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 19th 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.