental and medicinal value.

As an ornamental plant, the colour of L. chinense var. rubrum leaves are one of its most significant characteristics. Leaf coloration is determined by the pigment in the mesophyll cells, such as chlorophyll, carotenoids, and flavonoids [4, 5]. Flavonoids are comprised of chalcones, flavone, flavonol, and anthocyanins, which colour plants blue, pink, yellow, purple, and red [6, 7]. Anthocyanins are considered to be the main coloration pigments, while flavone and flavonol are the synergistic pigments [8, 9]. Anthocyanins have been proved to determine the appearance of colour in many fruits, flowers, and vegetables, such as Vaccinium corymbosum [10], Morella rubra [11], Centaurea cyanus [12], Primula vulgaris [13], and Allium cepa L (Onion) [14]. In addition, they also played a vital role in the physiological activities of plants and human health. They have been shown to alleviate the stress of cold, drought, and pests on plants [15,16,17], and also contributed to protecting the human body from oxidative stress, cancer, bacterial infection, and cardiovascular and neurodegenerative diseases [18, 19].

Anthocyanins, which are the product of the phenylpropanoid pathway, are the central issue in the study of plant colour, which is [20], and its synthetic pathway has been well-characterized in Arabidopsis thaliana and Petunia [21, 22]. First, phenylalanine is required as a substrate to be converted to cinnamic acid in the presence of phenylalanine ammonia-lyase (PAL) [23]. Cinnamic acid is then converted into various dihydroflavonols by a series of enzymes, such as cinnamate 4-hydroxylase (C4H), 4-coumarate-CoA ligase (4CL), chalcone synthase (CHS), chalcone isomerase (CHI), clavanone3-hydroxylase (F3H), flavonoid 3’-hydroxylase (F3’H) and flavonoid 3’5’ -hydroxylase (F3’5’H) [20, 24]. Subsequently, dihydroflavonol4-reductase (DFR) catalyses the conversion of dihydroflavonols to leucoanthocyanidins [25], which are finally converted into anthocyanins by anthocyanidin synthase (ANS) [26]. Anthocyanins are the end products of the anthocyanin synthesis pathway and are divided into six groups: cyanidin, pelargonidin, delphinidin, peonidin, petunidin, and malvidin [27, 28]. However, they are unstable in the cytoplasm and require further glycosylation (GT), methylation (MT), and acylation (AT) to be stored in vacuoles [29] (Fig. 1). The anthocyanin derivatives produced vary among plant species. For example, anthocyanin 3-O-glucoside in Ipomoea nil was glycosylated to form anthocyanin 3-O-sophoroside [30]; Vitis vinifera was pigmented through glycosylation and methylation to generate procyanidin-3-glucoside and paeoniflorin-3-glucoside [31]; Chrysanthemum×morifolium was pigmented by acylating cyanidin 3-O-glucoside to cyanidin-3-O-(6’’-malonylglucoside) [32].

Fig. 1
figure 1

Biosynthesis pathway of anthocyanin. PAL (phenylalanine ammonia-lyase); C4H (cinnamate 4-monooxygenase); 4CL (4-coumarate-CoA ligase); CHS (chalcone synthase); CHI (chalcone isomerase); F3H (flavanone 3-hydroxylase); F3’H (flavonoid 3’-hydroxylase); F3’5’H (flavonoid 3’,5’-hydroxylase); DFR (dihydroflflavonol 4-reductase); ANS (anthocyanidin synthase); GT (glucosyltransferases); MT (methyltransferases) and AT (acyltransferases)

Full size image

The transcription factor families MYB, bHLH, and WD40 are also involved in anthocyanin synthesis by regulating the expression of structural genes [33]. MYB is one of the most abundant family of transcription factors in higher plants, which is related to regulating secondary metabolism, cell morphogenesis and differentiation, signal transduction, and stress response [34,35,36,37]. They regulate the expression of the early genes in anthocyanin synthesis [38], such as PAL, C4H, 4CL, CHS, and CHI [39,40,41]. The bHLH transcription factors regulate anthocyanin synthesis by binding to either MYB transcription factors or WD40 proteins, or both of them to form an MBW protein complex [42,43,44]. Therefore, a thorough investigation is warranted to find out the relationship between leaf colour and anthocyanin species.

In this study, the transcriptional and metabolic data on L. chinense var. rubrum leaves of three colours (green, mosaic and purple) were compared to identify the key metabolites and genes that regulate leaf colour formation, to clarify the molecular and metabolic mechanisms underlying the different pigmentations and to provide a basis for colour improvement in ornamental plants.

Results

Leaf colour observation and pigment content determination

To understand the general colour characteristics of L. chinense var. rubrum leaves, leaves of three colours were observed quantitatively, anatomically, and microscopically. The colour of the leaves was consistent with that of the pigment cells or cell clusters inside. The three colours of leaves that were examined were: green leaves (GL), purple leaves (PL), and mosaic leaves (ML) (Fig. 2A1-C1). When observed at a magnification of 20×, the upper epidermal cells of ML leaves had a small amount of purple pigment, while those of PL leaves had a large amount of purple pigment, and those of GL leaves had a large amount of green pigment (Fig. 2A2-C2). Meanwhile, the transverse section was observed to show that the L. chinense var. rubrum leaves had typical structural characteristics such as upper epidermis, palisade tissue, sponge tissue, and lower epidermis (Fig. 2A3-C3). Chlorophyll and anthocyanin were mainly in the mesophyll cells of leaves. The mesophyll cells of ML leaves were purple and green, while those of PL leaves were mostly purple and those of GL leaves were green ( Fig. 2A3-C3).

Fig. 2
figure 2

Pictures show mesophyll cells and sections of ML, PL, and GL leaves in turn, scale bars = 100 μm. A1-C1 ML, PL, and GL leaves; A2-C2 Microscopic observation of leaf epithelial cell, scale bars = 100 μm; A3-C3 The anatomical structure of a transverse section of the blade, scale bars = 100 μm

Full size image

To further evaluate the leaf colour objectively, we used the CIELAB system to detail various leaf colour indexes (L*, a*, b*) and detect pigment contents. The L* (lightness) parameter varies from 100 (white) to 0 (black), A positive value of a* indicates more red than green, and a positive value of b* means more yellow than blue [45]. Chromatic value analysis showed that the L* and b* of GL leaves were significantly higher than those of ML leaves, while the a* value of ML and PL leaves was higher compared with GL leaves (Table 1).

We quantified the photosynthetic pigment and total anthocyanin contents in the leaves of the three samples (Fig. 3) (Additional file 1). For photosynthetic pigment, we found notable differences among the three samples. GL had the highest content of photosynthetic pigment, around 2-3-fold higher than ML and PL, while the corresponding differences between ML and PL were insignificant (Fig. 3A, B, C, D). PL had the highest anthocyanin content (Fig. 3E).

Fig. 3
figure 3

The X-axis indicates the name of the sample, and the Y-axis indicates the absolute content of the extracted fresh weight. A Absolute content of chlorophyll (A) B Absolute content of chlorophyll (B) C Absolute content of carotenoid concentration. D Absolute contents of total chlorophyll. E Absolute contents of total anthocyanins. The data represent six biological repeats and their average. Mean (± SE) with different lower letters are significantly different within the (mean separation by LSD and Duncan’s test at P < 0.05)

Full size image
Table 1 The leaf color difference values of ML, PL, and GL.
Full size table

Note

a-cuticle; b-epidermis from adaxial leaf surface; c-palisade issue; d-spongy issue.

Statistical analysis of metabolomic data

Physiological data showed that the anthocyanin contents in the different-coloured leaves of L. chinense var. rubrum differed significantly; however, the reason for this difference remains unclear. We profiled the metabolome of the three samples using the liquid chromatography-tandem mass spectrometry metabolomics approach. A total of 207 compounds were detected in the L. chinense var. rubrum leaf, which was grouped into eight classes: proanthocyanins, polyphenol, isoflavone, flavonol, flavanone, anthocyanins, flavonoids, and flavones (Additional file 2). The 207 metabolites were analyzed by principal component analysis to compare the metabolite compositions involved in the pigmentation of the leaves. The compositions of the three samples separated significantly in the first principal component (38.5% of the total variable) and the second principal component (26.2% of the total variable) (Fig. 4A), indicating that the ML, PL, and GL leaves had significant inter-group differences.

A total of 37, 35, and 41 differential metabolites were selected in GL vs. ML, GL vs. PL, and ML vs. PL, respectively (Additional file 3), with a total of 11 overlaps (Fig. 4B). The annotations of the different groups of metabolites in the various pathways of flavonoid biosynthesis (ko0094) and anthocyanin biosynthesis (ko00942) were shown in Additional file 4.

Fig. 4
figure 4

Differential metabolites from different leaves. A PCA score plot of three materials and numbers of potential markers for each leaf color. B Venn diagram shows the overlap** and cultivar-specific differential metabolites from ML, PL, and GL

Full size image

Anthocyanin content in the three-coloured leaves

Interestingly, the leaf colours are closely related to the content of anthocyanin-related metabolites. Using metabolomics, we isolated and identified 15 anthocyanins from leaf extracts cyanidin 3-O-glucoside, peonidin, cyanidin O-syringic acid, delphinidin, delphinidin 3-O-glucoside, cyanidin 3-O-rutinoside, cyanidin 3,5-O-diglucoside, pelargonin, petunidin 3-O-glucoside, pelargonidin 3-O-beta-D-glucoside, cyanidin, cyanidin 3-O-galactoside, petunidin 3,5-diglucoside, malvidin 3-acetyl-5-diglucoside, and peonidin 3-sophoroside-5-glucoside. Notably, the results of total anthocyanin content were consistent with those of anthocyanin content determination, following the trend PL > ML > GL leaves (Fig. 2E). The same kind of anthocyanins was examined from the three-coloured leaves, but their levels differed significantly (Table 2). The leaves of different colours might differ in anthocyanin biosynthesis or the expression of regulatory genes.

Table 2 Type and content of anthocyanins in leaves of ML, PL, and GL.
Full size table

Sample quality control (QC) analysis

Three standardized cDNA libraries were constructed from the RNA of GL, ML, and PL. After the cDNA library was cleaned and characterized, a total of 186,694,570,149,946,386 and 123,143,062 reads were obtained, respectively. The percentages of reads having Q20 (an error probability of 0.02%) were 98.09%, 97.48%, and 97.9% for GL, ML, and PL, respectively. The GC contents of the reads were approximately 43.73%, 43.98%, and 43.84%, respectively (Additional file 5). These clean reads were assembled into 231,810 unigenes ranging from 65 to 2135 bp in length (average 1271 bp) and an N50 of 2608 bp (Additional file 6). The sequencing quality covered the majority of expressed genes in GL, ML, and PL, providing a reference for further analysis. We compared the obtained sequences with the information in seven databases; 118,518 (17.88%) unigenes had homologues in the nr database, 96,572 (14.57%) in SwissProt, 115,058 (17.36%) in KEGG, 75,254 (11.36%) in KOG, 84,662 (12.77%) in GO, 88,001 (13.28%) in NT and 84,662 (12.77%) in Pfam databases, respectively (Additional file 7). We calculated the correlation coefficient of samples according to the fragments per kilobase of transcript per million mapped reads (FPKM) value to evaluate the reliability of the measured gene expression levels. The higher the similarity was, the closer the Pearson coefficient was to 1, indicating that the measurement was reliable (Fig. 5A).

Fig. 5
figure 5

Differential expression genes in different colors. A Thermal diagram of the correlation coefficient between leaves. The Pearson correlation coefficient is within [-1, 1], and the closer it is to 1 or -1, the stronger the positive/anti-linear relationship. B Venn diagram of DEGs

Full size image

The intersection of differentially expressed genes (DEGs) in three-coloured leaves

The Venn diagram more intuitively showed the overlap of DEGs in the three comparison groups (Fig. 5B). There were 5646 DEGs (3447 downregulated, 2199 upregulated) between the GL vs. ML group, 4217 DEGs (3539 downregulated, 678 upregulated) between the GL vs. PL group and 6836 DEGs (2613 downregulated, 4223 upregulated) between the ML vs. PL group (Additional file 8).

The results of GO database annotation, presented in Additional file 9, showed that DEGs in the three-coloured can be successfully annotated into three biological processes.

The KEGG database is pathway-related. To further study the biochemical pathways of these DEGs, they were mapped onto the KEGG database [46]. Notably, KEGG pathway enrichment analysis in the pairwise comparisons of DEGs between two groups highlighted several metabolic processes including flavone and flavanol biosynthesis (ko00942) and flavonoid biosynthesis (ko00941), which were ​closely related to anthocyanin synthesis (Additional file 10). These pathways provided insights into the metabolic processes underlying different leaf pigmentations in L. chinense var. rubrum.

Genes involved in anthocyanin biosynthesis

To further study the determinants of colour diversity in ML, PL, and GL, the anthocyanin synthesis metabolic pathway was emphasized. Anthocyanins played an important role in plant coloration. Therefore, pathways related to anthocyanin synthesis were screened out from 241 DEGs (Fig. 4B), among these nine DEGs showed significant changes in expression levels: one ANR (ANR1217), four CYP75A (CYP75A1815, CYP75A2846, CYP75A2909, and CYP75A1716) and four UFGTs (UFGT1876, UFGT1649, UFGT1839, and UFGT3273) (Additional file 11). As the key gene in the biosynthesis of delphinidin, CYP75A catalysed the conversion of its major substrate, dihydrokaempferol, to dihydromyricetin [47]. ANR [48] and UFGT [49] convert the substrate of anthocyanin into (-)- epicatechin and anthocyanin, respectively. In this study, we found that these nine genes (particularly UFGTs) were expressed considerably higher in PL than in ML and GL (Additional file 12). However, little is known about the UFGT gene of L. chinense var. rubrum and its function needs to be further studied.

Genes encoding transcription factors

Transcription factors participate in the synthesis and accumulation of metabolites by modulating the expression levels of structural genes. Therefore, we screened genes related to the flavonoid synthesis pathway by comparing transcription factors between the three-coloured leaves. Genes involved in flavonoid biosynthesis are usually regulated by MYB, bHLH, WD40, bZIP, and MADS-box transcription factors [50]. Nine transcription factors were selected from these families (Additional file 12), including two MYBs (MYB1057 and MYB1211), one MADS-box (MADS1235), two AP2-likes (AP2-like1779 and AP2-like2234), one bZIP (bZIP3720), two WD40s (WD2173 and WD1867) and one bHLH (bHLH1631). In this study, we found that AP2-like2224, bZIP3720, WD1867, WD2173, and bHLH1631 were all upregulated in PL, while the MYBs, MADS-box, and AP2-like1779 were upregulated in ML and GL.

Validation of transcriptome results by quantitative reverse transcription PCR (qRT-PCR)

To verify the expression levels of structural genes and transcription factors related to anthocyanin synthesis in L. chinense var. rubrum, we selected nine structural genes and nine transcription factors for qRT-PCR analysis (Additional file 12), and their correlation was evaluated (Additional file 13). ANR and CYP75A had higher expression levels in GL and ML than PL (Fig. 6A). Although ANR promotes the formation of (-)- epicatechin, it weakens the transformation ability of anthocyanins and leads to the formation of green leaves. ​In contrast, UFGT was highly expressed in PL leaves but had lower or no expression in ML and GL, respectively. Thus, we speculated that the upregulated expression of the UFGT gene might contribute to anthocyanin synthesis and regulate the formation of mosaic and purple leaves.

Fig. 6
figure 6

FPKM values calculated from the transcriptomic data, and transcriptional levels of flavonoid biosynthetic genes in the L. chinense var. rubrum detected by qRT-PCR analysis. The β-actin gene was used as an internal control. AANR, UFGT, and CYP75A. B MYB, MADS-box, AP2-like, bZIP, WD40 and bHLH. The control for relative expression GL was assigned the arbitrary value of 1.0. The data represent the six biological repeats and their average. Error bars represent the standard deviations of six biological replicates

Full size image

We verified expression trends of candidate transcription factors, such as AP2-like2224, bZIP3720, WD1867, WD2173, and bHLH1631; they were consistent with the UFGT gene expression trend with higher expression in the PL compared with GL and ML (Fig. 6B). At the same time, the Pearson correlation coefficient showed a strong correlation (R2 > 0.92) between the five transcription factors and the UFGT gene (Additional file 14), suggesting that these transcription factors were related to anthocyanin biosynthesis and regulated the appearance of leaf colour.

Discussion

In ornamental plants with colourful leaves, research on leaf colour has always been the focus since it affects the ornamental quality and commercial value of the plants. However, the mechanism of leaf coloration in L. chinense var. rubrum was still unclear, necessitating its study using the existing technique.

In this study, anatomic and microscopic observations, pigment content determination, flavonoid metabolomics, and transcriptome sequencing were performed on L. chinense var. rubrum leaves of three different colours. Anatomic observations under the microscope showed that the mesophyll cells in ML were a mix of purple and green, those in PL were all purple, and those in GL were all green. On this basis, the determination of L. chinense var. rubrum pigment content proved the existence of chlorophyll, carotenoids, and anthocyanins in plant leaves. We speculated that ML had the pigments chlorophyll and anthocyanin, PL had anthocyanin and GL had chlorophyll. A total of 207 flavonoid compounds were detected using metabolomics. DEGs related to pigmentation were found in the transcriptome of the three-coloured leaves. To screen the main components and candidate genes of pigmentation, we proposed a hypothetical biosynthetic pathway (Fig. 7).

Fig. 7
figure 7

Putative genes in the anthocyanin synthesis pathway and their expression level. Heatmaps were constructed based on log2 (FPKM) of leaves ML, PL, and GL

Full size image

Our pigment content results showed that chlorophyll and carotenoid content was low relative to anthocyanins (Fig. 2). Therefore, differences in the type and content of anthocyanins were considered to be the possible reason for the different colours of L. chinense var. rubrum leaves.

At the metabolic level, anthocyanin biosynthesis may be the main pathway involved in leaf pigmentation (Additional file 4). We established that the total anthocyanin content of GL was lower than that of ML and PL and the change in the trend of anthocyanin content corresponded with the change in leaf colour. The contents of cyanidin 3-O-glucoside, cyanidin O-syringic acid, cyanidin 3,5-O-diglucoside, pelargonidin, petunidin 3-O-glucoside, and peonidin 3-sophoroside-5-glucoside were significantly different in the three leaves while those of peonidin, delphinidin, and cyanidin 3-O-rutinoside were not. Therefore, we speculate that the changes in the levels of these anthocyanins influence the colour of L. chinense var. rubrum leaves.

According to the differential expression of genes involved in anthocyanin synthesis and the difference in the content of various anthocyanins, we speculated the reasons for the different colour of L. chinense var. rubrum leaves. Although CYP75A was upregulated in ML and GL to accumulate the raw materials needed for anthocyanin synthesis, anthocyanin was converted into (-)-epicatechin due to the upregulation of ANR. Moreover, the downregulation of the UFGT gene ensured that anthocyanin could not be converted into a stable form, resulting in reduced anthocyanin content in the leaves. The UFGT gene in the anthocyanin biosynthesis pathway is the key to the formation of different types of anthocyanins [51]. In Vitis vinifera, the upregulation of UFGT genes led to the accumulation of anthocyanins [52]. Similar observations were made regarding UFGT expression and anthocyanin accumulation in Malus pumila Mill [53], Litchi chinensis [59,60]. PpMYB forms the MBW complex and especially activates UFGT to regulate the biosynthesis of anthocyanins [79]. Differential gene expression analyses in the different-coloured leaves were performed using the DESeq R package. DEGs were identified as those genes that had |log2(fold change)|≥2, FDR value < 0.01. GO and KEGG pathway enrichment analysis of the DEGs was done using the GOseq R package-based hypergeometric distribution [80], which can adjust for gene length bias in DEGs.

qRT-PCR analysis

To further confirm the reliability of RNA-seq data in our differential expression analysis, the relative expression of nine structural genes in the flavonoid metabolic pathway was analyzed through qRT-PCR. Primers were designed using Primer Premier 5.0 software and the details are shown in Additional file 15. The β-actin gene was used as an internal reference gene and three biological replicates were set for each biological sample. The expression level of each gene in the list was calculated using the Livak method (delta-delta CT, 2−ΔΔCt)and expressed as the average standard deviation [81].

Integrative metabolomic and transcriptomic analysis

Transcriptomic and metabolomic data for L. chinense var. rubrum leaves with clear differences were used for analysis. Correlation analysis was carried out according to the metabolite content of different colors of leaves in metabolic data and the differential gene expression in transcriptome data.

Figure 1 Biosynthesis pathway of anthocyanin. PAL (phenylalanine ammonia lyase); C4H (cinnamate 4-monooxygenase); 4CL (4-coumarate-CoA ligase); CHS (chalcone synthase); CHI (chalcone isomerase); F3H (flavanone 3-hydroxylase); F3’H (flavonoid 3’-hydroxylase); F3’5’H (flavonoid 3’,5’-hydroxylase); DFR (dihydroflflavonol 4-reductase); ANS (anthocyanidin synthase); GT (glucosyltransferases); MT (methyltransferases) and AT (acyltransferases).

Figure 4 Differential metabolites from different leaves. A PCA score plot of three materials and numbers of potential markers for each leaf color. B Venn diagram shows the overlap** and cultivar-specific differential metabolites from ML, PL, and GL.

Figure 5 Differential expression genes in different colors. A Thermal diagram of the correlation coefficient between leaves. The Pearson correlation coefficient is within [-1, 1], and the closer it is to 1 or -1, the stronger the positive/anti-linear relationship. B Venn diagram of DEGs.