Two heading Chinese cabbage pure inbred lines, the brown-seeded line ‘92S105’ and the yellow-seeded line ‘91-125,’ provided by Chinese cabbage research group of Northwest A&F University, were crossed by artificial emasculation, then an F₂ population was developed from self-breeding of a single heterozygous F₁ individual plant. The F₂ segregation population and two parental lines were planted in the experimental station of Northwest A&F University, Yangling, China, in 2013. The seed coat colors of individuals in the F₂ population were divided into brown and yellow types by visual observation at the seed maturation stage. The F₃ family populations derived from brown-seeded individuals of F₂ by artificial self-pollination were planted in 2014, which were used to detect whether the genotypes of F₂ individuals at seed color loci is homozygous or heterozygous. A self-compatible and homozygous F₃ brown-seeded progeny B147 and yellow-seeded progeny B80 were selected and planted in 2015, which were used for collecting roots, shorted stems, rosette leaves, leaf petioles, flower stems, cauline leaves, flowers, buds, and different ripeness seeds of 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34 days after flowering (DAF). In addition, a new F₂ population which derived from the hybridization of the other one yellow-seeded line and one brown-seeded line of heading Chinese cabbage was used for testing the accuracy of co-segregation marker, which number of individual is 270 progeny plants.

Different ripeness period seeds of B147 and B80 at 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, 34 DAF were harvested and directly fixed in a FAA fixation solution (50% ethanol, 5% acetic acid, and 5% formaldehyde) for more than 24 h at 4°C. Then seeds were taken out from the fixation solution and were rinsed by distilled water. Seeds coat and embryo were separated by tweezers extrusion. Seeds coat were stained in 0.5% DMACA hydrochloric acid ethanol solution (0.5% DMACA, 15% hydrochloric acid, and 50% ethanol) for 3 h at room temperature and observed the changes of seed coat color.

Referenced as Qu et al. (2013) report, 10 μm thick sections of B147 and B80 seeds at 10, 14, 28 DAF age were cut with a rotary microtome (LEICA, Germany), then stained with TBO and observed with a upright microscope (Olympus, Japan). Three replications of each material were sectioned.

Total DNA of individuals were extracted from young fresh leaves by modified CTAB (cetyl trimethyl ammonium bromide) method (Porebski et al., 1997). The final DNA concentration was adjusted to 100 ng/μL with double distilled water. Total RNA was extracted from roots, shorted stems, rosette leaves, leaf petioles, flower stems, cauline leaves, flowers, buds, and different ripeness period seeds of brown-seeded line (B147) and yellow-seeded line (B80) with TaKaRa MiniBEST Plant RNA Extraction Kit (TaKaRa, Japan) according to the manufacturer’s instructions. The quality of RNA was assessed by electrophoresis with 0.8% agarose gel with staining in ethidium bromide solution. Genomic DNA elimination and the first-strand cDNA synthesis were performed with a PrimeScriptTM RT reagent Kit with gDNA Eraser (TaKaRa, Japan) following the manufacturer’s instructions and the final cDNA concentration was adjusted to 100 ng/μL with RNase free water.

The genomic DNA was randomly broken into short DNA fragments with enzymes and repaired to a flunt end. Then dA tails and sequencing joints were connected at both ends of the DNA fragment one after another. DNA fragments with the joint were purified by AMPure XP beads and 300–400 bp length fragments were selected for PCR amplification. The library needed to be purified and tested, and sequenced by Hiseq X10 PE150 sequencing machine. Whole genome re-sequencing was performed by Guangzhou GENE DENOVO Biotechnology Co. Ltd (Guangzhou, China).

The physical position of flanking markers linked with seed coat color were determined by the BLAST network online service and candidate gene prediction was performed using GBrowse tool from BRAD¹. The specific primers for the candidate gene were designed on the basis of the reference genome of B. rapa (Wang et al., 2011). The sequences of the candidate gene were cloned from the ‘92S105’ and ‘91-125.’ The PCR products were purified using a TaKaRa MiniBEST agarose Gel DNA Extraction Kit (TaKaRa, Japan) and used for TA cloning. The purified PCR products were introduced into the pMD18-T Simple Vector (TaKaRa, Japan) and transformed into Escherichia coli strain DH5α. The recombinant plasmids were sequenced by Sunny Biotechnology, Co. Ltd (Shanghai, China) and sequence alignment was performed with DANMAN software. The complete genomic DNA with intronless region of genomic sequences were submitted to the GenBank under the following accession numbers: 92S105 (KY929015) and 91-125 (KY929016). The function of the predicted gene was searched by the BLASTP tool from NCBI².

Based on the variation in sequences of TTG1 of brown seed (TTG1-B) and TTG1 of yellow seed (TTG1-Y), a specific pair primer, was designed and synthesized, which amplified a functional marker Brsc-yettg1. The PCR amplifications were performed for detecting the TTG1 variation in two parents and other yellow-seeded progeny plants.

The analysis of genes expression were performed with quantitative real-time PCR (qPCR). A house-keeping gene encoding glyceraldehyde-3-phosphate dehydrogenase (GAPDH, GO0048316) was used as the reference gene (Li et al., 2016). The qPCR reactions was performed with 20 μL volume reaction solution included 2 μL of cDNA template (100 ng/μL), 0.5 μL of each primers (10 μmol), 10 μL 2× RealStar Green Power mixture (TaKaRa, Japan), and 7 μL double distilled water in an iQTM5 Multicolor Real-Time PCR Detection System (Bio-Rad, United States). The relative expression levels of genes were measured by 2^-ΔΔC_t method (Livak and Schmittgen, 2001). The cDNA sequences of qPCR primers are listed in Table 1.

Different ripeness period seeds of B147 and B80 were selected to observe and take pictures (Figure 1). The seed coat color of B147 (Figure 1a) was pale green from 8 to 14 DAF, then became green at 16 DAF and emerged brown color at 26 DAF, finally became full brown at 32 DAF. The seed coat color of B80 (Figure 1b) was also pale green from 8 to 14 DAF and becoming green at 16 DAF, but emerging yellow color at 26 DAF, finally became full yellow color at 32 DAF.

Seed coats of B147 and B80 stained with DMACA were shown in Figures 1d,e. PAs in the seed coat specifically reacted with DMACA and produced blue products, therefore the shades of blue color may reflect PAs content in corresponding tissue. Seeds coat color of B147 stained with DMACA (Figure 1d) had no difference with control (Figure 1c) at 8 DAF to 12 DAF, but became light blue at 14 DAF, then became deep blue from 16 DAF to ripe seed. After DMACA staining, the seed coat color began to change at 14 DAF, which was earlier than the color change of ripe seed in natural at 26 DAF. This result indicated that PAs in seed coats of B147 has produced at 14 DAF, but its phenotype color can be investigated at 26 DAF. Compared with B147, seeds coat color of B80 stained with DMACA (Figure 1e) had no difference with control (Figure 1f) in all seed ripening stage from 8 to 34 DAF, which suggested that PAs was little or not produced in seed coats of B80 during seed ripening.

According to the color changes of B147 seed coat stained with DMACA above, seeds of B147 and B80 at three stage of 10 DAF, 14 DAF, 28 DAF were selected for histochemical investigation. TBO staining transverse sections of seeds revealed that flavonoids began to deposit in the hilum of both seeds at 10 DAF (Figures 2a,d). The blue color of B147 in hilum stained was deeper than B80, indicating that more accumulation of flavonoids was in B147. At 14 DAF, the blue color of seeds stained extended to palisade layer, pigment layer and embryo (Figures 2b,e) and color of palisade layer and pigment layer in B147 seed stained more intensely than B80, suggesting that more flavonoids had deposited in B147 than B80. At 28 DAF, the color of palisade layer and pigment layer in B147 seed stained deeper (Figure 2c), but little change in B80 seed (Figure 2f), which indicated that the accumulation of flavonoids had raised more significantly in the palisade and pigment layers of B147 seeds than in those of B80.

In general, Figure 2 showed that flavonoids gradually increased during seed maturation and showed significant differences in the seed coat of B147 and B80. The flavonoids deposited only in hilum in the seed of B147 and B80 at 10 DAF, then extended to the palisade layer and pigment layers at 14 DAF. The significant differences of flavonoids content in the seed coat between B147 and B80 occurred at 14 DAF, which suggested that more flavonoids deposited in B147 seed from this stage, but little change in B80 seed. These results were in agreement with the seed coat color change of B147 and B80 stained with DMACA.

The genome sequences of the parental line, the brown-seeded line ‘92S105’ and the yellow-seeded line ‘91-125,’ were obtained by whole genome re-sequencing (the statistics of whole genome re-sequencing were shown in Supplementary Table S1). Based on the previous study, the locus linked with seed coat color had been mapped to a region of 40.2 Kb in A06 chromosome (Ren et al., 2017). Thus, comparing with the B. rapa variety chiifu genome sequence (Wang et al., 2011), it was found that there are three large fragment deletion in this range of the 40.2 Kb region (Figure 3A) of ‘92S105,’ therefore, the actual physical distance between the marker SSR449a and SSR317 shorten to 26.2 Kb. Similarly, there are two large fragment deletion in this range of ‘91-125,’ and the actual physical distance between the marker SSR449a and SSR317 also shorten to 23 Kb. Based on whole genome sequence of ‘92S105,’ ‘91-125’ and Brassica database¹, there are three predicted genes in this corresponding chromosome region, TRANSPARENT TESTA GLABRA 1 (TTG1), DOWNY MILDEW RESISTANT 6 (DMR6) and a unknown protein (Bra009772) (Figure 3B). TTG1 is a transcription factor that belongs to the WD40-repeat family protein and is involved in formation of MBW complexes. Thus, TTG1 was considered as the candidate gene for seed coat color.

Compare the TTG1 genome sequence between ‘92S105’ and ‘91-125,’ then sequences of TTG1 were cloned, sequenced and named as TTG1-B and TTG1-Y for ‘92S105’ and ‘91-125,’ respectively. A 94-base deletion was found in the 96th base downstream of the initiation codon in the TTG1-Y, which caused the termination codon to appear at 297th base. Thus, the full length of TTG1 of yellow seed is 300 bp (sequence alignment result was showed in Supplementary Figure S1). Amino acid sequence prediction indicated that the full protein of 337 amino acids were encoded by TTG1-B, while a curtate protein of 99 amino acids was encoded by TTG1-Y (Supplementary Figure S2). The full protein of TTG1 includes four WD40 structures, which are located in the position of amino acids residues from 60th to 105th, 112th to 157th, 160th to 198th, 249th to 289th (Supplementary Figure S3), respectively. However, no WD40 structure was found in the curtate protein coded by TTG1-Y.

Based on the sequence difference between the TTG1-B and the TTG1-Y, a new pair of SSR primer, named as Brsc-yettg1, was designed and was used for testing the consistency between seed coat color phenotypes and product of Brsc-yettg1 in 270 F₂ individuals, which showed that the polymorphic bands of Brsc-yettg1 was completely consistent with the seed coat color, as a co-segregated marker (Supplementary Figure S4). This result further indicated that TTG1 is the candidate gene controlling yellow-seeded trait. The co-segregated markers, Brsc-yettg1, will be useful in MAS breeding of yellow-seeded Chinese cabbage.

TTG1 is thought to be the candidate gene controlling seed coat color, therefore, the TTG1 transcript level in different tissues (roots, shorted stems, rosette leaves, leaf petioles, flower stems, cauline leaves, flowers, buds) of B147 and B80 were analyzed by qPCR. Results showed that BrTTG1 expressed in all organs tested and had higher expression in B147 than in B80. In tissues of B147, the highest expression was in rosette leaf and the lowest expression was in bud (Figure 4A). The order of BrTTG1 expression level from high to low is rosette leaf > shorted stem > flower stem > cauline leaf > root > flower > leave petiole > bud. In tissues of B80, the highest expression was in shorted stem and the lowest expression was in flower stem (Figure 4A). The order of BrTTG1 expression level from high to low is shorted stem > rosette leaf > leave petiole > root > flower > cauline leaf > bud > flower stem in B80.

Further, BrTTG1 expression patterns in different ripeness seeds at 8, 10, 12, 14, 16, 18, 20, 22, 24, 26, 28, 30, 32, and 34 DAF of B147 and B80 were performed by qPCR. Results showed that the expression level of BrTTG1 in B147 was higher than that in B80 in the all developing stages. In the process of seed growth, the tendency of BrTTG1 expression appeared a wavelike in both of B147 and B80, which has no relationship with the changes of seed color in them. Among them, the expression of BrTTG1 in B147 was five times higher than that in B80 at 8 DAF. When the seeds grow to 30 DAF, there was no difference in the expression of BrTTG1 between B147 and B80 (Figure 4B).

Four LBGs (BrDFR, BrANS, BrANR1, and BrANR2) involved in flavonoids biosynthesis were measured in seeds at 30 DAF of B147 and B80. The relative transcription levels were showed in Figure 5 referenced by B80. The expression of BrDFR in B147 was approximate 400-fold of that in B80, and the expression of BrANS in B147 also increase 17-fold of that in B80. The expression of BrANR1 and BrANR2 in B147 were nearly 500-fold and 50-fold higher than those in B80, respectively. All these results indicated that the TTG1-Y mutation seriously debased the transcript level of LBGs in flavonoids pathway, and maybe results in the formation of yellow seed coat.