The white-flowered inbred line ‘JG800’ (WP) of B. juncea L., which exhibits a stable white petal phenotype, and ‘Caijie’ (CJ), a natural variant characterized by its apricot-flowered trait, were used as the parental material in this study. The CJ germplasm was originally collected from the coastal area of Wenzhou, Zhejiang Province, China, and self-pollinated for at least three generations under greenhouse conditions to ensure phenotypic stability. F₁ hybrids were generated by crossing WP with CJ, and the F₂ population was subsequently obtained by self-pollinating the F₁ plants. All materials were cultivated in a greenhouse set to 23°C with a 16-h light/8-h dark photoperiod for phenotypic evaluation.

Petal color in the F₂ population was scored based on the presence or absence of visible anthocyanin pigmentation. Petals exhibiting apricot or pink coloration were classified as anthocyanin-pigmented, whereas yellow and white petals were classified as anthocyanin-deficient. Segregation ratios in the F₂ population were tested for goodness-of-fit to expected Mendelian ratios using the chi-square (χ²) test. The chi-square value was calculated using the following formula:

where O_i is the observed number of individuals in phenotype class ‘I,’ E_i is the expected number of individuals in that class based on the genetic hypothesis, and Σ denotes the summation over all phenotypic classes. The degrees of freedom (df) were calculated as ‘n’ − 1, where ‘n’ is the number of phenotypic categories. A P-value > 0.05 was used as the threshold to indicate that the observed data did not deviate significantly from the expected genetic ratio, thus supporting the proposed genetic model.

The flower buds and fully opened flowers of the WP and CJ lines were collected at four distinct developmental stages: small buds (stage S1), medium buds (stage S2), large buds (stage S3), and fully opened flowers (stage S4). All samples were immediately frozen in liquid nitrogen and stored at −80°C for subsequent anthocyanin and transcriptome analyses. Anthocyanin profiles were analyzed using ultra-performance liquid chromatography–tandem mass spectrometry (UPLC–MS/MS). Extraction, detection, and quantification of anthocyanins were performed by METWARE Biotechnology Co., Ltd. (Wuhan, China; http://www.metware.cn/) according to standard protocols. For the transcriptome analysis, total RNA was extracted from the same set of samples. Library construction and RNA sequencing were performed by METWARE Biotechnology Co., Ltd., using the Illumina platform, followed by bioinformatics analysis of the generated data.

Two bulked DNA samples were constructed from the F₂ population based on flower phenotypes. The “Anth-bulk” comprised equal amounts of genomic DNA from 30 apricot- and 2 pink-flowered plants, reflecting the expected 15:1 segregation ratio. Similarly, the “NonAnth-bulk” was constructed from 30 yellow- and 2 white-flowered plants. Genomic DNA from two parental lines (WP_77 and CJ_72) was also included, with one individual per parent used as a reference. Genomic DNA was extracted from the fresh leaf tissue of all selected individuals using MolPure Plant DNA Kit (Yeasen Biotechnology, Cat. No. 18800ES50). Four DNA sequencing libraries (two parental and two bulk libraries) were constructed. Paired-end sequencing was performed on an Illumina NovaSeq Plus platform by Novogene Corporation (Beijing, China).

BSA-seq was performed using the Brassica juncea reference genome AU213 (http://brassicadb.cn/) (Yang et al., 2021; Chen et al., 2022). Initially, SNPs and InDels identified against this reference and passing GATK hard filtering were subjected to a genotype missing-rate filter using VCF tools (max-missing 0.75) (Danecek et al., 2011). This step retained only sites that were successfully genotyped in at least three of the four samples (two parents and two trait bulks). Key information, including chromosome position, alleles, genotype calls, and read depths, was extracted using GATK (McKenna et al., 2010). Subsequent analysis was conducted using the R package easyQTLseq: the maternal line CJ_72 and its high-trait bulk were designated as the high group, whereas the paternal line WP_77 and its low-trait bulk were designated as the low group, with the population type defined as F₂. Sliding-window statistics (SNP-index, Δ(SNP-index), and Euclidean distance (ED)) were calculated using window sizes of 1 or 2 Mb and a step size of 20 kb (Takagi et al., 2013). Windows containing fewer than 20 informative SNPs were excluded, and 95% and 99% confidence intervals were estimated by simulation under the null hypothesis of no QTL. Finally, genomic windows ranked in the top 1% by absolute |Δ(SNP-index)| values were defined as putative QTL regions.

Petals (~200 mg) were collected from S3 stage flower buds of F₂ individuals exhibiting four distinct colors (apricot, pink, yellow, and white) to validate the gene expression patterns. Samples were immediately frozen in liquid nitrogen for subsequent RNA extraction. Total RNA was isolated using Eastep Super Total RNA Extraction Kit (Promega). Genomic DNA was removed during the extraction to obtain high-quality RNA. First-strand cDNA was synthesized from 3 µg of total RNA using RevertAid First Strand cDNA Synthesis Kit (Thermo Scientific, Cat. No. K1622) in a 20 µL reaction volume, following the manufacturer’s instructions. The cDNA was diluted 50-fold with nuclease-free water. qRT-PCR was performed in 10 µL reaction volumes containing 4.6 µL of diluted cDNA template, 0.2 µL each of forward and reverse primers (10 µM), and 5.0 µL of KAPA SYBR FAST Universal qPCR Master Mix (Cat. No. KK4601). Reactions were performed on QuantStudio 3 Real-Time PCR System (Thermo Fisher Scientific). All primer sequences used in this study are listed in Supplementary Table 1.

The CJ germplasm is a natural variant possessing the apricot-flowered (APF) trait, which has been stabilized through successive selfing generations (Figure 1A). The WP was selected as the parent for hybridization to introduce a novel color variation into its progeny (Figure 1B). A previous study established that the white petal phenotype of WP results from simultaneous loss-of-function mutations in the functionally redundant xanthophyll esterase genes BjA02.XES and BjB04.XES disrupt carotenoid storage (Li et al., 2023). We crossed WP with the apricot-flowered parent CJ to generate F₁ and F₂ populations to determine the inheritance of the APF locus. The uniform apricot-flowered phenotype observed in both the F₁ and reciprocal F₁ (RF₁) populations demonstrates that this trait is controlled by a dominant nuclear gene at the APF locus, with no detectable cytoplasmic effects (Figures 1C, D).

In the F₂ population, 1,187 individuals were obtained and segregated into four distinct phenotypic classes: apricot, pink, yellow, and white (Figures 1E–H). Genetic analysis of the F₂ population confirmed that the APF locus is the single dominant gene controlling a novel pigmentation pathway. Two observations supported this conclusion. First, the segregation of apricot + pink versus yellow + white petals fits a 3:1 ratio (χ² = 1.11, P = 0.29), demonstrating monogenic dominant inheritance (Table 1). Second, when considering carotenoid accumulation independent of the APF locus, the segregation of carotenoid-pigmented (apricot and yellow) versus non-pigmented (pink and white) petals fitted a 15:1 ratio (Table 1). This indicates that the presence of carotenoids is controlled by two redundant genes, BjA02.XES and BjB04.XES (Li et al., 2023). Critically, the two pathways are genetically independent, as the overall segregation perfectly fits a 45:3:15:1 dihybrid ratio (χ² = 1.97, P = 0.58), confirming that no epistasis exists between the APF-mediated and carotenoid pathways (Table 1). In summary, we established that the APF locus is a single dominant gene controlling a novel pigmentation pathway that is genetically independent of the carotenoid biosynthesis pathway governed by redundant XES genes.

The apricot petal phenotype and visible red pigmentation are typically attributed to anthocyanin accumulation. For instance, the expression of BnaA07.PAP2 induces anthocyanin deposition in both the petals and anthers of B. napus (Ye et al., 2022). To ascertain the pigment composition of the petals, we conducted detailed observations of petal tissues at different developmental stages from the parental lines CJ and WP, as well as their hybrid progeny. Petal and calyx observations confirmed developmentally intensified red pigmentation in CJ and its apricot- or pink-flowered progeny but not in WP or its yellow and white descendants (Figures 2A, B; Supplementary Figure 1). The absence of red pigmentation in the anthers across all materials indicated that the APF locus acts in a spatially distinct manner and may have a different genetic basis than BnaA07.PAP2 (Figure 2A, B; Supplementary Figure 1).

We performed targeted metabolomic analysis using UPLC–MS/MS on the bud and flower samples of CJ and WP collected at four key developmental stages (S1–S4) to further quantify the composition and concentration of anthocyanins. We profiled 53 anthocyanins and their derivatives. The results revealed a consistent pattern of differential accumulation in CJ buds compared with that in WP during development. Particularly, the number of anthocyanins showing upregulated accumulation in CJ increased from 10 in S1 to 14 in S2, 16 in S3, and 21 in S4 (Figure 2C; Supplementary Table 2). Concurrently, the numbers of downregulated anthocyanins were 1, 3, 4, and 2 at S1 through S4, respectively, indicating that upregulation was the predominant pattern (Figure 2C; Supplementary Table 2). Targeted metabolomics revealed that the apricot flower phenotype is associated with a developmentally enhanced anthocyanin profile.

BSA-seq was performed to identify genomic regions contributing to apricot-flowered development. The two parental lines, along with two bulked pools, Anth-bulk and NonAnth-bulk, were subjected to whole-genome resequencing. NGS generated high-quality paired-end reads for all four samples. After quality filtering, >95% of reads from each sample were retained. Mapping rates to the B. juncea AU213 reference genome exceeded 99.13%, and duplication rates ranged from 14.63% to 18.62% (Supplementary Table 3). The mean sequencing depths for the parents were approximately 8–9×, whereas the two bulks exhibited depths of 31× and 33× (Supplementary Figure 2A). Joint genotyping identified ~5.39 million high-quality SNPs and ~1.57 million high-quality InDels across the parental lines and bulks. After applying hard filtering and missing-rate thresholds, ~3.80 million SNPs and ~1.07 million InDels were retained for BSA. The variants were uniformly distributed across the 18 chromosomes of the AA and BB subgenomes (Supplementary Figure 2B).

For each segregating site, the SNP-index (allele frequency of the alternative allele) was calculated for both the Anth-bulk and NonAnth-bulk. The Δ(SNP-index) was computed as the difference between bulk indices. Sliding-window analysis (window size = 2 Mb, step size = 20 kb) revealed pronounced genomic regions exhibiting strong deviations in Δ(SNP-index), suggesting linkage to APF loci (Figures 3A, B). A major QTL was detected on chromosome BB_Chr03, where ΔSNP-index values showed a strong deviation from expected Mendelian segregation, and consistently exceeded the 99% confidence interval threshold across an approximately 47.41 Mb region (10.38–57.79 Mb). The peak SNP was located at 23.48 Mb, where the ΔSNP-index reached its maximum (0.89). Using the top 1% ΔSNP-index threshold, a large continuous interval (Chr03: 18.98–28.74 Mb) was identified, corresponding to the major QTL for flower color. No other chromosomes contained extended high-ranking windows, indicating a single major-effect locus (Figures 3A, B). To complement Δ(SNP-index), ED between bulk allele-depth vectors was computed across all variants. Peaks in ED coincided with major Δ(SNP-index) peaks, further supporting candidate intervals (Figures 3A, B).

Collectively, both Δ(SNP-index) and ED analyses convergently identified strong signals on chromosome BB_Chr03 (18.98–28.74 Mb), representing major-effect genomic regions associated with apricot-flowered phenotype. These intervals span approximately 9.76 Mb and contain 1,406 annotated genes, including transcription factors and pigment biosynthetic genes (BjuOB03G41150 and BjuOB03G45920) with known roles in anthocyanin accumulation (Figures 3A, B).

RNA-seq was conducted on four developmental stages of flower buds (S1–S4) from the CJ and WP parental lines to refine the candidate gene pool (Figures 2A, B). High-quality sequencing was achieved, with 86.38–92.92% of reads per sample mapping to the B. juncea genome, and high replicate concordance validated data robustness (Supplementary Figure 3; Supplementary Table 4). Differential expression analysis was performed through pairwise transcriptome comparisons between the differently colored buds. Following the differential expression analysis, we focused on genes involved in the anthocyanin biosynthesis pathway, given their established role in governing the apricot-flowered phenotype (Supplementary Table 5). Early biosynthetic genes (EBGs) of the flavonoid pathway are largely non-differentially expressed. Nonetheless, several EBGs were significantly deregulated in CJ buds: 16 genes (including BjA04.PALa, BjB01.C4Ha, BjA05.4CLd, and BjA10.CHS) were upregulated, and a distinct set of 16 genes (including BjA05.PALa, BjA03.C4Ha, BjA07.4CL, BjB08.CHSa, and BjA09.CHIa) were downregulated (|log₂ Fold Change| > 1, Figure 4A). In contrast to EBGs, most late biosynthetic genes (LBGs) in CJ buds, including key enzymes such as FLS, DFR, ANS, UFGT, MT, and GST, were significantly upregulated across all four developmental stages (Figure 4A). Notably, we observed an upregulated expression of genes encoding both positive (PAP2 and TT8) and negative (MYBL2) regulatory components of the MBW complex in CJ buds (Figure 4A). This suggests that an altered balance within this master regulatory complex may be the key driver of the widespread upregulation of LBGs.

Given that our genetic analysis indicated APF is a dominant single-gene locus, we hypothesized that the dominant allele confers enhanced expression. Therefore, we focused on the subsequent screening of genes upregulated in the CJ line. Next, we analyzed the differential expression of 1,406 annotated genes within the APF candidate region. Across the four developmental stages, this interval contained 44–59 upregulated and 76–98 downregulated genes (Figure 4B). Notably, the anthocyanin-positive regulator BjB03.PAP2 was upregulated at all stages (Figure 4B), whereas its homolog BjA07.PAP2 was not detected (Figure 4A). For the negative regulator BjB03.MYBL2, expression did not differ between CJ and WP at stages S1 and S2 but was upregulated in CJ at S3 and S4 (Figure 4B). Its homolog BjA07.MYBL2 was upregulated in CJ at stages S1, S3, and S4 (Figure 4A). In line with this genetic model and the fact that differences in anthocyanin accumulation are established from the early stages (S1 and S2), the late and stage-specific upregulation of BjB03.MYBL2 was excluded as a candidate gene. Finally, to identify high-confidence candidates, we intersected the DEGs from all four stages with genes from the 9.76 Mb BSA-seq interval (Supplementary Table 6). This process yielded a refined set of 50 DEGs, comprising 23 upregulated and 27 downregulated genes (Figure 4C; Supplementary Table 7).

We performed a comparative annotation with Arabidopsis homologs to functionally characterize the 23 consistently upregulated candidates. This analysis identified BjuOB03G41150 (BjB03.PAP2) as a key candidate due to its known role in anthocyanin regulation, whereas the other 22 genes showed no links to flavonoid/anthocyanin pathways (Table 2). We quantified the expression of core MBW complex genes via qPCR in S3 petals of F₂ plants to validate the transcriptome data. BjB03.PAP2 expression was significantly higher in anthocyanin-rich apricots and pink petals than in yellow and white petals (Figure 5A). Its homolog, BjA07.PAP2, showed low and no differential expression across all petal types (Figure 5A). Similarly, the positive regulators BjB08.TT8 and BjA09.TT8 were significantly upregulated in apricot and pink petals (Figure 5B). In contrast, the expression of BjB02.TTG1 and BjA06.TTG1 did not differ significantly among petal types (Figure 5C). Both negative regulators, namely, BjB03.MYBL2 and BjA07.MYBL2, were expressed at significantly higher levels in apricot petals than in white petals (Figure 5D). The qPCR results were consistent with the transcriptomic data (Figure 4A). Collectively, functional annotation and independent expression validation strongly implicate BjB03.PAP2 as the most promising candidate gene for the APF locus.

We cloned and sequenced both the promoter and full-length genomic regions of this gene from the two parents to further investigate the molecular basis underlying the differential expression of the BjB03.PAP2 gene between the parental lines CJ and WP. Sequence comparisons revealed no variation across the three exons; however, a single SNP (A-to-T) was identified in intron 1 (Figure 5E; Supplementary Figure 4). Because this SNP is also present in the yellow-flowered reference genome AU213, it is unlikely to be responsible for the observed differences in flower color (Figure 5E).

Further analysis of the approximately 1,463 bp promoter region (based on the AU213 reference) identified an A-to-T substitution at position −860 bp in the high-anthocyanin parent CJ (Figure 5E). In contrast, the non-anthocyanin parent, WP, retained the A allele at this site, suggesting that this polymorphism may influence gene expression (Figure 5E; Supplementary Figure 5). Moreover, an interesting structural variation was detected upstream of the BjB03.PAP2 gene. At approximately −413 bp relative to the transcriptional start site in the reference genome, we identified a (TC)_n tandem repeat (Figure 5E). The reference genomes Tumida and AU213 carried an exceptionally long repeat of (TC)₈₄ and (TC)₁₆₄, respectively, whereas WP contained (TC)₇₆ (Figure 5E). A parallel pattern was observed for the low-expression homolog BjA07.PAP2, in which the (TC)_n repeat length in its promoter region varied between 41 and 63 units across the four accessions (Figure 5E). Notably, in the high-anthocyanin parent, CJ, this repeat was substantially shorter and consisted only of (TC)₁₅ (Figure 5E).

To investigate the evolutionary origin of these sequence variations, we conducted an analysis of the sequence features in the PAP2 promoter region across diploid progenitor species of Brassica. First, we identified a large-scale deletion within the PAP2 promoter region in all three available reference genomes of the diploid progenitor species B. nigra (Supplementary Figure 5B). This deletion spans the putative (TC)_n repeat region. Second, consistent presence of long (TC)_n repeats (64–74 units) was observed in the promoters of PAP2 homologs across all three B. rapa reference genomes (Supplementary Figure 5B), which were derived from accessions exhibiting anthocyanin-deficient petals. Given that long tandem repeats in promoter regions are potent regulators of gene expression (typically repressive), this marked difference in repeat length may significantly contribute to the differential expression of BjB03.PAP2 in CJ and WP.