Introduction

Perilla frutescens (L.) Britton as an herbaceous plant belongs to Lamiaceae family, and is widely cultivated throughout several Asian countries (Gaihre et al. 2021; Nakajima et al. 2015; Wang et al. 2020a, b). Now, it has been an economic crop and a source of medicine and spices in China, Korea, Japan, and other many countries (Tong et al. 2015). This species includes two cultivated types, viz. P. frutescens var. frutescens, and P. frutescens var. crispa according to their morphology and uses, and they have being freely cross-fertile with each other by artificial pollination. In China and Korea, P. frutescens var. frutescens with green leaf is an important oil crop and also used as vegetable crops in Korea, while P. frutescens var. crispa with purple leaf extensively cultivated in China and Japan, are more often used for medicinal properties, fish garnish and ingredient for various foods (Gaihre et al. 2021; Tong et al. 2015).

Leaf color variations are large in perilla, and therefore it is also used as an ornamental plant (Zhang et al. 2021). Perilla leaves contain many bioactive components including phenolic compounds, rosmarinic acid, caffeic acid, flavanone, anthocyanin, vitamins, amino acids, polysaccharides, and minerals (Nakajima et al. 2015; Deguchi and Ito 2020; Miki et al. 2015; Zheng et al. 2020). These compounds have various pharmacological activities, such as anti-oxidative activity, anti-inflammatory, anti-anxiety, anti-cancer activities, anti-depressive activities, and beneficial for gastrointestinal digestion (Wang et al. 2020a, b; Zheng et al. 2020; Kwak and Ju 2015). In addition, perilla is one of the plant species with the most abundant α-linolenic acid (Zhang et al. 2021). As traditional Chinese medicine, P. frutescens has been widely used in treating colds, coughs, vomiting, constipation, and food poisoning in clinical settings (Nakajima et al. 2015; Wang et al. 2020a, b; Kwak and Ju 2015). Recently, it was found that perilla leave extracts had ability of inhibiting severe acute respiratory syndrome coronavirus (SARS-CoV-2) replication by inactivating the virion (Tang et al. 2021). Previous studies showed that purple perilla leaves have high anti-oxidative activities than that of green perilla leaves due to the different concentrations of anthocyanins and flavonoids (Zheng et al. 2020). Therefore, deciphering the molecular mechanism of leaf coloration is benefit for exploring perilla function.

Metabolomics, differential display of mRNA, and transcriptome technology have been used to investigate green and red leaf coloration and a few regulated genes of leaf coloration have been identified in P. frutescens (Yamazaki et al. 2003, 2008). The regulation of anthocyanin (mainly malonylshisonin) production, but not that of other metabolites determined the red and green forma (Yamazaki et al. 2003). The key enzymes of anthocyanin synthesis including glutathione S-transferase (PfGST1) and chalcone isomerase (PfCHI) were preferentially expressed in the leaf of red perilla (Yamazaki et al. 2008). Tong et al. (2015) used RNA-seq sequenced four samples, including one weedy and one cultivar of var. frutescens, and two cultivars of varieties crispa with red and green leave color of Perilla. According to expression profiling between P. frutescens (L.) var. frutescens and var. crispa, 22, 962 differently expressed genes (DEGs) were detected in the weedy variety compared to the cultivars, 18 unigenes assigned to anthocyanin played important roles in determining leaf color were identified (Tong et al. 2015). A total of 77 unigenes involved in flavonoid biosynthesis were identified in purple and green perilla leaf by transcriptomic approaches, and highly expressed chalcone isomerase (PfCHS) enhanced accumulation of red perilla leaf (Zhang 2021). However, back leaf coloration is still unclear. The other key genes regulated leaf color, such as transcription factor genes, should be further identified for deciphering molecular mechanisms of perilla leaf color formation.

Chlorophyll, carotenoids and anthocyanins are the primary pigments that determine plant leaf color. In fact, the leaf color variations of perilla are extensive, and the leaf color contains green, red, purple, light purple, dark purple and etc.. Much more studies were focused on anthocyanins, but not considering chlorophyll, carotenoid, and other pigment change, together. The mechanism of leaf color formation should be further studied.

In this paper, four different perilla cultivars with four kinds of leaf colors were used as materials. The leaves of four cultivars in their strong growth stage were collected for pigment measurements and transcriptome sequencing. The photosynthesis rate and color aberration were also tested. The correlation between pigment content and leaf color change were analyzed, and key genes and related regulatory genes controlled leaf color formation were also obtained. The probable scheme of regulated both green and purple leaf color formation and back leaf coloration was postulated at last.

Materials and methods

Plant materials and sample collection

One cultivar ‘94’ from P. frutescens var. frutescens and three cultivars including ‘272’, ‘103’ and ‘M357’ from P. frutescens var. crispa were used as material in this work. The front and back leaf color were different in these four cultivars and their leaf colors were shown in Fig. 1. The front and back leaf colors of ‘94’ and ‘M357’ were completely green and completely purple, respectively. In ‘272’, the leaf color was front green and back light purple, while the leaf color of ‘103’ was front green with purplish vein at bottom and back dark purple (Fig. 1). 94-GG, 272-GLP, 103-GPP, and 357-PP were represented cultivars ‘94’, ‘272’, ‘103’, and ‘M375’ and their color abbreviation, respectively. The same abbreviations are used below in this work. The perilla seeds were introduced from Oil Crops Research Institute, Chinese Academy of Agricultural Sciences (CAAS), Wuhan city, Hubei Province, China, and Guizhou Academy of Agricultural Science, Guizhou city, Guizhou Province. After seedling culture, the two leaves’ stage seedlings were transplanted in the experimental field of the Institute of Vegetables and Flowers, CAAS in Beijing, China, on 20th-April-2021. The leaves in strong growth stage were collected on 6th-June-2021. In this stage, the leaf colors of the four cultivars have been steadily formed. The fresh leaves were immediately frozen in liquid nitrogen, and stored in a refrigerator at − 80 °C until use. Other two developmental leaf stages including original leaf color, and middle leaf color of the ‘272’ and ‘103’ were also collected on for chromatic parameters analysis on 1st-June-2021 and 3rd-June-2021, respectively.

Fig. 1
figure 1

The morphology and the front and back leaf color of four cultivars ‘94’, ‘272’, ‘103’, and ‘M357’

Full size image

Measurements of chlorophyll, carotenoid, anthocyanin, and flavonoid content

The pigment content was determined using a 722 spectrophotometer (Shanghai Jinghua Technology Instrument Co., Ltd, Shanghai, China) after extraction with 80% acetone. Concentration of Chl a, Chl b, Chl (a + b), and carotenoids in the obtained extracts was calculated according to the formulas (Ivanov et al. 2020).

The total anthocyanin content of leave extract was determined using a modified pH differential method according to previous reports in Rabino and Mancinelli (1986). The anthocyanin of leaf was extracted at 4 °C in acidic (1% HCl, w/v) methanol for 48 h. The absorbance of the supernatant was determined at 520 nm and 700 nm using a 722 spectrophotometer (Shanghai Jinghua Technology Instrument Co., Ltd, Shanghai, China) at pH 1.0 and 4.5 conditions with modified by HCl, respectively. The total anthocyanins content were calculated according Eqs. 1, and 2, respectively. For comparison purposes, pigment content was calculated as cyanidin 3-glucoside using an extinction coefficient (ε) of 29,600 L mol−1 cm−1 and molecular weight of anthocyanins 449.2. Measurements were replicated four times with means being reported.

$$ {\text{Absorbance }}\left( {\text{A}} \right)\, = \,\left( {{\text{A}}_{{52}0} --{\text{A}}_{{7}00} } \right){\text{pH1}}.0 - \left( {{\text{A}}_{{52}0} --{\text{A}}_{{7}00} } \right){\text{pH4}}.{5} $$
(1)
$$ {\text{Anthocyanin }}\left( {{\text{mg}}/{\text{L}}} \right)\, = \,{\text{A}}\, \times \,{\text{Mw}}\, \times \,{\text{DF}}\, \times \,{1}0^{3} /(\varepsilon \, \times \,{\text{L}}). $$
(2)

Mw is the molecular weight of anthocyanins (449.2 g/mol), DF is the dilution factor, ε is the extinction coefficient (29,600 L mol−1 cm−1), and L is the path length (1 cm).

The measurement of flavonoid content was referred to previous report (Matvieieva et al. 2019). Each sample was measured in triplicate.

Photosynthetic measurements of four perilla cultivars

The photosynthesis of four perilla cultivars were measured at 9:30 am–10:00 am during three continuous sunny days from 6th-June-2021 to 8th-June-2021, using portable photosynthesis system (CRAS-3, Lufthansa Technology Group, Beijing, China) that maintains photosynthetic photon flux density at 1,200 μmol m−2 s−1 and CO2 concentration at 400 μmol mol−1. Four photosynthetic parameters, namely net photosynthetic rate (Pn) estimated by CO2 uptake (it is also represented by “A” in the equipment), stomatal conductance (gs; it is also represented by “gsw” in the equipment), intercellular CO2 concentration (Ci), and transpiration rate (Tr; it is also represented by “E” in the equipment). The procedures were performed by manufacture of company. The three mature leaves of each plant were used for all above measurements, with three biological replicated for each samples.

Analysis of leaf chromatic aberration

The front and back leaves were measured with a hand-held color analyser (CM-2500D, Konica Minolta, Japan) in the room, according to the CIE L*a*b* system. The leaf colors were expressed as three main colors parameters, including L*, a* and b* value (McGuire 2019; Voss 1992). The color coordinate L* represents lightness, from black (0) to white (100), a* describes red (positive) to green (negative), and b* describes yellow (positive) to blue (negative). Chroma (C*) can be calculated based on the equations C* = (a*2 + b*2)1/2, and hue angle (h) were calculated using the formulae h = tan−1 (b*/a*). The experimental materials were collected from at least three plants and three points on each leave were measured.

RNA extraction, construction of the cDNA library, and transcriptome sequencing

Total RNA was isolated from fresh leaves of four perilla cultivars using a Trizol extraction kit (Invitrogen, USA), according to the instructions of the manufacturer. The concentration and integrity of RNA per sample was determined using an Agilent Bioanalyzer 2100 (Agilent Technologies, Santa Clara, CA, USA), and the quality of RNA was measured using a NanoDrop 2000 UV/Visible spectrophotometer (Thermo Scientific™, USA). The mRNA enrichment and cDNA library construction were referred to previous reports (Wang et al. 2019; Gao et al. 2020). A total of eight RNA-seq libraries (two libraries for each sample) were constructed and then sequenced using an Illumina HiSeqTM4000 (4 × 100 bp read length) platform at the Beijing Genomics Institute Company (Shenzhen, China). All obtained data are available at the NCBI Short Read Archive and the accession number is and SRP371629.

De novo transcriptome assembly and functional annotation

The clean reads were produced by removing the reads contained adapters, unknown nucleotides (Ns), low quality reads (phred score < 20), and terminal nucleotides in both 3’ and 5’ ends, as well as relatively short reads (< 50 nt) from the raw reads (Lohse et al. 2012). The high-quality clean reads were de novo assembled using Trinity software to construct unique consensus sequences (Grabherr et al. 2011).

All assembled unigenes were annotated in the following six databases: NCBI non-redundant protein database (NR), NCBI non-redundant nucleic acid sequence database (NT), Swiss-Prot protein sequence database (Swiss-Prot), Gene Ontology (GO), Kyoto Encyclopedia of Genes and Genomes (KEGG), and Clusters of EuKaryotic Orthologous Groups (KOG) using BLAST software (Altschul et al. 1990). Blast2GO was used to perform GO annotation (Conesa et al. 2005). The potential coding sequencing (CDS) of unigenes identification and the simple sequence repeats (SSRs) prediction was used the program TransDecoder and MISA software (http://pgrc.ipk-gatersleben.de/misa/), respectively.

Differentially expressed genes (DEGs) identification, and GO and KEGG enrichment

The expression levels of genes were calculated using fragments per kilobase of exon region per million mapped reads (FPKM) values. The FPKM value for each gene was determined based on the length of the gene and reads count mapped to this gene. The unigenes showed differential expression levels among different samples were calculated based on the ration of FPKM values. Meanwhile, the differently expressed genes (DEGs) were evaluated based on the genes with FPKM of > 1 in at least one sample. The DEGs between any two comparisons were identified using DEseq2, and Benjamini and Yekutieli’s method was used to adjust the resulting P-values, to minimize the false detection rate (FDR) (Benjamini and Yekutieli, 2005). A false discovery rate (FDR) ≤ 0.001, Q-values ≤ 0.05, P-values ≤ 0.001 and the absolute value of |log2(Fold change)|≥ 1 were used as the thresholds to determine the significance of the gene expression differences. Furthermore, the DEGs were subjected to Gene Ontology (GO) and KEGG pathway classification and functional enrichment analysis s. The hierarchical clustering of DEGs was constructed by Mev software (Wang et al. 2020a, b).

Validation of gene expression by quantitative real-time PCR (qRT-PCR)

A total of 14 interested DEGs related to chlorophyll and anthocyanins metabolisms were randomly selected to validate the expression levels through quantitative real-time PCR (qRT-PCR). The total RNA extraction, mRNA purification, cDNA synthesis, and qRT-PCR reaction were conducted according to Wang et al., (2012). The actin gene was selected as a reference gene (Jiang et al. 2020). The primers used in qRT-PCR are summarized in Table S1. All qRT-PCR reactions were carried out in triplicate using independent samples.

Statistical analysis

All the results in the experiment were repeated at least three times. The experimental data are expressed as mean ± the standard error (SE) and were conducted using one-way ANOVA, followed by Duncan’s multiple range tests at p < 0.05 level. The statistical significance among leaves with different colors was performed with the SPSS 22.0 (SPSS Institute, IBM, Endicott, NY, USA).

Results

Pigment levels in four perilla cultivars

The transplanted seedlings grew well in the experimental field and the front and back leave colors of ‘94’ and ‘M357’ kept steady with green and purple color, respectively, in all the growth stage (Fig. 1). The cultivar ‘272’ and ‘103’ underwent quick color changes, and then the leaf colors were steady in all growth stage (Fig. 1). Finally, the back of mature leaf in ‘272’ and ‘103’ turned light and dark purple, respectively. At same time, the front leaf color of ‘272’ was still green, and the front leaf color was green with light purplish vein at the bottom of leaf in ‘103’.

For chlorophyll and carotenoid contents, it was found notable difference among four cultivars. The total chlorophyll and carotenoid contents of 357-PP were highest among the four cultivars (Fig. 2A). For anthocyanin, the largest content was also in 357-PP, followed by 103-GPP, and the low contents were in 94-GG and 272-GLP (Fig. 2B). The anthocyanin content was around 20 fold higher in 357-PP than that of in 272-GLP (Fig. 2B). The total content of flavonoid was also high in 357-PP (Fig. 2C).

Fig. 2
figure 2

The contents of chlorophyll and carotenoid (A), anthocyanin (B), and total flavonoid (C) in different colored leaves of four perilla cultivars. 94-GG, 272-GLP, 103-GPP, and 357-PP were represented cultivar ‘94’, ‘272’, ‘103’, and ‘No.375’, and their leaf color abbreviation, respectively. The same abbreviations are used below. a, b, and c indicate significant differences at the P < 0.05 level

Full size image

Analysis of photosynthetic parameters in four perilla cultivars

Four photosynthetic parameters were shown in Fig. 3. For the net photosynthetic rate, 357-PP had highest level, followed by 103-GPP, and the lowest level was in 272-GLP. Concerning the Gs, it was closely matched the Pn curve and the highest level was in 357-PP. Similarly, the Tr and Gs curves had the same trends of Pn. According to the chlorophyll contents of four cultivars, perilla cultivar with higher chlorophyll contents has higher photosynthetic rate.

Fig. 3
figure 3

The comparison of net photosynthetic rate (Pn/A) (A), transpiration rate (Tr/E) (B), intercellular CO2 concentration (Ci) (C), and stomatal conductance (Gs) (D) among four perilla cultivars

Full size image

Identification of leave color in four perilla cultivars

The chromatic parameters of the front and back leaves of each samples were recorded using the CIE L*a*b* color system (Table 1). The values of L*a*b* of original leaf colors of 272-GLP, and 103-GPP were similar with those in 94-GG (Table 1). After leaf color steady, the values of a* b* in the front leaf colors were ranged from − 15.81 to 4.79, and 4.55 to 22.77, respectively, while they were ranged from − 14.12 to 8.76, and 2.19 to 22.16, respectively, in the back leaf colors (Table 1). The a* represents red and green color. The values of the chromatic parameters a* in 357-PP was distinct from those in other three cultivars, because it showed positive a* values (mean 4.79) in the front leaf and lowest (mean 2.19) among four perilla cultivars in back leaves. The L* values (mean 40.28, 48.86 in front and back, respectively) of the front and back leaves were high in green leaves of ‘94’, and low in front leave of ‘M357’ (mean 25.56) and back leave of ‘No.101’ (mean 27.21) and ‘M357’ (28.45). These samples show no significant differences in hue (Table 1). The lowest chroma value (C*) was in 6.61. As described, these measured results were consistent with the leaf pigmentation appearance.

Table 1 The CIE L*a*b*chromatic parameters in front and back leaf of four perilla cultivars
Full size table

The correlative analysis between pigment contents and the L*a*b* values in front leaf showed that chlorophyll, carotenoid, anthocyanin, and flavonoid were positive related to a* value, and they together determined purple leaf color formation (Table 2). In back leaf, only anthocyanin and flavonoid has positive correlative relationship with a* value (Table 2). It was indicated that back leaf color change was controlled by anthocyanin.

Table 2 The correlative analysis between different pigment contents and L*a*b* values of front and back leaf
Full size table

Tri-dimenstional graphics allowed evidencing that separation among the four leaf color types. The front leaf color analysis showed that 357-PP in one room, while the other cultivars in the same room. The original color of 272-GLP and 103-GPP and 94-GG was close in the same room, while the colors of their middle stage were close to their steady colors (Fig. 4A). For the back leaf color analysis, purple leaf of 103-GPP and 357-PP were in the same room and steady color of 272-GLP and color of middle stage in 103-GPP was in other room (Fig. 4B). As a whole, for 272-GLP and 103-GPP, the front leaf colors were relatively steady, and their back leaf colors were gradually change during growth and steady at last (Fig. 4).

Fig. 4
figure 4

The distribution of chromatic coordinates of front (A) and back (B) leaf in four perilla cultivars

Full size image

Illumina sequencing and de novo transcriptome assembly

In this work, 51.43 Gb nucleotides were produced by Illumina sequencing BGISEQ platform. After the removal of adapter sequences, ambiguous reads and low quality reads, total of 42.86 Gb clean reads were assembled into 128,041 unigenes, with average sequence length of 1,488 bp (Table S2). The N50 length, N90 length, and GC percentage of assembled unigenes were 2,229 bp, 807 bp, and 40.93%, respectively (Table S2). The top three sequence length distribution and percentage of the unigenes were 200–300 bp (15.04%), 300–400 bp (6.74%), and 400–500 bp (4.57%), respectively (Fig. S1A). A total of 13,714 unigenes were longer than 3000 bp (10.71%) (Fig. S1A). These results showed that the quality of sequencing data was good. BUSCO quality evaluation also showed assembled unigenes were complete and reliable, and complete(C) and single-copy(S) of all unigenes arrived at 96 (Fig. S1B). A total of 78,105 protein coding sequences (CDS) with 84,896,910 bp total length were detected using Transdecoder, and the 300–400 bp CDS (11.96%) were highly aligned (Fig. S2, Table S3). The maximum CDS length reached to 15,303 bp, and a total of 2,201 CDS were longer than 3000 bp (2.82%) (Table S3).

Functional annotation of perilla leaf transcriptomes

The assembled unigenes were annotated to seven public databases, including nonredundant protein sequence database (Nr), nucleotide sequence database (Nt), Swiss-prot (Swissprot), Kyoto encyclopedia of genes and genomes (KEGG), clusters of orthologous groups of proteins (KOG/COG), Pfam, and gene ontology (GO). In total, 98,086 unigenes (76.61%) could be annotated in at least on database, while there were 37,219 unigenes (29.07%) successfully annotated in all seven databases (Fig. 5; Table 3). The venn diagram of Nr, Nt, Swissprot, KEGG, and GO database showed that 44,972 unigenes were annotated in all the five public databases (Fig. 5A). The species showed the best BlastX matches were Sesamum indicum which took up 45.39%, the second- and third-ranked species Handroanthus impetiginosus and Erythranthe guttata was contributed to 22.38%, and 12.87%, respectively (Fig. 5B). For GO assignments, 69,508 unigenes matched to Nr database were further divided into 38 major functional groups. Of all the GO term categories, binding (34,680), membrane part (20,741), and cellular process (14,769) were the highest ranking in the three GO categories molecular function, cellular component, and biological process, respectively (Fig. 5C). A total of 74,907 unigenes were mapped to KEGG pathways, and global and overview maps (15,265, 20.38%), carbohydrate metabolism (6,206, 8.28%), and translation (5,472, 7.31%) were occupied the top three ranking (Fig. 5D). The 73,203 unigenes were classified into 25 KOG classifications. General function prediction only (15,060, 20.57%) had the most representation, followed by signal transduction mechanisms (9,836, 13.44%) and function unknown (6,570, 8.98%) (Fig. 5E).

Fig. 5
figure 5

Venn diagrams of the five public databases annotation including Nr, Nt, Swissprot, GO, and KEGG (A), Species distribution (B), Gene Ontology (GO) assignments (C), KEGG pathways distribution (D), and KOG functional classification (E) for perilla transcriptome unigenes

Full size image
Table 3 The annotations of perilla leaf unigenes against the seven public databases
Full size table

Gene structure analysis suggested that 3,577 simple sequence repeats (SSRs) were detected in 26,807 unigenes (Fig. S3A). In addition, 3,577 transcript factors were predicted in this study (Fig. S3B) Principle components analysis (PCA) showed that four cultivars were clustered different positions and purple leaf ‘M357’ was apart from the other three perilla cultivars (Fig. S4A). Gene expression profiles of accumulative bar diagram composite column-diagram showed that many unigenes (FPKM values = 1–10) were distributed in M357-PP (Fig. S4B). These results indicated that transcriptome sequencing could be used to identify the leave color regulated genes in perilla. A total of 78,700 unigenes were expressed in all the four cultivars were seen in Fig. S4C.

Analysis of differentially expressed genes (DEGs) with different pairwise comparisons

In order to explore the differentially expressed genes (DEGs) among four perilla cultivars with different leaf colors, the gene expression levels between each pairwise comparison were studied, and then DEGs were identified by different pairwise comparisons (Fig. S5; Fig. S6). The highest number of DEGs was appeared in 272-GLP versus 357-PP (35,381) among six pairwise comparisons, with 17,824 and 17,577 unigenes up- and down-regulated, respectively (Fig. S5). In contrast, the lowest number of DEGs was observed in 272-GLP versus 103-GPP (19,376), with 13,369 and 12,088 unigenes up- and down-regulated, respectively (Fig. S5). The venn diagrams of any three pairwise comparison showed that highest number of DEGs (9,670) were common to the 357-PP versus 94-GG, 357-PP versus 272-GLP and 357-PP versus 103-GPP comparisons, and the lowest number of DEGs were common in 94-GG versus 103-GPP, 103-GPP versus 272-GLP and 94-GG versus 272-GLP comparisons (Fig. S6A-D). Furthermore, much more DEGs (1,795) were observed in many purple leaf attended pairwise comparisons including 357-PP versus 94-GG, 357-PP versus 103-GPP, 103-GPP versus 272-GLP and 357-PP versus 272-GLP than those pairwise comparisons with many green leaf attended comparisons, ‘94-GG versus 272-GLP’, ‘94-GG versus 357-PP’, ‘94-GG versus 103-GPP’ and 103-GPP versus 272-GLP (Fig. S6E-F). A total of 611 DEGs were common to the 357-PP versus 94-GG, 357-PP versus 103-GPP, 357-PP versus 272-GLP, 94-GG versus 272-GLP, and 94-GG versus 103-GPP comparisons (Fig. S6G). These results suggested that the genes regulated green and purple color had more significant difference. Notably, many DEGs were appeared in the pairwise comparisons contained purple leaf of 357-PP.

GO functional and KEGG pathway enrichment of DEGs and their expression patterns analysis

GO functional enrichment analysis of DEGs revealed that of top three main enrichment of the GO among five comparisons viz. 357-PP versus 94-GG, 357-PP versus 103-GPP, 357-PP versus 272-GLP, 94-GG versus 103-GPP, and 103-GPP versus 272-GLP, were ribosome (GO:0005840), and translation (GO:0006412) and structural constituent of ribosome (GO:0003735), respectively (Fig. S7). In 94-GG versus 272-GLP, the top three ranking GO functional enrichment were ADP binding (GO:0043531), photosystem I (GO:0009522), and photosynthesis, light harvesting (GO:0009765), respectively (Fig. S7E). The GO functional enrichment further analysis showed that the categories of photosynthesis (GO:0015979), chloroplast thylakoid membrane (GO:0009535), photosystem I (GO:0009522), and II (GO:0009523), chloroplast (GO:0009507), were all enriched among the above six pairwise comparisons. The categories of photosystem I reaction center (GO:0009538), and L-phenylalanine catabolic process (GO:0006559) were specifically distributed in 357-PP versus 103-GPP. We also found the GO functional enrichment of pairwise comparisons with green leaf cultivar was different with those with purple leaf cultivar. These results were also suggested that the content, synthesis and metabolim of chlorophyll, and photosynthesis ability were different in four perilla cultivars with different leave colors.

A total of 50 DEGs related to photosystem I and II, photosynthesis, light harvesting, chloroplast thylakoid membrane, chlorophyll binding, chloroplast, chloroplast envelope, photosynthesis, photosystem I reaction center, ferredoxin-NADP + reductase activity and PS II associated light-harvesting complex II catabolic process, and photosynthesis were identified and summarized in Table S4.

It is interested that the DEGs enriched in photosynthesis, light harvesting in photosystem I, such as chlorophyll a/b-binding protein genes (CL125.Contig9_All, Unigene136_All, Unigene127_All, Unigene128_All) and photosystem I reaction center, photosystem I reaction center subunit genes (CL10496.Contig2_All, CL5429.Contig4_All, CL238.Contig4_All, CL12334.Contig1_All, and CL13338.Contig3_All) showed up-regulated expression pattern in 357-PP (Table S4). These results suggested that purple leaf of perillia have strong ability of light harvesting and photosynthesis in photosystem I, compared with those of green or green plus purple leaves. However, ferredoxin-NADP + reductase activity related genes showed down-regulated expression pattern in purple leaves. The key gene controlled chlorophyll synthesis, such as magnesium-chelatase subunit ChlD (Unigene12797_All) CPO (CL5607.Contig4_All, CL2268.Contig3_All), PPOX (CL3522.Contig4_All, CL1474.Contig27_All), and MgCh (CL9498.Contig1_All), showed up-regulated in 94-GG and 103-GPP. At the same time, the gene chlorophyll M (CHLM, Unigene17397_All), chlorophyll 27 (CHL27, CL1169.Contig13_All), and protochlorophyllide reductase (POR, CL374.Contig6_All) control chlorophyll synthesis were also up-regulated in 357-PP, while the Chlorophyll synthase (CHLG, CL5824.Contig3_All) were down-regulated in 357-PP. However, the MCS/SGR/NYC1 (CL5222.Contig3_All, CL3099.Contig4_All, CL3287.Contig1_All, CL3709.Contig1_All, CL6196.Contig1_All, and CL3865.Contig1_All), with PPH unigenes (Unigene19598_All and Unigene19603_All), and pheophorbide a oxygenase (PAO) (CL9496.Contig2_All) played role in chlorophyll degradation was down- and up-regulated, respectively, in 357-PP. In chlorophyll recycling process, chlorophyll a and b could not be easily recycling in purple leaf 357-PP, according to lower expression of CAO (CL1364.Contig13_All), CBR (CL3043.Contig4_All), and HCAR (CL6020.Contig12_All) (Fig. S8A). As a whole, it was deduced that more chlorophyll synthesis in green leaf, while the chlorophyll degradation rate was slow at the beginning, and then fast in purple leaf 357-PP.

In chloroplast enrichment, zeaxanthin epoxidase (ZEP, CL950.Contig28_All) regulated carotenoid synthesis also showed up-regulated expression in 357-PP (Fig. S8B, Table S4). It was suggested that carotenoid may be played in leaf color formation.

KEGG pathway enrichment analyses of DEGs with p < 0.001 as the threshold were also performed in six pairwise comparisons. Like GO functional enrichment, 94-GG versus 272-GLP had a little difference with other five pairwise comparisons. In 357-PP versus 94-GG, 357-PP versus 272-GLP, and 357-PP versus 103-GPP comparisons, the top one to five ranking pathways all include ribosome (ko03010), carbon fixation in photosynthetic organisms (ko00710), photosynthesis-antenna proteins (ko00196), and photosynthesis (ko00195), flavonoid biosynthesis (ko00941, ranked sixth in 357-PP versus 94-GG) (Fig. S9A-C). Notably, porphyin and chlorophyll metabolism (ko00860) in 357-PP versus 94-GG was ranked fifth, and it was also enriched in 357-PP versus 272-GLP, and 357-PP versus 103-GPP (Fig. S9A). It was further suggested that chlorophyll metabolism was different among four cultivars with four kinds of leaf colors. Phenylpropaniod biosynthesis (ko00940), stilbenoid, diarylheptanoid and gingerol biosynthesis (ko00945) were also common in the above three pairwise comparisons. In 94-GG versus 272-GLP, 94-GG versus 103-GPP, and 103-GPP versus 272-GLP comparisons, carbon fixation in photosynthetic organisms (ko00710), photosynthesis (ko00195), porphyin and chlorophyll metabolism (ko00860), photosynthesis-antenna proteins (ko00196), phenylpropaniod biosynthesis (ko00940), and flavonoid biosynthesis (ko00941) were common in these three comparisons (Fig. S9D-F). Flavone and flavonol biosynthesis (ko00944), and isoflavonoid biosynthesis (ko00943) were specific observed in 357-PP versus 94-GG and 94-GG versus 272-GLP, and 357-PP versus 103-GPP, respectively comparisons.

A total of 17 DEGs related to carotenoid biosynthesis (ko00906) were also identified in five pairwise comparisons (Fig. S8B, Table S4). The DEGs including 9-cis epoxycarotenoid dioxygenase (NCED, Unigene214_All), carotenoid cleavage dioxygenase (CCD7, CL14084.Contig2_All), and zeaxanthin epoxidase (CL950.Contig28_All), showed much more expression in purple leaf, compared with those in green leaf. The up-regulation of beta-carotene isomerase D27 (DWRF27, Unigene32590_All), phytoene synthase 2 (PSY2, CL6263.Contig2_All), cytochrome P450 CYP97A41 (CYP97A41, CL12610.Contig3_All), 15-cis-zeta-carotene isomerase (ZIISO, CL3672.Contig2_All), and cytochrome P450 CYP97A41 (CL12610.Contig3_All) was showed in green leaf (Fig. 8B). The kinds of carotenoids were different in purple leaf 357-PP and in green leaf 94-GG, according to these genes’ expression. It was deduced that the neoxanthin and carlactone, and lutein were highly accumulated in purple leaf 357-PP and in green leaf 94-GG, respectively.

The DEGs enriched by KEGG pathway related to chlorophyll metabolism, flavonoid biosynthesis, and photosynthesis were analyzed and summarized in Table S4. Most of DEGs enriched photosynthesis, porphyrin and chlorophyll metabolism (ko00860), and photosynthesis—antenna proteins (ko00196) were same as those DEGs in GO functional enrichment. Especially, chlorophyll breakdown gene MCS/SGR/NYC1 (CL5222.Contig3_All, CL3099.Contig4_All and CL3709.Contig1_All) highly expressed in 94-GG suggested fast chlorophyll degradation was in green leaf.

A total of 13 and 24 DEGs related to phenylpropanoid biosynthesis (ko00940), and flavonoid biosynthesis (ko00941) were screened out in KEGG enrichment, respectively. They were took part in perilla leaf color formation. The PAL (CL13983.Contig3_All), C4H (Unigene27773_All, Unigene14839_All, CL6772.Contig2_All), C3H (CL11971.Contig9_All) HCT (CL13936.Contig3_All, Unigene10775_All, and Unigene5083_All) and COMT (Unigene17835_All) all showed up-down expression pattern in 94-GG, 272-GLP, 103-GPP and 357-PP (Fig. 8C, Table S4). Much more expression of PAL, C4H, C3H, and HCT would be caused much more caffeic acid synthesis in green and front green and back purple leaves in 94-GG, 272-GLP, and 103-PP, compared with that of only purple leaf 357-PP. Much more caffeic acid synthesis might be promoted rosomarinic acid synthesis. Rosomarinic acid synthesis gene, RAS (CL13662.Contig2_All) also showed high expression in 272-GLP and 94-GG (Fig. 8C; Table S4). Many CCR (CL14054.Contig2_All, CL1293.Contig1_All), 4CL (CL12689.Contig5_All, Unigene15858_All, Unigene9488_All, and CL3950.Contig8_All) unigenes showed up-regulation in purple cultivar ‘M357’. At the same time, the key anthocyanins synthesized genes CHS (Unigene15383_All) and CHI (CL3453.Contig1_All) all highly expressed in 357-PP.

As a whole, the genes related to flavonoid 3'-hydroxylase (F3’H), flavonoid 3',5'-methyltransferase-like (F3’,5’H), dihydroflavonol 4-reductase (DFR), and anthocyanidin synthase/leucoanthocyanidin dioxygenase (LODX/ANS) showed specific expression in 103-GPP and 357-PP. Highly expressed of these above genes induced much more anthocyanins accumulation. Much more expression of F3’,5’H gene could be promoted delphinidin-3-glycoside accumulation and further cause leaf of 357-PP turning purple (Fig. 8C). Based on F3’H, F3H, and FSII gene expression, it was suggested that much more flavones, flavonals, and isoflavones might synthesized in green leaf, compared with that in the purple leaf (Fig. 8C, Table S4).

The hierarchical clustering analysis of key DEGs involved in chlorophyll metabolism and flavonoid biosynthesis

According to the results of venn diagrams and DEGs enriched GO function and KEGG pathway (Fig. S4, Fig. S9), all the genes referred to chlorophyll biosynthesis and metabolisms were searched. Meanwhile, the expression patterns of these genes according to FPKM of RNA-seq data in four perilla cultivars were analyzed, and the relationship between their expression patterns and leaf color formation was investigated. The 23 chlorophyll synthesis and degradation related genes (59 unigenes or contigs) were identified in our work (Table 4). For this study, without those genes with a FPKM value of less than 1, a total of 55 unigenes or contigs were different expressed and probably related to leaf color formation. The hierarchical clustering thermogram of 55 DEGs was shown in Fig. 6A. The different four cultivars could be easily separated by expression levels of 42 unigenes, and it was indicated that these genes may be regulated leaf coloration.

Table 4 The DEGs regulated chlorophyll metabolic pathway
Full size table
Fig. 6
figure 6

The heatmap of the relative expression levels of the top highly differentially expressed genes (DEGs) in chlorophyll (A) and flavonoid (B) metabolism. The color scale (0.0–31.75) represents the expression levels of DEGs. The hierarchical clustering of unigenes and samples is shown in the dendrograms on the top and side of the heat map using the complete linkage approach

Full size image

At the same time, the 20 flavonoid synthesis related genes were also analyzed according to different pairwise comparisons with different groups, and a total of 102 unigenes or contigs were identified (Table 5). Without those genes with a FPKM value of less than 1, the remaining 73 unigenes or contigs were different expressed in four perilla cultivars and they were probably related to perilla leaf color formation. Their hierarchical clustering thermogram was shown in Fig. 6B. According to heatmap cluster, the relationship of cultivar 272 and 94, and 103 and 357 was close, respectively (Fig. 6B). It was suggested that these flavonoid synthesis related genes were regulated perilla leaf color.

Table 5 DEGs in flavonoid biosynthesis pathway
Full size table

The important transcription factor genes involved in perilla leaf color regulation

Transcription factors (TFs) play an important role in plant development and secondary metabolism. In this paper, a total of 28 TF genes were identified and they have different expression patterns in four perilla cultivars. It includes one AP10/EREBP gene, one bZIP gene, one MADS box gene SVP, one TCP gene, four MYB gene, one zf-HD gene, two NAC21/22 genes, three WRKY genes, four WD40 genes, and seven bHLH genes. The AP10/EREBP, bHLH35-like, bHLH29, bHLH137-like, bZIP, SVP, MYB-like, MYB113, NAC21/22, WD40 repeat containing protein 23, U4/U6 small nuclear ribonucleoprotein PRP4-like protein, and autophagy-related protein 18b were up-regulated and the remained 14 TFs were down-regulated (Fig. 7, Table 6). The relationship of four cultivars constructed by heatmap was similar with that constructed by flavonoid. It was suggested that these TFs might control flavonoid accumulation, and effect leaf color formation. The bHLH35-like (CL1476.Contig1_All), and WD repeat-containing protein 43 (CL4426.Contig1_All) and zinc-finger homeodomain protein 3 (CL13045.Contig26_All) were specifically expressed in 272-GLP and 103-GPP, respectively. Previous study showed that MYB113 regulated perilla leaf color, and this TF was also identified in our work. Meanwhile, MYB61 (CL3951.Contig1_All) and MYB-like (CL12881.Contig13_All) showed same expression pattern as MYB113 viz. up-regulated in deep color leaves. TF genes including AP10/EREBP (CL1914.Contig26_All), bZIP (CL403.Contig9_All), SVP (CL11571.Contig4_All), NAC21/22 (Unigene3776_All and Unigene4273_All), and bHLH137-like (Unigene631) also highly expressed in purple leaf 357-PP (Fig. 7B). It was suggested that these TFs may regulated perilla leaf color formation.

Fig. 7
figure 7

The heatmap of the relative expression levels of the top highly differentially expressed genes (DEGs) (A), and expression patterns of the important DEGs according to FPKM analysis in four samples (B). The color scale (0.0–7.28) represents the expression levels of DEGs. The hierarchical clustering of unigenes and samples is shown in the dendrograms on the top and side of the heat map using the complete linkage approach

Full size image
Table 6 Different expressed transcript factors in four perilla cultivars
Full size table

Validation of candidate DEGs by qRT-PCR

In order to further verify the reliability of the RNA-Seq data, 14 genes from the DEGs related to the flavonoid synthesis and cholophyll metabolism were randomly selected, and validated them using qRT-PCR. The qRT-PCR results showed that most of these selected genes had similar expression patterns with those identified in the RNA sequencing data (Fig. 8), which indicated that the transcriptome sequencing data had high repeatability and accuracy. The average of these correlation values was 0.613, and it was indicated that there was a high correlation between the results of RNA-seq and qRT-PCR. Meanwhile, 42.86% of these selected genes had a correlation value of over 0.81. The correlation coefficient was more than 0.933 in flavonoid synthesis gene 4CL, F3H, and ANR. These results indicated that the relative levels of transcripts produced from the RNA-seq were in a good agreement with values obtained by qRT-PCR.

Fig. 8
figure 8

The expression patterns of the 14 randomly selected DEGs confirmation of RNA-seq data by qRT-PCR analysis

Full size image

Discussion

Perilla as herb and oil plant was widely used in East Asia. Due to much more leaf color variations, and it was also used as an ornamental plant. Investigation of molecular mechanism of leaf coloration might be applied for perilla breeding with desirable color. Previous studies showed that anthocyanin metabolism and related gene expression affected leaf color formation. However, only a few anthocyanin-related genes and transcription factors were identified in green and red perilla. The mechanism of leaf coloration should be further studied, especially, back leaf color variations. Generally, leaf color formation was determined three pigments including chlorophyll, carotenoid and flavonoid. However, the reported papers were not investigated the effect of chlorophyll, and carotenoid on leaf coloration. In our paper, four different perilla leaf colors were used as materials to compare the contents of three pigments and then identify leaf coloration genes by RNA-seq technology.

Difference analysis of three pigments in four perilla cultivars

Most of reports suggested that purple leaf has high contents of flavonoid/anthocyanin, and low content of carotenoid and chlorophyll (Lu et al. 2017; Shen et al. 2018). The red and green perilla forms were only regulated by anthocyanin production, but not that of other metabolites, such as inorganic anions, organic anions and amino acids (Yamazaki et al. 2003). However, it was found that both the total contents of chlorophyll and the content of chlorophyll a were all high in 357-PP, the completely purple leaf. Meanwhile the Pn was also high in M357-PP than those of the other three cultivars. The contents of anthocyanin and carotenoid were also largest in 357-PP, while anthocyanin was rich in 103-GPP with back purple leaf (Fig. 2). Thus, high contents of chlorophyll, carotenoid, and anthocyanin mixed together promote full purple leaf formation, while the content of anthocyanin was determined back leaf color formation. The correlation between the content of rosmarinic acid, and the total content of total anthocyanins was week (Deguchi and Ito 2020). The expression of rosmarinic acid synthase gene (CL13662.Contig2_All) was indeed low in high content of anthocyanins in green with dark purple and purple leaf in perilla. Green leaf might be accumulated much more rosmarinic acid in perilla. In fact, there was no difference of plant biological mass in four perilla cultivars (data was not shown). These results further suggested that perilla 357-PP with high Pn, and high contents of carotenoid, anthocyanin and flavonoid could be used as ornamental, edible, and medicinal plant.

Perilla leaf color regulated gene was identified by transcriptome sequencing

Saito et al. (1999) isolated anthocyanidin synthase (ANS) gene 2-oxoglutarate-dependent oxygenase and verified its expression in green and red leaf perilla, for the first time. Then, Sompornpailin et al. (2002) isolated the gene PFWD (contained a WD repeat) controlled anthocyanin biosynthesis in perilla. Differential display of mRNA technology had been used to identify several genes encoding anthocyanin-biosynthetic enzymes and presumptive regulatory proteins (Yamazaki et al. 2003, 2008). Recently, transcriptome sequencing technology as a convenient tool was used to identify interested genes (Wang et al. 2018, 2019; Gao et al. 2020). In perilla, only a few papers were used RNA-seq to identify leaf color regulated genes, and the genes controlled back leaf color formations are still unknown.

Tong et al. (2015) used de novo assembly transcriptome sequencing generated 48,009 contigs with average length of 873 bp and a N50 length of 904 bp and identified the DEGs between P. frutescens (L.) var. frutescens and var. crispa. The unigenes number and average length in our work are larger than those reported studies in perilla (Tong et al. 2015; Jiang et al. 2020). Higher expression of PfGST1 and PfCHI1 was found in red perilla (Tong et al. 2015), and high expression of CHS gene enhanced the accumulation of flavonoids in red perilla (Jiang et al. 2020). They also deduced transcription factors including MYB, Myc-like, AP2/ERFs, WRKY, bHLH, MADS, WD40, and NACs, and flavonoid synthesized related genes including PAL, C4H, CHS, F3H, F3’H, DFR, ANS, UFGT, and UGT75C1 involved in red leaf color formation (Tong et al. 2015; Jiang et al. 2020). The above flavonoid synthesized related genes together with flavonol synthase (FLS) were linked to fruit pigmentation in the olive and redness of leaf pigmentation in Acer rubrum (Chen et al. 2019; Iaria et al. 2016). AmDFR and AmPAL1 were significant genes involved in the anthocyanin metabolic pathway and regulated leaf color change in A. pictum subsp. Momo (Ge et al. 2019). In Paphiopedilum hirsutissimum, F3H and CHS controlled essentially important steps in anthocyanin biosynthesis pathway and regulated flower color variation (Liu et al. 2018). It was deduced that those above genes related to anthocyanin synthesis may be realized green and purple leaf coloration in perilla.

In our work, the expression levels of these above genes were also found up-/down- regulated. According to the expression patterns of flavonoid related genes, it was found that the five genes, F3’H, F3H, DFR, ANS, and F3’,5’H were the key genes control green and purple leaf coloration, and back leaf color change. Almost all of the five genes did not expressed in green leaf, and up-regulated expression during with back leaf color deepened. Especially, the gene F3’,5’H specifically expressed in purple leaf 357-PP and back purple 103-GPP was found for the first time in perilla green and purple leaf, and it is very important gene induced delphinidin-3 glycoside synthesis. Therefore, delphinidin-3 glycoside is played role in perilla purple leaf coloration. In water lily (Nymphaea colorata), F3’,5’H was controlled blue-petal formation (Zhang et al. 2020). This is further suggested that F3’,5’H controlled purple leaf color formation in perilla. As a whole, according to the expression levels of flavonoid related genes suggested that 94-GG, 272-GLP, and 103-GPP can also synthesize flavones and flavonols, while anthocyanin synthesis was hampered in green leaf 94-GG.

Except of anthocyanin biosynthetic genes, the chlorophyll metabolic genes including BoHEMA1, BoCRD1, BoPORC1, BoPORC2, BoCAO, and BoCLH1 were showed different expression levels in purple and white inner leaf pigmentation in ornamental cabbage (Brassica oleracea var. acephala) (Jin et al. 2018). Their results indicated that the purple inner leaves of purple ornamental cabbage result from a high level of anthocyanin biosynthesis, a high level of chlorophyll degradation and an extremely low level of chlorophyll biosynthesis (Jin et al. 2018). A total of 21 DEGs related to chlorophyll biosynthesis including GluTR/HemA, HemC/HMBS, ChlD, ChlH, POR, ChlG and CAO were up-regulated, and 10 to degradation containing MCS/SGR/NYC1, and PAO were down-regulated in green leaf color of Pennisetum setaceum ‘Rubrum’, compared with those in purple leaf (Zhu et al. 2020). Therefore, in the last stage of chlorophyll biosynthesis and chlorophyll degradation were essentially steps for leaf color formation (from green to purple). However, from purple color to green color change, up-regulated Glu-TR expression induced chlorophyll accumulation in bicolor leaf of ornamental kale (Ren et al. 2019). In different perilla leaves with four kinds of colors, the expression patterns of CHLM, CHL27, POR, DVR, CAO, and HCAR (in the last stage of chlorophyll biosynthesis), and NYC1, and PAO (chlorophyll degradation) were consistent with those of Pennisetum setaceum ‘Rubrum’. These genes were related to green and purple leaf color formation in perilla.

Carotenoids are one of the three largest classes of pigments and are also colorants that range from colorless to yellow, orange, and red, with variations reflected in many species (Chen et al. 2019). The contents of chlorophyll, anthocyanin, and carotenoid together determined leaf pigmentation in A. rubrum (Chen et al. 2019). Leaf pigmentation from green to red and then yellow, contents of carotenoids was down-regulated and the rate-limiting enzyme PSY were highly concentrated in yellow and green leaf. In perilla, PSY2 was highly expressed in green leaf 94-GG. However, DEGs, zeaxanthin epoxidase (ZEP) and violaxanthin de-epoxidase (VDE) at the downstream stage of carotenoid biosynthetic pathway were up- and down-regulated expression, respectively, in purple leaf 357-GG, and the two genes promoted violaxanthin accumulation. High quantities of carotenoids accumulated in 357-PP caused purple leaf color formation. The down-up-down expression patterns of 15-cis-zeta-carotene isomerase (ZIISO), beta-carotene isomerase D27, 15-cis-zeta-carotene isomerase, cytochrome P450 CYP97A41 (CYP97A) suggested that more lutein accumulation in green leaf 94-GG.

Transcription factors regulated perilla leaf coloration

Among various transcription factors, MYB and bHLH are most significant regulators of flower or leaf variations, and MYB and bHLH transcription factors were also identified in our work. R2R3-type MYB transcription factors forming MYB-bHLH-WD10 complex regulate anthocyanin biosynthesis (Liu et al. 2018). R2R3 MYB transcript factor, MYB113, and MYB110a controls purple spot formation at the base of flower petals, and flower petal color in tetraploid cotton, and kiwifruit, respectively (Zhang et al. 2021; Abid et al. 2022). Zhang et al. (2021) found MYB113 played the pivotal role in perilla leaf coloration. The MYB113 was also identified in our work, and MYB-like gene (CL12881.Contig13_All) had similar expression pattern with MYB113. It was deduced that MYB113 and MYB-like gene ((CL12881.Contig13_All) were two important MYB transcript factors controlled perilla leaf coloration. The yeast two-hybrid system showed PFWD strong interacted with MYB-like protein and finally, enhanced anthocyanin production in perilla (Sompornpailin et al. 2002). The hierarchical clustering thermogram showed that bHLH29 (CL3060.Contig3_All) and bHLH137-like (Unigene631_All) were clustered in the same branch with MYB113, and they may be formed MYB-bHLH-WD10 complex to regulated anthocyanin synthesis (Fig. 7).

In apple, the expression level of the NAC (no apical meristem) transcription factor, MdNAC52, which controlled MdMYB9 and MdMYB11 expression, was increased during apple coloration, and overexpressed MdNAC52 gene promoted anthocyanin accumulation (Sun et al. 2019). In our work, the NAC21/22 (Unigene3776_All) was also showed high expression in purple perilla leaf of 357-PP, and could be advanced anthocyanin accumulation in purple leaf. In kiwifruit, overexpression of the kiwifruit SVP3 gene interfered with the expression of MYB110a, which encodes the key anthocyanin biosynthetic step, F3GT1, and finally suppressed anthocyanin biosynthesis in petals (Wu et al. 2014). Unlike kiwifruit SVP3, the SVP gene may be a positive regulator of anthocyanin synthesis. The bZIP transcription factor HY5 could directly recognize and bind the promoter of chalcone synthase 1, chalcone synthase 2, and dihydroflavonol 4-reductase to induce anthocyanin biosynthesis in tomato (Liu et al. 2018). The bZIP also showed high expression in purple perilla leaves, and it might be promoted anthocyanin accumulation. The PbWRKY75 promotes anthocyanin synthesis by activating PbDFR, PbUFGT, and PbMYB10a in pear (Cong et al. 2018). Like pear WRKY75, WRKY1 (CL6289.Contig1_All) and WRKY33 (CL12462.Contig1_All) all advanced anthocyanin synthesis. TCP15 attenuates anthocyanin accumulation after short time exposition to high light intensity (Ivana et al. 2016), which was consistent with our result. Thus, TCP15 was another important transcription factor in regulating perilla leaf coloration. According to expression pattern of AP10/EREBP, it was a positive regulator of anthocyanin synthesis.

The probable regulated scheme of perilla leaf coloration

The completely purple leaf color formation was determined by the high contents of chlorophyll, carotenoid, flavonoid, and anthocyanin, while different deepened back leaf color was caused by anthocyanin accumulation (Fig. 9). The contents of these pigments were caused by expression levels of these important genes of pigment synthesis and metabolism. In purple leaf, chlorophyll synthesis was lower than that of green leaf. However, down-regulated expression of MCS/SGR/NYC1NYC1 may be delayed initial cholorophyll degradation. The initially high content of chlorophyll kept high photosynthesis rate of 357-PP. However, chlorophyll degradation was fast throughout up-regulation of chlorophyll degradation genes PPH, PAO, and RCCR at last. The chlorophyll degradation was unmasked purple color of 357-PP. The actively accumulated carotenoid by high expression of the regulator gene ZEP may be took part in light harvesting and leaf coloration in 357-PP. Much more expression of F3’H, F3’,5’H, F3H, DFR, and LDOX/ANS promoted anthocyanin accumulation, and induced purple leaf color and back leaf purple color formation. Transcription factor genes, MYB113, MYB-like, bHLH29, bHLH137-like, NAC21/22, WRKY1, WRKY33, SVP, and bZIP, and TCP15 played roles in anthocyanin accumulation, together with the above flavonoid synthesized related genes. According to previous study, it was deduced that TCP15 negatively regulated anthocyanin accumulation, and leaf prolonged light exposure affected expression level of TCP15, and finally caused the lower surface of perilla leaf purple color formation.

Fig. 9
figure 9

The postulate scheme of leaf color variations regulated by anthocyanin, chlorophyll, and carotenoid, in perilla. Heatmaps were generated using long2FC based on FPKM values and the color scale (0.0–233.13) represents the expression levels of DEGs

Full size image

Conclusions

In this study, comparative transcriptome was used to explore genes regulated green, purple and biocolor leaf in perilla. The differences of chlorophyll synthesis, and degradation, carotenoid accumulation, and anthocyanin synthesis in leaves with different colors caused perilla leaf coloration. These genes CHLM, CHL27, POR, DVR, CAO, NYC1, PPH, PAO, and RCCR controlled chlorophyll synthesis and degradation. The up- and down-regulated ZEP and VDE expression at the downstream stage of carotenoid biosynthetic pathway induced xanthoxin accumulation. The up-regulation of anthocyanin biosynthesis genes F3’H, F3’,5’H, F3H, DFR, and LDOX/ANS is probably leading to perllia full purple leaf and back leaf coloration, while transcription factors MYB113, MYB-like, bHLH29, bHLH137-like, NAC21/22, WRKY1, WRKY33, SVP, and bZIP, and TCP15 together with the above genes might positively and negatively regulated anthocyanin synthesis, respectively. Further functional characterization of these identified genes could be further deciphered the mechanism of perilla leaf coloration.