Inbred lines of the pale-yellow parent (20M-293, P₁) and yellow parent (20M-294, P₂) were used in all experiments. P₁ was crossed with P₂ to produce the hybrid F₁ then the F₁ was self-pollinated to obtain the F₂ population. The F₁ plants were also backcrossed with 20M-293 to produce a BC₁ population. All materials were grown in a greenhouse at the Beijing Vegetable Research Center.

Candidate genes were identified according to the BSA method (Takagi et al., 2013). Fifty yellow and 50 pale-yellow individuals from the F₂ population were selected for each pool. Equal amounts of DNA from the selected individuals were then combined and DNA libraries were prepared using a TruSeq Library Construction Kit before sequencing on the HiSeq™ PE150 platform (Illumina, San Diego, United States).

The raw reads were filtered by removing reads with adapters, reads with an N value greater than 10%, and low-quality reads with a quality value Q ≤ 10. The clean reads were then aligned to the Brassica rapa reference genome V3.0 (Zhang L. et al., 2018) using Burrows-Wheeler Aligner (BWA) (Li and Durbin, 2010). Detection of SNPs and InDels was carried out using a GATK software toolkit (McKenna et al., 2010) with default parameters. The average ΔSNP-index across 10 chromosomes in the B. rapa genome was measured individually using the sliding-window approach, with a 2 Mb window size and a 10 kb sliding-window step.

High−quality SNP candidates between two parents were selected for KASP marker development. For each SNP, two allele-specific forward primers and one common reverse primer were designed by LGC Genomics (Supplementary Table 1). The amplifications and analysis methods were performed based on the previously published guidelines (Li et al., 2019).

Total genomic DNA was extracted from fresh petals using the cetyl trimethylammonium bromide (CTAB) method (Chen and Ronald, 1999). The reference sequences of BraA02g037160.3C and BraA02g037170.3C were obtained from the B. rapa database, and used to design gene-specific and promoter primers (Supplementary Table 2). The cloning and sequencing of candidate genes were performed as described previously (Li et al., 2021).

The sequences of BraA02g037170.3C were then aligned using ClustalW¹ and displayed using ESPript² The cis- and trans-acting elements of the BraA02g037170.3C promoter region were also predicted using the PlantCARE³ online analysis software program.

Total RNA was extracted from petals of the two parent lines using the RNAprep Pure Plant Plus Kit (TIANGEN, Beijing, China). First-strand cDNA, which was used for real-time PCR (RT-PCR) and quantitative real-time PCR (qRT-PCR), was synthesized using the PrimeScript RT Reagent Kit with gDNA Eraser (TaKaRa, Japan). Specific primers of BraA02g037160.3C, BraA02g037170.3C, and two phytoene desaturase (PDS) genes for RT-PCR and qRT-PCR were designed using the Primer Premier 6.0 software. PCR amplification and conditions were as described previously (Li et al., 2020). Gene expression was normalized using the housekeeping gene GAPDH. Reactions were performed with three biological replicates and three technical replicates, and transcript levels in different colored flowers were analyzed using the 2^{–Δ Δ Ct} method (Livak and Schmittgen, 2001). Primer sequences are listed in Supplementary Table 3.

We detected carotenoid contents by Metware⁴ based on the AB SCIEX QTRAP 6500 liquid chromatography with tandem mass spectrometry (LCMS/MS) platform. Three biological replicates of each parent were assessed. In brief, samples from pale-yellow and yellow petals were freeze-dried then stored at -80°C until use. The dried samples were then homogenized in a grinder before extracting 50 mg of dried powder with a solution of n-hexane: acetone: ethanol. An internal standard was then added and the extract was centrifuged for 20 min at room temperature. The supernatants were then collected and the above steps were repeated to re-extract the residue. The extract was then evaporated to dryness under nitrogen flow and reconstituted in methanol: MTBE solution then filtered through a 0.22 μm filter for LC-MS analysis.

The analytical conditions were as follows: column, YMC C30 (3 μm, 100 mm × 2.0 mm); solvent system, methanol: acetonitrile (1:3, V/V) with 0.01% BHT and 0.1% formic acid (A), methyl tert-butyl ether with 0.01% BHT (B); gradient program, started at 0% B (0–3 min), increased to 70% B (3–5 min), then increased to 95% B (5–9 min), finally ramped back to 0% B (10–11 min); flow rate, 0.8 mL/min; temperature, 28°C; injection volume: 2μL.

AB 6500 Q TRAP LC/MS/MS System, equipped with an APCI Heated Nebulizer interface, operating in a positive ion mode, and controlled by Analyst 1.6.3 software (AB Sciex). The APCI source operation parameters were as follow: ion source, APCI +; source temperature 350°C; curtain gas (CUR) was set at 25.0 psi. DP and CE for individual MRM (Multiple Reaction Monitoring) transition was done with further DP and CE optimization. A specific set of MRM transitions were monitored for each period according to the carotenoids eluted within this period.

To determine the efficiency of the VIGS technique in B. rapa, the PDS gene, which can be easily used to score the photobleached phenotype, was used as a visual marker. The hairpin sequence and plasmid construction of BrPDS were performed based on the previously published protocol (Yu et al., 2018). An appropriate fragment of 40 nt of BrPYP was then selected and a 80-nt palindromic DNA fragment of with SnaBI restriction sites and homologous arms on both sides was designed (TTAAACTACTTAGCTCCGTCTTCGGGCTCTCTCCTGTCT CGAGACAGGAGAGAGCCCGAAGACGGAGCTAAGTAGTT TAA). The palindromic DNA fragment was produced by Nanjing Genscript Technology Company (Nanjing, China) and linked to pTYs using the ClonExpress Entry One Step Cloning Kit (Vazyme) (pTYsBrPDS and pTYsBrPYP). The plasmid was then transformed into competent Super Stbl3 (Beijing, Huayueyang) on an Ampicillin (50 μg/ml) resistant plate then cultured overnight before selecting single clones. The clones were then placed in 1 L LB liquid medium and shaken at 225 rmp at a temperature of 30°C for 16 h. The plasmid sample was extracted using a Plasmid Giga Kit (OMEGA, United States) then the working concentration was adjusted to 300 μg/μl. A 5 μl leaf sample was then selected, rubbed with quartz sand, left to stand for 1 min then washed with water. The second vaccination was carried out 4–5 days later. Plants used for the VIGS assay were grown in a greenhouse with a temperature regimen of 25°C day/18°C night. Bacterial solution was then infiltrated into the leaves to obtain the silencing effect.

A sequence approximately 2 kb upstream of the gene start codon was used as a template for the cloned gene promoter sequence. Specific PCR primers were designed, and the target fragment was amplified then connected to the PBI121 (HindIII/BamhI) vector via homologous recombination. After transforming DH5α, a single clone was selected then sequenced before extracting the plasmid and transforming it into Agrobacterium GV3101. After plaque detection, A. thaliana was transformed using the floral dipping method (Clough and Bent, 1998). Seeds from the transformed Arabidopsis plants were then screened on 1/2 MS medium supplemented with 1.5% (w/v) sucrose and 0.8% (w/v) agar containing 25 mg/L of kanamycin (Yue et al., 2019). GUS staining of positive plants was performed using GUS staining solution at 37°C for 24 h with a GUSBlue KIT (Beijing Huayueyang, China). They were then decolorized with 75% ethanol and observed under a stereoscope (Olympus, Japan).

Natural mutant flowering Chinese cabbage plants with visually distinguishable pale-yellow flowers were found in the field (Figures 1A,B). To clarify the inheritance pattern, we constructed two populations using a pale-yellow parent (P₁) and yellow parent (P₂) (Figures 1A,B). As expected, the flowers of F₁ plants derived from a reciprocal cross between the two parents were all yellow. Of 442 F₂ individuals, 343 had yellow flowers and 99 had pale yellow flowers. A χ²-test subsequently confirmed the F₂ segregation ratio to be 3:1 (χ²_0.05 = 3.84; P > 0.05; Table 1). Furthermore, flower color was also examined in a BC₁ population derived from F₁ individuals backcrossed with the pale yellow-flowered parent, revealing an expected Mendelian inheritance ratio of 1 yellow: 1 pale yellow (χ²_0.05 = 3.84; P > 0.05; Table 1). These results indicated that the pale-yellow flower trait is controlled by a single recessive BrPYP gene (Table 1).

Since carotenoids are biosynthesized and stored in chromoplasts, TEM was used to determine whether the pale-yellow color was due to changes in the morphology of the chromoplasts. Accordingly, chromoplasts in the yellow petals were well developed and surrounded by numerous fully developed PGs (Figures 1C,E). However, in contrast, chromoplasts in the pale-yellow petals only contains far fewer well-developed PGs (Figures 1D,F).

To examine whether the petal coloration was due to a reduction in carotenoid accumulation, carotenoid components were analyzed in three pale-yellow and three yellow flowers. Ultra-performance liquid chromatography-tandem mass spectrometry (LC-MS/MS) analysis was used to identify non-esterified and esterified carotenoids (Table 2). As a result, a total of 34 carotenoid compounds were differently accumulated between yellow and pale-yellow petals (Table 2), 28 and 25 of which were found in the pale-yellow and yellow petals, respectively. Of these, 22 compounds were less abundant in the pale-yellow petals, including nine compounds that were not detected (Table 2).

A comparison of carotenoid profiles between the pale-yellow and yellow petals demonstrated abundant esterified carotenoids in the yellow petals. Moreover, carotenoid esters, such as lutein caprate, violaxanthin dilaurate, violaxanthin-myristate-laurate, 5,6epoxy-luttein dilaurate, lutein dilaurate, and lutein laurate, were highly abundant or found only in the yellow petals (Table 2). Taken together, these results suggest that yellow petals accumulate more carotenoid esters (namely, lutein caprate, and violaxanthin dilaurate) than pale-yellow petals, resulting in greater carotenoid accumulation and subsequent yellow pigmentation.

To identify candidates of the BrPYP gene, 50 pale yellow and 50 yellow individuals were selected from the F₂ population and used to construct extreme pools for bulk segregation analysis (BSA). Genome resequencing of the pale-yellow parent (P₁), yellow parent (P₂) and two DNA pools (pale-yellow and yellow bulk) resulted in 202,531,626, 208,423,334, 209,006,868, and 207,592,868 high-quality clean reads, respectively (Supplementary Table 4). Meanwhile, 98.12, 98.35, 98.52, and 98.57% of the clean reads (Supplementary Table 5) were, respectively, aligned to the reference genome. Using a single-nucleotide polymorphism (SNP) index of 0.99 as the threshold, we were able to preliminarily map BrPYP to a candidate region of 2.76Mb (24,960,000–27,720,000) on chromosome A02 (Figure 2A).

To fine map BrPYP, we used both BC₁ and F₂ individuals to conduct Kompetitive Allele-Specific PCR (KASP) genotyping analysis. SNPs were acquired based on the resequencing results with 50 SNP markers in the candidate region initially used to identify recombinant individuals (Figure 2B). Subsequently, 15 SNP markers exhibiting polymorphisms were detected between the pale-yellow bulk and yellow bulk (Supplementary Table 1). Finally, BrPYP was localized to a 101.79 kb region between SNP markers A02_25962798 and A02_26064584 (Figure 2B). Specifically, the SNP marker within the BraA02g037170.3C, A02_26057775 gene co-segregated with the pale-yellow flower trait.

To identify BrPYP in flowering Chinese cabbage, comparative gene annotation was carried out in Arabidopsis thaliana. Overall, 13 genes were identified in the candidate region, of which nine possessed functional annotations suggestive of Arabidopsis homologs (Table 3). However, among these, no carotenoid biosynthesis genes were identified. We therefore screened each gene according to previously published data, revealing two homologs of Phytyl ester synthase 2 (PES2) as candidates. PES2 encodes a carotenoid modifying protein, which plays a vital role in the production of xanthophyll esters in anthers and petals (Ariizumi et al., 2014).

Quantitative real-time PCR (qRT-PCR) was subsequently performed to determine the expression level of BraA02g037160.3C and BraA02g037170.3C in the petals of the pale-yellow and yellow parent. Expression of both genes was lower in the pale-yellow mutant than the yellow parent (Figures 3A,B). However, in comparison, expression of BraA02g037170.3C was very high in the yellow parent and six-fold higher than in the pale-yellow parent. Meanwhile, expression of BraA02g037160.3C was only two-fold higher in the yellow parent than the pale-yellow parent, and was very low in both (Figure 3). To further characterize the difference between the two parents, the promoter of BraA02g037170.3C was subsequently amplified and sequenced. Sequence alignment revealed a 1,148 bp deletion starting 1,078 bp upstream of the promoter in the pale-yellow parent, with several other SNPs and InDels (Figures 3C,D). Analysis of Cis-acting regulatory elements further revealed that the deleted sequence contained multiple light-responsive motifs, including the TCT-motif, chs-CMA1a, MRE, GT1-motif, G-box, and ACE (Supplementary Table 6).

Analysis of cDNA sequences of BraA02g037170.3C from both parents indicated a 105-bp deletion in the fifth exon of the pale-yellow parent, resulting in a 35-amino acid deletion (from 219 to 253) (Supplementary Figure 1a). Furthermore, protein analysis demonstrated two domains based on the NCBI Conserved Domain Database (Supplementary Figure 1b): alpha/beta hydrolases and lysophospholipid acyltransferases (LPLATs), both of which are involved in glycerophospholipid biosynthesis. Alpha/beta hydrolases are a functionally diverse superfamily consisting of proteases, lipases, peroxidases, epoxide hydrolases, and esterases. Interestingly, the 35-amino acid deletion in pale-yellow allele was in alpha/beta hydrolases domain (Supplementary Figure 1), which may also affect the function of BrPYP protein.

Virus-induced gene silencing (VIGS) is a transient gene knockdown system that has successfully been used to characterize genes with different functions in B. rapa. We therefore used the VIGS approach to determine whether BrPYP can induce pale-yellow petals. Loss of chlorophyll pigmentation, caused by silencing of the PDS gene, was used as a visual marker to demonstrate the usefulness of this technique in efficient silencing of B. rapa genes (Figure 4A). Notably, both pTYsBrPDS and pTYs showed the same yellow petals as the yellow petal parent (Figures 4A,B). BrPYP gene silencing was subsequently monitored during flowering, revealing that successful silencing resulted in the pale-yellow phenotype (Figure 4C). The enlarged images of these petals were shown in Figures 4D–F.

qRT-PCR was also performed to further confirm the function of PYP at the transcriptional expression level mediated by VIGS (Figure 4G). The relative expression demonstrated that the BrPDS genes were successfully reduced in the leaves of plants infected with pTYsBrPDS (Supplementary Figure 2). Accordingly, transcript levels of BrPYP were substantially reduced both in the leaves and petals of pTYsBrPYP compared with pTYsBrPDS, pTYs and the yellow parent control (Figure 4G). Meanwhile, transcript levels of the BrPYP gene in BrPYP-silenced plants, which were similar to the pale-yellow parent, showed an approximate 70% reduction compared with the PDS-inoculated plants. This further confirms that BraA02g037170.3C represents the BrPYP gene responsible for the pale-yellow petal trait in flowering Chinese cabbage.

To determine whether the promoter of BrPYP functions in carotenoid esterification, we constructed transgenic Arabidopsis plants harboring beta-glucuronidase (GUS) driven by the synthetic promoter BrPYP (Figure 5). We first analyzed the expression pattern of BrPYP:GUS in the Arabidopsis seedlings, revealing strong GUS activity in the leaves under the yellow petal promoter (BrPYP^Y::GUS) compared with the pale-yellow petal promoter (BrPYP^PY::GUS) (Figures 5A,B). As expected, GUS expression was also stronger in the petals of Arabidopsis with BrPYP^Y::GUS (Figures 5C,D), indicating that variation in the promoter of pale-yellow petals results in a decrease in BrPYP expression.

This study demonstrated that the chromoplasts in the pale-yellow petals of flowering Chinese cabbage only contains far fewer well-developed PGs compared to the yellow petals (Figure 1). Chromoplasts are a type of plastid that participate in carotenoid pigment biosynthesis and storage through multiple catalytic steps (Sun et al., 2018; Hermanns et al., 2020). An increase in the number and size of mature chromoplasts leads to an accumulation of carotenoids, resulting in colorful petals that then act to attract pollinators (Li and Yuan, 2013). The PGs in chromoplasts are much bigger than those in chloroplasts and contain more xanthophyll esters, which are formed during the degradation of carotenoids (Mulisch and Krupinska, 2013). However, irregularly shaped PGs, and disorganized and electron-lucent membrane structures lead to a decrease in carotenoid, as found in carnations, tomato, and B. juncea (Ariizumi et al., 2014; Zhang X. et al., 2018; Iijima et al., 2020). In this study, far fewer well-developed PGs were observed in the pale-yellow petals of flowering Chinese cabbage, suggesting that insufficient PGs reduces accumulation of carotenoid esters.

(E/Z)-Phytoene and carotenoid esters, such as lutein caprate, violaxanthin dilaurate, violaxanthin-myristate-laurate, 5,6epoxy-luttein dilaurate, lutein dilaurate and lutein laurate, were highly accumulated or only found in the yellow flowering Chinese cabbage petals (Table 2). Carotenoids can be acylated by different fatty acids, generating a large number of possible carotenoid ester compounds (Schweiggert et al., 2016; Mariutti and Mercadante, 2018). Lutein esters, such as lutein caprate, 5,6epoxy-luttein dilaurate, lutein dilaurate and lutein laurate, are also the well-known predominant carotenoids in petals of marigold (Rodrigues et al., 2019) and tomato (Watanabe et al., 2021). Carotenoid esters are also generated by acylation of fatty acids (Fas), and esterification does not change the carotenoid chromophore; thus, all carotenoid esters present the same UV-VIS spectrum and color of free-carotenoid (Mercadante et al., 2017; Rodrigues et al., 2019). Interestingly, the pale-yellow flowering Chinese cabbage petals observed here had slightly more lutein dimyristate, violaxanthin myristate, violaxanthin-myristate-palmitate, and violaxanthin dimyristate than those of the yellow petals (Table 2). However, the absolute content of carotenoids in the pale-yellow petals was not very high, possibly resulting in the pale-yellow rather than pure white coloring. Based on the carotenoid profile detailed above, these findings suggest that the presence of lutein caprate and violaxanthin dilaurate are mainly responsible for the obvious color differences between yellow and pale-yellow flowering Chinese cabbage petals.

In many plants, yellow petals generally contain high concentrations of carotenoids, consisting largely of esterified lutein (Meng et al., 2019). A decease in carotenoids accumulation and subsequent alteration in petal color from yellow to white or pale yellow can result from carotenoid biosynthesis, degradation or sequestration (Ariizumi et al., 2014; Meng et al., 2019). In Medicago truncatula, disruption of the anthocyanin-related R2R3-MYB protein WP1, which functions as a transcriptional activator directly regulating expression of the carotenoid biosynthetic genes MtLYCe and MtLYCb, was found to cause a reduction in lutein accumulation (Meng et al., 2019). Meanwhile, the tetratricopeptide repeat protein RCP2 was found to regulate carotenoid biosynthesis and chromoplast compartment size in Monkeyflowers, suggesting that loss-of-function mutations in RCP2 damage the entire carotenoid biosynthetic pathway (Stanley et al., 2020). Plastid-localized CCDs are known to play significant roles in carotenoid degradation in many plants (Hai et al., 2012; Gonzalez-Jorge et al., 2013; Xu et al., 2019), which is negatively correlated with yellow pigment intensity, and is also part of the proteome of PGs. In addition to biosynthesis and degradation, carotenoid levels are also modulated by sequestration. Although the mechanisms involved in carotenoid sequestration remain elusive, several lines of evidence suggest the involvement of the phytyl ester synthase (PES) gene (Ariizumi et al., 2014; Yang et al., 2021). Disruption of PES caused a loss of xanthophyll esters, resulting in a decrease in total carotenoid levels and disruption of chromoplast development in the petals of plants (Ariizumi et al., 2014; Yang et al., 2021).

Our study strongly suggests that the BrPYP gene encodes an enzyme protein that controls the pale-yellow petal color in flowering Chinese cabbage (Figure 2). BrPYP is a homolog of Arabidopsis phytyl ester synthases 2 (PES2, At3g26840), a chloroplastic acyltransferase with phytyl ester synthesis and diacylglycerol acyltransferase activities (Lippold et al., 2012). Furthermore, BrPYP mapped to the same gene with BrWF3, which is a candidate gene responsible for white pigmentation in Chinese cabbage (Brassica rapa L. ssp. pekinensis) (Yang et al., 2021). However, different variations were identified between BrPYP and BrWF3. In the coding sequence of BrWF3, a SNP deletion was speculated to be key variant, causing a frameshift mutation and a premature stop codon (Yang et al., 2021). Meanwhile, we identified a functional 1,148 bp deletion in the promoter region of BrPYP that reduces promoter activity and expression level, and then verified by VIGS. As Caixin and Chinese cabbage are different subspecies of B. rapa, this could be different alleles between these two subspecies.

Fatty acid phytyl esters are a type of neutral lipid, the main compound of the PG core (Besagni and Kessler, 2013). Furthermore, characterization of phytyl ester synthases previously confirmed that PES is the main contributor to thylakoid decomposition and PG enlargement during chloroplast senescence and nitrogen deprivation through their role in the biosynthesis of fatty acid phytate and triacylglycerol (Lippold et al., 2012). Moreover, a homolog of PES in tomato was found to be involved in the formation of abundant carotenoid esters in chromoplasts, also affecting petal color (Ariizumi et al., 2014). The reduction in accumulation of esterized carotenoids in the pyp mutant was therefore expected. Taken together, these findings suggest that BrPYP forms more xanthophyll esters during the breakdown of carotenoids.