Integrated Phenotypic Physiology and Transcriptome Analysis Revealed the Molecular Genetic Basis of Anthocyanin Accumulation in Purple Pak-Choi

Abstract

Purple Pak-choi is rich in anthocyanins, which have both ornamental and edible health functions, and has been used more and more widely in facility cultivation. In order to further clarify the molecular mechanism of purple Pak-choi, two Pak-choi inbred lines (‘PQC’ and ‘HYYTC’) were selected for the determination of pigment content and transcriptome analysis, and the key genes controlling the formation of purple character in leaves of Pak-choi were discovered. The results of pigment determination showed that the anthocyanin content of ‘PQC’ was 0.29 mg/g, which was about 100 times than ‘HYYTC’; The chlorophyll content of ‘HYYTC’ was 2.25 mg/g, while ‘PQC’ only contained 1.05 mg/g. A total of 20 structural genes related to anthocyanin biosynthesis and 28 transcriptional regulatory genes were identified by transcriptome analysis. Weighted gene co-expression network analysis (WGCNA) was used to construct the weight network analysis map of 14 genes. The results showed that the cinnamate hydroxylase gene (BraA04002213, BrC4H3), flavanone-3- hydroxylase (BraA09004531, BrF3H1), and chalcone synthetase (BraA10002265, BrCHS1) were the core genes involved in the anthocyanin synthesis pathway of purple Pak-choi. The results identified the key genes controlling the formation of purple leaf traits, which laid a foundation for further analysis of the molecular mechanism of anthocyanin accumulation in purple Pak-choi and provided a theoretical basis for leaf color regulation.

1. Introduction

Pak-choi (Brassica campestris ssp. Chinensis) is an important cruciferous leaf vegetable crop, which is widely cultivated in the middle and lower reaches of the Yangtze River in China, because of its rich germplasm resources, short growth cycle, and easy cultivation. As a main type of Pak-choi, purple Pak-choi is rich in anthocyanins and has the functions of both ornamental and edible health care, it is increasingly favored by people. However, in the production and cultivation, the purple character is often affected by external environment such as temperature and light, resulting in uneven leaf coloring and dull color, which is mainly due to the influence of environmental conditions on the expression of key genes of anthocyanin metabolism pathway [1,2,3,4]. Effective leaf color regulation was performed on the DFR gene of sweet potato and the PAP1 gene of tobacco by means of RNAi and gene overexpression, respectively, indicating that leaf color metabolic engineering improvement through the regulation of key anthocyanin genes is feasible [5].

Anthocyanin is an important category of secondary metabolic substances in plants, belonging to a kind of flavonoid [6]. Anthocyanin biosynthesis originated from the branch of the flavonoid pathway. These enzymes are encoded by biosynthetic structural genes which include early biosynthetic genes (EBGs) and late biosynthetic genes (LBGs). The EDGs including CHS (encoding chalcone synthase), CHI (encoding chalcone isomerase), and F3H (encoding flavanone 3-hydroxylase) are involved in the production of precursors. The LBGs are involved in the production of colored anthocyanins, such as DFR (encoding dihydroflavonol 4-reductase) and UFGT (encoding UDP-glucose: flavonoid-3-O-glucosyltransferase). Although the anthocyanin biosynthesis pathway is relatively conserved and well-studied in model plants such as Arabidopsis, and many genes related to the synthesis pathway have been cloned, there are few tudies on the mining and expression of related genes in Chinese cabbage vegetables. Previous research was conducted for the transcriptomic analysis of green and purple Chinese cabbage in which 36 genes were screened for up-regulated expression [7]. Late anthocyanin synthesis genes DFR, ANS, and UFGT may play a more critical role in the anthocyanin biosynthesis of purple Pak-choi concluded by transcriptome analysis [8].

Despite some beneficial attempts made by predecessors, the regulatory mechanism and regulatory network of anthocyanin synthesis related to the formation of purple leaf traits in Pak-choi are still unclear, and the key genes controlling anthocyanin synthesis and accumulation in purple Pak-choi need to be further studied and confirmed. In this study, the purple leaf line ‘PQC’ and green leaf line ‘HYYTC’ were selected to analyze gene differential expression at the transcriptional level. Weighted gene co-expression network analysis (WGCNA) and weighted network analysis were used to identify gene modules closely related to anthocyanin synthesis. The aim was to dig out the key genes involved in anthocyanin biosynthesis and regulation of Pak-choi, and initially construct the anthocyanin molecular regulatory network related to purple leaf character formation, which laid a foundation for in-depth analysis of the molecular mechanism of anthocyanin accumulation in purple Pak-choi and provided a theoretical basis for the color regulation in facility cultivation.

2. Materials and Methods

2.1. Experimental Material

After seeding, the two materials ‘PQC’ and ‘HYYTC’ were placed in an incubator at 25 °C for germination promotion. After the germination, the seedlings were cultivated in plate substrate using Pindstrup Mosebrug A/S, Denmark. After sowing, they were placed in a light incubator. The following procedure was carried out: light culture at 22 °C for 12 h and dark culture at 18 °C for 12 h. The cotyledon extension stage (T1), cotyledon flattening stage (T2), and two true leaf stages (T3) were sampled, respectively. Cotyledon was selected at T1 and T2, and the second true leaf was selected as the sample at T3. Each sample was randomly collected from five seedlings, and six samples were collected. All samples were repeated three times, totaling eighteen samples. All samples were frozen in liquid nitrogen at −80 °C for RNA extraction (Figure 1).

2.2. Determination of Pigment

The two materials were sown separately in the plastic greenhouse and transplanted into the insect-proof net room at the seedling age of 30 d. After the transplantation for 45 d, the largest leaf free from disease and insect infestation was selected and placed in the refrigerator at 4 °C for subsequent tests.

Anthocyanin content determination: 0.1 g of leaf tissue was weighted and placed in Erlenmeyer flasks then 10 mL, 0.1 mol/L hydrochloric alcohol solution was added and soaked in a water bath at 60 °C for 30 min then poured into a 25 mL volumetric bottle, and the volume was fixed to 25 mL. With 0.1 mol/L hydrochloric ethanol solution as the control, the optical density of the extracted solution at 530 nm, 620 nm, and 650 nm was determined. The optical density of anthocyanins ODλ = (OD530 − OD620) − 0.1 (OD650 − OD620); Anthocyanin content (nmol/g) = ODλ/ε × V/m × 106, anthocyanin content (mg/g) = anthocyanin content (nmol/g) × M × 10⁻³ [9].

Determination of chlorophyll content: 0.2 g of leaf tissue was weighted into a mortar, 5 mL 95% ethanol was added into the mortar, the mixture was ground into a homogenate, filtered into a 25 mL volumetric bottle, and 95% ethanol was used to set the volume to 25 mL. The absorbance was measured by spectrophotometer at 649 nm and 665 nm with 95% ethanol as the control. Chlorophyll a content Ca = 13.95A665 − 6.88A649, chlorophyll b content Cb = 24.96A649 − 7.32A665. Chlorophyll content (mg/g) = (C × V × N)/(W × 1000) [10].

2.3. RNA Extraction, cDNA Library Construction and Quality Control

Trizol kit (Invitrogen Company, Waltham, MA, USA) was used to extract total RNA from samples, and Qiaquick kit (Qiagen, Shanghai, China) was used to purify cDNA. RNA concentration and integrity were evaluated using a NanoDrop 2000 (Thermo Fisher Scientific, Waltham, MA, USA) spectrophotometer and polyacrylamide gel electrophoresis, respectively. mRNA was divided into fragments by fragmentation buffer and amplified by PCR to build a fragment sequencing library. The library was sequenced using Illumina Hi-Seq™ 2000. The original image data files obtained by Illumina Hi-Seq™ 2000 were converted into raw reads by base recognition analysis. Clean reads were obtained from the sequenced raw reads after the removal of splices and low-quality duplicate reads.

2.4. Screening Differentially Expressed Genes

Gene expression was obtained by calculating FPKM [11]. In the screening process, FC (log2 fold change) > 2 and FDR (false discovery rate) < 0.05 were used as the selection criteria for differential genes. Differential expressions between samples were analyzed by DEG-Seq [12].

2.5. Enrichment Analysis of Differentially Expressed Genes GO and KEGG

The GO annotation was carried out by a specific perl script using the TopGo R 2.5 software package to perform Fisher significance enrichment tests with the assembled transcript as the background. Kobas 3.0 software was used for KEGG enrichment analysis, and the R 2.10 software package ggplot was used for visualization of the enrichment analysis point plot.

2.6. Functional Annotation of Differentially Expressed Genes

The obtained clean reads were compared to the Chinese cabbage reference genome (http://brassicadb.cn/#/Download/, accessed on 6 May 2023). Differentially expressed genes were identified by BLAST (BLAST: Basic Local Alignment Search Tool nih.gov, accessed on 6 May 2023). The sequence was compared with the protein database to obtain annotation information. TMM was used to standardize read count data [13].

2.7. Weighted Gene Co-Expression Network Analysis

The association network of all differentially expressed genes was constructed using WGCNA in R language. The adjacency matrix is generated by calculating the correlation between all the genes, and the soft threshold β is chosen according to the scale-free topological criteria, using the adjacency matrix to calculate the topological overlap matrix (TOM). The dynamic cutting algorithm was used for gene clustering and module division, and the truncation height of 0.25 was used to merge the branches into the final module. The value of the module characteristic gene (ME) was calculated, and the correlation degree between the module and anthocyanin was estimated by ME value.

2.8. Quantitative RT-qPCR Verification

Nine differentially expressed anthocyanin structural genes and transcription factor genes (BrPAL, BrC4H3, BrF3H, BrCHS, BrCHI, BrDFR, BrUFGT, BrANS, BrMYB44, and BrTT8) were selected for RT-qPCR verification. The sequence of primers used for the above gene expression is shown in

Table 1. The primer sequences used for RT-qPCR.

No	Gene ID	Gene Name	Forward Primer (5′–3′)	Reverse Primer (5′–3′)
1	BraA04000661	BrPAL	AGCAACATAACCAAGATG	TCTCAGATTCTCCTCAAG
2	BraA04002213	BrC4H3	TGAGGAAACGCTTGCAGT	GGCCTGAGGATAGGGATG
3	BraA05002651	Br4CL4	ATCTTTCCTCGCCGTGGTTT	CTCCGGCGAAATCTTAGGCT
4	BraA09004531	BrF3H	ATTCATTGTCTCTAGTCATCTTC	CCGTGAGTAGTCTCTGTT
5	BraA10002265	BrCHS	TATCCTGACTACTACTTC	CTCCTTTAGAAACTCTTC
6	BraA03005399	BrANS	TCCTGATTCCATTGTGAT	TCCTAACCTTCTCCTTATTC
7	BraA06000554	BrUFGT	GTAATGTATCCGTGGTTAG	GGTAGAGGTTAAGAGGTT
8	BraA08002374	BrMYB44	TTATGAGACGGAGAATGT	TACCTCTTCCTTCCTAAC
9	BraA09002835	BrTT8	AGACGAAGAAGAAGTAGA	CCTCCATTAGATTCATCAT
10	-	BrActin	GTTGCTATCCAGGCTGTTC	AGCGTGAGGAAGAGCATAAC

. The qPCR amplification reaction system was based on the SYBR Premix Ex Taq kit with single-strand cDNA as the template. The system was 20.0 μL, including Taq 10.0 μL, template 2.0 μL, positive and negative primes 1.0 μL and ddH₂O 6.0 μL. The qPCR reaction condition was set at 95 °C for 5 min; 95 °C 10 s; 60 °C for 30 s; and 72 °C for 15 s, a total of 40 cycles. The 2^−△△Ct method was used for relative quantitative analysis of the data [14].

3. Results

3.1. Observation of Leaf Color

In the cotyledon flattening stage, the color of the cotyledon of ‘PQC’ began to appear, and the distribution was uneven. The purple pigment mainly accumulated on the true leaves at the stage of two true leaves. In the adult stage, the adult plants of ‘PQC’ are more colorful, with dark purple on the front of the leaves and slightly purple on the back, while the leaves of ‘HYYTC’ are green on the front and back of the leaves at all stages (Figure 2).

3.2. Determination of Pigment Content in Leaves

The results showed that the anthocyanin content of the two materials was significantly different. The anthocyanin content of ‘PQC’ was 0.29 mg/g, and that of ‘HYYTC’ was only 0.002 mg/g. The chlorophyll content in ‘HYYTC’ was as high as 2.25 mg/g, while the chlorophyll content in ‘PQC’ was significantly lower than that in ‘HYYTC’ (Figure 3).

3.3. Sequencing Results and Comparison Statistics

Total RNA from leaves was extracted and used to construct 18 cDNA libraries. A total of 155.04 Gb of Clean Data were obtained from 18 samples sequenced. After removing low-quality reads and connectors, 337,561,131 pairs of clean reads were obtained, and between 22 million and 33 million pairs of clean reads were obtained for each library. The Q30 of each sample used for assembly was higher than 92.71%, indicating that the sequencing results met the quality requirements for subsequent assembly analysis (Table S1). Using the V2.5 version of the Chinese cabbage genome as the reference genome, the comparison rate between the reads of each sample and the reference genome ranged from 87.48 to 90.29%, among which the ratio of ‘PQC’ was 89.22–90.04%, and that of ‘HYYTC’ was 87.48–90.29% (

Table 2. Overview of the transcriptome sequencing.

Samples	Clean Reads	Clean Bases	GC Content (%)	≥Q30 (%)
PQC-T1-1	26,549,215	7,928,202,428	48.92%	92.83%
PQC-T1-2	26,180,713	7,829,647,066	48.76%	93.19%
PQC-T1-3	34,780,519	10,381,289,642	48.64%	93.28%
PQC-T2-1	27,200,992	8,115,863,242	48.72%	93.43%
PQC-T2-2	26,021,139	7,764,834,504	48.67%	93.38%
PQC-T2-3	22,706,919	6,782,398,766	48.59%	93.38%
PQC-T3-1	22,221,865	6,636,130,982	48.23%	92.71%
PQC-T3-2	27,585,040	8,220,152,954	48.90%	93.62%
PQC-T3-3	28,884,210	8,615,366,610	48.63%	93.46%
HYYTC-T1-1	33,792,919	10,064,723,406	48.75%	93.10%
HYYTC-T1-2	27,771,004	8,285,974,048	48.88%	93.08%
HYYTC-T1-3	32,973,684	9,829,969,670	49.10%	92.72%
HYYTC-T2-1	27,844,329	8,323,582,204	48.66%	93.43%
HYYTC-T2-2	31,916,361	9,526,775,132	48.57%	93.13%
HYYTC-T2-3	32,197,152	9,602,692,572	49.28%	93.10%
HYYTC-T3-1	29,935,983	8,932,854,782	48.45%	93.06%
HYYTC-T3-2	28,258,874	8,433,086,514	48.74%	93.16%
HYYTC-T3-3	32,788,267	9,761,637,592	48.49%	93.07%

).

3.4. Identification of Differentially Expressed Genes

By comparing the FPKM values of ‘PQC’ and ‘HYYTC’ in three different leaf growth stages, the expression patterns of differential genes (DEGs) were determined. 2632 (up-regulation 1098, down-regulation 1534), 3147 (up-regulation 1427, down-regulation 1720), and 1920 (up-regulation 697, down-regulation 1223) DEGs were identified in the three comparison groups of HYYTC-T1 vs. PQC-T1 (G1), HYYTC-T2 vs. PQC-T2 (G2), and HYYTC-T3 vs. PQC-T3 (G3), respectively (Figure 4A). Through Wayne plot analysis, out of a total of 5070 DEGs, a total of 738 (308 up-regulation, 427 down-regulation) were identified. Three of the genes had inconsistent expression patterns in each of the three comparison groups, which may be the result of specific expression at different reproductive periods (Figure 4B–D).

3.5. Clustering Analysis

A cluster analysis of 738 gene expression profiles from purple material ‘PQC’ and green material ‘HYYTC’ at three different periods showed that compared with green material, the up-regulated gene expression of purple material was less than that of down-regulated gene expression (Figure 5A). According to trend expression analysis, DEG was divided into nine gene clusters, with significant differences in gene expression between cluster 0 and cluster 8 (Figure 5B). In cluster 0, there are 264 genes in total. With the growth process, the expression of related genes in purple material ‘PQC’ is higher than that in green material ‘HYYTC’, and these genes are mainly concentrated in transcription, ribosomal structure, and biotransformation (BraA03005834, BraA04002715, and BraA05001180). Amino acid transport and metabolism (BraA09005771, BraA05000655, and BraA06001513) and post-translational modification, protein turnover, companion (BraA01001595, BraA01003552 and BraA05001325). Cluster 8 contains a total of 400 genes. With the development of the growth period, the green material ‘HYYTC’ showed a higher transcription level than the purple material ‘PQC’. These genes are mainly involved in post-translational modification, protein turnover, chaperones (BraA01002075, BraA03000073, and BraA08002394), and signal transduction mechanisms (BraA05004531, BraA07002138 and BraA09003227). In addition, carbohydrate transport and metabolism (BraA02003506, BraA03000585, and BraA03000676) were found to be highly expressed in ‘HYYTC’. Other DEGs are grouped into clusters 1–7 according to their expression patterns.

3.6. GO Analysis and KEGG Enrichment Pathway Analysis

738 common DEGs were enriched in the single-organism biosynthesis process (GO:0044711), organonitrogen compound metabolic process (GO:1901564), and phosphorus metabolic process (GO: 0006793) (Figure 6). In addition, we found that some genes were also enriched in the anthocyanin-containing compound biosynthetic process (GO:009718, GO:0046283), flavonoid metabolic process (GO:0009812), and flavonoid biosynthetic process (GO:0009812); these results indicate that the purple formation of purple Pak-choi may be related to some genes of the specific GO pathway.

KEGG pathways in G1, G2, and G3 were analyzed. There were 528, 694, and 386 DEGs annotated by KEGG pathways, while 87 pathway classes were enriched from 738 common DEGs. As shown in Figure 7, the flavonoid biosynthesis pathway (ko00941), anthocyanin biosynthesis (ko00942), phenylalanine metabolism (ko00940), and glutathione metabolism (ko00480) were the four pathways with the most significant differences. Song (2020) pointed out anthocyanin accumulation is the main reason for the formation of the purple leaf of Pak-choi, and the above four pathways are closely related to anthocyanin synthesis [15]. There were 6 DEGs related to the phenylalanine metabolic pathway, 10 DEGs related to flavonoid biosynthesis, 2 DEGs related to anthocyanin biosynthesis, and 8 DEGs related to glutathione metabolism.

The phenylalanine biosynthesis pathway is the precursor of the anthocyanin metabolism pathway. Phenylalanine synthetase gene BraA04000661 (BrPAL2.2) and Cinnamate hydroxylase genes BraA04002213 (BrC4H3) and BraA03001710 (BrC4H5) were up-regulated in these three comparison groups. These differential genes may promote anthocyanin synthesis and accumulation by producing more anthocyanin biosynthesis precursors. Although the expression of the ligase gene BraA05002651 (Br4CL4), which is used to synthesize 4-coumaryl CoA, is down-regulated in purple materials, BraA05001886 (Br4CL1), BraA05002646 (Br4CL2), and BraA07003006 (Br4CL3) were expressed normally in the two materials. The expressions of early anthocyanin biosynthesis genes such as hydroxylase gene BraA09004531 (BrF3H), chalcone isomase gene BraA09004891 (BrCHI1), and chalcone synthase gene (BraA10002265, BrCHS1; BraA03000633, BrCHS2) were up-regulated in the three phases of purple material. Late anthocyanin biosynthesis genes such as anthocyanin dioxygenase genes (BraA0100144, BrANS1; BraA03005399, BrANS2), the dihydroflavonol-4-reductase gene BraA09002044 (BrDFR), and the flavonoid 3-O-glucosidtransferase gene (BraA10000963, BrUGT79B1.1; BraA06000554, BrUGT79B1.2) were significantly up-regulated in the three comparison groups. The expression patterns of anthocyanin transport genes such as BraA01003440 (BrTT12), BraA10002029 (BrTT19.1), BraA02000754 (BrTT19.2), BraA08004000 (BrGSTF6), BraA09000406 (Br5MAT), and BraA08003959 (Br3AT1) are all up-regulated (

Table 3. Identification of structure genes and transportation genes of anthocyanin synthesis.

Gene ID	FPKM						HYYTC -T1 Vs PQC -T1	HYYTC -T2 Vs PQC -T2	HYYTC -T3 Vs PQC -T3	Gene Annotation
	PQC -T1	PQC -T2	PQC -T3	HYYTC -T1	HYYTC -T2	HYYTC -T3
BraA04000661	30.43	42.55	137.89	3.74	11.90	23.96	Up	Up	Up	BrPAL2.2
BraA04002213	11.58	11.90	68.38	4.13	5.83	24.98	Up	Up	Up	BrC4H3
BraA03001710	33.88	36.42	164.03	2.87	4.41	11.99	Up	Up	Up	BrC4H5
BraA05002651	1.17	1.28	3.43	12.08	10.99	17.32	Down	Down	Down	Br4CL4
BraA08002920	25.22	12.86	25.15	8.83	5.78	5.52	Up	Up	Up	BrCCoAOMT
BraA09004531	108.59	85.32	306.09	18.84	23.84	96.13	Up	Up	Up	BrF3H1
BraA10002265	55.68	43.29	299.84	8.87	9.53	108.21	Up	Up	Up	BrCHS1
BraA03000633	14.07	14.17	139.38	0.70	1.72	14.58	Up	Up	Up	BrCHS2
BraA09004891	10.96	18.55	86.07	1.42	1.94	8.72	Up	Up	Up	BrCHI1
BraA09002044	160.86	115.98	443.50	5.45	7.75	7.46	Up	Up	Up	BrDFR
BraA01001444	109.91	85.16	393.62	3.57	5.76	36.51	Up	Up	Up	BrANS1
BraA03005399	12.68	13.28	60.08	0.99	0.76	0.81	Up	Up	Up	BrANS2
BraA10000963	21.80	16.19	50.35	0.60	0.66	1.34	Up	Up	Up	BrUGT79B1.1
BraA06000554	55.17	40.02	132.54	1.41	1.96	4.30	Up	Up	Up	BrUGT79B1.2
BraA01003440	17.90	21.31	39.02	1.80	3.30	0.94	Up	Up	Up	BrTT12
BraA02000754	83.23	70.92	205.42	3.90	3.33	23.59	Up	Up	Up	BrTT19.2
BraA10002029	22.01	19.89	80.82	0.44	0.88	0.56	Up	Up	Up	BrTT19.1
BraA08004000	4.66	8.18	34.57	0.48	0.51	0.30	Up	Up	Up	BrGSTF6
BraA08003959	59.03	38.90	149.81	1.72	1.26	4.69	Up	Up	Up	Br3AT1
BraA09000406	26.01	17.14	74.53	0.50	0.59	0.38	Up	Up	Up	Br5MAT

).

3.7. Transcription Factor Identification

Several transcription factors (TFs) have been found to be closely related to anthocyanin synthesis in Arabidopsis. In the previous analysis, we found that some DEGs belong to genes that encode transcription factors. To better understand which transcription factors are involved in regulating anthocyanin biosynthesis, the differentially expressed transcription factors were analyzed in this study. By comparing 738 DEG sequences with PlantTFDB, a total of 28 transcription factors belonging to 18 transcription factor families were identified. Among them, MYB-related, bHLH, and NAC types are the main transcription factor families, each containing three or more DEGs (

Table 4. Identification of transcription factors.

Gene ID	FPKM						HYYTC -T1 Vs PQC -T1	HYYTC -T2 Vs PQC -T2	HYYTC -T3 Vs PQC -T3	Gene Annotation
	PQC -T1	PQC -T2	PQC -T3	HYYTC -T1	HYYTC -T2	HYYTC -T3
BraA08002374	1.43	2.48	2.02	2.68	6.98	6.28	Down	Down	Down	MYB44
BraA05004112	0.39	1.47	17.01	2.41	5.91	36.21	Down	Down	Down	REVEILLE 8
BraA05001641	0.33	1.02	0.29	13.47	10.87	4.40	Down	Down	Down	REVEILLE 2
BraA02002130	1.39	0.77	1.86	3.33	5.06	6.84	Down	Down	Down	MYB1R1
BraA05003486	4.19	7.22	4.97	1.55	1.81	0.73	Up	Up	Up	ABCG29
BraA10003119	3.05	1.97	2.87	0.04	0.15	0.29	Up	Up	Up	FLA10
BraA08002359	1.83	10.69	2.05	0.60	0.90	0.24	Up	Up	Up	GARP-G2-like
BraA01004300	2.70	3.61	3.35	13.46	12.6	9.96	Down	Down	Down	TRIHELIX
BraA09002835	2.27	2.91	11.05	0.55	0.75	1.45	Up	Up	Up	TT8
BraA01002210	10.20	15.05	11.72	2.97	4.44	2.44	Up	Up	Up	bHLH3
BraA03006442	0.20	0	0.11	21.55	12.22	11.01	Down	Down	Down	BEE 2
BraA09000011	0.33	0.19	0.78	1.13	1.77	4.05	Down	Down	Down	UNE10
BraA07001006	0	0	0	1.48	1.49	2.34	Down	Down	Down	C3H
BraA07003043	0.79	0.77	1.95	3.21	5.40	7.33	Down	Down	Down	C2H2
BraA02000673	0.19	1.09	7.61	1.70	3.55	22.18	Down	Down	Down	C2C2-CO-like
BraA04002035	1.51	3.18	4.07	0.44	0.13	1.04	Up	Up	Up	C2H2
BraA09002226	0	0	0	28.04	24.80	28.55	Down	Down	Down	NAC69
BraA03000407	0	0	0	5.17	5.02	5.83	Down	Down	Down	NAC82
BraA07000730	1.44	1.05	1.26	9.42	6.79	5.31	Down	Down	Down	NAC101
BraA02004460	12.68	11.73	19.29	4.94	4.55	7.41	Up	Up	Up	MADS-MIKC
BraA10001525	4.74	9.59	5.03	39.14	47.88	16.02	Down	Down	Down	AP2/ERF-ERF
BraA05004664	3.71	3.00	4.79	14.32	12.83	15.55	Down	Down	Down	TCP
BraA09002413	0.31	0.12	0.73	2.15	2.51	2.33	Down	Down	Down	NF-YB
BraA08001844	1.12	7.81	1.11	3.22	20.05	9.09	Down	Down	Down	WRKY18
BraA03002344	0.20	0.07	0.03	7.21	7.87	2.27	Down	Down	Down	bZIP34
BraA09005994	0.02	0	0.14	2.45	1.69	2.99	Down	Down	Down	BES1
BraA04002319	3.07	7.84	71.47	23.13	41.41	149.16	Down	Down	Down	DBB
BraA08002660	0.17	0.05	0.37	0.94	0.72	1.68	Down	Down	Down	HB-HD-ZIP

).

For the three types of transcription factors, MYB, bHLH, and WD40, a total of eight MYBs and three bHLHs were detected in this study. Of the eight DEGs that code MYB, BraA08002374 (MYB44), BraA05004112 (REVEILLE 8), BraA05001641 (REVEILLE 2), BraA02002130 (MYB1R1), and BraA01004300 (TRIHELIX) were down-regulated in the three comparison groups, while BraA05003486 (FLA10), BraA10003119 (ABCG29), and BraA08002359 (GARP-G2-like) were the opposite. Among the three DEGs encoding bHLH, BraA09002835 (TT8) and BraA01002210 (bHLH3) were up-regulated in the three comparison groups, while BraA03006442 (BEE 2) was the opposite. In addition, no differential genes annotated as the WD40 family were found.

In addition to MYB and bHLH transcription factors, 17 transcription factors were identified in this study. Only BraA02004460 (MADS) and BraA04002035 (C2H2) were up-regulated in three phases of purple material, while other genes were down-regulated. In particular, BraA03002344 (bZIP34) and BraA08001844 (WRKY18) are more down-regulated.

3.8. Identification of WGCNA Modules of Genes Related to Anthocyanin Metabolism

The weighted gene co-expression network analysis (WGCNA) is a method to analyze target genes at the network level, which mainly constructs weighted association networks with differentially expressed genes and further screens for target genes. In this study, eight modules were identified from the RNA-seq data (Figure 8A), and correlation analysis was conducted between the module feature genes of the eight modules and different samples. Correlation analysis between modules and traits showed that the absolute value of the correlation coefficient between module ‘MEblack’ and two true leaf stages of purple material (PQC-T3) was the highest (r = 0.98), and the correlation was significant p = 7 × 10⁻⁴, while the other modules were weakly correlated or not correlated with the treatment relationships (Figure 8B). Since purple material has the most purple leaf color and the highest anthocyanin content in the two true leaf stages, it is believed that the gene in the module “black” has a high correlation with the anthocyanin metabolism pathway. KEGG enrichment analysis was performed on 784 genes contained in module “Black”. The enrichment was mainly in physiological metabolism, and a total of 50 pathway categories were enriched, among which phenylalanine metabolism and flavonoid biosynthesis were the most important pathways, and 16 and 13 genes were enriched, respectively (Figure 9). In order to further investigate whether the 784 genes discovered by WGCNA share differential genes with the previous 738 genes, it was found through Venn diagram analysis that there were 76 common DEGs and these genes were up-regulated in three comparison groups (Table S2). Interestingly, 76 genes contain all previously identified anthocyanin synthesis structure genes and transport genes, including two transcription factors BraA09002835 (TT8) and BraA02004460 (MADS-MIKC). In addition, 784 genes were newly identified that may be involved in anthocyanin biosynthesis. Genes such as BraA04002675 (BrPAL1), BraA09004715 (BrPAL2.1), BraA02000559 (BrCHS3), BraA05001212 (Peroxidase 19), BraA10000387 (Peroxidase 3), BraA10002815 (Cytochrome P450), and BraA07003905 (BrCOMT) were up-regulated in the two comparison groups (G2 and G3). The anthocyanin accumulation-related gene enriched in module ‘black’ was identified by WGCNA, which confirms the accuracy of WGCNA data analysis.

In order to further evaluate the importance of 76 shared genes in the anthocyanin metabolism pathway, association nodes with connectivity (weight value) >0.60 were selected. Import data into Cytoscape software (version 3.10.2), co-expression network construction, and core gene mining were carried out, and the weight network analysis diagram of 14 genes was constructed. Cinnamate hydroxylase gene (BraA04002213, BrC4H3), flavanone-3-hydroxylase (BraA09004531, BrF3H1), and Chalcone synthetase (BraA10002265, BrCHS1) were selected as the core genes of anthocyanin biosynthesis pathway of purple Pak-choi (Figure 10).

3.9. RT-qPCR Verification

To verify the results of RNA-seq sequencing, nine differentially expressed genes related to anthocyanin biosynthesis and regulation were selected for quantitative RT-qPCR detection in purple material ‘PQC’ and green material ‘HYYTC’. The quantitative RT-qPCR results of the nine genes were similar to the expression patterns of transcriptome data. The results of transcriptome analysis were reliable (Figure 11).

4. Discussion

4.1. Structural Genes and Transport Genes Affecting Anthocyanin Synthesis in Purple Pak-Choi

A previous study showed the anthocyanin synthesis structural genes (BrPAL1.2, BrPAL2.2, BrC4H5, BrFLS1, BrFLS3, BrDFR, BrANS1, BrANS2, BrCOMT1, BrCCoAOMT, BrUGT75C1, Brugt79B1.1, and BrUGT79B1.2) and anthocyanin transport genes (BrTT12, BrTT19.1, BrTT19.2, BrATPase 1, and BrATPase 10) were differentially expressed in purple Pak-choi variety ‘Zi Zuan’ and green Pak-choi variety ‘Jing Guan’ in their five leaf stage [8]. However, in our study, BrFLS3, BrCOMT1, BrUGT75C1, and BrATPase 1 were not differentially expressed, and BrFLS1 was not even expressed in T1, T2, and T3. Meanwhile, eight differentially expressed genes, including early anthocyanin synthesis genes (BrC4H3, BrF3H1, BrCHI1, BrCHS1, and BrCHS2) and anthocyanin transport genes (BrGSTF6, Br5MAT, and Br3AT1), were newly identified. Although the expression of BraA05002651 (Br4CL4) is down-regulated in purple materials, there was no differential expression of Br4CL1, Br4CL2, and Br4CL3 in the two materials. BraA05001886 (Br4CL1), BraA05002646 (Br4CL2), and BraA07003006 (Br4CL3) were expressed normally in the two materials. This indicates that the Br4CL4 gene is not the rate-limiting enzyme gene for anthocyanin accumulation in purple materials. The stable expression of other genes can provide sufficient substrates for the synthesis of subsequent anthocyanin precursors. On this basis, the cinnamate hydroxylase gene (BraA04002213, BrC4H3), flavanone-3-hydroxylase gene (BraA09004531 and BrF3H1), and chalcone synthetase gene (BraA10002265 and BrCHS1) are the core genes in the module significantly related to anthocyanin biosynthesis by WGCNA and weight network analysis. BraA10002265 was a candidate gene for the purple-red trait gene in purple-red Chinese cabbage by transcriptome sequencing and gene mapping [7]. In four purple Pak-choi materials with different anthocyanin content, the expression of the above two genes was significantly correlated with anthocyanin content. The correlation coefficients between gene expression of BraA09004531, BraA10002265, and anthocyanin content reached 0.9234 and 0.6904, which further indicated that BrF3H1 plays a more critical role in anthocyanin metabolism [15].

4.2. Transcription Factors Affecting Anthocyanin Accumulation in Purple Pak-Choi

Five transcription factors related to anthocyanin synthesis and regulation were found in Pak-choi, including MYB08, MYB12, CPC, TT8, and LBD39 in their five-leaf stage [8]. But only BraA09002835 (BrTT8) was identified in this study, and the other four genes were not differentially expressed in three treatments. BrTT8 was significantly up-regulated in purple Pak-choi material ‘Zi He’, and the allogenic expression promoted the transcription of some anthocyanin biosynthesis genes in tomato regenerated buds, further proving that BrTT8 played an important role in anthocyanin biosynthesis [16]. bHLH often forms complexes with MYB and WD40 to coordinate the biosynthesis of anthocyanins, among the 8 MYB-related DEGs identified, especially the expression level of BraA08002374 (BrMYB44) was significantly down-regulated in purple Pak-choi. The increase in the StMYB44 expression level may down-regulate the expression of DFR promoter activity, suggesting that the transcription level of BrMYB44 may play a similar role in the anthocyanin metabolism pathway of Pak-choi. Therefore, whether BrTT8 and BrMYB44 form a ternary complex to jointly regulate anthocyanin metabolism is worthy of further study.

In addition to being directly controlled by MBW, it has been reported that anthocyanin synthesis is also regulated by other transcription factors and regulatory genes, such as PIF3, HY5, COP1, WRKY, WIP, MADS, JAZ, SPL, and bZIP domains, which affect anthocyanin biosynthesis [17,18]. In this study, with the exception of BraA02004460 (MADS) and BraA04002035 (C2H2), the remaining transcription factors were identified as negative regulatory factors, and their expression was inhibited in purple materials. In particular, BraA03002344 (bZIP34) and BraA08001844 (WRKY18) were less expressed in purple materials. Studies have shown that the bZIP family acts primarily as positive regulators of anthocyanin biosynthesis. But in Tu’s report, it was mentioned that that knocking out one allele of VvbZIP36 promotes the accumulation of anthocyanins in grapevine [19]. Additionally, BraA03002344 (bZIP34) may be involved in anthocyanin biosynthesis as a negative regulatory gene. WRKY is one of the most characteristic of plant TF, which regulates various plant processes related to development, physiology, metabolism, and plant defense. The overexpression of MdWRKY11 in apple callus could significantly promote anthocyanin accumulation [20]. However, there are currently no studies on WRKY18 in anthocyanins accumulation, which is worth investigating. This study shows that all of the above genes are potential transcription factors regulating anthocyanin biosynthesis.

5. Conclusions

In this study, via RNA-seq analysis of two Pak-choi materials with different anthocyanin contents, a total of 20 anthocyanin biosynthesis-related structural genes and 28 transcriptional regulatory genes were identified, and the weight network analysis of 14 genes was constructed. The results showed that the cinnamate hydroxylase gene (BraA04002213 and BrC4H3), flavanone-3-hydroxylase (BraA09004531 and BrF3H1), and chalcone synthetase (BraA10002265 and BrCHS1) were the core genes involved in the anthocyanin synthesis pathway of purple Pak-choi. The above results identified the key genes and metabolic pathways controlling the purple character of leaves, laid a foundation for further in-depth analysis of the molecular mechanism of anthocyanin accumulation in purple Pak-choi, and provided theoretical support for its application in facility cultivation and breeding.