Brassica napus (rapeseed, oilseed rape, Canola) is an oilseed crop of global economic significance. Over 50 million tonnes of rapeseed were produced in 2007 from an area of 30 million hectares [1]. In addition both tuberous (swede or rutabaga) and leafy forms (fodder rape and kale) of the species are grown as vegetables for human consumption and animal fodder. Oilseed B. napus has only achieved economic importance in the past forty years following an intensive breeding programme to decrease nutritionally undesirable components of the oil and meal, and to increase yields. Attention initially focused on reducing levels of erucic acid in the seed oil, and then reducing levels of aliphatic glucosinolate in the meal to make it more palatable and safer for livestock. More recently, varieties yielding oils suitable for conversion to biodiesel and industrial lubricants have been developed. As with other crops, ongoing breeding programmes aim to increase overall harvestable yield and quality, with resistance to crop pests and pathogens as major targets. Whilst successful, the collateral effect of these improvements has been the production of elite varieties that possess only a fraction of the genetic diversity available in the wider Brassica genepools. This is causing increasing concern, particularly with respect to lack of resistance to insect and other pests. Sources of new alleles that can easily be transferred into elite breeding lines are required in order to maintain and increase yield, provide new functional and adaptive disease resistance loci, and refine oil qualities to serve a variety of nutritional and industrial purposes.

The relationships between the six major cultivated Brassica species were originally described by U [2], who associated the diploid species B. rapa, B. oleracea and B. nigra L. with the amphiploids B. juncea L., B. carinata A. Br. and B. napus. Each of the amphiploids contains a combination of two diploid genomes. B. napus (n = 19) contains both the A and C genomes of its two progenitors, B. rapa (A genome, n = 10) and B. oleracea and related wild species (C genome, n = 9). Recreation of stable B. napus by crossing B. rapa and B. oleracea is difficult, but possible. Re-synthesized B. napus may be generated by crossing both diploid and tetraploid B. rapa and B. oleracea, although only a very small proportion of the attempted pollinations result in a viable hybrid plant [3]. In a current research context, the production of hybrids from diploid parents is commonly enhanced through embryo rescue or somatic hybridisation [4] although such plants often have a relatively high frequency of chromosomal rearrangements and reduced fertility [5, 3].

The nature, direction and geographic location of the initial hybridisation events that led to the generation of B. napus remain unclear. B. napus is thought to be a relatively new species, since the earliest reliable documented record appears only 500 years ago, and although feral populations are common, no truly wild populations have been recorded ([6] and references therein). The location or locations of the original hybridisations is also unclear, as is whether they occurred in a wild or domesticated context. Both B. rapa and C genome species (particularly B. oleracea) have wide geographic ranges and geographically distinct centres of diversity. The earliest molecular studies suggested that the maternal parent of B. napus was likely to be B. oleracea, due to similarities in restriction patterns of their chloroplast genomes [7]. Subsequent RFLP analysis indicated that B. montana, a C genome relative of B. oleracea, had an identical chloroplast type to B. napus, and supported the contention that the maternal parent was not A genome B. rapa [8].

The identities of the A and C genome subtaxa involved in the original hybridisation that led to the formation of B. napus are not known, although [9] suggested that the genotype of the original A genome parent could be closely related to that of a B. rapa accession 'Spring Broccoli Raab'. However, the authors noted that post speciation introgression of B. napus genome fragments into the B. rapa accession could also explain their findings. Other studies have detected introgression of B. rapa into different B. napus genotypes [10]. Evidence based on either chloroplast or nuclear markers has suggested that B. napus appears to have resulted from several independent hybridisation events [9, 8]. More recently, diversity in the Brassica chloroplast genome was characterised using nine microsatellite markers [11]. The study found 10 different haplotypes in the 15 B. napus individuals tested, although none of these haplotypes were shared by any other A or C genome species. Whilst of interest, such studies are restricted in value due to limited sampling and use of different marker systems that makes direct comparison or compilation impossible.

In order to establish a more robust basis for clarifying the origins of B. napus, and in particular to ascertain the species which was the likely maternal parent, we used both nuclear and chloroplast molecular markers. It was hoped that this approach would also provide baseline evidence to clarify the possible polyphyletic origins of the species. We carried out a detailed sampling of the genepools of the potential A and C genome donor species by surveying a total of 367 accessions representing 15 species. This included 94 B. napus accessions, together with 10 accessions of B. montana, the putative maternal ancestor of B. napus [8]. Representatives of B. nigra (B genome, n = 8) and the amphiploids B. carinata (BC genome, n = 17) and B. juncea (AB genome n = 18) were also included for comparison.

In total we sampled 198 accessions representing 6 Brassica species (B. napus, B. rapa, B. carinata, B. juncea, B. nigra and B. montana) in order to determine diversity using chloroplast SSRs. Data are directly comparable with the 171 samples of B. oleracea and related C genome species previously described in [12]. Where possible, we selected accessions that had already been used in other published studies, particularly with regard to the B. napus and B. rapa accessions used in [8]. The B. napus accessions represent all B. napus crop types, with an emphasis on oilseeds due to their prevalence in an agricultural setting (global area sown and opportunities for gene flow to populations of related wild species). DNA was extracted from young leaves of a single seedling or single seeds as described in [12]. Most accessions were represented by a single sample. Tests on three individuals of five different B. montana accessions failed to detect more than one haplotype per accession (data not shown). Although it is likely that intra-accession chloroplast polymorphism is present, particularly within wild accessions, this is compensated by the large number of accessions sampled. Six primer pairs were used to amplify chloroplast SSRs, with PCR products visualised and scored on an ABI 3100 Genetic Analyzer (Applied Biosystems) following the methods described in [12].

A subset of 93 accessions were analysed for nuclear genome diversity using AFLPs using the restriction enzymes EcoRI and MseI. This subset was chosen to be representative of the chloroplast haplotypes detected using the SSRs, whilst including a more thorough representation of particular groups of interest such as B. rapa brocoletto and B. montana. The number of accessions was limited to 93 in this analysis to avoid potential problems when combing data derived from different PCR reactions and to allow for control samples. Methods were based on those described by [13] except that we used a fluorescent detection system. A pre-selective step was carried out using the primers 5'-GACTGCGTACCAATTCA-3'and 5'-GATGAGTCCTGAGTAAC-3'which anneal to the EcoRI and MseI adapters respectively. Selective amplifications with two primer pairs were then carried out, the primer sequences being identical to the pre-selective pairs with the addition of EcoRI-AAC-3'/MseI-CAG-3'and EcoRI-AAG-3'/MseI-CAA-3'. The EcoRI selective primer was labelled with FAM at the 5'end. Fragments were sized using an ABI 3100 Genetic Analyzer (Applied Biosystems) with a Genescan Rox 500 internal size standard and then scored using GeneMarker (Softgenetics) software. We examined peak heights across each trace and any peaks with a height less than the mean were regarded as absent. Traces from samples which had an overall mean peak height <200 relative fluorescence units (rfu) were disregarded in order to ensure only high quality data were analysed. As a control, AFLP analysis was carried out on six replicates of the same B. nigra sample to assess the robustness of the data.

The diversity of B. rapa and B. napus was assessed using Nei's measure of gene diversity H [14]) based on the frequencies of the haplotypes present in each species. The AFLP data were analysed using Principal Coordinates Ordination (PCO) as implemented in the programme PAST (Øyvind Hammer and David Harper, available from http://folk.uio.no/ohammer/past/) using the DICE similarity metric. Nei's H was also calculated for B. rapa, B. oleracea and related C genome species, and B. napus using AFLP-SURV [15]. A Neighbour-Joining tree (Figure 1) based on the genetic distance measure of Link et al. [16] was constructed from the AFLP data using the software package TreeCon [17]. Support for the tree was assessed using 100 bootstrap replicates. Chloroplast haplotype data were also mapped onto this tree.

The chloroplast haplotypes found in B. napus effectively rule out most of the C genome species from the maternal lineage of nearly all of the B. napus samples tested. Only three samples (all kale types) had a haplotype commonly associated with B. oleracea. A small number of B. napus samples shared haplotype A:01 with both B. rapa and B. hilarionis. The most prevalent haplotype in B. napus (A:06) occurs elsewhere in only a small number of B. rapa samples. Three of these A:06 B. rapa accessions are from the UK and were collected from wild or weedy populations occurring within or alongside B. napus oilseed rape fields. The remaining two samples are 'brocoletto' types from Italy. There are two explanations for the co-occurrence of A:06 in B. rapa and B. napus. One is that the original donor of A:06 was not sampled in this study, and that any occurrence of A:06 in B. rapa is the result of recent or historical introgression from B. napus into B. rapa. An introgressed origin of the A:06 chloroplast is suggested as this haplotype is more common in UK populations which exist in close proximity to oilseed rape [18]. Additionally, B. napus was historically grown very widely in the UK in the 19^th Century as a fodder crop (swedes and turnips occupied at least 598 k ha in England in 1881; [19]), providing further opportunity for introgression.

We sampled most of the geographical and morphological diversity of B. rapa. Outside of the UK, haplotype A:06 only occurred in two of the eight brocoletto accessions tested. The AFLP analysis shows that these accessions are not recent hybrids with B. napus or B. oleracea since they cluster as expected with the rest of the B. rapa samples in the PCO plot and the NJ tree (figures 1 and 2). The AFLP analysis is sensitive to inter-specific hybrids as demonstrated by the single weedy B. rapa individual ('1' on Figure 2a - chloroplast haplotype A:04) which falls between the B. rapa and B. napus clusters in the PCO plot. We tested this B. rapa sample further with three nuclear CAPS (cleaved amplified polymorphic sequence) markers and three nuclear SSRs. These markers revealed the presence of both A and C genome alleles (data not shown). Interestingly, the three UK wild/weedy individuals with the A:06 haplotype do not have C genome markers as tested by the AFLPs. They also cluster with the rest of the B. rapa accessions. We conclude that either the introgression event must have been followed by many generations of back crossing to B. rapa, or that the C genome fragments were lost rapidly within a few generations.

The relatively recent origin of B. napus as a species is supported by the reduced diversity as determined by Nei's measure of gene diversity within B. napus, compared to the A and C genome ancestral genepools. It is difficult at present to be certain whether the original hybridisation event(s) involved more than one chloroplast haplotype. However, multiple hybridisation events are indicated by the presence of three different cytoplasms among the B. napus samples tested. Again, this could alternately be explained by post-speciation introgression. This is indeed the case for some spring oilseed types and fodder rape types where B. rapa and B. oleracea are documented to have been used in B. napus breeding programmes [3, 20]. A:01 and C:01 haplotypes also occur in non-oilseed B. napus crop types, namely rape kales and fodder kales, where crop improvement programmes may also at least be responsible in part for their presence. However, only two accessions of these kale and fodder types with A:01 or C:01 chloroplasts are classed as 'advanced cultivars' (i.e. resulting from a formal modern crop improvement programme). The remainder are 'traditional varieties or landraces' so are less likely to have been developed from deliberate inter-specific crossing. As with B. rapa, the AFLP analysis does not differentiate between the different cytoplasmic types of B. napus, with a single exception. A sample from the spring oilseed variety 'Comet' (A:01 haplotype) falls outside of the well defined cluster of B. napus samples, and in fact groups within the C genome cluster. The reasons for this are unclear and further testing would be required for confirmation of this result. The presence of haplotype A:01 in B. hilarionis as reported by [12] is intriguing and may shed light on the common ancestry of A and C genome species. Although it is possible that B. hilarionis represents a source of the A:01 chloroplast in B. napus this is unlikely since B. hilarionis is endemic to Cyprus.

The B. nigra sample included in the AFLP analysis provided both an outgroup for clustering analysis and a control. Overall reproducibility of the five replicates was 97.1% as only three bands out of a total of 102 were scored differently. The chloroplast haplotype (B:01) found in the three different B. nigra accessions is very distinct from those found in the A or the C genome, and the PCO analysis of AFLP data shows that the other Brassica samples tested are more similar to each other than they are to B. nigra. This in agreement with the findings of [21], in their investigation of the phylogeny of the Brassicaceae.

The tree constructed from the AFLP data shows good discrimination between B. napus, the C genome species, and B. rapa. However, intra-specific relationships remain mostly unclear. This is probably due to the relatively low marker to sample ratio (102:83). Future studies may improve resolution through the use of massively parallel sequencing technologies rather than the anonymous markers produced using AFLP. However, intra-specific relationships were not the primary focus of this study, and the PCO and NJ tree clearly indicate that intra-specific cytoplasmic differences are not always associated with whole genome diversity.

In contrast with [8], we did not find that any of the ten B. montana accessions tested shared a chloroplast haplotype with B. napus. We were not able to test exactly the same accession used by Song and Osborn as no further seed was available. A further 'B. montana' accession (not the same as that used by Song and Osborn) was originally included in our study. However subsequent taxonomic verification based on plant morphology revealed it to be incorrectly identified, and indeed it was very similar in appearance to B. napus. The AFLP data for this accession also indicated it was B. napus. Even though none of our B. montana accessions shared the A:06 chloroplast haplotype with B. napus, three of them did possess the next most closely related haplotype (C:06 - see [12]). It is possible that the RFLP markers used by [8] did not distinguish between these chloroplast genomes.

The B. oleracea chloroplast type has been detected previously in another B. napus accession, 'New Zealand Rawara' and this was proposed as further evidence to support a polyphyletic origin for B. napus [8]. We tested the ploidy level of one individual of this accession using flow cytometry, and discovered that it was in fact B. oleracea. However, three other verified B. napus samples in our study did contain the C:01 chloroplast common in B. oleracea. We did not find more than one of the B. rapa chloroplast haplotypes in B. napus, unlike [8] who detected two. Our sampling strategy was based on maximising the coverage of the A and C genome genepools by only testing one individual per accession. As with material derived from most ex situ genetic resource collections, many of the accessions are sampled from wild populations, local selections and open-pollinated varieties, and as such one would expect a degree of variation within accessions. In addition to cases of mis-identification as demonstrated above, there is always potential for apparent differences between studies to result from within-accession variation arising either from natural diversity or contamination of seed lots. We minimised these factors through using either ploidy analysis or by visual taxonomic confirmation of plants. Confirmation of the taxomomy of accessions will be facilitated in future by the ongoing efforts in genetic resource collections to provide online visual records of mature plants.

Multiple hybridisation events consistent with a polyphyletic origin were also indicated by the results of [9] who found that a sample of B. napus 'asparagus kale' differed in RFLP profile from other B. napus tested, suggesting an additional diploid parental genotype. We also found that five out of the six asparagus kale accessions in our study had the A:01 chloroplast haplotype typical of B. rapa. Most of these are traditional varieties and unlikely to have been selected through formal crop improvement. Such evidence suggests that B. napus may indeed have multiple origins. In addition, [9] also found that a B. rapa accession ('spring broccoli raab' - another name for the brocoletto crop type) shared a unique marker with the majority of B. napus samples in their study. This marker was absent from all other potential diploid progenitors, including B. oleracea. Interestingly, the spring broccoli raab sample tested was the only B. rapa to possess markers more commonly associated with the C genome. The authors suggest that the presence of (presumably introgressed) C genome fragments may have facilitated the inter-specific hybridisation which lead to the formation of a stable B. napus.

A recent study also based on chloroplast SSRs did not find any haplotypes in common between B. napus, B. oleracea and B. rapa [11]. Five varieties of B. napus were tested using nine SSRs and the authors detected eleven unique haplotypes, indicating that their nine markers detected a much higher level of intra-accession diversity than our six. Since the B. napus haplotypes were much more similar to those found in B. rapa than B. oleracea, the authors suggested that B. rapa was a much more likely maternal progenitor for B. napus than B. oleracea. This supports the findings of our study. However, as the authors indicated, SSR markers mutate at a relatively high rate, leading to the possibility of homoplasy and parallel origins of allele size. This, in addition to the hybridized origins of a significant portion of Brassica breeding material means that chloroplast SSRs alone may not provide sufficient information for conclusions to be drawn on maternal ancestry.

Artificial (re-synthesised) B. napus is known to undergo a relatively high frequency (compared to natural B. napus) of genomic rearrangements, including non-reciprocal translocations, due to pairing between homeologous chromosomes at meiosis [5]. Evidence has accumulated through several studies that a genetic factor regulating chromosome pairing is present in B. napus [22]. Control of chromosome pairing is required in order prevent the formation of unbalanced gametes and aneuploid progenies which reduced fertility. Identification of the closest extant relatives of the original A and C genome genotypes involved in the initial hybridisations leading to B. napus should allow closer investigations of these mechanisms and enable the resynthesis of a more meiotically stable artificial B. napus.

Our study indicates that it is highly unlikely that B. oleracea or any of the C genome species are closely related to the maternal progenitor of most B. napus accessions. The detection of two other chloroplast SSR haplotypes at low frequencies in B. napus does suggest that multiple hybridisation events involving different maternal ancestors may have occurred. However, the use of inter-specific hybrids (and re-synthesised B. napus) in modern crop improvement programmes is most likely responsible for some of the observed diversity. Natural post-speciation introgression (or chloroplast capture) is also a possibility since B. napus (A:06) chloroplasts are observed at a frequency of 0.12 in some B. rapa populations sympatric with oilseed rape fields [18]. B. napus samples with 'atypical' cytoplasm are not usually distinguishable from those harbouring the prevalent chloroplast haplotype in terms of nuclear genome diversity. Our results are consistent with those of [9] who suggested that B. rapa 'spring broccoli raab' may be the closest extant relative of the maternal ancestor of B. napus. Our study based on chloroplast haplotypes and 102 AFLP markers provides further support to their inference, based on 38 nuclear and 6 chloroplast RFLP probes, since the prevalent B. napus haplotype was also found in two additional accessions of the same crop type.

Given that no truly wild populations of B. napus have been documented, it seems reasonable to suggest that the initial hybridisations must have occurred in a cultivated context rather than a wild setting [6]. If the A:06 chloroplast haplotype detected in the two brocoletto accessions is not the result of chloroplast capture, then it is possible to envisage hybridisation occurring between brocoletto and B. oleracea crops growing in the same location. The brocoletto crop type originates from southern Italy, and a similar crop is also grown in Portugal. In both of these areas, it is highly likely that brocoletto would have been cultivated alongside B. oleracea crops such as kales, cabbages and broccolis, providing the necessary opportunities for inter-specific crosses to occur. Further work on the genetic diversity of the brocoletto crop type is required to verify such speculation.

Future genetic improvement of B. napus crops (e.g. focusing on abiotic stress tolerance, pest and disease resistance and other yield increases) will depend to a large degree on utilising the diversity present within the ancestral A and C genepools. It is clear that hybridisation between B. rapa and B. oleracea is very rare in nature, and knowing which genotypes of the parental species were involved will allow a greater understanding of the mechanisms and genetic factors controlling the creation of stable amphiploids, and this will facilitate the incorporation of novel alleles from the wider Brassica genepool.

Charlotte Allender is the Assistant Manager of the Genetic Resources Unit at Warwick HRI which maintains globally significant vegetable seed collections. Her research interests centre on the processes and partitioning of genetic diversity in crops and their wild relatives. Graham King has wide experience of quantitative genetics underlying crop trait improvement. His group at Rothamsted Research are involved in comparative genomics, epigenetics and developmental biology with a focus on seed development of oilseed brassicas. He leads the UK Oilseed Rape Genetic Improvement Network and his group host the http://www.brassica.info website.