Introduction

After the construction of the first substantial linkage map of B. oleracea, with isozyme loci, pioneered by Arus and Orton (1983), several others have followed using a variety of molecular markers (for review see Quiros 2000 and Quiros and Paterson 2004). Little effort was spent at that time to align these maps across laboratories and to perpetuate the mapping populations used for their construction. However, it was possible to assign some of the linkage groups to their respective chromosomes with alien addition lines (Hu and Quiros 1991; Heneen and Jorgensen 2001). Sebastian et al. (2000) established a consensus map for this species based on perpetuated individuals from two double haploid populations, (cauliflower × Brussels sprouts and broccoli × kale) constructed with 547 RFLP, AFLP and SSR markers . Its nine linkage groups have been now physically assigned to their respective chromosomes by Howell et al. (2002) using FISH. Furthermore, these chromosomes have now been aligned with the linkage groups for both the A and C genomes in B. napus by Bohuon et al. (1996), Lowe et al. (2004) and Piquemal et al. (2005). With the availability of EST sequences from Arabidopsis, these were used to construct several maps allowing partial comparison of the A. thaliana genome with the B. oleracea genome (Kowalski et al. 1994; Lan et al. 2000; Babula et al. 2003). This task was also accomplished by Li et al. (2003) using cDNA polymorphisms to construct a linkage map in B. oleracea, followed by comparative physical mapping to A. thaliana. Parkin et al. (2005) has now aligned all linkage groups of B. napus to A. thaliana with RFLP markers.

We report the construction of a high-density genetic map based on the broccoli × cauliflower F2 population used by Li et al. (2003), adding various types of PCR-based markers and sequences of known genes. Each linkage group has been assigned to their respective chromosomes based on common markers with the Sebastian et al. (2000) and Piquemal et al. (2005) maps and to the chromosomes of A. thaliana. Further, the map was used to determine QTLs for curd formation, which segregates in this population. Assignment of the linkage groups of this map to their respective C genome chromosomes adds over 1,000 new markers as mapping tools for B. oleracea and B. napus. The contribution of a substantial amount of new markers from our map will increase the efficiency of marker-assisted selection and map-based gene cloning in B. oleracea and B. napus.

Materials and methods

F2 mapping population

To construct this map, we used the same F2 segregating population as that used by Li et al. (2003) to construct a transcriptome map based on cDNA-SRAPs in B. oleracea. It was developed by crossing two double haploid lines (broccoli “Early-Big” and cauliflower “An-Nan Early”), then selfing the F1 to make 143 F2 plants, which were used as parents to generate inbreds by single seed descent. In addition to the existing cDNA markers (Li et al. 2003), we added genomic SRAP markers, SSR markers, B. oleracea BAC clone sequences (B40L6, B59A4, B59C4) and sequences corresponding to 11 known B. oleracea genes.

Genetic markers

A total of 170 primer pairs, including 87 SRAP primers labeled with IRDye 800 or IRDye 700 fluorescent dyes combined with various unlabeled primers (Table 1), were used to amplify genomic DNA in the F2 population following the protocol of Li and Quiros (2001). The sequences of these primers have been published by Sun et al. (2007) (in press). The PCR products were run in 5% polyacrylamide with the Li-Cor Global IR2 4200 sequencing system.

Table 1 Primer pairs for the SRAP markers used in this study. Primer sequences reported in Sun et al. (2007) (in press)
Full size table

Public SSR primer sequences were obtained from http://brassica.bbsrc.ac.uk/cgi-bin/ace/searches/browser/BrassicaDB and some from published papers (Sebastian et al. 2000; Smith and King 2000). A total of 50 SSR primer pairs were screened between two parents. Of these, 24 SSR primer pairs showing polymorphism between two parents were run in the F2 population.

Map construction

The map was constructed with the program Joinmap 3.0 (LOD score from 4.0 to 8.0). SRAP markers from genomic DNA were developed for this map (starting with M or S on the actual map). These were combined with 155 cDNA SRAP markers (Li et al. 2003), 26 SSR markers (starting with OL on the map, or named NGA248, LS107, sORA21b, MB4), three BAC end sequences: B40L6 (corresponding to A. thaliana At5g23400), B59A4 (At2g03240), and B59C4 (At4g29905) and 11 B. oleracea genes as follows: glucosinolate pathway: BoGSL-ALK, BoGSL-ELONG, BoGSL-PROa , BoGSL-PROb, BoCS-lyase, BoGS-OH, BoCYP79F1, BoS-GT; resistance to cotyledon stage downy mildew: BoDM1; and inflorescence development: BoCAL ,and BoAP1. Primer sequences used to map these genes are shown in Table 2.

Table 2 Primer sequences for 11 genes included in the map
Full size table

Alignment of our map to existing maps and assignment of linkage groups to specific chromosomes

A total of 26 SSR markers and 77 SRAP markers were compared with the current B. napus and B. oleracea maps (Sebastian et al. 2000; Howell et al. 2002; Lowe et al. 2004; Piquemal et al. 2005; Qiu et al. 2006). Following the rationale of Li et al. (2003), 155 cDNA markers, 11 known gene sequences and three BAC-end sequences were use for alignment with Arabidopsis chromosomes.

QTL mapping of curd phenotype

We visually scored each of the F2 plants in the greenhouse and F3 families in the field and greenhouse for inflorescence type in three major classes: 1 = broccoli-like, 2 = intermediate and 3 = cauliflower-like based on the scoring system of Labate et al. (2006). QTL determination was carried out by composite interval mapping with the software WinQTLCart 2.0 (Zeng. 1994). The threshold LOD of 2.50 was selected based on a 5% significance level determined by 300 permutations.

Results

Construction of the map

A 1,257-marker high-density genetic map of Brassica oleracea was constructed and spanned 703 cM in nine linkage groups designated LG1–LG9. Most of these markers were randomly distributed throughout the map (Fig. 1). It included a total of 1,062 genomic SRAP dominant markers generated with the 170 primer pairs. On an average, 6.2 SRAP polymorphic markers were produced per primer pair. All 155 cDNA SRAP markers produced by Li et al. (2003) in the same population were integrated into the nine linkage groups of this map. Of the 50 SSR primer pairs screened between two parents, only 24 showed polymorphism. The 24 SSR primer pairs found to be polymorphic between the two parents generated 26 SSR markers. Of these, 26 markers were mapped into the nine linkage groups.

Fig. 1
figure 1figure 1figure 1

Linkage groups LG1–LG9 for the B. oleracea map and their correspondence to the C genome chromosomes (C1–C9). cM are shown on the left side. Bars on the right indicate homologous segments to chromosomes 1–5 of A. thaliana

Full size image

The eight glucosinolate genes, BoDM1, BoCAL and BoAP1 and three BAC clone end sequences were scored as co-dominant markers. BoGSL-ELONG and BoDM1 were mapped on BoLG2, 0.84 cM apart. Giovanelli et al. (2002) reported RAPD marker OMP-750 linked at 3 cM to BoDM1 in B. oleracea. After sequencing of this marker, we found conservation of the chromosome segment containing BoGSL-ELONG and OMP-750 in A. thaliana. The OMP-750 Arabidopsis homolog is in BAC MOP9 and is 53 Kb apart from At5 g23020, which corresponds to the BoGSL-ELONG gene (Gao et al. 2005) (Fig. 2). Downstream from the BoGSL-ELONG ortholog, there is a putative disease resistance gene, At5 g23400. Screening of the B. oleracea “Early Big” BAC library produced clone B57M17 harboring the gene corresponding to At5 g23400. Sequencing of this gene in B. oleracea revealed that it has 1,764 bases, it is intronless and has 86% identity with At5 g23400 (data not shown). A marker developed from sequencing the Brassica homolog to this gene on this BAC clone, co-segregated with BoGSL-ELONG, revealed further conservation of the chromosomal segment in both species. The mapping progeny did not segregate for downy mildew resistance, therefore it was not possible to confirm whether the At5 g23400 homolog is the true gene for BoDM1. Genes BoCAL and BoS-GT mapped on BoLG3, BoGSLPROa, BoGSL-PROb (previously named BoGSL-PRO and BoGSL-PROL, respectively; (Gao et al. 2006) and BoCYP79F1 mapped on BoLG5. The first two genes were only 0.027 cM apart. They are duplicated gene members of the MAM (methylthioalkylmalate synthase) gene family (Gao et al. 2006). BoAP1-a and BoCS-lyase mapped on BoLG 6. BoGSL-ALK and BoGS-OH mapped on BoLG9 5.4 cM apart. These two genes are members of the AOP (2- oxoacid-dependent dioxygenases) family. The latter corresponds to the A. thaliana ortholog (at2 g25450) and has not been previously mapped on Brassica.

Fig. 2
figure 2

Maps for BoGSL-ELONG and BoDM1. Top, genetic map; bottom, physical map.OPM-750 is RAPD marker reported by Giovannelli et al. (2002)

Full size image

Alignments to C genome linkage groups from other B. oleracea and B. napus maps

Using 22 common SSR markers, we were able to align the linkage groups of our map to the genome specific groups of B. oleracea and B. napus. The nine linkage groups BoLG1–BoLG9 on our map are equivalent to the B. oleracea linkage groups O1–O9 (Sebastian et al. 2000; Howell et al. 2002) and B. napus linkage group N11–N19, respectively (Bohuon et al. 1996; Lowe et al. 2004; Piquemal et al. 2005) (Table 3). Of the genomic SRAP markers in our map, 77 had also been included in an ultradense B. napus map constructed by Sun et al. (2007) (in press). Thus, it was possible to align both maps, confirming our previous assignment of our linkage groups to the N11 through N19 standardized groups.

Table 3 Marker statistics for B. oleracea map and number of aligned markers with other genetic maps
Full size table

Based on the physical assignment of the linkage groups in the map developed by Sebastian et al. (2000) and Howell et al. (2002), our linkage groups BoLG1–BoLG9 are analogous to their linkage groups, which corresponds to the C genome chromosomes 8, 5, 1, 2, 6, 9, 4, 7 and 3, respectively (Fig. 1).

Alignment of B. oleracea linkage groups and the Arabidopsis thaliana physical map

Our map was compared to the Arabidopsis physical map with 11 gene sequences, three BAC-end sequences, and 155 cDNA markers. As expected , the alignment between the B. oleracea linkage groups and chromosomes of A. thaliana (Fig. 1) fully agrees with that reported by Li et al. (2003). However, there was only partial agreement with the alignment reported for the C genome chromosomes of B. napus and A. thaliana chromosomes by Parkin et al. (2005). Lack of total agreement between the two reports is not unexpected considering that none of the two maps are complete.

QTL mapping of curd phenotype

The frequency distribution for inflorescence type, characterized as broccoli-like versus exhibiting curd formation in the mapping populations is shown in Fig. 3. Although in general it followed a bell-shaped curve, it was skewed toward the broccoli phenotype. Three chromosomal regions for curd formation were detected in this population by statistically significant QTLs. Two QTLs in BoLG1 explained 21 and 6% of the variation, respectively. These two were 17 cM apart with non-overlapping confidence intervals. The third QTL was on BoLG6 associated with BoAP1-a and explained 15% variation. No QTLs were detected in the linkage group regions containing the BoCAL-a and BoGSL-ELONG genes (Table 4).

Fig. 3
figure 3

Histogram showing phenotype distribution for inflorescence type in F2 mapping population. (phenotype 1 = broccoli; phenotype 2 = intermediate; phenotype 3 = cauliflower)

Full size image
Table 4 Chromosomal location of segments involved in curd development detected by QTL analysis
Full size table

Discussion

The number of markers, linkage group coverage and density reported in our map and its alignment to the B. oleracea maps of Sebastian et al. (2000) and Howell et al. (2002), and to the B. napus maps of Lowe et al. (2004), Piquemal et al. (2005), Parkin et al. (2003, 2005), Qiu et al. (2006) and Sun et al. (2007) (in press) add a significant number of markers to the C genome useful for marker assisted selection and map-based cloning in both species. Basically, we have added 1,257 markers to the B. oleracea maps (Table 3). The level of polymorphism of genomic SRAP markers between broccoli and cauliflower is high, so similar levels are expected between other more divergent Brassica crops, such as broccoli and kale (Li and Quiros 2001). Although most of these markers are dominant, they could be quite efficient for marker-assisted selection when associated in repulsion phase to genes targeted for selection (Haley et al. 1994). Regarding the usefulness of a high-marker density map for map-based-cloning, it can be estimated based on genome size and map length, that 1 cM corresponds to approximately 800 Kb in B. oleracea. However, this value must be taken conservatively considering the variation in density along the chromosomes. An example of this is the positional cloning of gene BoGSL-ALK tagged with a marker at 1.4 cM, but at a physical distance of less than 100 Kb (Li and Quiros 2003).

Further, anchoring this map to the A. thaliana physical map is an important asset because the latter could serve as a useful source of additional markers to saturate specific segments carrying a gene(s) of interest in B. oleracea. Discrepancies of alignment between the C genome chromosomes of B. oleracea and B. napus might reflect chromosomal structural changes during alloploidization and stabilization of B. napus (Song et al. 1995).

Several linkages described in this paper are of particular interest. For the downy mildew resistance gene BoDM1, we populated the BoLG 2 region with three markers including the gene BoGSL-ELONG. Some of these markers should prove more useful than others previously described (Giovannelli et al. 2002) for marker-assisted selection to develop cotyledon stage downy mildew resistance. Another interesting linkage on BoLG9 was for the AOP gene family members BoGSL-ALK and BoGSL-OH. These genes act sequentially in the side chain modification of aliphatic glucosinolates, the first directing desaturation to produce alkenyl glucosinolates and the second one their subsequent hydroxylation (Li and Quiros 2003). In A. thaliana, there are three AOP genes in triplicate, GS-OH (AOP3), GS-ALK (AOP2) and AOP1 (unknown function) (Gao et al. 2004). Similar to A. thaliana, in B. oleracea, BoGSL-ALK and BoAOP1 are next to each other, but both are duplicated in tandem and the sequence corresponding to the gene GS-OH is absent (Gao et al. 2004). Evidently in B. oleracea, although genes BoGSL-OH and BoGSL-ALK lay on the same chromosome they are not contiguous as in A. thaliana. Another conserved linkage is between the MAM gene family members BoGSL-PROa and its duplicate BoGSL-PROb involved in the synthesis of 3 carbon side chain glucosinolates (Gao et al. 2006). These are homologs of the same Arabidopsis gene (At1 g18500) located at the top of chromosome 1.

Among the glucosinolate pathway genes that we mapped, BoGSL-ALK, BoGSL-ELONG and BoGSL-PRO have been previously cloned and their function assessed (Gao et al. 2003, 2004, 2006; Li and Quiros 2002, 2003). For the rest of the genes, we only mapped their sequences. Although there were enough differences in their sequences in both parental plants for each of these genes to follow their segregation in the progeny, we could not follow segregation for their glucosinolate phenotype. This was due to the fact that the phenotypes of these genes are based more on glucosinolate amount and not quality, and the broccoli and cauliflower parents of our population have similar amounts for most of the glucosinolates controlled by these genes. The only exception was for gene BoGSL-OH, whose phenotype is presence/absence of the glucosinolate progoitrin. As expected, a major QTL for the presence of this glucosinolate was found in the map location for gene BoGSL-OH on LG 9 (data not shown). In any case, the sequences and flanking markers of the genes that could not be associated with specific glucosinolate segregation in our mapping population could be used in other populations, segregating for glucosinolate amount and composition. Application of marker-assisted selection in these populations will be helpful for the development of plants with specific glucosinolate profiles and content.

Although the intention of this paper was not to perform an exhaustive QTL analysis for curd formation, as was reported by Lan and Paterson (2000), we took advantage of the fact that we used two double haploid plants as the parents of the mapping population to do a general analysis to compare with the results from previous studies. Our phenotypic analysis involved only visual scoring for inflorescence type and did not include detailed measurements as done by Lan and Paterson (2000). The three chromosome segments detected by QTL analysis in our population explain 42% of the total phenotypic variation for inflorescence type. Two of these segments were on LG1, approximately 17 cM apart, but their confidence intervals did not overlap, thus indicating their independence. The other segment was on LG6. Only the latter fell on a major gene predicted to be involved in curd formation in cauliflower, BoAP-a1 (Smith and King 2000; Purugganan et al. 2000). The peak of this QTL is located at 57.19 cM and the BoAP1 sequence is located at 56.8 cM on LG6, which is in agreement with previous reports indicating that this gene plays a role in inflorescence architecture. We did not find any association with the BoCAL-a sequence, which is another predicted gene of similar function (Smith and King 2000; Purugganan et al. 2000). Thus, our results agree with those of Labate et al. (2006), who found that the BoCAL-a gene actually provides very little contribution to the cauliflower phenotype. Additionally, Lan and Paterson (2000) detected at least 67 loci, distinguishing broccoli from cauliflower in a much more exhaustive analysis, including not only of curd morphology, but also of size, shape and other related traits. Thus, it is clear from Lan and Paterson (2000), Labate et al. (2006) and from our study that additional genes must be involved in curd development.