Brown-seeded inbred B147 and ttg1 yellow-seeded inbred B80 mutants were used as plant materials, and their seeds were sown in a greenhouse in the winter of 2016 for vernalization. Seedlings were subsequently transplanted and transferred to a plastic shed in the spring of 2017 for artificial pollination during blossom in Yangling, Shaanxi province, China. Seeds of different development stages were collected at 10, 14, 18, 22, 26, 30, and 34 days after flowering (DAF), frozen in liquid nitrogen, and immediately stored in a −80°C freezer (Sanyo, Japan). Three biological replicates were analyzed for each sample.

Leaves from tobacco (Nicotiana benthamiana) and yellow-seeded inbred B80 B. rapa ttg1 mutant plants were used for transient expression. Tobacco and B. rapa seeds were sown separately in the culture soil after soaking for 24 h. Fifteen-day-old tobacco seedlings were transplanted into plastic bowls (10 cm ×10 cm) and placed in the culture room under a 16-h light/8-h dark photoperiod, 22°C temperature and 60% humidity. Fourteen-day-old B. rapa seedlings were used for the transient expression analysis.

Genomic DNA was extracted from fresh young leaves using a modified version of the cetyltrimethylammonium bromide method (Porebski et al., 1997). Total RNA extraction and cDNA synthesis were performed as previously described (Ren et al., 2017). Gene expression analysis was performed using quantitative real-time PCR (RT-qPCR), and a housekeeping gene encoding glyceraldehyde-3-phosphate dehydrogenase (GAPDH; GO0048316) was used as a reference gene (Li et al., 2016). The RT-qPCR reaction was performed in 20-µL volume in a Roche LightCycler 480 Real-Time PCR Detection System with three replications as previously described (Ren et al., 2017). Relative expression level of genes was calculated using the 2^−ΔΔCt method (Livak and Schmittgen, 2001). The specificity of primers ( Supplementary Table 1 ) used for the amplifications was confirmed by melting curve analysis.

Based on the location of the ATG initiation codon of each gene in the BRAD database, 1,500- to 2,000-bp sequences upstream the ATG were screened to design promoter cloning primer ( Supplementary Table 2 ) using Primer Premier 5.0. PCR amplification procedure was performed in a 25-µL volume according to previously described instructions. A tail A was added to the purified PCR amplification products, which were subsequently cloned to the pMD19-T simple vector and transformed to Escherichia coli strain DH5α for sequencing. Promoter function was predicted using PlantCARE database online tools. Visualization of the cis-acting elements was performed using the online software Gene Structure Display Server.

For the Y1H assays, promoters were inserted into the pAbAi vector, and, then, pAbAi-promoter recombination plasmids were transformed into the Y1H Gold yeast strain for aureobasidin A (AbA; Clontech, USA) concentration screening. Confirmed transformants were spread on SD/-Ura, SD/-Ura/AbA (150 ng/mL), and SD/-Ura/AbA (300 ng/mL) culture media, and the optimal AbA concentration was determined on the basis of colony growth.

The TTG1-ORF was integrated into the pGADT7 vector, and the pGADT7-TTG1 recombinant plasmid was transformed into a recombinant yeast strain containing pAbAi promoters. Recombinant transformants were spread on SD/-Leu/AbA culture medium, whereas SD/-Leu without AbA was used as a control. The interaction between BrTTG1 and its promoter was assessed on the basis of the colony growth. The combination of pGADT7-p53 + pAbAi-p53 was used as a positive control, and the combination of empty pGADT7 + pAbAi-promoters was used as a negative control. All primers used for recombinant plasmid construction in the Y1H assay are listed in Supplementary Table 3 .

For the dual-luciferase assays, BrTTG1-ORF and promoters were introduced into pGreenII 62-SK and pGreenII 0800-LUC vectors by homologous recombination. The recombinant plasmids were then separately transformed into GV3101. pGreenII 62-SK-BrTTG1 was used as an effector, and promoter-pGreenII 0800-LUC was used as a reporter. All primers used for the construction of recombinant plasmids in the dual-luciferase assay are listed in Supplementary Table 4 .

Six independent tobacco (Nicotiana benthamiana) leaves were transfected during the transient expression assays. The transfected plants were kept in the dark overnight and then transferred to the culture room for 2 days. Luciferase activity was evaluated using a TransDetect Double-Luciferase Reporter Assay Kit (Promega, USA), and the regulatory relationship between TTG1 and the promoters was determined on the basis of the ratio of firefly luciferase activity to sea Renilla luciferase activity.

Based on the distribution of cis-acting elements in the promoter sequences, several fragments of different lengths containing HindIII and SaII restriction sites were cloned and introduced into the pcambia1391-GUS vector by homologous recombination individually. All primers used for recombinant plasmid construction in promoter activity assays are listed in Supplementary Table 5 .

Transient expression was conducted in tobacco leaves, and six independent leaves were transfected as described above. The promoter: GUS-transfected leaves were dyed by soaking in β-glucuronidase (GUS) staining buffer [0.05% m/v 5-bromo-4-chloro-3-indolyl-β-glucuronide, 100 mM NaH₂PO₄·H₂O, 10 mM EDTA, 0.5 mM K₄FE(CN)₆·3H₂O, 0.1% Triton] and vacuum penetration for 30 min followed by staining at 37°C for 24 h. The stained leaves were soaked in 70% alcohol several times until they were completely decolorized, and photographs were taken. The reaction of GUS with 4-methylumbelliferyl β-D-glucuronide (MUG) resulted in the production of the fluorescent substance 4-methylumbelliferone (MU), and its content was determined by a fluorospectrophotometer. Thus, following the method described by Jefferson et al. (1987), the GUS content was quantified on the basis of the MU content of the plant total protein per minute. The protein content was determined using the Bradford method (Bradford, 1976). Our strategy followed the ones reported by Debeaujon et al. (2003); Thévenin et al. (2012), and Xu et al. (2014) with minor modifications. TTG1-binding motif enrichment was predicted using the MEME-ChiP (Bailey and Machanick, 2012).

The CDS of TTG1 was inserted into the pVBG2307 vector by homologous recombination, and the recombinant plasmid pVBG2307-TTG1 was transformed into GV3101 for transient injection into 14-day-old B. rapa cotyledons. GV3101 cells with an optical density (OD₆₀₀) set to approximately 0.600 using a UV spectrophotometer (Thermo Fisher Scientific, USA) were injected into the leaves. Transformed plants were kept in the dark for 12 h, and, then, the light conditions were changed to 16-h light/8-h dark photoperiod for 2 days. Three individually transformed cotyledons and their corresponding leaves were collected for gene expression analysis and PA content determination.

Total PA extraction and content analysis were performed using the Plant PA Content Detection kit (Solarbio, China, BC1355). A quantity of 100 mg of dried transformed leaves was collected and dissolved in 1 mL of extract buffer, and, then, the ultrasonic method (300 W, 25°C, 30 min) was used for extraction followed by a centrifugation at 12,000 rpm at 25°C for 10 min. The supernatant was collected and diluted in 1 mL of extraction buffer, and the absorbance was measured at 500 nm.

All data calculations were performed using Microsoft Excel 2019, and column diagrams were drawn using Origin 2021. Significance analysis was performed using SPSS version 20.0.

Seeds from brown-seeded inbred B147 and yellow-seeded inbred B80 plants collected at seven different developmental stages were used to predict the structural genes involved in the TTG1-dependent flavonoid biosynthesis pathway. Expression level analysis showed that 11 of the 22 structural genes had significant higher expression levels in B147 than in B80 at seven different developmental stages (10, 14, 18, 22, 26, 30, and 34 days after flowering) ( Figure 1 , Supplementary Figure 1 ). These were CHS-Bra008792, CHS-Bra006224, DFR-Bra027457, LDOX-Bra013652, LDOX-Bra019350, BAN-Bra021318, BAN-Bra031403, TT12-Bra003361, TT19-Bra008570, TT19-Bra023602, and AHA10-Bra016610. We speculated that these TTG1-dependent genes probably regulate seed coat PA formation in B. rapa.

Based on the reference B. rapa Chiifu genome from the BRAD database, a total of 11 candidate target gene promoters of brown-seeded B147 plants were cloned and analyzed. These were proCHS-Bra008792, proCHS-Bra006224, proDFR-Bra027457, proLDOX-Bra013652, proLDOX-Bra019350, proBAN-Bra021318, proBAN-Bra031403, proTT12-Bra003361, proTT19-Bra008570, proTT19-Bra023602, and proAHA10-Bra016610 and had a length of 1,222 bp, 2,085 bp, 1,582 bp, 2,000 bp, 1,357 bp, 1,768 bp, 1,431 bp, 1,235 bp, 1,116 bp, and 1,249 bp, respectively. All promoter sequences are provided as Supplementary Materials . Analysis of these 11 promoter sequences revealed 98, 131, 77, 123, 140, 71, 102, 127, 100, 77, and 76 cis-acting elements, respectively. These elements are implicated in light response, circadian regulation, jasmonic acid response, salicylic acid response, regulation of zeolin metabolism, abscisic acid response, auxin-response, defense and stress responses, and anaerobic induction and constitute core elements of transcription initiation, promoter, and enhancer regions ( Figure 2 ).

Furthermore, considering that PA formation is regulated by the MYB and bHLH/MYC transcription factors, we analyzed their binding sites separately. Results revealed no MYC-binding sites in proTT12-Bra003361, whereas the remaining 10 promoters contained both MYB and MYC-binding sites.

Eleven pAbAi promoter recombinant plasmid autoactivation assays were performed. Five recombinant yeast strains (pAbAi-CHS-Bra008792, pAbAi-BAN-Bra021318, pAbAi-BAN-Bra031403, pAbAi-LDOX-Bra019350, and pAbAi-TT19-Bra008570) failed to grow in the SD/-Ura/AbA (150 ng mL⁻¹) culture medium. Six recombinant yeast strains, including pAbAi-CHS-Bra006224, pAbAi-DFR-Bra027457, pAbAi-LDOX-Bra013652, pAbAi-TT12-Bra003361, pAbAi-TT19-Bra023602, and pAbAi-AHA10-Bra016610, failed to grow in the SD/-Ura/AbA (300 ng mL⁻¹) culture medium ( Supplementary Figure 2 ). Based on the autoactivation activity assay results, the interaction of TTG1 with the candidate target gene promoters was assessed using Y1H assays in two types of culture media. The results showed that TTG1 bound to proCHS-Bra008792, proDFR-Bra027457, proTT12-Bra003361, proTT19-Bra008570, proTT19-Bra023602, and proAHA10-Bra016610 ( Figures 3A, B ). A dual-luciferase assay was performed to confirm that TTG1 directly interacts with the above six promoters and induces their expression ( Figures 3C-I ).

To characterize the DNA regions involved in the regulation of proCHS-Bra008792, proDFR-Bra027457, proTT12-Bra003361, proTT19-Bra008570, proTT19-Bra023602, and proAHA10-Bra016610 activity by TTG1, transient transformation was performed in tobacco leaves. Based on the cis-acting element location in the promoter sequences, several fragments of different lengths were cloned into the pcambia1391-GUS vector for transient transformation, using GUS as a reporter. The key deletion fragments were characterized on the basis of GUS staining and the drastic changes in activity. GUS activity kept decreasing from proCHS-Bra008792-1 (−1,222 bp) to proCHS-Bra008792-4 (−143 bp) with the blue color of GUS staining gradually becoming lighter. Removal of an additional 278-bp fragment (proCHS-Bra008792-4) led to a strong decrease (~88.7%) in activity. These results indicate that this fragment contains the minimal information necessary to drive transcriptional activity. Intriguingly, proCHS-Bra008792-5 led to a stronger increase (~ 6.43-fold) compared to the GUS activity of proCHS-Bra008792-4, suggesting that some regions upstream of proCHS-Bra008792-5 have a negative impact on proCHS-Bra008792 activity ( Figure 4A ). Among the seven fragments of proDFR-Bra027457, GUS activity was higher than that of proDFR-Bra027457-1 (−1,541 bp) and proDFR-Bra027457-5 (−319 bp) and then decreased in proDFR-Bra027457-6 (−194 bp). Removal of an additional fragment of 125 bp (proDFR-Bra027457-6) led to a strong decrease (~81.2%) in the activity ( Figure 4B ), suggesting that this fragment contains the minimal information necessary to drive transcriptional activity. Analysis of six fragments of proTT12-Bra003361 led to the identification of a 219-bp promoter fragment from proTT12-Bra003361-4 (−392 bp) to proTT12-Bra003361-5 (−173 bp) ( Figure 4C ). Analysis of the proTT19-Bra008570 identified a 218-bp promoter fragment from proTT19-Bra008570-5 (−316 bp) to proTT19-Bra008570-6 (−98 bp), which led to a strong decrease (~95.9%) in activity ( Figure 4D ). Analysis of proTT19-Bra023602 identified a 219-bp promoter fragment from proTT19-Bra023602-4 (−665 bp) to proTT19-Bra023602-5 (−360 bp), which led to a strong decrease (~73.8%) in activity. The removal of an additional fragment of 173 bp of proTT19-Bra023602-5 led to a strong increase (~3.57-fold) in the GUS activity, suggesting that some regions upstream of proTT19-Bra023602-6 have a negative impact on proTT19-Bra023602-5 activity ( Figure 4E ). Five fragments were analyzed in proAHA10-Bra016610, and a 217-bp fragment deletion from proAHA10-Bra016610-3 (−495 bp) to proAHA10-Bra016610-4 led to a GUS activity decrease, whereas a 218-bp fragment deletion from proAHA10-Bra016610-4 (−316 bp) to proAHA10-Bra016610-5 led to a GUS activity increase ( Figure 4F ), indicating that the 218-bp fragment upstream of proAHA10-Bra016610-5 could possibly have an opposite effect with the 217-bp fragment upstream of proAHA10-Bra016610-4 in terms of driving the transcriptional activity.

Based on the GUS activity that was more than 5.0 MU/min/ug, motif enrichment analysis of promoter fragments was performed using MEME-ChiP to determine the sequence of TTG1-binding motifs. Five motifs were significantly enriched, including three novel motifs, a DOF-binding motif, and an NAC-binding motif ( Figure 5A ). Among these motifs, motif 2 (RTWWGTRGM) occurred one or two times per promoter; thus, novel motif 2 was considered the candidate TTG1-binding motif. The position of motif 2 in the promoters was analyzed and is shown in Figure 5B . Based on the combined results of GUS staining and GUS activity, fragments containing motif 2 exhibited higher GUS activity in proCHS-Bra008792-3 (−240 bp ~ −232 bp), proDFR-Bra027457-5 (−246 bp ~ −238 bp), proTT12-Bra003361-4 (−287 bp ~ −279 bp), and proTT19-Bra008570-5 (−215 bp ~ −207 bp). Interestingly, two motif 2 were detected in proTT19-Bra023602-5 (−287 bp to −279 bp and −166 bp to −158 bp) and proAHA10-Bra016610-3 (−456 bp to −448 bp and −148 bp to −140 bp), whose GUS activity differed. The proTT19-Bra023602-5 (−360 bp) containing two motif 2 showed lower GUS activity (2.07 ± 0.05 MU/min/ug), whereas proTT19-Bra023602-6 (−187 bp) containing one motif 2 showed higher GUS activity (9.48 ± 2.46 MU/min/ug). The proAHA10-Bra016610-3 (−495 bp) containing two motif 2 showed higher GUS activity (11.29 ± 0.32 MU/min/ug), whereas proAHA10-Bra016610-4 (-278 bp) containing one motif 2 showed lower GUS activity (1.82 ± 0.23 MU/min/ug).

To clarify the effect of BrTTG1 on B. rapa seed coat color and PA formation, a BrTTG1 overexpression assay in B. rapa seedling cotyledons was carried out. The PAs in untransformed cotyledons were 2.78 ± 0.88 mg/g DW, whereas PAs in transformed cotyledons were 33.80 ± 3.30 mg/g DW. The PAs in the transformed seedling cotyledons were significantly higher than those in the control ( Figure 6A ). In addition, the expression levels of eight genes involved in the flavonoid biosynthesis pathway and Glutathione S-transferases were determined using qRT-PCR. Five target genes including CHS-Bra008792, DFR-Bra027457, TT19-Bra008570, TT19-Bra023602, and AHA10-Bra016610 exhibited a significantly increased expression than that in control, whereas LDOX-Bra13652, BAN-Bra021318, and BAN-Bra031403 exhibited no significant changes ( Figure 6B ).