A consensus map of rapeseed (Brassica napusL.) based on diversity array technology markers: applications in genetic dissection of qualitative and quantitative traits

Rapeseed (also known as canola and oilseed rape, Brassica napus L., 2n = 4× = 38; genomes AACC) is the second largest oilseed crop after soybean and provides a valuable rotational crop for farmers in many parts of the world. During the past 15 years, global oilseed rape production has doubled to >60 million tonnes (http://faostat.fao.org/; data sourced April 2012), due to the high demand for healthy vegetable oil, feedstock and a renewable source for the biodiesel industry. This amphidiploid Brassica species most likely originated as a result of repeated natural hybridization and genome doubling between the monogenomic diploid species Brassica rapa (2n = 2× = 20, genome AA) and Brassica oleracea (2n = 2× = 18, genome CC) along the Mediterranean coastline in Southern Europe, and was probably selected as an oilseed crop only 300–400 years ago [1–3].

Selection of useful variation in crop plants has been a major thrust of early farmers since the dawn of agriculture. In modern Brassica breeding programs, classical genetic analysis and molecular genetic approaches have been used to improve our understanding of the inheritance of various qualitative and quantitative loci and causative genes, to estimate their number, position and genetic effects, and to identify DNA-based markers associated with traits of agronomic importance [4].

Several linkage maps have been constructed from recombination data in B. napus mapping populations, which depict the distances between loci, as well as their order on a chromosome. These linkage maps have been based on a range of marker systems such as restriction fragment length polymorphisms (RFLPs), randomly amplified polymorphic DNAs (RAPD), amplified fragment length polymorphisms (AFLPs), simple sequence repeats (SSRs), sequence- tagged sites (STSs), sequence-related amplified polymorphisms (SRAPs) and single nucleotide polymorphisms (SNPs) [5–14]. These maps have been used to uncover qualitative and quantitative trait loci (QTLs) underlying traits of agricultural importance, for linkage disequilibrium assessment, candidate gene identification, marker-assisted selection, and genetic evolutionary studies [14–19]. The number of marker loci in these genetic linkage maps varied from 219 to 13,551, where the marker density was dependent upon the level of polymorphism between the parental lines of mapping populations and type of marker system used for detecting polymorphisms.

In order to increase marker density for predictive molecular breeding and map-based cloning of genes, dense consensus genetic maps have been developed for various crops including B. napus[14, 20, 21]. Such consensus maps are developed by mapping several common markers from each linkage group, in different mapping populations, and anchoring the linkage maps with consensus markers. However, most B. napus genetic maps have been based either upon low-throughput marker assays or on different marker systems, rendering them difficult to utilise to compare the map locations of markers and therefore their linkages with traits of interest across populations.

Diversity Array Technology (DArT) markers do not require prior DNA sequence information and can simultaneously genotype a large number of SNPs and insertion/deletion polymorphisms across the genome in a single assay, providing a low-cost option for genotyping [22]. The DArT marker system was first used in rice, and has since been extensively utilised for genetic applications such as genetic diversity analysis, construction of genetic linkage maps, and QTL/linkage analysis in different crops [23, 24]. Recently, a 3072-clone B. napus DArT array was used to generate a linkage map using a B. napus doubled haploid (DH) population derived from Lynx-/Monty-028DH [13]. This information was used to identify QTLs for flowering time, resistance to blackleg disease caused by Leptosphaeria maculans, and resistance to pod shatter [25–27].

Comparative mapping studies have shown extensive co-linearity between the genomes of Brassica and Arabidopsis, which diverged from a common ancestor approximately 20 MYA [28–30]. The genome of B. rapa has recently been sequenced [31], whilst the genomes of B. oleracea and B. napus are currently being annotated and analysed and are expected to become publicly available in 2013. These advances in structural and functional genomics have provided us with an opportunity to align the genetic DArT markers with the sequenced genomes of B. rapa and B. oleracea. This allows the identification of markers for marker-assisted selection in rapeseed improvement programs.

In the present study, we constructed a consensus map of rapeseed based on DArT marker datasets produced using six DH B. napus populations. These populations were derived from Ag-Castle/Topas (AT), BLN2762/Surpass400 (BS), Lynx-037DH/Monty-028DH (LM), Maxol*1/Westar-10 (MW), Skipton/Ag-Spectrum (SAS), and Tapidor/Ningyou7 (TN). The parents of the mapping populations originated from Australia, Canada, China and Europe, and are currently being used internationally in rapeseed germplasm enhancement programs. We physically mapped the B. napus DArT sequences using genome sequences from scaffolds of B. rapa and B. oleracea. DArT marker loci associated with agronomic traits, including various components of phenology, plant architecture, seed and oil quality, nutrient uptake and mobilisation, and resistance to blackleg disease and abiotic stresses were also identified.

In this study, we developed the first consensus genetic map of the rapeseed genome based on DArT markers. The development of this consensus map will provide a platform to compare chromosomal locations of markers across rapeseed populations and facilitate identification of simple and complex inherited trait-marker associations, comprehensive assessment of genetic diversity, and whole genome selection in rapeseed breeding programs. The consensus map consisted of 1,359 markers spanning all 19 B. napus chromosomes, with an average marker density of one marker per 1.46 cM. This corresponds to approximately 880 Mbp of the B. napus genome, 94% of the mapped AC genome was contained within intervals <10 cM. These observations suggest that this consensus map will be suitable for various applications including detection of quantitative traits in rapeseed improvement programs, as in most QTL analyses a 10-cM interval between marker loci is commonly used for regression analysis. This consensus map based on the sequenced DArT markers will also allow the positional cloning of the causative genes controlling phenotypic variation, estimation of linkage disequilibrium at the individual chromosome/genome level, and identification of genomic rearrangements such as translocations that occurred during the period between the genome triplication of Brassica and the divergence of B. rapa and B. oleracea and their distant relative model plant, Arabidopsis.

Different marker systems were used in each population, for example, the LM population was mapped with DArT markers, it had the highest density of SSR markers. These SSR markers (developed by the Agiculture and Agri-Food Canada Consortium) were not used in any other population undertaken in this study except for the TN population. The TN population was also mapped with a variety of markers based on SNPs and CS, which were not used in any other population. Furthermore, DArT and SSR markers amplify intra- and inter-homoalleles as a result of chromosomal duplications between homoeologous sequences among subgenomes and between paralogous duplicated sequences and other rearrangements such as in the TN (Additional files 4 and 11) and SAS populations [35, 48], which makes it difficult to compare individual component maps. Of the DArT clones that were useful for mapping across the six segregating populations in this study, relatively few loci were common between populations. This could be due to utilisation of a range of DH populations derived from genetically diverse parental lines used in order to map the DArT clones across the B. napus genome. For example, the cultivar Surpass 400 is derived from several backcrosses into cultivated B. napus from B. rapa ssp. sylvestris. The introgression of chromosomal segments from B. rapa into Ningyou7 – one of the parental lines of the TNDH population - has also been recently reported [49]. We could not ascertain the map positions of some DArT loci or their order on certain chromosomes (for example in the MW population, LG I to VI, Additional file 1) due to the absence of shared markers among mapping populations. However, marker order was generally consistent within some chromosomes where more than four shared markers were present for assessment of statistical significance [21] and there was a global conservation of co-linearity between the corresponding linkage groups of integrated maps and the A (B. rapa) and C (B. oleracea) genome scaffolds. On some chromosomes no polymorphism was found between the parental lines of the DH populations, for example on chromosome C8 in the populations MW and BS. These findings suggest that the DArT sequences in the genomic representations were conserved.

The results presented here demonstrate that the consensus map was longer (1,987.2 cM) than the six individual maps with the exception of the LM population. The LM map contained a higher proportion of distorted and distinct loci [13], than most of the populations. The recombination frequency can also vary with genetic background [48]. In addition, scoring errors tend to be higher with SSR markers which show extensive stuttering and multiple bands (alleles), as compared to DArT markers. These biological and non-biological factors could have skewed the map estimates.

In general, the DArT markers were well-distributed across the genome, however, certain chromosomal regions showed extensive clustering such as on chromosomes A9 and C2 (Additional file 4). This may indicate the presence of gene-rich regions and uneven distribution of recombination events along chromosomes, or it may suggest that the DArT markers are preferentially surveying DNA polymorphisms that are unevenly distributed along chromosomes. Similar results have been reported in other crops including in B. napus[14, 49, 50]. Uneven distribution of DArT markers between the A and C genomes may indicate genome evolution events of B. napus from its diploid progenitor species or uneven representation of DArT clones from source species used in the array development. The DArT markers derived from genomic representations are known to introduce some degree of redundancy [22]. However, in this study, genomic representations were prepared using the CNG methylation sensitive restriction enzyme PstI, which generates low and single copy DNA fractions in plants.

In comparison to the non-DArT markers (11.4%), a low proportion (2.27%) of multi-locus DArT markers was found. Hybridisation-based markers (such as DArT) select against multi-locus markers, because hybridisation intensities contributed by different loci are difficult to resolve in the DArT allele calling process and such markers are normally scored as monomorphic. A similar frequency (~1.8%) of multi-locus DArT markers has been reported in rye, sorghum, and barley [22, 50, 51]. Existence of multi-locus markers as a result of DNA (DArT) hybridisation from orthologous and non-homologous regions of rapeseed, and multiple hits between DArT sequences and the sequenced scaffolds of B. rapa and B. oleracea, provide strong evidence for the presence of intra- and inter-chromosomal duplicated loci in B. napus. Previous studies have reported that most of the approximately 1.2 Gb genome of B. napus[52] comprises genome sequences from the two progenitor species, which exhibit significant co-linearity. However, homoeologous recombination plays a major role in chromosome rearrangements, such as duplications and reciprocal translocations [53–57]. In certain cases, the physical position of DArT markers on the B. rapa and B. oleracea scaffolds did not correspond with their genetic positions on the linkage maps of segregating populations as reported previously [58]. This could have been caused by large translocations within the mapping populations, mapping inaccuracies and/or errors in assembling genome scaffolds. However, there was a good correspondence between genetic ordering and the current genome sequence assemblies. Approximately 20% of the DArT marker sequences did not have a significant match to B. rapa and B. oleracea. This may be due to the incomplete genome coverage in the current genomic scaffolds or some of them may represent novel loci that may have evolved in B. napus as a result of inter-specific recombination between B. rapa and B. oleracea. However, a more likely explanation is biological (structural) variation among ancestral genotypes involved in the development of B. napus cultivars used for mapping, as compared to B. rapa and B. oleracea genotypes used for sequencing the A and C genomes [58]. It is also possible that the reference accessions used to build current A and C genome scaffold assemblies lack some genomic regions that are present in the wider gene-pools.

Discrepancies such as inversions and translocations were identified when comparing the consensus map and the individual component maps, as have been reported previously in B. napus[59] and in comparisons of the A genome in B. napus and B. rapa[58]. Apart from genome rearrangements, this could be due to the smaller size of the populations used here [60], and variation in the number of recombination events in different regions of the plant genomes [61].

A recent study has shown good co-linearity of DArT and non-DArT markers between the genetic linkage map of the SAS population and the assembled pseudomolecules of B. rapa[27]. As the sequenced genome scaffolds of other A and C genome Brassica species become available in the public domain, DArT sequences (available on http://www.diversityarrays.com/) can be used for comparative genomic analysis, identification of QTLs for predictive breeding (as shown in this study) and for the identification of candidate genes linked to traits of agricultural importance.

In this study, we related DArT markers with the majority of QTLs that have been identified in six breeding populations, however this can be extended to other traits and populations of B. napus, as well as to related A and C genome Brassica species. For example, several DArT markers were related to the positions of QTLs associated with flowering and quality components in the TN and SAS populations [9, 15, 27, 37, 40]. Two QTLs for flowering time were mapped in the genomic regions harbouring homoeologues of FLOWERING TIME LOCUS C of Arabidopsis, BnFLC.A3 on chromosomes A3 and Bn.FLC.A10 on chromosome A10 [15, 27, 40]. In the LM population, we identified four QTLs associated with 16:0, 18:1, 18:2 and 18:3 fatty acids on chromosomes A1, A5 and C4 (Table 4). Among them, two major QTLs associated with 18:1 (oleic acid) and 18:2 (linoleic acid) were mapped on A5 within a region containing FAD2 gene (Table 4). Localisation of these QTLs suggested that FAD2 controls variation for both oleic acid and linoleic acid content in this population. The FAD2 of the endoplasmic reticulum encodes ω-6 desaturases which is responsible for conversion of oleic acid to linoleic acid by inserting a double bond at the ω-6 position [62].

The parents of mapping populations used in this study such as Ag-Castle, Maxol, Monty, Skipton, Ag-Spectrum, and Surpass 400 are known to harbour resistance both at the seedling (race-specific, qualitative) and adult plant stages (race-specific and/or race-non-specific) to L. maculans[26, 32, 63]. DArT markers were mapped in the vicinity of a genomic region associated with Rlm1, Rlm3, and Rlm4 genes for resistance to L. maculans on the chromosome A7 [32, 33]. It was interesting to note that some of the loci associated with correlated traits such as seed yield, pod number, seed weight, flowering time, seed number and plant height were localised on the same genomic regions (A1, A2, A3, A7, C3 and C6) on the consensus map, which may have been due to pleiotropic effects of a QTL on different traits and/or tight linkage between multiple traits and QTLs [64].

We constructed a high-density consensus map of B. napus utilising 1,359 DArT and non-DArT based markers. This consensus map was useful in locating loci associated with various traits of agronomic importance. The sequences of the DArT clones are publicly available (http://www.diversityarrays.com/) and provide a valuable resource in predicted genomic breeding, map-based gene cloning, and comparative analysis studies of A and C genomes Brassica species. The development of the integrated consensus map described in conjunction with the physical locations of B. napus DArT markers on the genome scaffolds, and the identification of molecular markers flanking genomic regions associated with agronomic traits, will empower rapeseed breeding programs to identify candidate/causative genes controlling genetic variation for such traits. This will enhance selection efficiency, especially of quantitative traits governed by a large number of genes and influenced with G × E interaction via marker-assisted selection.