Background

Ornamental kale (Brassica oleracea L. var. acephala) is a popular ornamental plant cultivated worldwide owing to its variegated, colorful leaves [1] and strong resistance to cold. The leaves of ornamental kale are diverse and colorful: the edges are emerald green, dark-green, gray-green, or yellow-green, and the center can be white, yellow, pink, red, purple or other complex color varieties. The leaf pattern of ornamental kale is classified as foliage leaves, round leaves, cracked leaves, wave leaves and other types. Furthermore, purple-leaf ornamental kale, which is mainly due to various anthocyanin components [2], is reported to possess strong antioxidant capacity and thus potential benefits to human health [3, 4].

Some leaf color genes have been mapped in Brassica crops. For example, the red leaf color trait is reportedly controlled by the single dominant gene Re in ornamental kale, which was mapped to C09 between SSR markers C9Z90 and C9Z94, with genetic distances of 0.3 cM and 2.0 cM, respectively [5]. In Brassica juncea, the purple leaf gene (PL) was mapped between SRAP markers ME7EM9 and ME2EM2, with genetic distances of 1.2 cM and 5.5 cM, respectively [6]. A purple leaf dominant gene (BrPur) in Brassica rapa was assigned to a locus between InDel markers BVRCPI613 and BVRCPI431, with a genetic interval of 0.6 cM [7]. An incomplete dominant gene (BnaA.PL1) for the purple leaf trait in Brassica napus has been mapped, and the possible candidate gene is predicted to encode adenosine 5’-phosphosulfate reductase [8]. Zhu et al. found the pink leaf color trait in ornamental kale to be controlled by a single semi-dominant gene mapped to C03 between SSR marker Ni2C12 and co-dominant SCAR marker Boac04, with genetic distances of 0.6 cM and 2.4 cM, respectively [9].

With the development of molecular biology and the release of the B. oleracea draft genome [10, 11], uncovering the genetic basis of important traits has become a focus of theoretical and applied studies in this species. To date, many genes/QTLs have been mapped in Brassica oleracea using InDel markers, including the Fusarium wilt resistance gene FOC1 [12, 13], the petal color gene cpc-1 [14], the yellow-green gene ygl-1 [15], and QTLs associated with heading traits [16], head-splitting resistance [17, 18], Diamondback (Plutella xylostella) moth resistance [19], and resistance to diseases such as Sclerotinia sclerotiorum [20], black rot [21, 22], and clubroot [23].

In this study, kale inbred line W1827 (with white leaves) was crossed to inbred line P1835 (with purple leaves) to construct F2 and backcross (BC) populations, which were used for genetic analysis and fine mapping of the gene for the purple leaf trait. Genes located within the fine-mapping region were analyzed. These findings will lay a foundation for the cloning and validation of BoPr and will facilitate elucidation of the molecular mechanism for purple leaf formation in B. oleracea L. var. acephala.

Methods

Plant materials

The female kale inbred line W1827 (P1, white leaf, Fig. 1a) was crossed with male kale inbred line P1835 (P2, purple color, Fig. 1b) to generate the F1 population. The F2 population was obtained via self-pollination of F1 plants; BC1P1 and BC1P2 populations were created by backcrossing of F1 × W1827 and F1 × P1835, respectively. All materials were from Cabbage and Broccoli Research Group, Institute of Vegetables and Flowers (IVF), Chinese Academy of Agricultural Sciences (CAAS).

Fig. 1
figure 1

Phenotypes of the parental lines. a W1827. b P1835

Full size image

Genetic analysis and InDel marker development

Leaf color was identified visually. The segregation ratios of the F2 and BC1 populations were analyzed using a Chi-square test with SAS software.

The ‘02–12’ reference genome of cabbage was downloaded from BRAD (http://brassicadb.org) [10]; the TO1000 sequence was obtained from the genome sequence (http://plants.ensembl.org/Brassica_oleracea) [11]. The sequencing-by-synthesis method was used to re-sequence the parents at approximately 10× coverage over the entire genome [24]. This work was completed at Beijing Genomics Institute (BGI) (Shenzhen, China). For preliminary mapping, the re-sequencing data for both parents was mapped to the ‘02–12’ reference genome of B. oleracea (http://brassicadb.org) to detect 3-5 bp insertion-deletion mutations between the parents. InDel primers were then designed. Due to the possible assembly error of the 02–12 reference genome, new primers in the preliminary mapping region were designed based on the TO1000 reference genome. Primers were designed to have amplicon lengths of 100–200 bp, GC contents of 40–50% and Tm values of 52–62 °C.

DNA extraction and bulked segregant analysis (BSA)

Genomic DNA was extracted from fresh leaves using a modified cetyltrimethylammonium bromide (CTAB) protocol [25]. The concentration of DNA was estimated using a spectrophotometer (BioDrop, UK) and diluted to 40–50 ng/μL.

Ten purple-leaf F2 individuals and ten white-leaf F2 individuals were selected to construct two pools using the BSA method [26]. Polymorphic InDel markers between the parents were used to screen these pools, and polymorphic markers between the pools were used to examine recombination with all white-leaf individuals in the F2 and BC1P1 populations.

The 20 μL PCR reaction mixture contained 4 μL DNA template (40–50 ng/μL), 2 μL 10× PCR buffer (Mg2+ included), 1.6 μL dNTPs (2.5 mM each), 0.8 μL forward primer (10 μM), 0.8 μL reverse primer (10 μM), 0.4 μL Taq DNA polymerase (2.5 U/μL) and 10.4 μL ddH2O. Reactions were performed in a thermal cycler as follows: 94 °C for 5 min; 35 cycles of 94 °C for 30 s, 55 °C for 30 s and 72 °C for 45 s; 72 °C for 10 min. Amplicons were separated by 8% PAGE at 160 V for 1.5 h, followed by silver staining.

Data analysis

For each marker, individuals with the W1827 allele were categorized as ‘a’, and individuals with the P1835 allele were categorized as ‘b’; those with the F1 allele were categorized as ‘h’. The Kosambi mapping function was used to calculate genetic distances between markers [27], and the genetic map was constructed using MapDraw [28].

To identify probable genes associated with purple leaves, genes located within the candidate interval were analyzed based on annotations from the B. oleracea reference genome ‘TO1000’.

Analysis of the candidate gene

To amplify the candidate gene Bo9g058630, primers were designed in DNAMAN 7.0 using the sequence of Bo9g058630. PCR amplification was performed using the Q5 Ultra High Fidelity DNA polymerase. The PCR amplification conditions followed the manufacturer’s specifications. The annealing temperature was determined using the NEB Tm calculator (New England Biolabs, Inc.), and the elongation step was based on a rate of 30 s/kb. The candidate gene sequence was determined by Sanger sequencing of the PCR amplicons.

Results

Genetic analysis of leaf color of W1827 with P1835

Sixteen F1 individuals generated by crossing W1827 with P1835 exhibited purple leaves, indicating that the purple leaf color trait is dominant over the white-leaf trait. In the F2 population, 258 of 1008 individuals displayed white leaves and 750 purple leaves. The Chi-square test confirmed this segregation ratio to be 3:1. In the BC1P1 population, 1040 of 2034 individuals showed purple leaves and 994 white leaves, with the Chi-square test confirming the segregation ratio to be 1:1. The leaves of all 200 individuals in the BC1P2 population were purple (Table 1). These results indicated that the purple leaf trait is controlled by a single dominant gene, which we termed BoPr.

Table 1 The Chi-square (χ 2) goodness-of-fit test ratios of leaf color segregation in BC and F2 populations
Full size table

Preliminary mapping of BoPr

Illumina paired-end sequencing generated reads of 9.2 Gb for kale inbred line W1827 and 9.9 Gb for line P1835. A total of 211 pairs of InDel marker primers were designed by comparing the whole-genome re-sequencing data with the sequence of the ‘02–12’ reference genome.

These InDel marker primers were used to screen for polymorphisms between the parents. Ultimately, 58 of the 211 pairs of primers revealed polymorphisms, with a polymorphic rate of 27.5%. The polymorphisms were further screened in two the bulks of the F2 population, and only BoID4814, BoID4826 and BoID4837 exhibited polymorphisms between the two bulks. The three markers were preliminarily located on chromosome C09 according to the marker location in the reference genome.

Subsequently, 41 new additional InDel markers near the three polymorphic primers were designed, 26 of which showed polymorphisms. Only co-dominant markers with clear and stable amplification were chosen for further analysis in the F2 and BC1P1 populations. A genetic map comprising 16 InDel markers (Table 2) was constructed using MapDraw [26] (Fig. 2a). InDel markers M17 and BoID4714 were found to be closest to BoPr, flanking the gene at genetic distances of 4.3 cM and 1.5 cM, respectively. Based on marker locations in the reference genome ‘02–12’, BoPr is located on chromosome C09. The interval between the two markers is 280 kb (6,312,350–6,592,994 bp).

Table 2 The InDel primer sequences used in this study
Full size table
Fig. 2
figure 2

a Preliminary mapping of BoPr, units: cM. b Fine mapping of BoPr, units: cM. c Fine mapping using the BC1P1 population with 2034 plants. The BoPr gene was delimited to an interval between BRID485 and BRID490, with an estimated length of 44.8 kb, seven genes were annotated between markers BRID485 and BRID490 based on the reference genome sequence. The genetic structure for each recombinant type is depicted as white for homozygous white leaf alleles, black rectangles for heterozygous alleles, respectively. The relative positions of markers on C09 were determined according to the TO1000 genome sequence

Full size image

Fine mapping of the purple leaf gene

The 280 kb interval spans two scaffolds (Scaffold000195_P2, Scaffold000050), and no polymorphic marker between the parents was found in this narrow interval. Moreover, the genetic distance for this 280 kb interval was calculated to be 5.8 cM, 12 times higher than the usual level (~600 kb/cM). Thus, another reference genome, ‘TO1000’ [11], was used to determine the accuracy of this interval. Blast alignment analysis showed the physical distance between markers M17 (7,590,215–7,590,234 bp) and BoID4714 (20,327,191–20,327,211 bp) to be approximately 12.73 Mb in the ‘TO1000’ reference genome, indicating possible assembly errors in the reference genome ‘02–12’ or ‘TO1000’. To determine whether this 12.73 Mb interval is linked to the BoPr gene, 70 pairs of InDel marker primers were designed based on the ‘TO1000’ reference genome. Polymorphisms between the two bulks were found for 12 of the 70 pairs, and these were used to analyze the BC1P1 and F2 populations. A genetic map comprising 12 InDel markers (Table 2) was constructed using MapDraw (Fig. 2b). InDel markers BRID485 and BRID490 were found to be the closest to BoPr, flanking the gene at genetic distances of 0.10 cM and 0.20 cM, respectively. Based on marker locations in the reference genome ‘TO1000’, the interval between BRID485 and BRID490 is 44.8 kb (17,102,497–17,147,250 bp).

Prediction and analysis of the candidate gene

Based on the ‘TO1000’ reference genome (http://plants.ensembl.org/Brassica_oleracea) (Parkin et al. 2014), seven genes were identified in the 44.8 kb region (Table 3). According to domain annotations from InterPro and BLASTX (best hit) analyses, four of these seven genes have not been annotated (Table 3). The other three genes are as follows: Bo9g058600 (homologous gene AT5G42790) encoding the largest proteasomal subunit; Bo9g058660 (homologous gene AT5G42810) encoding inositol tetra-pentaphosphate 2-kinase; Bo9g058630 (homologous gene AT5G42800) encoding dihydroflavonol 4-reductase (DFR), which catalyzes conversion of dihydroquercetin to leucocyanidin in anthocyanin biosynthesis [29]. Thus, we tentatively designated Bo9g058630 as the candidate gene controlling the purple leaf color in ornamental kale.

Table 3 Annotation of B. oleracea genes in the candidate region
Full size table

Primer pairs were designed spanning the full length of Bo9g058630 (Table 4), and PCR was performed using genomic DNA of W1827 and P1835 as the template. Based on agarose gel electrophoresis (Fig. 3), amplicons of approximately 4.0 kb for W1827 and approximately 2.0 kb for P1835 were obtained. Sequencing revealed that the full-length Bo9g058630 sequence in purple-leaf P1835 is 1580 bp, whereas that in white-leaf W1827 is 3856 bp. Compared with Bo9g058630 in P1835, a 124 bp deletion, a 2400 bp insertion at nucleotide 68 and four SNPs are present in W1827 (sequences of W1827 and P1835 are supplied in Additional file 1: Supplementary 1). Polymorphism of the candidate gene Bo9g058630 was further confirmed in the segregating population.

Table 4 Primers used for full-length candidate gene cloning
Full size table
Fig. 3
figure 3

PCR products of amplicon Bo630 (isolated by agarose gel electrophoresis); M represents the DNA ladder, Pr is the inbred line P1835 with purple leaves, Wh is the inbred line W1827 with white leaves

Full size image

Discussion

Mapping of the purple leaf gene

Purple-leaf traits have been studied in several species, such as rice [30, 31], tetraploid wheat [32], Brassica rapa [7], and carrot [9, 33]. Some genes related to anthocyanins, which might be responsible for purple leaf color, have been reported in Brassica, including BrPur in B. rapa [7], BnaA.PL1 in B. napus [8], and the Re and Pi genes in B. oleracea [5, 9].

Ren et al. [5] reported that markers C9Z90 (C09:18,377,796 bp) and C9Z94 (C09:6,872,051 bp) are tightly linked to Re, which controls the red-leaf phenotype in kale [5]. The red-leaf phenotype is similar with the purple phenotype of the parents used in the present study. Based on reference genome ‘TO1000’, our results show that C9Z90 is located between InDel markers BRID517 (17,879,063 bp) and BRID530 (18,485,633 bp) and that C9Z94 is located between InDel markers BoID4704 (6,418,130 bp) and BoID4707 (7,177,017 bp). This indicates that the mapping interval in Ren et al. [5] is larger than ours. Furthermore, the distance between the Re gene and the closest markers are 2.0 cM for C9Z94 and 0.3 cM for C9Z90 [5] but 6.1 cM for BoID4704 and 1.1 cM for BRID517 in our mapping population, which may be caused by different mapping populations. In our study, however, the BoPr gene was successfully fine mapped to a 44.8-kb interval based in the ‘TO1000’ reference genome.

For the mapping population, F2 and BC1 populations are usually constructed for mapping important agronomic traits. In this study, for markers far away from the candidate gene, the recombination rate of the F2 population was larger than that of the BC1P1 population; however, the opposite case was true for markers close to the candidate region (Additional file 2: Table S1). The closest flanking markers BRID485 and BRID490 were all screened from the BC1P1 population, with no recombinants in the F2 population. Furthermore, the rate of recessive individuals in the BC1P1 populations was 1/2, which was twice that of the F2 population. Therefore, we used the large BC1P1 population to fine map the candidate gene.

Possible assembly errors in cabbage reference genome ‘02–12’

With the development of technology, large amounts of data are produced by genome sequencing, though genome assemblies based on these data are often woefully incomplete. Although the Nipponbare reference sequence (RefSeq) exhibits the best quality among crop genomes, it still contains many assembly errors and gaps [34, 35]. Genome sequencing has been completed in three species of Brassica: B.napus [36], B. rapa [37] and B. oleracea [10, 11]. Some assembly problems have been reported. For example, Lee et al. [23] revised the ‘02–12’ reference genome assembly when they mapped clubroot resistance QTLs using genotyping-by-sequencing; new positions for 27 v-blocks and 10 s-blocks and several inversions of some alphabetically named blocks in reference genome ‘02–12’ and two mis-anchored sequence scaffolds in reference genome ‘TO1000’ were identified. In addition, Liu et al. [15] found possible assembly errors in ‘02–12’ while mapping a yellow-green leaf mutant.

In this study, the recombination rate of markers M17 and BoID4714 in ‘02–12’ [10] (48 kb/cM) was found to be almost 12 times higher than the usual level in the cabbage genome (~600 kb/cM) in preliminary mapping. In ‘TO1000’, the primary mapping interval was found to be 12.7 Mb (~1760 kb/cM). Based on the ‘TO1000’ genome, the BoPr gene was fine mapped to a 44.8 kb interval, but according to primary mapping, this 44.8 kb is not located in the 280 kb region but is on an unanchored scaffold (Scaffold000035_P2) of ‘02–12’. Thus, the mapping results suggested possible assembly error and the potential location of the unanchored Scaffold000035_P2 in the 02–12 reference genome. However, this scaffold is only 1.5 Mb. More unanchored scaffolds could be detected within this interval in the future if PacBio, a third-generation sequencing platform, is applied to improve the reference genome of ‘02–12’. These results will contribute to the improvement of the cabbage genome and provide guidance for map-based cloning of other genes in this region.

The DFR gene in anthocyanin biosynthetic pathways

The biosynthetic pathways of anthocyanins and related genes have been well characterized, and DFR is a key enzyme in the anthocyanin biosynthetic pathway [38]. In characterizing DFR genes, Ahmed et al. found that these genes are associated with cold and freezing stresses in B. rapa [39]. Guo et al. identified 73 anthocyanin biosynthetic genes, locating the DFR gene on chromosome A09, named BrDFR (Bra027457), in B. rapa [40].

In our study, Bo9g058630 was found to be homologous to Arabidopsis thaliana AT5G42800, which encodes DFR. DFRs utilizing the three primary dihydroflavonol substrates have been characterized. Expression of DFR clones together with plant-specific 4-coumaroyl: CoA ligase, chalcone synthase, chalcone isomerase, and flavanone 3-hydroxylase in E. coli resulted in the synthesis of various levels of pelargonidin and could be used for engineering of other bioactive flavonoids, such as flavan-3-ols [41]. The purple-leaf trait of ornamental kale is largely attributed to various anthocyanin components [2], and sequence alignment showed that the function of Bo9g058630 in the white-leaf line W1835 is disrupted by two InDels. Furthermore, we designed a co-dominant marker based on sequence differences between W1835 and P1827, and this marker co-segregated with the phenotype of all recessive individuals in the F2 and BC1 populations. Thus, we suggest that Bo9g058630 is the most likely candidate gene for the purple-leaf trait. However, further work involving transformation is needed to verify whether the function of this gene is responsible for purple leaves in kale.

Conclusions

Inheritance of purple leaves in kale was found to be controlled by a single dominant gene, BoPr, and this gene was mapped to a 44.8 kb interval (reference genome ‘TO1000’) on chromosome C09. InDel markers BRID485 and BRID490 are closest to BoPr, flanking the gene at genetic distances of 0.1 cM and 0.2 cM. In the fine-mapping region, only Bo9g058630 showed high homology to AT5G42800 (dihydroflavonol reductase), which was identified as a candidate gene for BoPr. This study lays a foundation for the cloning of the gene BoPr as well as further function analyses.